Release minimap2-2.29 (r1283)

This speeds up alignment a lot

.github/workflows/ci.yaml

@@ -7,7 +7,8 @@ on:
   pull_request:
 
 jobs:
-  build:
+  build-linux-x8664:
+    name: Linux x86_64
     runs-on: ubuntu-latest
     strategy:
       matrix:
@@ -15,7 +16,53 @@ jobs:
 
     steps:
     - name: Checkout minimap2
-      uses: actions/checkout@v2
+      uses: actions/checkout@v4
 
     - name: Compile with ${{ matrix.compiler }}
-      run:  make CC=${{ matrix.compiler }}
+      run: |
+        make CC=${{ matrix.compiler }}
+        file minimap2 | grep x86-64
+
+  build-linux-aarch64:
+    name: Linux aarch64
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        compiler: [gcc]
+
+    steps:
+    - name: Checkout
+      uses: actions/checkout@v4
+
+    - name: Compile with ${{ matrix.compiler }}
+      uses: uraimo/run-on-arch-action@v2
+      with:
+        arch: aarch64
+        distro: ubuntu22.04
+        githubToken: ${{ github.token }}
+        dockerRunArgs: |
+          --volume "${PWD}:/minimap2"
+        install: |
+          apt-get update -q -y
+          apt-get install -q -y make ${{ matrix.compiler }} zlib1g-dev file
+        run: |
+          cd /minimap2
+          make CC=${{ matrix.compiler }} arm_neon=1 aarch64=1 -j
+          file minimap2 | grep aarch64
+
+  build-mac-arm64:
+    name: Mac ARM64
+    runs-on: macos-14
+    strategy:
+      matrix:
+        compiler: [clang]
+
+    steps:
+    - name: Checkout minimap2
+      uses: actions/checkout@v4
+
+    - name: Compile with ${{ matrix.compiler }}
+      run: |
+        make CC=${{ matrix.compiler }} arm_neon=1 aarch64=1 -j
+        file minimap2 | grep arm64
+        

.travis.yml

@@ -1,24 +0,0 @@
-matrix:
-  include:
-    - language: c
-      compiler: gcc
-      script: make
-    - language: c
-      compiler: clang
-      script: make
-    - arch: arm64
-      language: c
-      compiler: gcc
-      script: make arm_neon=1 aarch64=1
-    - language: python
-      python: "2.7"
-      before_install: pip install cython
-      script: python setup.py build_ext
-    - language: python
-      python: "3.5"
-      before_install: pip install cython
-      script: python setup.py build_ext
-    - language: python
-      python: "3.9"
-      before_install: pip install cython
-      script: python setup.py build_ext

FAQ.md

@@ -2,7 +2,7 @@
 
 Without `-a`, `-c` or `--cs`, minimap2 only finds *approximate* mapping
 locations without detailed base alignment. In particular, the start and end
-positions of the alignment are impricise. With one of those options, minimap2
+positions of the alignment are imprecise. With one of those options, minimap2
 will perform base alignment, which is generally more accurate but is much
 slower.
 

Makefile

@@ -2,12 +2,16 @@ CFLAGS=		-g -Wall -O2 -Wc++-compat #-Wextra
 CPPFLAGS=	-DHAVE_KALLOC
 INCLUDES=
 OBJS=		kthread.o kalloc.o misc.o bseq.o sketch.o sdust.o options.o index.o \
-			lchain.o align.o hit.o seed.o map.o format.o pe.o esterr.o splitidx.o \
+			lchain.o align.o hit.o seed.o jump.o map.o format.o pe.o esterr.o splitidx.o \
 			ksw2_ll_sse.o
 PROG=		minimap2
 PROG_EXTRA=	sdust minimap2-lite
 LIBS=		-lm -lz -lpthread
 
+ifneq ($(aarch64),)
+	arm_neon=1
+endif
+
 ifeq ($(arm_neon),) # if arm_neon is not defined
 ifeq ($(sse2only),) # if sse2only is not defined
 	OBJS+=ksw2_extz2_sse41.o ksw2_extd2_sse41.o ksw2_exts2_sse41.o ksw2_extz2_sse2.o ksw2_extd2_sse2.o ksw2_exts2_sse2.o ksw2_dispatch.o
@@ -26,12 +30,12 @@ endif
 
 ifneq ($(asan),)
 	CFLAGS+=-fsanitize=address
-	LIBS+=-fsanitize=address
+	LIBS+=-fsanitize=address -ldl
 endif
 
 ifneq ($(tsan),)
 	CFLAGS+=-fsanitize=thread
-	LIBS+=-fsanitize=thread
+	LIBS+=-fsanitize=thread -ldl
 endif
 
 .PHONY:all extra clean depend
@@ -111,8 +115,9 @@ esterr.o: mmpriv.h minimap.h bseq.h kseq.h
 example.o: minimap.h kseq.h
 format.o: kalloc.h mmpriv.h minimap.h bseq.h kseq.h
 hit.o: mmpriv.h minimap.h bseq.h kseq.h kalloc.h khash.h
-index.o: kthread.h bseq.h minimap.h mmpriv.h kseq.h kvec.h kalloc.h khash.h
-index.o: ksort.h
+index.o: kthread.h bseq.h minimap.h mmpriv.h kseq.h ksw2.h kalloc.h kvec.h
+index.o: khash.h ksort.h
+jump.o: mmpriv.h minimap.h bseq.h kseq.h
 kalloc.o: kalloc.h
 ksw2_extd2_sse.o: ksw2.h kalloc.h
 ksw2_exts2_sse.o: ksw2.h kalloc.h

NEWS.md

@@ -1,3 +1,170 @@
+Release 2.29-r1283 (18 April 2025)
+----------------------------------
+
+Notable changes to minimap2:
+
+ * New feature: added the `splice:sr` preset for short RNA-seq read alignment.
+   Users may use `-j` to specify known gene annotation to improve spliced
+   alignment close to the ends of short reads. Also added `--write-junc` and
+   `--pass1` for 2-pass short-read RNA-seq alignment.
+
+ * Experimental feature: read splice scores from a file specified by `--spsc`
+   and consider the scores during base alignment. The feature makes it possible
+   to apply advanced splice models and to improve spliced alignment.
+
+ * Change: adjusted the mapping quality calculation for spliced alignment.
+
+ * Bugfixes: a) missing overlap alignment when base alignment is requested
+   (#969); b) incorrect summary information for long genomes (#1192); c)
+   missing parameter check for `--score-N` (#1226).
+
+ * Improvement: a) warn about absent junction files (#1229); b) report an error
+   if a wrong preset prefixed with "splice" is specified (#589).
+
+Notable changes to mappy:
+
+ * Improvement: allow passing read name (#1260)
+
+ * Improvement: exposed score for ambiguous bases (#1240)
+
+Minimap2 now supports short/long genomic/RNA-seq read alignment along with
+contig alignment and all-vs-all read overlapping. It produces identical genomic
+long-read or contig alignment to v2.27. Short genomic read alignment and the
+mapping quality of long RNA-seq read alignment may slightly differ in very rare
+cases.
+
+(2.29: 18 April 2025, r1283)
+
+
+
+Release 2.28-r1209 (27 March 2024)
+----------------------------------
+
+Notable changes to minimap2:
+
+ * Bugfix: `--MD` was not working properly due to the addition of `--ds` in the
+   last release (#1181 and #1182).
+
+ * New feature: added an experimental preset `lq:hqae` for aligning accurate
+   long reads back to their assembly. It has been observed that `map-hifi` and
+   `lr:hq` may produce many wrong alignments around centromeres when accurate
+   long reads (PacBio HiFi or Nanopore duplex/Q20+) are mapped to a diploid
+   assembly constructed from them. This new preset produces much more accurate
+   alignment. It is still experimental and may be subjective to changes in
+   future.
+
+ * Change: reduced the default `--cap-kalloc` to 500m to lower the peak
+   memory consumption (#855).
+
+Notable changes to mappy:
+
+ * Bugfix: mappy option struct was out of sync with minimap2 (#1177).
+
+Minimap2 should output identical alignments to v2.27.
+
+(2.28: 27 March 2024, r1209)
+
+
+
+Release 2.27-r1193 (12 March 2024)
+----------------------------------
+
+Notable changes to minimap2:
+
+ * New feature: added the `lr:hq` preset for accurate long reads at ~1% error
+   rate. This was suggested by Oxford Nanopore developers (#1127). It is not
+   clear if this preset also works well for PacBio HiFi reads.
+
+ * New feature: added the `map-iclr` preset for Illumina Complete Long Reads
+   (#1069), provided by Illumina developers.
+
+ * New feature: added option `-b` to specify mismatch penalty for base
+   transitions (i.e. A-to-G or C-to-T changes).
+
+ * New feature: added option `--ds` to generate a new `ds:Z` tag that
+   indicates uncertainty in INDEL positions. It is an extension to `cs`. The
+   `mgutils-es6.js` script in minigraph parses `ds`.
+
+ * Bugfix: avoided a NULL pointer dereference (#1154). This would not have an
+   effect on most systems but would still be good to fix.
+
+ * Bugfix: reverted the value of `ms:i` to pre-2.22 versions (#1146). This was
+   an oversight. See fcd4df2 for details.
+
+Notable changes to paftools.js and mappy:
+
+ * New feature: expose `bw_long` to mappy's Aligner class (#1124).
+
+ * Bugfix: fixed several compatibility issues with k8 v1.0 (#1161 and #1166).
+   Subcommands "call", "pbsim2fq" and "mason2fq" were not working with v1.0.
+
+Minimap2 should output identical alignments to v2.26, except the ms tag.
+
+(2.27: 12 March 2024, r1193)
+
+
+
+Release 2.26-r1175 (29 April 2023)
+----------------------------------
+
+Fixed the broken Python package. This is the only change.
+
+(2.26: 25 April 2023, r1173)
+
+
+
+Release 2.25-r1173 (25 April 2023)
+----------------------------------
+
+Notable changes:
+
+ * Improvement: use the miniprot splice model for RNA-seq alignment by default.
+   This model considers non-GT-AG splice sites and leads to slightly higher
+   (<0.1%) accuracy and sensitivity on real human data.
+
+ * Change: increased the default `-I` to `8G` such that minimap2 would create a
+   uni-part index for a pair of mammalian genomes. This change may increase the
+   memory for all-vs-all read overlap alignment given large datasets.
+
+ * New feature: output the sequences in secondary alignments with option
+   `--secondary-seq` (#687).
+
+ * Bugfix: --rmq was not parsed correctly (#1010)
+
+ * Bugfix: possibly incorrect coordinate when applying end bonus to the target
+   sequence (#1025). This is a ksw2 bug. It does not affect minimap2 as
+   minimap2 is not using the affected feature.
+
+ * Improvement: incorporated several changes for better compatibility with
+   Windows (#1051) and for minimap2 integration at Oxford Nanopore Technologies
+   (#1048 and #1033).
+
+ * Improvement: output the HD-line in SAM output (#1019).
+
+ * Improvement: check minimap2 index file in mappy to prevent segmentation
+   fault for certain indices (#1008).
+
+For genomic sequences, minimap2 should give identical output to v2.24.
+Long-read RNA-seq alignment may occasionally differ from previous versions.
+
+(2.25: 25 April 2023, r1173)
+
+
+
+Release 2.24-r1122 (26 December 2021)
+-------------------------------------
+
+This release improves alignment around long poorly aligned regions. Older
+minimap2 may chain through such regions in rare cases which may result in
+missing alignments later. The issue has become worse since the the change of
+the chaining algorithm in v2.19. v2.23 implements an incomplete remedy. This
+release provides a better solution with a X-drop-like heuristic and by enabling
+two-bandwidth chaining in the assembly mode.
+
+(2.24: 26 December 2021, r1122)
+
+
+
 Release 2.23-r1111 (18 November 2021)
 -------------------------------------
 

README.md

@@ -14,13 +14,15 @@ cd minimap2 && make
 # use presets (no test data)
 ./minimap2 -ax map-pb ref.fa pacbio.fq.gz > aln.sam       # PacBio CLR genomic reads
 ./minimap2 -ax map-ont ref.fa ont.fq.gz > aln.sam         # Oxford Nanopore genomic reads
-./minimap2 -ax map-hifi ref.fa pacbio-ccs.fq.gz > aln.sam # PacBio HiFi/CCS genomic reads (v2.19 or later)
-./minimap2 -ax asm20 ref.fa pacbio-ccs.fq.gz > aln.sam    # PacBio HiFi/CCS genomic reads (v2.18 or earlier)
+./minimap2 -ax map-hifi ref.fa pacbio-ccs.fq.gz > aln.sam # PacBio HiFi/CCS genomic reads (v2.19+)
+./minimap2 -ax lr:hq ref.fa ont-Q20.fq.gz > aln.sam       # Nanopore Q20 genomic reads (v2.27+)
 ./minimap2 -ax sr ref.fa read1.fa read2.fa > aln.sam      # short genomic paired-end reads
 ./minimap2 -ax splice ref.fa rna-reads.fa > aln.sam       # spliced long reads (strand unknown)
-./minimap2 -ax splice -uf -k14 ref.fa reads.fa > aln.sam  # noisy Nanopore Direct RNA-seq
-./minimap2 -ax splice:hq -uf ref.fa query.fa > aln.sam    # Final PacBio Iso-seq or traditional cDNA
-./minimap2 -ax splice --junc-bed anno.bed12 ref.fa query.fa > aln.sam  # prioritize on annotated junctions
+./minimap2 -ax splice -uf -k14 ref.fa reads.fa > aln.sam  # noisy Nanopore direct RNA-seq
+./minimap2 -ax splice:hq -uf ref.fa query.fa > aln.sam    # PacBio Kinnex/Iso-seq (RNA-seq)
+./minimap2 -ax splice --junc-bed=anno.bed12 ref.fa query.fa > aln.sam  # use annotated junctions
+./minimap2 -ax splice:sr ref.fa r1.fq r2.fq > aln.sam     # short-read RNA-seq (v2.29+)
+./minimap2 -ax splice:sr -j anno.bed12 ref.fa r1.fq r2.fq > aln.sam
 ./minimap2 -cx asm5 asm1.fa asm2.fa > aln.paf             # intra-species asm-to-asm alignment
 ./minimap2 -x ava-pb reads.fa reads.fa > overlaps.paf     # PacBio read overlap
 ./minimap2 -x ava-ont reads.fa reads.fa > overlaps.paf    # Nanopore read overlap
@@ -38,7 +40,8 @@ man ./minimap2.1
     - [Map long noisy genomic reads](#map-long-genomic)
     - [Map long mRNA/cDNA reads](#map-long-splice)
     - [Find overlaps between long reads](#long-overlap)
-    - [Map short accurate genomic reads](#short-genomic)
+    - [Map short genomic reads](#short-genomic)
+    - [Map short RNA-seq reads](#short-rna-seq)
     - [Full genome/assembly alignment](#full-genome)
   - [Advanced features](#advanced)
     - [Working with >65535 CIGAR operations](#long-cigar)
@@ -74,8 +77,8 @@ Detailed evaluations are available from the [minimap2 paper][doi] or the
 Minimap2 is optimized for x86-64 CPUs. You can acquire precompiled binaries from
 the [release page][release] with:
 ```sh
-curl -L https://github.com/lh3/minimap2/releases/download/v2.23/minimap2-2.23_x64-linux.tar.bz2 | tar -jxvf -
-./minimap2-2.23_x64-linux/minimap2
+curl -L https://github.com/lh3/minimap2/releases/download/v2.29/minimap2-2.29_x64-linux.tar.bz2 | tar -jxvf -
+./minimap2-2.29_x64-linux/minimap2
 ```
 If you want to compile from the source, you need to have a C compiler, GNU make
 and zlib development files installed. Then type `make` in the source code
@@ -139,12 +142,15 @@ parameters at the same time. The default setting is the same as `map-ont`.
 ```sh
 minimap2 -ax map-pb  ref.fa pacbio-reads.fq > aln.sam   # for PacBio CLR reads
 minimap2 -ax map-ont ref.fa ont-reads.fq > aln.sam      # for Oxford Nanopore reads
+minimap2 -ax map-iclr ref.fa iclr-reads.fq > aln.sam    # for Illumina Complete Long Reads
 ```
 The difference between `map-pb` and `map-ont` is that `map-pb` uses
 homopolymer-compressed (HPC) minimizers as seeds, while `map-ont` uses ordinary
-minimizers as seeds. Emperical evaluation suggests HPC minimizers improve
+minimizers as seeds. Empirical evaluation suggests HPC minimizers improve
 performance and sensitivity when aligning PacBio CLR reads, but hurt when aligning
-Nanopore reads.
+Nanopore reads. `map-iclr` uses an adjusted alignment scoring matrix that
+accounts for the low overall error rate in the reads, with transversion errors
+being less frequent than transitions.
 
 #### <a name="map-long-splice"></a>Map long mRNA/cDNA reads
 
@@ -168,9 +174,8 @@ or the last exons.
 
 Minimap2 rates an alignment by the score of the max-scoring sub-segment,
 *excluding* introns, and marks the best alignment as primary in SAM. When a
-spliced gene also has unspliced pseudogenes, minimap2 does not intentionally
-prefer spliced alignment, though in practice it more often marks the spliced
-alignment as the primary. By default, minimap2 outputs up to five secondary
+spliced gene also has unspliced pseudogenes, minimap2 slightly prefers
+the spliced alignment. By default, minimap2 outputs up to five secondary
 alignments (i.e. likely pseudogenes in the context of RNA-seq mapping). This
 can be tuned with option **-N**.
 
@@ -201,6 +206,10 @@ bonus score (tuned by `--junc-bonus`) if an aligned junction matches a junction
 in the annotation. Option `--junc-bed` also takes 5-column BED, including the
 strand field. In this case, each line indicates an oriented junction.
 
+**Note:** `--junc-bed` is intended for long noisy RNA-seq reads only.
+Applying the option to short RNA-seq reads would increase run time with little
+improvement to junction accuracy.
+
 #### <a name="long-overlap"></a>Find overlaps between long reads
 
 ```sh
@@ -213,7 +222,7 @@ the overlapping mode because it is slow and may produce false positive
 overlaps. However, if performance is not a concern, you may try to add `-a` or
 `-c` anyway.
 
-#### <a name="short-genomic"></a>Map short accurate genomic reads
+#### <a name="short-genomic"></a>Map short genomic reads
 
 ```sh
 minimap2 -ax sr ref.fa reads-se.fq > aln.sam           # single-end alignment
@@ -226,8 +235,18 @@ be paired if they are adjacent in the input stream and have the same name (with
 the `/[0-9]` suffix trimmed if present). Single- and paired-end reads can be
 mixed.
 
-Minimap2 does not work well with short spliced reads. There are many capable
-RNA-seq mappers for short reads.
+#### <a name="short-rna-seq"></a>Map short RNA-seq reads
+
+```sh
+minimap2 -ax splice:sr ref.fa reads-se.fq.gz > aln.sam           # single-end
+minimap2 -ax splice:sr ref.fa r1.fq.gz r2.fq.gz > aln.sam        # paired-end
+minimap2 -ax splice:sr -j anno.bed ref.fa r1.fq r2.fq > aln.sam  # use annotation
+# 2-pass alignment
+minimap2 -x splice:sr -j anno.bed --write-junc ref.fa r1.fq r2.fq > junc.bed
+minimap2 -ax splice:sr -j anno.bed --pass1=junc.bed ref.fa r1.fq r2.fq > aln.sam
+```
+The new preset `splice:sr` was added in v2.29. It functions similarly to `sr`
+except that it performs spliced alignment.
 
 #### <a name="full-genome"></a>Full genome/assembly alignment
 
@@ -350,6 +369,11 @@ If you use minimap2 in your work, please cite:
 > Li, H. (2018). Minimap2: pairwise alignment for nucleotide sequences.
 > *Bioinformatics*, **34**:3094-3100. [doi:10.1093/bioinformatics/bty191][doi]
 
+and/or:
+
+> Li, H. (2021). New strategies to improve minimap2 alignment accuracy.
+> *Bioinformatics*, **37**:4572-4574. [doi:10.1093/bioinformatics/btab705][doi2]
+
 ## <a name="dguide"></a>Developers' Guide
 
 Minimap2 is not only a command line tool, but also a programming library.
@@ -399,5 +423,6 @@ mappy` or [from BioConda][mappyconda] via `conda install -c bioconda mappy`.
 [manpage]: https://lh3.github.io/minimap2/minimap2.html
 [manpage-cs]: https://lh3.github.io/minimap2/minimap2.html#10
 [doi]: https://doi.org/10.1093/bioinformatics/bty191
-[smide]: https://github.com/nemequ/simde
+[doi2]: https://doi.org/10.1093/bioinformatics/btab705
+[simde]: https://github.com/nemequ/simde
 [unimap]: https://github.com/lh3/unimap

align.c

@@ -6,6 +6,8 @@
 #include "mmpriv.h"
 #include "ksw2.h"
 
+#define MM_MAX_QLEN_FLANK 100
+
 static void ksw_gen_simple_mat(int m, int8_t *mat, int8_t a, int8_t b, int8_t sc_ambi)
 {
 	int i, j;
@@ -21,6 +23,18 @@ static void ksw_gen_simple_mat(int m, int8_t *mat, int8_t a, int8_t b, int8_t sc
 		mat[(m - 1) * m + j] = sc_ambi;
 }
 
+static void ksw_gen_ts_mat(int m, int8_t *mat, int8_t a, int8_t b, int8_t transition, int8_t sc_ambi)
+{
+	assert(m == 5);
+	ksw_gen_simple_mat(m, mat, a, b, sc_ambi);
+	if (transition == 0 || transition == b) return;
+	transition = transition > 0? -transition : transition;
+	mat[0 * m + 2] = transition;  // A->G
+	mat[1 * m + 3] = transition;  // C->T
+	mat[2 * m + 0] = transition;  // G->A
+	mat[3 * m + 1] = transition;  // T->C
+}
+
 static inline void mm_seq_rev(uint32_t len, uint8_t *seq)
 {
 	uint32_t i;
@@ -246,7 +260,7 @@ static void mm_update_extra(mm_reg1_t *r, const uint8_t *qseq, const uint8_t *ts
 	if (p == 0) return;
 	mm_fix_cigar(r, qseq, tseq, &qshift, &tshift);
 	qseq += qshift, tseq += tshift; // qseq and tseq may be shifted due to the removal of leading I/D
-	r->blen = r->mlen = 0;
+	r->blen = r->mlen = 0, r->is_spliced = 0;
 	for (k = 0; k < p->n_cigar; ++k) {
 		uint32_t op = p->cigar[k]&0xf, len = p->cigar[k]>>4;
 		if (op == MM_CIGAR_MATCH) {
@@ -280,17 +294,16 @@ static void mm_update_extra(mm_reg1_t *r, const uint8_t *qseq, const uint8_t *ts
 			if (s < 0) s = 0;
 			toff += len;
 		} else if (op == MM_CIGAR_N_SKIP) {
-			toff += len;
+			r->is_spliced = 1, toff += len;
 		}
 	}
-	p->dp_max = (int32_t)(max + .499);
+	p->dp_max = p->dp_max0 = (int32_t)(max + .499);
 	assert(qoff == r->qe - r->qs && toff == r->re - r->rs);
 	if (is_eqx) mm_update_cigar_eqx(r, qseq, tseq); // NB: it has to be called here as changes to qseq and tseq are not returned
 }
 
-static void mm_append_cigar(mm_reg1_t *r, uint32_t n_cigar, uint32_t *cigar) // TODO: this calls the libc realloc()
+void mm_enlarge_cigar(mm_reg1_t *r, uint32_t n_cigar) // TODO: this calls the libc realloc()
 {
-	mm_extra_t *p;
 	if (n_cigar == 0) return;
 	if (r->p == 0) {
 		uint32_t capacity = n_cigar + sizeof(mm_extra_t)/4;
@@ -302,6 +315,13 @@ static void mm_append_cigar(mm_reg1_t *r, uint32_t n_cigar, uint32_t *cigar) //
 		kroundup32(r->p->capacity);
 		r->p = (mm_extra_t*)realloc(r->p, r->p->capacity * 4);
 	}
+}
+
+static void mm_append_cigar(mm_reg1_t *r, uint32_t n_cigar, const uint32_t *cigar)
+{
+	mm_extra_t *p;
+	if (n_cigar == 0) return;
+	mm_enlarge_cigar(r, n_cigar);
 	p = r->p;
 	if (p->n_cigar > 0 && (p->cigar[p->n_cigar-1]&0xf) == (cigar[0]&0xf)) { // same CIGAR op at the boundary
 		p->cigar[p->n_cigar-1] += cigar[0]>>4<<4;
@@ -313,25 +333,31 @@ static void mm_append_cigar(mm_reg1_t *r, uint32_t n_cigar, uint32_t *cigar) //
 	}
 }
 
-static void mm_align_pair(void *km, const mm_mapopt_t *opt, int qlen, const uint8_t *qseq, int tlen, const uint8_t *tseq, const uint8_t *junc, const int8_t *mat, int w, int end_bonus, int zdrop, int flag, ksw_extz_t *ez)
+static void mm_align_pair(void *km, const mm_mapopt_t *opt, int qlen, const uint8_t *qseq, int tlen, const uint8_t *tseq, const uint8_t *junc,
+						  const int8_t *mat, int w, int end_bonus, int zdrop, int ksw_flag, ksw_extz_t *ez)
 {
 	if (mm_dbg_flag & MM_DBG_PRINT_ALN_SEQ) {
 		int i;
-		fprintf(stderr, "===> q=(%d,%d), e=(%d,%d), bw=%d, flag=%d, zdrop=%d <===\n", opt->q, opt->q2, opt->e, opt->e2, w, flag, opt->zdrop);
+		fprintf(stderr, "===> q=(%d,%d), e=(%d,%d), bw=%d, ksw_flag=%d, zdrop=%d, end_bonus=%d <===\n", opt->q, opt->q2, opt->e, opt->e2, w, ksw_flag, opt->zdrop, end_bonus);
 		for (i = 0; i < tlen; ++i) fputc("ACGTN"[tseq[i]], stderr);
 		fputc('\n', stderr);
 		for (i = 0; i < qlen; ++i) fputc("ACGTN"[qseq[i]], stderr);
 		fputc('\n', stderr);
 	}
-	if (opt->max_sw_mat > 0 && (int64_t)tlen * qlen > opt->max_sw_mat) {
+	if (opt->transition != 0 && opt->b != opt->transition)
+		ksw_flag |= KSW_EZ_GENERIC_SC;
+	if (opt->max_sw_mat > 0 && (int64_t)tlen * qlen > opt->max_sw_mat) { // too much memory; skip alignment
 		ksw_reset_extz(ez);
 		ez->zdropped = 1;
-	} else if (opt->flag & MM_F_SPLICE)
-		ksw_exts2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, opt->q2, opt->noncan, zdrop, opt->junc_bonus, flag, junc, ez);
-	else if (opt->q == opt->q2 && opt->e == opt->e2)
-		ksw_extz2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, w, zdrop, end_bonus, flag, ez);
-	else
-		ksw_extd2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, opt->q2, opt->e2, w, zdrop, end_bonus, flag, ez);
+	} else if (opt->flag & MM_F_SPLICE) { // spliced alignment
+		assert((ksw_flag & KSW_EZ_SPLICE_FOR) == 0 || (ksw_flag & KSW_EZ_SPLICE_REV) == 0);
+		if (!(opt->flag & MM_F_SPLICE_OLD)) ksw_flag |= KSW_EZ_SPLICE_CMPLX;
+		ksw_exts2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, opt->q2, opt->noncan, zdrop, end_bonus, opt->junc_bonus, opt->junc_pen, ksw_flag, junc, ez);
+	} else if (opt->q == opt->q2 && opt->e == opt->e2) { // affine gap
+		ksw_extz2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, w, zdrop, end_bonus, ksw_flag, ez);
+	} else { // dual affine gap
+		ksw_extd2_sse(km, qlen, qseq, tlen, tseq, 5, mat, opt->q, opt->e, opt->q2, opt->e2, w, zdrop, end_bonus, ksw_flag, ez);
+	}
 	if (mm_dbg_flag & MM_DBG_PRINT_ALN_SEQ) {
 		int i;
 		fprintf(stderr, "score=%d, cigar=", ez->score);
@@ -341,6 +367,45 @@ static void mm_align_pair(void *km, const mm_mapopt_t *opt, int qlen, const uint
 	}
 }
 
+static int mm_align_sr_rna(void *km, const mm_mapopt_t *opt, int qlen, const uint8_t *qseq, int tlen, const uint8_t *tseq, const uint8_t *junc, uint8_t *tseq2, uint8_t *junc2,
+							const int8_t *mat, int w, int end_bonus, int zdrop, int ksw_flag, ksw_extz_t *ez)
+{
+	int32_t ilen = opt->q2 * 2, tlen2 = qlen * 2 + ilen;
+	int32_t i, ll = 0, lr = 0, nn = 0, n_ins = 0;
+	if (!(opt->flag & MM_F_SPLICE)) return 0; // only for spliced alignment
+	if (qlen > MM_MAX_QLEN_FLANK || qlen * 2 + ilen > tlen) return 0; // the query sequence can't be too long and the target sequence must be long enough
+	for (i = 0; i < qlen; ++i) // exact match length from the left
+		if (qseq[i] == tseq[i] && qseq[i] < 4)
+			++ll;
+	for (i = 0; i < qlen; ++i) // exact match length from the right
+		if (qseq[qlen - 1 - i] == tseq[tlen - 1 - i] && qseq[qlen - 1 - i] < 4)
+			++lr;
+	if (qlen - (ll + lr) > 9) return 0; // qlen may be smaller than ll+lr
+	memcpy(tseq2, tseq, qlen);
+	memset(&tseq2[qlen], 4, ilen);
+	memcpy(&tseq2[qlen + ilen], &tseq[tlen - qlen], qlen);
+	if (junc) {
+		memcpy(junc2, junc, qlen);
+		memset(&junc2[qlen], 0, ilen);
+		memcpy(&junc2[qlen + ilen], &junc[tlen - qlen], qlen);
+	}
+	if (!(opt->flag & MM_F_SPLICE_OLD)) ksw_flag |= KSW_EZ_SPLICE_CMPLX;
+	ksw_exts2_sse(km, qlen, qseq, tlen2, tseq2, 5, mat, opt->q, opt->e, opt->q2, opt->noncan, zdrop, end_bonus, opt->junc_bonus, opt->junc_pen, ksw_flag, junc2, ez);
+	if (ez->zdropped) return 0;
+	if ((ez->cigar[0]&0xf) != KSW_CIGAR_MATCH || (ez->cigar[ez->n_cigar-1]&0xf) != KSW_CIGAR_MATCH) return 0;
+	for (i = 0; i < ez->n_cigar; ++i) { // count the number of introns in the alignment
+		if ((ez->cigar[i]&0xf) == KSW_CIGAR_N_SKIP)
+			++nn;
+		else if ((ez->cigar[i]&0xf) == KSW_CIGAR_INS)
+			++n_ins;
+	}
+	if (nn != 1 || n_ins > 0) return 0; // the heuristic only works when there is exactly one intron
+	for (i = 0; i < ez->n_cigar; ++i)
+		if ((ez->cigar[i]&0xf) == KSW_CIGAR_N_SKIP)
+			ez->cigar[i] += (tlen - tlen2) << 4;
+	return 1;
+}
+
 static inline int mm_get_hplen_back(const mm_idx_t *mi, uint32_t rid, uint32_t x)
 {
 	int64_t i, off0 = mi->seq[rid].offset, off = off0 + x;
@@ -570,12 +635,19 @@ static void mm_fix_bad_ends_splice(void *km, const mm_mapopt_t *opt, const mm_id
 	}
 }
 
+static inline void mm_get_junc(const mm_idx_t *mi, int32_t ctg, int32_t st, int32_t en, int32_t rev, uint8_t *junc)
+{
+	if (mi->spsc) mm_idx_spsc_get(mi, ctg, st, en, rev, junc);
+	else if (mi->I) mm_idx_bed_junc(mi, ctg, st, en, junc);
+	else memset(junc, 0, en - st);
+}
+
 static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int qlen, uint8_t *qseq0[2], mm_reg1_t *r, mm_reg1_t *r2, int n_a, mm128_t *a, ksw_extz_t *ez, int splice_flag)
 {
-	int is_sr = !!(opt->flag & MM_F_SR), is_splice = !!(opt->flag & MM_F_SPLICE);
+	int is_sr = !!(opt->flag & MM_F_SR), is_splice = !!(opt->flag & MM_F_SPLICE), is_sr_rna = (!!(opt->flag & MM_F_SR_RNA) && is_splice);
 	int32_t rid = a[r->as].x<<1>>33, rev = a[r->as].x>>63, as1, cnt1;
-	uint8_t *tseq, *qseq, *junc;
-	int32_t i, l, bw, bw_long, dropped = 0, extra_flag = 0, rs0, re0, qs0, qe0;
+	uint8_t *tseq, *qseq, *junc, *tseq2 = 0, *junc2 = 0;
+	int32_t i, l, bw, bw_long, dropped = 0, ksw_flag = 0, rs0, re0, qs0, qe0;
 	int32_t rs, re, qs, qe;
 	int32_t rs1, qs1, re1, qe1;
 	int8_t mat[25];
@@ -584,7 +656,7 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 
 	r2->cnt = 0;
 	if (r->cnt == 0) return;
-	ksw_gen_simple_mat(5, mat, opt->a, opt->b, opt->sc_ambi);
+	ksw_gen_ts_mat(5, mat, opt->a, opt->b, opt->transition, opt->sc_ambi);
 	bw = (int)(opt->bw * 1.5 + 1.);
 	bw_long = (int)(opt->bw_long * 1.5 + 1.);
 	if (bw_long < bw) bw_long = bw;
@@ -610,9 +682,10 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 	assert(cnt1 > 0);
 
 	if (is_splice) {
-		if (splice_flag & MM_F_SPLICE_FOR) extra_flag |= rev? KSW_EZ_SPLICE_REV : KSW_EZ_SPLICE_FOR;
-		if (splice_flag & MM_F_SPLICE_REV) extra_flag |= rev? KSW_EZ_SPLICE_FOR : KSW_EZ_SPLICE_REV;
-		if (opt->flag & MM_F_SPLICE_FLANK) extra_flag |= KSW_EZ_SPLICE_FLANK;
+		if (splice_flag & MM_F_SPLICE_FOR) ksw_flag |= rev? KSW_EZ_SPLICE_REV : KSW_EZ_SPLICE_FOR;
+		if (splice_flag & MM_F_SPLICE_REV) ksw_flag |= rev? KSW_EZ_SPLICE_FOR : KSW_EZ_SPLICE_REV;
+		if (opt->flag & MM_F_SPLICE_FLANK) ksw_flag |= KSW_EZ_SPLICE_FLANK;
+		if (mi->spsc) ksw_flag |= KSW_EZ_SPLICE_SCORE;
 	}
 
 	/* Look for the start and end of regions to perform DP. This sounds easy
@@ -697,6 +770,12 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 	tseq = (uint8_t*)kmalloc(km, re0 - rs0);
 	junc = (uint8_t*)kmalloc(km, re0 - rs0);
 
+	if (is_sr_rna) {
+		int32_t max_tlen2 = MM_MAX_QLEN_FLANK * 2 + opt->q2 * 2;
+		tseq2 = Kmalloc(km, uint8_t, max_tlen2 * 2);
+		junc2 = tseq2 + max_tlen2;
+	}
+
 	if (qs > 0 && rs > 0) { // left extension; probably the condition can be changed to "qs > qs0 && rs > rs0"
 		if (opt->flag & MM_F_QSTRAND) {
 			qseq = &qseq0[0][qs0];
@@ -705,11 +784,11 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 			qseq = &qseq0[rev][qs0];
 			mm_idx_getseq(mi, rid, rs0, rs, tseq);
 		}
-		mm_idx_bed_junc(mi, rid, rs0, rs, junc);
+		mm_get_junc(mi, rid, rs0, rs, !!(ksw_flag&KSW_EZ_SPLICE_REV), junc);
 		mm_seq_rev(qs - qs0, qseq);
 		mm_seq_rev(rs - rs0, tseq);
 		mm_seq_rev(rs - rs0, junc);
-		mm_align_pair(km, opt, qs - qs0, qseq, rs - rs0, tseq, junc, mat, bw, opt->end_bonus, r->split_inv? opt->zdrop_inv : opt->zdrop, extra_flag|KSW_EZ_EXTZ_ONLY|KSW_EZ_RIGHT|KSW_EZ_REV_CIGAR, ez);
+		mm_align_pair(km, opt, qs - qs0, qseq, rs - rs0, tseq, junc, mat, bw, opt->end_bonus, r->split_inv? opt->zdrop_inv : opt->zdrop, ksw_flag|KSW_EZ_EXTZ_ONLY|KSW_EZ_RIGHT|KSW_EZ_REV_CIGAR, ez);
 		if (ez->n_cigar > 0) {
 			mm_append_cigar(r, ez->n_cigar, ez->cigar);
 			r->p->dp_score += ez->max;
@@ -721,14 +800,14 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 	re1 = rs, qe1 = qs;
 	assert(qs1 >= 0 && rs1 >= 0);
 
-	for (i = is_sr? cnt1 - 1 : 1; i < cnt1; ++i) { // gap filling
+	for (i = is_sr? cnt1 - 1 : 1; i < cnt1; ++i) { // gap filling; for short genomic reads, fill from the first seed to the last
 		if ((a[as1+i].y & (MM_SEED_IGNORE|MM_SEED_TANDEM)) && i != cnt1 - 1) continue;
 		if (is_sr && !(mi->flag & MM_I_HPC)) {
 			re = (int32_t)a[as1 + i].x + 1;
 			qe = (int32_t)a[as1 + i].y + 1;
 		} else mm_adjust_minier(mi, qseq0, &a[as1 + i], &re, &qe);
 		re1 = re, qe1 = qe;
-		if (i == cnt1 - 1 || (a[as1+i].y&MM_SEED_LONG_JOIN) || (qe - qs >= opt->min_ksw_len && re - rs >= opt->min_ksw_len)) {
+		if (i == cnt1 - 1 || (a[as1+i].y&MM_SEED_LONG_JOIN) || (qe - qs >= opt->min_ksw_len && re - rs >= opt->min_ksw_len)) { // gap filling
 			int j, bw1 = bw_long, zdrop_code;
 			if (a[as1+i].y & MM_SEED_LONG_JOIN)
 				bw1 = qe - qs > re - rs? qe - qs : re - rs;
@@ -740,21 +819,29 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 				qseq = &qseq0[rev][qs];
 				mm_idx_getseq(mi, rid, rs, re, tseq);
 			}
-			mm_idx_bed_junc(mi, rid, rs, re, junc);
-			if (is_sr) { // perform ungapped alignment
+			mm_get_junc(mi, rid, rs, re, !!(ksw_flag&KSW_EZ_SPLICE_REV), junc);
+			if (is_sr || (is_sr_rna && qe - qs == re - rs)) { // perform ungapped alignment
+				int32_t max_gapped_score = (qe - qs - 2) * opt->a - 2 * (opt->q + opt->e);
 				assert(qe - qs == re - rs);
 				ksw_reset_extz(ez);
 				for (j = 0, ez->score = 0; j < qe - qs; ++j) {
-					if (qseq[j] >= 4 || tseq[j] >= 4) ez->score += opt->e2;
+					if (qseq[j] >= 4 || tseq[j] >= 4) ez->score += opt->sc_ambi > 0? -opt->sc_ambi : opt->sc_ambi;
 					else ez->score += qseq[j] == tseq[j]? opt->a : -opt->b;
 				}
-				ez->cigar = ksw_push_cigar(km, &ez->n_cigar, &ez->m_cigar, ez->cigar, MM_CIGAR_MATCH, qe - qs);
+				if (ez->score > max_gapped_score)
+					ez->cigar = ksw_push_cigar(km, &ez->n_cigar, &ez->m_cigar, ez->cigar, MM_CIGAR_MATCH, qe - qs);
+				else
+					mm_align_pair(km, opt, qe - qs, qseq, re - rs, tseq, junc, mat, bw1, -1, opt->zdrop, ksw_flag|KSW_EZ_APPROX_MAX, ez);
 			} else { // perform normal gapped alignment
-				mm_align_pair(km, opt, qe - qs, qseq, re - rs, tseq, junc, mat, bw1, -1, opt->zdrop, extra_flag|KSW_EZ_APPROX_MAX, ez); // first pass: with approximate Z-drop
+				int32_t skip_full = 0;
+				if (is_sr_rna)
+					skip_full = mm_align_sr_rna(km, opt, qe - qs, qseq, re - rs, tseq, junc, tseq2, junc2, mat, bw1, -1, opt->zdrop, ksw_flag|KSW_EZ_APPROX_MAX, ez);
+				if (!skip_full)
+					mm_align_pair(km, opt, qe - qs, qseq, re - rs, tseq, junc, mat, bw1, -1, opt->zdrop, ksw_flag|KSW_EZ_APPROX_MAX, ez); // first pass: with approximate Z-drop
 			}
 			// test Z-drop and inversion Z-drop
 			if ((zdrop_code = mm_test_zdrop(km, opt, qseq, tseq, ez->n_cigar, ez->cigar, mat)) != 0)
-				mm_align_pair(km, opt, qe - qs, qseq, re - rs, tseq, junc, mat, bw1, -1, zdrop_code == 2? opt->zdrop_inv : opt->zdrop, extra_flag, ez); // second pass: lift approximate
+				mm_align_pair(km, opt, qe - qs, qseq, re - rs, tseq, junc, mat, bw1, -1, zdrop_code == 2? opt->zdrop_inv : opt->zdrop, ksw_flag, ez); // second pass: lift approximate
 			// update CIGAR
 			if (ez->n_cigar > 0)
 				mm_append_cigar(r, ez->n_cigar, ez->cigar);
@@ -792,8 +879,8 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 			qseq = &qseq0[rev][qe];
 			mm_idx_getseq(mi, rid, re, re0, tseq);
 		}
-		mm_idx_bed_junc(mi, rid, re, re0, junc);
-		mm_align_pair(km, opt, qe0 - qe, qseq, re0 - re, tseq, junc, mat, bw, opt->end_bonus, opt->zdrop, extra_flag|KSW_EZ_EXTZ_ONLY, ez);
+		mm_get_junc(mi, rid, re, re0, !!(ksw_flag&KSW_EZ_SPLICE_REV), junc);
+		mm_align_pair(km, opt, qe0 - qe, qseq, re0 - re, tseq, junc, mat, bw, opt->end_bonus, opt->zdrop, ksw_flag|KSW_EZ_EXTZ_ONLY, ez);
 		if (ez->n_cigar > 0) {
 			mm_append_cigar(r, ez->n_cigar, ez->cigar);
 			r->p->dp_score += ez->max;
@@ -816,11 +903,12 @@ static void mm_align1(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int
 			mm_idx_getseq(mi, rid, rs1, re1, tseq);
 			qseq = &qseq0[r->rev][qs1];
 		}
-		mm_update_extra(r, qseq, tseq, mat, opt->q, opt->e, opt->flag & MM_F_EQX, !(opt->flag & MM_F_SR));
+		mm_update_extra(r, qseq, tseq, mat, opt->q, opt->e, opt->flag & MM_F_EQX, !(is_sr || is_sr_rna));
 		if (rev && r->p->trans_strand)
 			r->p->trans_strand ^= 3; // flip to the read strand
 	}
 
+	if (tseq2) kfree(km, tseq2);
 	kfree(km, tseq);
 	kfree(km, junc);
 }
@@ -842,7 +930,7 @@ static int mm_align1_inv(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, i
 	if (ql < opt->min_chain_score || ql > opt->max_gap) return 0;
 	if (tl < opt->min_chain_score || tl > opt->max_gap) return 0;
 
-	ksw_gen_simple_mat(5, mat, opt->a, opt->b, opt->sc_ambi);
+	ksw_gen_ts_mat(5, mat, opt->a, opt->b, opt->transition, opt->sc_ambi);
 	tseq = (uint8_t*)kmalloc(km, tl);
 	mm_idx_getseq(mi, r1->rid, r1->re, r2->rs, tseq);
 	qseq = r1->rev? &qseq0[0][r2->qe] : &qseq0[1][qlen - r2->qs];
@@ -875,7 +963,7 @@ static int mm_align1_inv(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, i
 	}
 	r_inv->rs = r1->re + t_off;
 	r_inv->re = r_inv->rs + ez->max_t + 1;
-	mm_update_extra(r_inv, &qseq[q_off], &tseq[t_off], mat, opt->q, opt->e, opt->flag & MM_F_EQX, !(opt->flag & MM_F_SR));
+	mm_update_extra(r_inv, &qseq[q_off], &tseq[t_off], mat, opt->q, opt->e, opt->flag & MM_F_EQX, !(opt->flag & (MM_F_SR|MM_F_SR_RNA)));
 	ret = 1;
 end_align1_inv:
 	kfree(km, tseq);
@@ -917,14 +1005,14 @@ double mm_event_identity(const mm_reg1_t *r)
 static int32_t mm_recal_max_dp(const mm_reg1_t *r, double b2, int32_t match_sc)
 {
 	uint32_t i;
-	int32_t n_gap = 0, n_gapo = 0, n_mis;
+	int32_t n_gap = 0, n_mis;
 	double gap_cost = 0.0;
 	if (r->p == 0) return -1;
 	for (i = 0; i < r->p->n_cigar; ++i) {
 		int32_t op = r->p->cigar[i] & 0xf, len = r->p->cigar[i] >> 4;
 		if (op == MM_CIGAR_INS || op == MM_CIGAR_DEL) {
 			gap_cost += b2 + (double)mg_log2(1.0 + len);
-			++n_gapo, n_gap += len;
+			n_gap += len;
 		}
 	}
 	n_mis = r->blen + r->p->n_ambi - r->mlen - n_gap;
@@ -976,24 +1064,36 @@ mm_reg1_t *mm_align_skeleton(void *km, const mm_mapopt_t *opt, const mm_idx_t *m
 	n_a = mm_squeeze_a(km, n_regs, regs, a);
 	memset(&ez, 0, sizeof(ksw_extz_t));
 	for (i = 0; i < n_regs; ++i) {
-		mm_reg1_t r2;
+		mm_reg1_t r2; // only used for inversion
 		if ((opt->flag&MM_F_SPLICE) && (opt->flag&MM_F_SPLICE_FOR) && (opt->flag&MM_F_SPLICE_REV)) { // then do two rounds of alignments for both strands
-			mm_reg1_t s[2], s2[2];
-			int which, trans_strand;
+			mm_reg1_t s[2], s2[2], *r;
 			s[0] = s[1] = regs[i];
-			mm_align1(km, opt, mi, qlen, qseq0, &s[0], &s2[0], n_a, a, &ez, MM_F_SPLICE_FOR);
-			mm_align1(km, opt, mi, qlen, qseq0, &s[1], &s2[1], n_a, a, &ez, MM_F_SPLICE_REV);
-			if (s[0].p->dp_score > s[1].p->dp_score) which = 0, trans_strand = 1;
-			else if (s[0].p->dp_score < s[1].p->dp_score) which = 1, trans_strand = 2;
-			else trans_strand = 3, which = (qlen + s[0].p->dp_score) & 1; // randomly choose a strand, effectively
-			if (which == 0) {
+			mm_align1(km, opt, mi, qlen, qseq0, &s[0], &s2[0], n_a, a, &ez, MM_F_SPLICE_FOR); // assume the transcript is on the + strand of the genome
+			if ((opt->flag&MM_F_SR_RNA) && regs[i].qe - regs[i].qs == regs[i].re - regs[i].rs && s[0].qe - s[0].qs == s[0].re - s[0].rs && s[0].qs == 0 && s[0].qe == qlen) {
 				regs[i] = s[0], r2 = s2[0];
-				free(s[1].p);
+				regs[i].p->trans_strand = 0;
 			} else {
-				regs[i] = s[1], r2 = s2[1];
-				free(s[0].p);
+				int which, trans_strand;
+				mm_align1(km, opt, mi, qlen, qseq0, &s[1], &s2[1], n_a, a, &ez, MM_F_SPLICE_REV); // assume the transcript on the - strand
+				if (s[0].p->dp_score > s[1].p->dp_score) which = 0, trans_strand = 1;
+				else if (s[0].p->dp_score < s[1].p->dp_score) which = 1, trans_strand = 2;
+				else trans_strand = 3, which = (qlen + s[0].p->dp_score) & 1; // randomly choose a strand, effectively
+				if (which == 0) {
+					regs[i] = s[0], r2 = s2[0];
+					free(s[1].p);
+				} else {
+					regs[i] = s[1], r2 = s2[1];
+					free(s[0].p);
+				}
+				r = &regs[i];
+				r->p->trans_strand = trans_strand;
+				if (r->is_spliced) {
+					if (trans_strand == 1 || trans_strand == 2) // this is an *approximate* way to tell if there are splice signals.
+						r->p->dp_max += (opt->a + opt->b) + ((opt->a + opt->b) >> 1);
+					else if (trans_strand == 3)
+						r->p->dp_max -= opt->a + opt->b;
+				}
 			}
-			regs[i].p->trans_strand = trans_strand;
 		} else { // one round of alignment
 			mm_align1(km, opt, mi, qlen, qseq0, &regs[i], &r2, n_a, a, &ez, opt->flag);
 			if (opt->flag&MM_F_SPLICE)
@@ -1011,7 +1111,7 @@ mm_reg1_t *mm_align_skeleton(void *km, const mm_mapopt_t *opt, const mm_idx_t *m
 	kfree(km, qseq0[0]);
 	kfree(km, ez.cigar);
 	mm_filter_regs(opt, qlen, n_regs_, regs);
-	if (!(opt->flag&MM_F_SR) && !opt->split_prefix && qlen >= opt->rank_min_len) {
+	if (!(opt->flag&(MM_F_SR|MM_F_SR_RNA|MM_F_ALL_CHAINS)) && !opt->split_prefix && qlen >= opt->rank_min_len) {
 		mm_update_dp_max(qlen, *n_regs_, regs, opt->rank_frac, opt->a, opt->b);
 		mm_filter_regs(opt, qlen, n_regs_, regs);
 	}

cookbook.md

@@ -31,8 +31,8 @@ To acquire the data used in this cookbook and to install minimap2 and paftools,
 please follow the command lines below:
 ```sh
 # install minimap2 executables
-curl -L https://github.com/lh3/minimap2/releases/download/v2.23/minimap2-2.23_x64-linux.tar.bz2 | tar jxf -
-cp minimap2-2.23_x64-linux/{minimap2,k8,paftools.js} .  # copy executables
+curl -L https://github.com/lh3/minimap2/releases/download/v2.29/minimap2-2.29_x64-linux.tar.bz2 | tar jxf -
+cp minimap2-2.29_x64-linux/{minimap2,k8,paftools.js} .  # copy executables
 export PATH="$PATH:"`pwd`                               # put the current directory on PATH
 # download example datasets
 curl -L https://github.com/lh3/minimap2/releases/download/v2.10/cookbook-data.tgz | tar zxf -

format.c

@@ -119,6 +119,7 @@ int mm_write_sam_hdr(const mm_idx_t *idx, const char *rg, const char *ver, int a
 {
 	kstring_t str = {0,0,0};
 	int ret = 0;
+	mm_sprintf_lite(&str, "@HD\tVN:1.6\tSO:unsorted\tGO:query\n");
 	if (idx) {
 		uint32_t i;
 		for (i = 0; i < idx->n_seq; ++i)
@@ -138,10 +139,48 @@ int mm_write_sam_hdr(const mm_idx_t *idx, const char *rg, const char *ver, int a
 	return ret;
 }
 
-static void write_cs_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq, const mm_reg1_t *r, char *tmp, int no_iden, int write_tag)
+static void write_indel_ds(kstring_t *str, int64_t len, const uint8_t *seq, int64_t ll, int64_t lr) // write an indel to ds; adapted from minigraph
 {
-	int i, q_off, t_off;
-	if (write_tag) mm_sprintf_lite(s, "\tcs:Z:");
+	int64_t i;
+	if (ll + lr >= len) {
+		mm_sprintf_lite(str, "[");
+		for (i = 0; i < len; ++i)
+			mm_sprintf_lite(str, "%c", "acgtn"[seq[i]]);
+		mm_sprintf_lite(str, "]");
+	} else {
+		int64_t k = 0;
+		if (ll > 0) {
+			mm_sprintf_lite(str, "[");
+			for (i = 0; i < ll; ++i)
+				mm_sprintf_lite(str, "%c", "acgtn"[seq[k+i]]);
+			mm_sprintf_lite(str, "]");
+			k += ll;
+		}
+		for (i = 0; i < len - lr - ll; ++i)
+			mm_sprintf_lite(str, "%c", "acgtn"[seq[k+i]]);
+		k += len - lr - ll;
+		if (lr > 0) {
+			mm_sprintf_lite(str, "[");
+			for (i = 0; i < lr; ++i)
+				mm_sprintf_lite(str, "%c", "acgtn"[seq[k+i]]);
+			mm_sprintf_lite(str, "]");
+		}
+	}
+}
+
+static void write_cs_ds_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq, const mm_reg1_t *r, char *tmp, int no_iden, int is_ds, int write_tag)
+{
+	int i, q_off, t_off, q_len = 0, t_len = 0;
+	if (write_tag) mm_sprintf_lite(s, "\t%cs:Z:", is_ds? 'd' : 'c');
+	for (i = 0; i < (int)r->p->n_cigar; ++i) {
+		int op = r->p->cigar[i]&0xf, len = r->p->cigar[i]>>4;
+		if (op == MM_CIGAR_MATCH || op == MM_CIGAR_EQ_MATCH || op == MM_CIGAR_X_MISMATCH)
+			q_len += len, t_len += len;
+		else if (op == MM_CIGAR_INS)
+			q_len += len;
+		else if (op == MM_CIGAR_DEL || op == MM_CIGAR_N_SKIP)
+			t_len += len;
+	}
 	for (i = q_off = t_off = 0; i < (int)r->p->n_cigar; ++i) {
 		int j, op = r->p->cigar[i]&0xf, len = r->p->cigar[i]>>4;
 		assert((op >= MM_CIGAR_MATCH && op <= MM_CIGAR_N_SKIP) || op == MM_CIGAR_EQ_MATCH || op == MM_CIGAR_X_MISMATCH);
@@ -167,14 +206,42 @@ static void write_cs_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq
 			}
 			q_off += len, t_off += len;
 		} else if (op == MM_CIGAR_INS) {
-			for (j = 0, tmp[len] = 0; j < len; ++j)
-				tmp[j] = "acgtn"[qseq[q_off + j]];
-			mm_sprintf_lite(s, "+%s", tmp);
+			if (is_ds) {
+				int z, ll, lr, y = q_off;
+				for (z = 1; z <= len; ++z)
+					if (y - z < 0 || qseq[y + len - z] != qseq[y - z])
+						break;
+				lr = z - 1;
+				for (z = 0; z < len; ++z)
+					if (y + len + z >= q_len || qseq[y + len + z] != qseq[y + z])
+						break;
+				ll = z;
+				mm_sprintf_lite(s, "+");
+				write_indel_ds(s, len, &qseq[y], ll, lr);
+			} else {
+				for (j = 0, tmp[len] = 0; j < len; ++j)
+					tmp[j] = "acgtn"[qseq[q_off + j]];
+				mm_sprintf_lite(s, "+%s", tmp);
+			}
 			q_off += len;
 		} else if (op == MM_CIGAR_DEL) {
-			for (j = 0, tmp[len] = 0; j < len; ++j)
-				tmp[j] = "acgtn"[tseq[t_off + j]];
-			mm_sprintf_lite(s, "-%s", tmp);
+			if (is_ds) {
+				int z, ll, lr, x = t_off;
+				for (z = 1; z <= len; ++z)
+					if (x - z < 0 || tseq[x + len - z] != tseq[x - z])
+						break;
+				lr = z - 1;
+				for (z = 0; z < len; ++z)
+					if (x + len + z >= t_len || tseq[x + z] != tseq[x + len + z])
+						break;
+				ll = z;
+				mm_sprintf_lite(s, "-");
+				write_indel_ds(s, len, &tseq[x], ll, lr);
+			} else {
+				for (j = 0, tmp[len] = 0; j < len; ++j)
+					tmp[j] = "acgtn"[tseq[t_off + j]];
+				mm_sprintf_lite(s, "-%s", tmp);
+			}
 			t_off += len;
 		} else { // intron
 			assert(len >= 2);
@@ -186,6 +253,52 @@ static void write_cs_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq
 	assert(t_off == r->re - r->rs && q_off == r->qe - r->qs);
 }
 
+static inline void revcomp_splice(uint8_t s[2])
+{
+	uint8_t c = s[1] < 4? 3 - s[1] : 4;
+	s[1] = s[0] < 4? 3 - s[0] : 4;
+	s[0] = c;
+}
+
+void mm_write_junc(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r)
+{
+	int32_t i, t_off, swritten = 0;
+	s->l = 0;
+	if (!r->is_spliced || r->p == 0) return; // no junctions
+	if (r->p->trans_strand != 1 && r->p->trans_strand != 2) return; // no preferred strand
+	for (i = 0, t_off = r->rs; i < (int)r->p->n_cigar; ++i) {
+		int op = r->p->cigar[i]&0xf, len = r->p->cigar[i]>>4;
+		if (op == MM_CIGAR_MATCH || op == MM_CIGAR_EQ_MATCH || op == MM_CIGAR_X_MISMATCH || op == MM_CIGAR_DEL) {
+			t_off += len;
+		} else if (op == MM_CIGAR_N_SKIP) { // intron
+			uint8_t donor[2], acceptor[2];
+			int32_t score1 = 0, score2 = 0, rev;
+			assert(len >= 2);
+			rev = (r->p->trans_strand == 2) ^ r->rev;
+			if (!rev) {
+				mm_idx_getseq(mi, r->rid, t_off, t_off + 2, donor);
+				mm_idx_getseq(mi, r->rid, t_off + len - 2, t_off + len, acceptor);
+			} else {
+				mm_idx_getseq(mi, r->rid, t_off, t_off + 2, acceptor);
+				mm_idx_getseq(mi, r->rid, t_off + len - 2, t_off + len, donor);
+				revcomp_splice(donor);
+				revcomp_splice(acceptor);
+			}
+			//fprintf(stderr, "%c%c-%c%c\n", "ACGTN"[donor[0]], "ACGTN"[donor[1]], "ACGTN"[acceptor[0]], "ACGTN"[acceptor[1]]);
+			if (donor[0] == 2 && donor[1] == 3) score1 = 3;
+			else if (donor[0] == 2 && donor[1] == 1) score1 = 2;
+			else if (donor[0] == 0 && donor[1] == 3) score1 = 1;
+			if (acceptor[0] == 0 && acceptor[1] == 2) score2 = 3;
+			else if (acceptor[0] == 0 && acceptor[1] == 1) score2 = 1;
+			if (swritten) mm_sprintf_lite(s, "\n");
+			else swritten = 1;
+			mm_sprintf_lite(s, "%s\t%d\t%d\t%s\t%d\t%c", mi->seq[r->rid].name, t_off, t_off + len, t->name, score1 + score2, "+-"[rev]);
+			t_off += len;
+		}
+	}
+	assert(t_off == r->re);
+}
+
 static void write_MD_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq, const mm_reg1_t *r, char *tmp, int write_tag)
 {
 	int i, q_off, t_off, l_MD = 0;
@@ -217,7 +330,7 @@ static void write_MD_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq
 	assert(t_off == r->re - r->rs && q_off == r->qe - r->qs);
 }
 
-static void write_cs_or_MD(void *km, kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, int no_iden, int is_MD, int write_tag, int is_qstrand)
+static void write_cs_ds_or_MD(void *km, kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, int no_iden, int is_MD, int is_ds, int write_tag, int is_qstrand)
 {
 	extern unsigned char seq_nt4_table[256];
 	int i;
@@ -244,7 +357,7 @@ static void write_cs_or_MD(void *km, kstring_t *s, const mm_idx_t *mi, const mm_
 		}
 	}
 	if (is_MD) write_MD_core(s, tseq, qseq, r, tmp, write_tag);
-	else write_cs_core(s, tseq, qseq, r, tmp, no_iden, write_tag);
+	else write_cs_ds_core(s, tseq, qseq, r, tmp, no_iden, is_ds, write_tag);
 	kfree(km, qseq); kfree(km, tseq); kfree(km, tmp);
 }
 
@@ -255,7 +368,7 @@ int mm_gen_cs_or_MD(void *km, char **buf, int *max_len, const mm_idx_t *mi, cons
 	str.s = *buf, str.l = 0, str.m = *max_len;
 	t.l_seq = strlen(seq);
 	t.seq = (char*)seq;
-	write_cs_or_MD(km, &str, mi, &t, r, no_iden, is_MD, 0, is_qstrand);
+	write_cs_ds_or_MD(km, &str, mi, &t, r, no_iden, is_MD, 0, 0, is_qstrand);
 	*max_len = str.m;
 	*buf = str.s;
 	return str.l;
@@ -277,7 +390,7 @@ static inline void write_tags(kstring_t *s, const mm_reg1_t *r)
 	if (r->id == r->parent) type = r->inv? 'I' : 'P';
 	else type = r->inv? 'i' : 'S';
 	if (r->p) {
-		mm_sprintf_lite(s, "\tNM:i:%d\tms:i:%d\tAS:i:%d\tnn:i:%d", r->blen - r->mlen + r->p->n_ambi, r->p->dp_max, r->p->dp_score, r->p->n_ambi);
+		mm_sprintf_lite(s, "\tNM:i:%d\tms:i:%d\tAS:i:%d\tnn:i:%d", r->blen - r->mlen + r->p->n_ambi, r->p->dp_max0, r->p->dp_score, r->p->n_ambi);
 		if (r->p->trans_strand == 1 || r->p->trans_strand == 2)
 			mm_sprintf_lite(s, "\tts:A:%c", "?+-?"[r->p->trans_strand]);
 	}
@@ -299,15 +412,18 @@ static inline void write_tags(kstring_t *s, const mm_reg1_t *r)
 	if (r->split) mm_sprintf_lite(s, "\tzd:i:%d", r->split);
 }
 
-void mm_write_paf3(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, void *km, int64_t opt_flag, int rep_len)
+void mm_write_paf4(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, void *km, int64_t opt_flag, int rep_len, int n_seg, int seg_idx)
 {
 	s->l = 0;
+	mm_sprintf_lite(s, "%s", t->name);
+	if ((opt_flag & MM_F_FRAG_MODE) && n_seg >= 2 && seg_idx >= 0)
+		mm_sprintf_lite(s, "/%d", seg_idx + 1);
 	if (r == 0) {
-		mm_sprintf_lite(s, "%s\t%d\t0\t0\t*\t*\t0\t0\t0\t0\t0\t0", t->name, t->l_seq);
+		mm_sprintf_lite(s, "\t%d\t0\t0\t*\t*\t0\t0\t0\t0\t0\t0", t->l_seq);
 		if (rep_len >= 0) mm_sprintf_lite(s, "\trl:i:%d", rep_len);
 		return;
 	}
-	mm_sprintf_lite(s, "%s\t%d\t%d\t%d\t%c\t", t->name, t->l_seq, r->qs, r->qe, "+-"[r->rev]);
+	mm_sprintf_lite(s, "\t%d\t%d\t%d\t%c\t", t->l_seq, r->qs, r->qe, "+-"[r->rev]);
 	if (mi->seq[r->rid].name) mm_sprintf_lite(s, "%s", mi->seq[r->rid].name);
 	else mm_sprintf_lite(s, "%d", r->rid);
 	mm_sprintf_lite(s, "\t%d", mi->seq[r->rid].len);
@@ -325,12 +441,17 @@ void mm_write_paf3(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const
 		for (k = 0; k < r->p->n_cigar; ++k)
 			mm_sprintf_lite(s, "%d%c", r->p->cigar[k]>>4, MM_CIGAR_STR[r->p->cigar[k]&0xf]);
 	}
-	if (r->p && (opt_flag & (MM_F_OUT_CS|MM_F_OUT_MD)))
-		write_cs_or_MD(km, s, mi, t, r, !(opt_flag&MM_F_OUT_CS_LONG), opt_flag&MM_F_OUT_MD, 1, !!(opt_flag&MM_F_QSTRAND));
+	if (r->p && (opt_flag & (MM_F_OUT_CS|MM_F_OUT_DS|MM_F_OUT_MD)))
+		write_cs_ds_or_MD(km, s, mi, t, r, !(opt_flag&MM_F_OUT_CS_LONG), !!(opt_flag&MM_F_OUT_MD), !!(opt_flag&MM_F_OUT_DS), 1, !!(opt_flag&MM_F_QSTRAND));
 	if ((opt_flag & MM_F_COPY_COMMENT) && t->comment)
 		mm_sprintf_lite(s, "\t%s", t->comment);
 }
 
+void mm_write_paf3(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, void *km, int64_t opt_flag, int rep_len)
+{
+	mm_write_paf4(s, mi, t, r, km, opt_flag, rep_len, 0, 0);
+}
+
 void mm_write_paf(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, void *km, int64_t opt_flag)
 {
 	mm_write_paf3(s, mi, t, r, km, opt_flag, -1);
@@ -369,14 +490,16 @@ static void write_sam_cigar(kstring_t *s, int sam_flag, int in_tag, int qlen, co
 		clip_len[0] = r->rev? qlen - r->qe : r->qs;
 		clip_len[1] = r->rev? r->qs : qlen - r->qe;
 		if (in_tag) {
-			int clip_char = (sam_flag&0x800) && !(opt_flag&MM_F_SOFTCLIP)? 5 : 4;
+			int clip_char = (((sam_flag&0x800) || ((sam_flag&0x100) && (opt_flag&MM_F_SECONDARY_SEQ))) &&
+							 !(opt_flag&MM_F_SOFTCLIP)) ? 5 : 4;
 			mm_sprintf_lite(s, "\tCG:B:I");
 			if (clip_len[0]) mm_sprintf_lite(s, ",%u", clip_len[0]<<4|clip_char);
 			for (k = 0; k < r->p->n_cigar; ++k)
 				mm_sprintf_lite(s, ",%u", r->p->cigar[k]);
 			if (clip_len[1]) mm_sprintf_lite(s, ",%u", clip_len[1]<<4|clip_char);
 		} else {
-			int clip_char = (sam_flag&0x800) && !(opt_flag&MM_F_SOFTCLIP)? 'H' : 'S';
+			int clip_char = (((sam_flag&0x800) || ((sam_flag&0x100) && (opt_flag&MM_F_SECONDARY_SEQ))) &&
+							 !(opt_flag&MM_F_SOFTCLIP)) ? 'H' : 'S';
 			assert(clip_len[0] < qlen && clip_len[1] < qlen);
 			if (clip_len[0]) mm_sprintf_lite(s, "%d%c", clip_len[0], clip_char);
 			for (k = 0; k < r->p->n_cigar; ++k)
@@ -451,7 +574,7 @@ void mm_write_sam3(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, int se
 		if (cigar_in_tag) {
 			int slen;
 			if ((flag & 0x900) == 0 || (opt_flag & MM_F_SOFTCLIP)) slen = t->l_seq;
-			else if (flag & 0x100) slen = 0;
+			else if ((flag & 0x100) && !(opt_flag & MM_F_SECONDARY_SEQ)) slen = 0;
 			else slen = r->qe - r->qs;
 			mm_sprintf_lite(s, "%dS%dN", slen, r->re - r->rs);
 		} else write_sam_cigar(s, flag, 0, t->l_seq, r, opt_flag);
@@ -492,7 +615,7 @@ void mm_write_sam3(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, int se
 			mm_sprintf_lite(s, "\t");
 			if (t->qual) sam_write_sq(s, t->qual, t->l_seq, r->rev, 0);
 			else mm_sprintf_lite(s, "*");
-		} else if (flag & 0x100) {
+		} else if ((flag & 0x100) && !(opt_flag & MM_F_SECONDARY_SEQ)){
 			mm_sprintf_lite(s, "*\t*");
 		} else {
 			sam_write_sq(s, t->seq + r->qs, r->qe - r->qs, r->rev, r->rev);
@@ -532,8 +655,8 @@ void mm_write_sam3(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, int se
 				}
 			}
 		}
-		if (r->p && (opt_flag & (MM_F_OUT_CS|MM_F_OUT_MD)))
-			write_cs_or_MD(km, s, mi, t, r, !(opt_flag&MM_F_OUT_CS_LONG), opt_flag&MM_F_OUT_MD, 1, 0);
+		if (r->p && (opt_flag & (MM_F_OUT_CS|MM_F_OUT_DS|MM_F_OUT_MD)))
+			write_cs_ds_or_MD(km, s, mi, t, r, !(opt_flag&MM_F_OUT_CS_LONG), opt_flag&MM_F_OUT_MD, !!(opt_flag&MM_F_OUT_DS), 1, 0);
 		if (cigar_in_tag)
 			write_sam_cigar(s, flag, 1, t->l_seq, r, opt_flag);
 	}

hit.c

@@ -55,7 +55,7 @@ mm_reg1_t *mm_gen_regs(void *km, uint32_t hash, int qlen, int n_u, uint64_t *u,
 	mm_reg1_t *r;
 	int i, k;
 
-	if (n_u == 0) return 0;
+	if (n_u <= 0) return 0;
 
 	// sort by score
 	z = (mm128_t*)kmalloc(km, n_u * 16);
@@ -279,7 +279,7 @@ int mm_filter_strand_retained(int n_regs, mm_reg1_t *r)
 	int i, k;
 	for (i = k = 0; i < n_regs; ++i) {
 		int p = r[i].parent;
-		if (!r[i].strand_retained || r[i].div < r[p].div * 5.0f) {
+		if (!r[i].strand_retained || r[i].div < r[p].div * 5.0f || r[i].div < 0.01f) {
 			if (k < i) r[k++] = r[i];
 			else ++k;
 		}
@@ -418,16 +418,19 @@ static void mm_set_inv_mapq(void *km, int n_regs, mm_reg1_t *regs)
 	kfree(km, aux);
 }
 
-void mm_set_mapq(void *km, int n_regs, mm_reg1_t *regs, int min_chain_sc, int match_sc, int rep_len, int is_sr)
+void mm_set_mapq2(void *km, int n_regs, mm_reg1_t *regs, int min_chain_sc, int match_sc, int rep_len, int is_sr, int is_splice)
 {
 	static const float q_coef = 40.0f;
 	int64_t sum_sc = 0;
 	float uniq_ratio;
-	int i;
+	int i, n_2nd_splice = 0;
 	if (n_regs == 0) return;
-	for (i = 0; i < n_regs; ++i)
+	for (i = 0; i < n_regs; ++i) {
 		if (regs[i].parent == regs[i].id)
 			sum_sc += regs[i].score;
+		else if (regs[i].is_spliced)
+			++n_2nd_splice;
+	}
 	uniq_ratio = (float)sum_sc / (sum_sc + rep_len);
 	for (i = 0; i < n_regs; ++i) {
 		mm_reg1_t *r = &regs[i];
@@ -440,13 +443,18 @@ void mm_set_mapq(void *km, int n_regs, mm_reg1_t *regs, int min_chain_sc, int ma
 			pen_cm = pen_s1 < pen_cm? pen_s1 : pen_cm;
 			subsc = r->subsc > min_chain_sc? r->subsc : min_chain_sc;
 			if (r->p && r->p->dp_max2 > 0 && r->p->dp_max > 0) {
-				float identity = (float)r->mlen / r->blen;
-				float x = (float)r->p->dp_max2 * subsc / r->p->dp_max / r->score0;
+				float x, identity = (float)r->mlen / r->blen;
+				if (is_sr && is_splice)
+					x = (float)r->p->dp_max2 / r->p->dp_max; // ignore chaining score; for short RNA-seq reads, unspliced chaining score tends to be higher
+				else
+					x = (float)r->p->dp_max2 * subsc / r->p->dp_max / r->score0;
 				mapq = (int)(identity * pen_cm * q_coef * (1.0f - x * x) * logf((float)r->p->dp_max / match_sc));
 				if (!is_sr) {
 					int mapq_alt = (int)(6.02f * identity * identity * (r->p->dp_max - r->p->dp_max2) / match_sc + .499f); // BWA-MEM like mapQ, mostly for short reads
 					mapq = mapq < mapq_alt? mapq : mapq_alt; // in case the long-read heuristic fails
 				}
+				if (is_splice && is_sr && r->is_spliced && n_2nd_splice == 0)
+					mapq += 10;
 			} else {
 				float x = (float)subsc / r->score0;
 				if (r->p) {

index.c

@@ -12,6 +12,7 @@
 #include "bseq.h"
 #include "minimap.h"
 #include "mmpriv.h"
+#include "ksw2.h"
 #include "kvec.h"
 #include "khash.h"
 
@@ -32,7 +33,7 @@ typedef struct mm_idx_bucket_s {
 } mm_idx_bucket_t;
 
 typedef struct {
-	int32_t st, en, max; // max is not used for now
+	int32_t st, en, cnt;
 	int32_t score:30, strand:2;
 } mm_idx_intv1_t;
 
@@ -41,6 +42,11 @@ typedef struct mm_idx_intv_s {
 	mm_idx_intv1_t *a;
 } mm_idx_intv_t;
 
+typedef struct mm_idx_jjump_s {
+	int32_t n, m;
+	mm_idx_jjump1_t *a;
+} mm_idx_jjump_t;
+
 mm_idx_t *mm_idx_init(int w, int k, int b, int flag)
 {
 	mm_idx_t *mi;
@@ -65,11 +71,17 @@ void mm_idx_destroy(mm_idx_t *mi)
 			kh_destroy(idx, (idxhash_t*)mi->B[i].h);
 		}
 	}
+	if (mi->spsc) free(mi->spsc);
 	if (mi->I) {
 		for (i = 0; i < mi->n_seq; ++i)
 			free(mi->I[i].a);
 		free(mi->I);
 	}
+	if (mi->J) {
+		for (i = 0; i < mi->n_seq; ++i)
+			free(mi->J[i].a);
+		free(mi->J);
+	}
 	if (!mi->km) {
 		for (i = 0; i < mi->n_seq; ++i)
 			free(mi->seq[i].name);
@@ -99,7 +111,7 @@ const uint64_t *mm_idx_get(const mm_idx_t *mi, uint64_t minier, int *n)
 
 void mm_idx_stat(const mm_idx_t *mi)
 {
-	int n = 0, n1 = 0;
+	int64_t n = 0, n1 = 0;
 	uint32_t i;
 	uint64_t sum = 0, len = 0;
 	fprintf(stderr, "[M::%s] kmer size: %d; skip: %d; is_hpc: %d; #seq: %d\n", __func__, mi->k, mi->w, mi->flag&MM_I_HPC, mi->n_seq);
@@ -117,8 +129,8 @@ void mm_idx_stat(const mm_idx_t *mi)
 				if (kh_key(h, k)&1) ++n1;
 			}
 	}
-	fprintf(stderr, "[M::%s::%.3f*%.2f] distinct minimizers: %d (%.2f%% are singletons); average occurrences: %.3lf; average spacing: %.3lf; total length: %ld\n",
-			__func__, realtime() - mm_realtime0, cputime() / (realtime() - mm_realtime0), n, 100.0*n1/n, (double)sum / n, (double)len / sum, (long)len);
+	fprintf(stderr, "[M::%s::%.3f*%.2f] distinct minimizers: %ld (%.2f%% are singletons); average occurrences: %.3lf; average spacing: %.3lf; total length: %ld\n",
+			__func__, realtime() - mm_realtime0, cputime() / (realtime() - mm_realtime0), (long)n, 100.0*n1/n, (double)sum / n, (double)len / sum, (long)len);
 }
 
 int mm_idx_index_name(mm_idx_t *mi)
@@ -192,6 +204,7 @@ int32_t mm_idx_cal_max_occ(const mm_idx_t *mi, float f)
 	if (f <= 0.) return INT32_MAX;
 	for (i = 0; i < 1<<mi->b; ++i)
 		if (mi->B[i].h) n += kh_size((idxhash_t*)mi->B[i].h);
+	if (n == 0) return INT32_MAX;
 	a = (uint32_t*)malloc(n * 4);
 	for (i = n = 0; i < 1<<mi->b; ++i) {
 		idxhash_t *h = (idxhash_t*)mi->B[i].h;
@@ -656,10 +669,17 @@ int mm_idx_alt_read(mm_idx_t *mi, const char *fn)
 	return n_alt;
 }
 
+/***************
+ * BED reading *
+ ***************/
+
 #define sort_key_bed(a) ((a).st)
 KRADIX_SORT_INIT(bed, mm_idx_intv1_t, sort_key_bed, 4)
 
-mm_idx_intv_t *mm_idx_read_bed(const mm_idx_t *mi, const char *fn, int read_junc)
+#define sort_key_end(a) ((a).en)
+KRADIX_SORT_INIT(end, mm_idx_intv1_t, sort_key_end, 4)
+
+static mm_idx_intv_t *mm_idx_bed_read_core(const mm_idx_t *mi, const char *fn, int read_junc, int min_sc)
 {
 	gzFile fp;
 	kstream_t *ks;
@@ -668,7 +688,7 @@ mm_idx_intv_t *mm_idx_read_bed(const mm_idx_t *mi, const char *fn, int read_junc
 
 	fp = fn && strcmp(fn, "-")? gzopen(fn, "r") : gzdopen(fileno(stdin), "r");
 	if (fp == 0) return 0;
-	I = (mm_idx_intv_t*)calloc(mi->n_seq, sizeof(*I));
+	I = CALLOC(mm_idx_intv_t, mi->n_seq);
 	ks = ks_init(fp);
 	while (ks_getuntil(ks, KS_SEP_LINE, &str, 0) >= 0) {
 		mm_idx_intv_t *r;
@@ -689,7 +709,7 @@ mm_idx_intv_t *mm_idx_read_bed(const mm_idx_t *mi, const char *fn, int read_junc
 					t.en = atol(q);
 					if (t.en < 0) break;
 				} else if (i == 4) { // BED score
-					t.score = atol(q);
+					t.score = *q >= '0' && *q <= '9'? atol(q) : -1;
 				} else if (i == 5) { // strand
 					t.strand = *q == '+'? 1 : *q == '-'? -1 : 0;
 				} else if (i == 9) {
@@ -705,7 +725,8 @@ mm_idx_intv_t *mm_idx_read_bed(const mm_idx_t *mi, const char *fn, int read_junc
 				++i, q = p + 1;
 			}
 		}
-		if (id < 0 || t.st < 0 || t.st >= t.en) continue;
+		if (id < 0 || t.st < 0 || t.st >= t.en) continue; // contig ID not found, or other problems
+		if (min_sc > 0 && t.score < min_sc) continue;
 		r = &I[id];
 		if (i >= 11 && read_junc) { // BED12
 			int32_t st, sz, en;
@@ -738,14 +759,44 @@ mm_idx_intv_t *mm_idx_read_bed(const mm_idx_t *mi, const char *fn, int read_junc
 	return I;
 }
 
+static mm_idx_intv_t *mm_idx_bed_read_merge(const mm_idx_t *mi, const char *fn, int read_junc, int min_sc)
+{
+	long n = 0, n0 = 0;
+	int32_t i;
+	mm_idx_intv_t *I;
+	I = mm_idx_bed_read_core(mi, fn, read_junc, min_sc);
+	if (I == 0) return 0;
+	for (i = 0; i < mi->n_seq; ++i) {
+		int32_t j, j0, k;
+		mm_idx_intv_t *intv = &I[i];
+		n0 += intv->n;
+		radix_sort_bed(intv->a, intv->a + intv->n); // sort by st
+		for (j = 1, j0 = 0; j <= intv->n; ++j) { // sort by st and then by end
+			if (j == intv->n || intv->a[j].st != intv->a[j0].st) {
+				radix_sort_end(intv->a + j0, intv->a + j);
+				j0 = j;
+			}
+		}
+		for (j = 1, j0 = 0, k = 0; j <= intv->n; ++j) { // merge intervals with the same (st, en)
+			if (j == intv->n || intv->a[j].st != intv->a[j0].st || intv->a[j].en != intv->a[j0].en) {
+				intv->a[k] = intv->a[j0];
+				intv->a[k++].cnt = j - j0;
+				j0 = j;
+			}
+		}
+		intv->a = REALLOC(mm_idx_intv1_t, intv->a, k);
+		intv->n = intv->m = k;
+		n += k;
+	}
+	if (mm_verbose >= 3)
+		fprintf(stderr, "[%s] read %ld introns, %ld of which are non-redundant\n", __func__, n0, n);
+	return I;
+}
+
 int mm_idx_bed_read(mm_idx_t *mi, const char *fn, int read_junc)
 {
-	int32_t i;
 	if (mi->h == 0) mm_idx_index_name(mi);
-	mi->I = mm_idx_read_bed(mi, fn, read_junc);
-	if (mi->I == 0) return -1;
-	for (i = 0; i < mi->n_seq; ++i) // TODO: eliminate redundant intervals
-		radix_sort_bed(mi->I[i].a, mi->I[i].a + mi->I[i].n);
+	mi->I = mm_idx_bed_read_merge(mi, fn, read_junc, -1);
 	return 0;
 }
 
@@ -773,3 +824,244 @@ int mm_idx_bed_junc(const mm_idx_t *mi, int32_t ctg, int32_t st, int32_t en, uin
 	}
 	return left;
 }
+
+/*********************************
+ * Reading junctions for jumping *
+ *********************************/
+
+#define sort_key_jj(a) ((a).off)
+KRADIX_SORT_INIT(jj, mm_idx_jjump1_t, sort_key_jj, 4)
+
+#define sort_key_jj2(a) ((a).off2)
+KRADIX_SORT_INIT(jj2, mm_idx_jjump1_t, sort_key_jj2, 4)
+
+static void sort_jjump(mm_idx_jjump_t *jj2)
+{
+	int32_t j0, j, k;
+	if (jj2 == 0 || jj2->n == 0) return;
+	radix_sort_jj(jj2->a, jj2->a + jj2->n);
+	for (j0 = 0, j = 1; j <= jj2->n; ++j) {
+		if (j == jj2->n || jj2->a[j0].off != jj2->a[j].off) {
+			radix_sort_jj2(jj2->a + j0, jj2->a + j);
+			j0 = j;
+		}
+	}
+	// the actual merge
+	for (j0 = 0, j = 1, k = 0; j <= jj2->n; ++j) {
+		if (j == jj2->n || jj2->a[j0].off != jj2->a[j].off || jj2->a[j0].off2 != jj2->a[j].off2) {
+			int32_t t, cnt = 0;
+			uint16_t flag = 0;
+			for (t = j0; t < j; ++t) cnt += jj2->a[t].cnt, flag |= jj2->a[t].flag;
+			jj2->a[k] = jj2->a[j0];
+			jj2->a[k].cnt = cnt;
+			jj2->a[k++].flag = flag;
+			j0 = j;
+		}
+	}
+	jj2->n = k;
+	jj2->a = REALLOC(mm_idx_jjump1_t, jj2->a, k);
+}
+
+static mm_idx_jjump_t *mm_idx_bed2jjump(const mm_idx_t *mi, const mm_idx_intv_t *I, uint16_t flag)
+{
+	int32_t i;
+	mm_idx_jjump_t *J;
+	J = CALLOC(mm_idx_jjump_t, mi->n_seq);
+	for (i = 0; i < mi->n_seq; ++i) {
+		int32_t j, k;
+		const mm_idx_intv_t *intv = &I[i];
+		mm_idx_jjump_t *jj = &J[i];
+		jj->n = intv->n * 2;
+		jj->a = CALLOC(mm_idx_jjump1_t, jj->n);
+		for (j = k = 0; j < intv->n; ++j) {
+			jj->a[k].off = intv->a[j].st, jj->a[k].off2 = intv->a[j].en, jj->a[k].cnt = intv->a[j].cnt, jj->a[k].strand = intv->a[j].strand, jj->a[k++].flag = flag;
+			jj->a[k].off = intv->a[j].en, jj->a[k].off2 = intv->a[j].st, jj->a[k].cnt = intv->a[j].cnt, jj->a[k].strand = intv->a[j].strand, jj->a[k++].flag = flag;
+		}
+		sort_jjump(jj);
+	}
+	return J;
+}
+
+static mm_idx_jjump_t *mm_idx_jjump_merge(const mm_idx_t *mi, const mm_idx_jjump_t *J0, const mm_idx_jjump_t *J1)
+{
+	int32_t i;
+	mm_idx_jjump_t *J2;
+	J2 = CALLOC(mm_idx_jjump_t, mi->n_seq);
+	for (i = 0; i < mi->n_seq; ++i) {
+		int32_t j, k;
+		const mm_idx_jjump_t *jj0 = &J0[i], *jj1 = &J1[i];
+		mm_idx_jjump_t *jj2 = &J2[i];
+		jj2->n = jj0->n + jj1->n;
+		jj2->a = CALLOC(mm_idx_jjump1_t, jj2->n);
+		for (j = k = 0; j < jj0->n; ++j) jj2->a[k++] = jj0->a[j];
+		for (j = 0; j < jj1->n; ++j) jj2->a[k++] = jj1->a[j];
+		sort_jjump(jj2);
+	}
+	return J2;
+}
+
+int mm_idx_jjump_read(mm_idx_t *mi, const char *fn, int flag, int min_sc)
+{
+	int32_t i, j, n_anno = 0, n_misc = 0;
+	mm_idx_intv_t *I;
+	mm_idx_jjump_t *J;
+	if (mi->h == 0) mm_idx_index_name(mi);
+	I = mm_idx_bed_read_merge(mi, fn, 1, min_sc);
+	J = mm_idx_bed2jjump(mi, I, flag);
+	for (i = 0; i < mi->n_seq; ++i) free(I[i].a);
+	free(I);
+	if (mi->J) {
+		mm_idx_jjump_t *J2;
+		J2 = mm_idx_jjump_merge(mi, mi->J, J);
+		for (i = 0; i < mi->n_seq; ++i) {
+			free(mi->J[i].a); free(J[i].a);
+		}
+		free(mi->J); free(J);
+		mi->J = J2;
+	} else mi->J = J;
+	for (i = 0; i < mi->n_seq; ++i) {
+		for (j = 0; j < mi->J[i].n; ++j)
+			if (mi->J[i].a[j].flag & MM_JUNC_ANNO) ++n_anno;
+			else ++n_misc;
+	}
+	if (mm_verbose >= 3)
+		fprintf(stderr, "[%s] there are %d annotated and %d other splice positions in the index\n", __func__, n_anno, n_misc);
+	return 0;
+}
+
+static int32_t mm_idx_jump_get_core(int32_t n, const mm_idx_jjump1_t *a, int32_t x) // similar to mm_idx_find_intv()
+{
+	int32_t s = 0, e = n;
+	if (n == 0) return -1;
+	if (x < a[0].off) return -1;
+	while (s < e) {
+		int32_t mid = s + (e - s) / 2;
+		if (x >= a[mid].off && (mid + 1 >= n || x < a[mid+1].off)) return mid;
+		else if (x < a[mid].off) e = mid;
+		else s = mid + 1;
+	}
+	assert(0);
+}
+
+const mm_idx_jjump1_t *mm_idx_jump_get(const mm_idx_t *db, int32_t cid, int32_t st, int32_t en, int32_t *n)
+{
+	mm_idx_jjump_t *s;
+	int32_t l, r;
+	*n = 0;
+	if (cid >= db->n_seq || cid < 0 || db->J == 0) return 0;
+	if (en < 0 || en > db->seq[cid].len) en = db->seq[cid].len;
+	s = &db->J[cid];
+	if (s->n == 0) return 0;
+	l = mm_idx_jump_get_core(s->n, s->a, st);
+	r = mm_idx_jump_get_core(s->n, s->a, en);
+	*n = r - l;
+	return &s->a[l + 1];
+}
+
+/****************
+ * splice score *
+ ****************/
+
+typedef struct mm_idx_spsc_s {
+	uint32_t n, m;
+	uint64_t *a; // pos<<56 | score<<1 | acceptor
+} mm_idx_spsc_t;
+
+int32_t mm_idx_spsc_read(mm_idx_t *idx, const char *fn, int32_t max_sc)
+{
+	gzFile fp;
+	kstring_t str = {0,0,0};
+	kstream_t *ks;
+	int32_t dret, j;
+	int64_t n_read = 0;
+
+	fp = fn && strcmp(fn, "-") != 0? gzopen(fn, "rb") : gzdopen(0, "rb");
+	if (fp == 0) return -1;
+	if (idx->h == 0) mm_idx_index_name(idx);
+	if (max_sc > 63) max_sc = 63;
+	idx->spsc = Kcalloc(0, mm_idx_spsc_t, idx->n_seq * 2);
+	ks = ks_init(fp);
+	while (ks_getuntil(ks, KS_SEP_LINE, &str, &dret) >= 0) {
+		mm_idx_spsc_t *s;
+		char *p, *q, *name = 0;
+		int32_t i, type = -1, strand = 0, cid = -1, score = -1;
+		int64_t pos = -1;
+		for (i = 0, p = q = str.s;; ++p) {
+			if (*p == '\t' || *p == 0) {
+				int c = *p;
+				*p = 0;
+				if (i == 0) {
+					name = q;
+				} else if (i == 1) {
+					pos = atol(q);
+				} else if (i == 2) {
+					strand = *q == '+'? 1 : '-'? -1 : 0;
+				} else if (i == 3) {
+					type = *q == 'D'? 0 : *q == 'A'? 1 : -1;
+				} else if (i == 4) {
+					score = atoi(q);
+					break;
+				}
+				if (c == 0) break;
+				q = p + 1, ++i;
+			}
+		}
+		if (i < 4) continue; // not enough fields
+		if (score > max_sc) score = max_sc;
+		if (score < -max_sc) score = -max_sc;
+		cid = mm_idx_name2id(idx, name);
+		if (cid < 0 || type < 0 || strand == 0 || pos < 0) continue; // FIXME: give a warning!
+		s = &idx->spsc[cid << 1 | (strand > 0? 0 : 1)];
+		Kgrow(0, uint64_t, s->a, s->n, s->m);
+		if (pos > 0 && pos < idx->seq[cid].len) { // ignore scores at the ends
+			s->a[s->n++] = (uint64_t)pos << 8 | (score + KSW_SPSC_OFFSET) << 1 | type;
+			++n_read;
+		}
+	}
+	ks_destroy(ks);
+	gzclose(fp);
+	for (j = 0; j < idx->n_seq * 2; ++j) {
+		mm_idx_spsc_t *s = &idx->spsc[j];
+		if (s->n > 0)
+			radix_sort_64(s->a, s->a + s->n);
+	}
+	if (mm_verbose >= 3)
+		fprintf(stderr, "[M::%s] read %ld splice scores\n", __func__, (long)n_read);
+	return 0;
+}
+
+static int32_t mm_idx_find_intv(int32_t n, const uint64_t *a, int64_t x)
+{
+	int32_t s = 0, e = n;
+	if (n == 0) return -1;
+	if (x < a[0]>>8) return -1;
+	while (s < e) {
+		int32_t mid = s + (e - s) / 2;
+		if (x >= a[mid]>>8 && (mid + 1 >= n || x < a[mid+1]>>8)) return mid;
+		else if (x < a[mid]>>8) e = mid;
+		else s = mid + 1;
+	}
+	assert(0);
+}
+
+int64_t mm_idx_spsc_get(const mm_idx_t *db, int32_t cid, int64_t st, int64_t en, int32_t rev, uint8_t *sc)
+{
+	const mm_idx_spsc_t *s;
+	if (cid >= db->n_seq || cid < 0 || db->spsc == 0) return -1;
+	if (en < 0 || en > db->seq[cid].len) en = db->seq[cid].len;
+	memset(sc, 0xff, en - st);
+	s = &db->spsc[cid << 1 | (!!rev)];
+	if (s->n > 0) {
+		int32_t j, l, r;
+		l = mm_idx_find_intv(s->n, s->a, st);
+		r = mm_idx_find_intv(s->n, s->a, en);
+		for (j = l + 1; j <= r; ++j) {
+			int64_t x = (s->a[j]>>8) - st;
+			uint8_t score = s->a[j] & 0xff;
+			assert(x <= en - st);
+			if (x == en - st) continue;
+			if (sc[x] == 0xff || sc[x] < score) sc[x] = score;
+		}
+	}
+	return en - st;
+}

jump.c

@@ -0,0 +1,201 @@
+#include <stdio.h>
+#include "mmpriv.h"
+#include "kalloc.h"
+
+#define MM_MIN_EXON_LEN 20
+
+static int32_t mm_jump_check(void *km, const mm_idx_t *mi, int32_t qlen, const uint8_t *qseq0, const mm_reg1_t *r, int32_t ext, int32_t is_left) // TODO: check close N
+{
+	int32_t clip, clen, e = !r->rev ^ !is_left; // 0 for left of the alignment; 1 for right
+	uint32_t cigar;
+	if (!r->p || r->p->n_cigar <= 0) return -1; // only working with CIGAR
+	clip = e == 0? r->qs : qlen - r->qe;
+	cigar = r->p->cigar[is_left? 0 : r->p->n_cigar - 1];
+	clen = (cigar&0xf) == MM_CIGAR_MATCH? cigar>>4 : 0;
+	if (clen <= ext) return -1;
+	if (is_left) {
+		if (clip >= r->rs) return -1; // no space to jump
+	} else {
+		if (clip >= mi->seq[r->rid].len - r->re) return -1; // no space to jump
+	}
+	return 0;
+}
+
+static uint8_t *mm_jump_get_qseq_seq(void *km, int32_t qlen, const uint8_t *qseq0, const mm_reg1_t *r, int32_t is_left, int32_t ql0, uint8_t *qseq)
+{
+	extern unsigned char seq_nt4_table[256];
+	int32_t i, k = 0;
+	if (!r->rev) {
+		if (is_left)
+			for (i = 0; i < ql0; ++i)
+				qseq[k++] = seq_nt4_table[(uint8_t)qseq0[i]];
+		else
+			for (i = qlen - ql0; i < qlen; ++i)
+				qseq[k++] = seq_nt4_table[(uint8_t)qseq0[i]];
+	} else {
+		if (is_left)
+			for (i = qlen - 1; i >= qlen - ql0; --i) {
+				uint8_t c = seq_nt4_table[(uint8_t)qseq0[i]];
+				qseq[k++] = c >= 4? c : 3 - c;
+			}
+		else
+			for (i = ql0 - 1; i >= 0; --i) {
+				uint8_t c = seq_nt4_table[(uint8_t)qseq0[i]];
+				qseq[k++] = c >= 4? c : 3 - c;
+			}
+	}
+	return qseq;
+}
+
+static void mm_jump_split_left(void *km, const mm_idx_t *mi, const mm_mapopt_t *opt, int32_t qlen, const uint8_t *qseq0, mm_reg1_t *r, int32_t ts_strand)
+{
+	uint8_t *tseq = 0, *qseq = 0;
+	int32_t i, n, l, i0, m, mm0;
+	int32_t i0_anno = -1, n_anno = 0, mm0_anno = 0, i0_misc = -1, n_misc = 0, mm0_misc = 0;
+	int32_t ext = 1 + (opt->b + opt->a - 1) / opt->a + 1;
+	int32_t clip = !r->rev? r->qs : qlen - r->qe;
+	int32_t extt = clip < ext? clip : ext;
+	const mm_idx_jjump1_t *a;
+
+	if (mm_jump_check(km, mi, qlen, qseq0, r, ext + MM_MIN_EXON_LEN, 1) < 0) return;
+	a = mm_idx_jump_get(mi, r->rid, r->rs - extt, r->rs + ext, &n);
+	if (n == 0) return;
+
+	for (i = 0; i < n; ++i) { // traverse possible jumps
+		const mm_idx_jjump1_t *ai = &a[i];
+		int32_t tlen, tl1, j, mm1, mm2;
+		assert(ai->off >= r->rs - extt && ai->off <= r->rs + ext);
+		if (ts_strand * ai->strand < 0) continue; // wrong strand
+		if (ai->off2 >= ai->off) continue; // wrong direction
+		if (ai->off - ai->off2 < 6) continue; // intron too small
+		if (ai->off2 < clip + ext) continue; // not long enough
+		if (tseq == 0) {
+			tseq = Kcalloc(km, uint8_t, (clip + ext) * 2); // tseq and qseq are allocated together
+			qseq = tseq + clip + ext;
+			mm_jump_get_qseq_seq(km, qlen, qseq0, r, 1, clip + ext, qseq);
+		}
+		tl1 = clip + (ai->off - r->rs);
+		tlen = mm_idx_getseq2(mi, 0, r->rid, ai->off, r->rs + ext, &tseq[tl1]);
+		assert(tlen == r->rs + ext - ai->off);
+		tlen = mm_idx_getseq2(mi, 0, r->rid, ai->off2 - tl1, ai->off2, tseq);
+		assert(tlen == tl1);
+		for (j = 0, mm1 = 0; j < tl1; ++j)
+			if (qseq[j] != tseq[j] || qseq[j] > 3 || tseq[j] > 3)
+				++mm1;
+		for (mm2 = 0; j < clip + ext; ++j)
+			if (qseq[j] != tseq[j] || qseq[j] > 3 || tseq[j] > 3)
+				++mm2;
+		if (mm1 == 0 && mm2 <= 1) {
+			if (ai->flag & MM_JUNC_ANNO)
+				i0_anno = i, mm0_anno = mm1 + mm2, ++n_anno; // i0 points to the rightmost i
+			else
+				i0_misc = i, mm0_misc = mm1 + mm2, ++n_misc;
+		}
+	}
+	if (n_anno > 0) m = n_anno, i0 = i0_anno, mm0 = mm0_anno;
+	else m = n_misc, i0 = i0_misc, mm0 = mm0_misc;
+	kfree(km, tseq);
+
+	l = m > 0? a[i0].off - r->rs : 0; // may be negative
+	if (m == 1 && clip + l >= opt->jump_min_match) { // add one more exon
+		mm_enlarge_cigar(r, 2);
+		memmove(r->p->cigar + 2, r->p->cigar, r->p->n_cigar * 4);
+		r->p->cigar[0] = (clip + l) << 4 | MM_CIGAR_MATCH;
+		r->p->cigar[1] = (a[i0].off - a[i0].off2) << 4 | MM_CIGAR_N_SKIP;
+		r->p->cigar[2] = ((r->p->cigar[2]>>4) - l) << 4 | MM_CIGAR_MATCH;
+		r->p->n_cigar += 2;
+		r->rs = a[i0].off2 - (clip + l);
+		if (!r->rev) r->qs = 0;
+		else r->qe = qlen;
+		r->blen += clip, r->mlen += clip - mm0;
+		r->p->dp_max0 += (clip - mm0) * opt->a - mm0 * opt->b;
+		r->p->dp_max += (clip - mm0) * opt->a - mm0 * opt->b;
+		if (!r->is_spliced) r->is_spliced = 1, r->p->dp_max += (opt->a + opt->b) + ((opt->a + opt->b) >> 1);
+	} else if (m > 0 && a[i0].off > r->rs) { // trim by l; l is always positive
+		r->p->cigar[0] -= l << 4 | MM_CIGAR_MATCH;
+		r->rs += l;
+		if (!r->rev) r->qs += l;
+		else r->qe -= l;
+	}
+}
+
+static void mm_jump_split_right(void *km, const mm_idx_t *mi, const mm_mapopt_t *opt, int32_t qlen, const uint8_t *qseq0, mm_reg1_t *r, int32_t ts_strand)
+{
+	uint8_t *tseq = 0, *qseq = 0;
+	int32_t i, n, l, i0, m, mm0;
+	int32_t i0_anno = -1, n_anno = 0, mm0_anno = 0, i0_misc = -1, n_misc = 0, mm0_misc = 0;
+	int32_t ext = 1 + (opt->b + opt->a - 1) / opt->a + 1;
+	int32_t clip = !r->rev? qlen - r->qe : r->qs;
+	int32_t extt = clip < ext? clip : ext;
+	const mm_idx_jjump1_t *a;
+
+	if (mm_jump_check(km, mi, qlen, qseq0, r, ext + MM_MIN_EXON_LEN, 0) < 0) return;
+	a = mm_idx_jump_get(mi, r->rid, r->re - ext, r->re + extt, &n);
+	if (n == 0) return;
+
+	for (i = 0; i < n; ++i) { // traverse possible jumps
+		const mm_idx_jjump1_t *ai = &a[i];
+		int32_t tlen, tl1, j, mm1, mm2;
+		assert(ai->off >= r->re - ext && ai->off <= r->re + extt);
+		if (ts_strand * ai->strand < 0) continue; // wrong strand
+		if (ai->off2 <= ai->off) continue; // wrong direction
+		if (ai->off2 - ai->off < 6) continue; // intron too small
+		if (ai->off2 + clip + ext > mi->seq[r->rid].len) continue; // not long enough
+		if (tseq == 0) {
+			tseq = Kcalloc(km, uint8_t, (clip + ext) * 2); // tseq and qseq are allocated together
+			qseq = tseq + clip + ext;
+			mm_jump_get_qseq_seq(km, qlen, qseq0, r, 0, clip + ext, qseq);
+		}
+		tl1 = clip + (r->re - ai->off);
+		tlen = mm_idx_getseq2(mi, 0, r->rid, r->re - ext, ai->off, tseq);
+		assert(tlen == ai->off - (r->re - ext));
+		tlen = mm_idx_getseq2(mi, 0, r->rid, ai->off2, ai->off2 + tl1, &tseq[clip + ext - tl1]);
+		assert(tlen == tl1);
+		for (j = 0, mm2 = 0; j < clip + ext - tl1; ++j)
+			if (qseq[j] != tseq[j] || qseq[j] > 3 || tseq[j] > 3)
+				++mm2;
+		for (mm1 = 0; j < clip + ext; ++j)
+			if (qseq[j] != tseq[j] || qseq[j] > 3 || tseq[j] > 3)
+				++mm1;
+		if (mm1 == 0 && mm2 <= 1) {
+			if (ai->flag & MM_JUNC_ANNO) {
+				if (i0_anno < 0) i0_anno = i, mm0_anno = mm1 + mm2;
+				++n_anno;
+			} else {
+				if (i0_misc < 0) i0_misc = i, mm0_misc = mm1 + mm2;
+				++n_misc;
+			}
+		}
+	}
+	if (n_anno > 0) m = n_anno, i0 = i0_anno, mm0 = mm0_anno;
+	else m = n_misc, i0 = i0_misc, mm0 = mm0_misc;
+	kfree(km, tseq);
+
+	l = m > 0? r->re - a[i0].off : 0; // may be negative
+	if (m == 1 && clip + l >= opt->jump_min_match) { // add one more exon
+		mm_enlarge_cigar(r, 2);
+		r->p->cigar[r->p->n_cigar - 1] = ((r->p->cigar[r->p->n_cigar - 1]>>4) - l) << 4 | MM_CIGAR_MATCH;
+		r->p->cigar[r->p->n_cigar] = (a[i0].off2 - a[i0].off) << 4 | MM_CIGAR_N_SKIP;
+		r->p->cigar[r->p->n_cigar + 1] = (clip + l) << 4 | MM_CIGAR_MATCH;
+		r->p->n_cigar += 2;
+		r->re = a[i0].off2 + (clip + l);
+		if (!r->rev) r->qe = qlen;
+		else r->qs = 0;
+		r->blen += clip, r->mlen += clip - mm0;
+		r->p->dp_max0 += (clip - mm0) * opt->a - mm0 * opt->b;
+		r->p->dp_max += (clip - mm0) * opt->a - mm0 * opt->b;
+		if (!r->is_spliced) r->is_spliced = 1, r->p->dp_max += (opt->a + opt->b) + ((opt->a + opt->b) >> 1);
+	} else if (m > 0 && r->re > a[i0].off) { // trim by l; l is always positive
+		r->p->cigar[r->p->n_cigar - 1] -= l << 4 | MM_CIGAR_MATCH;
+		r->re -= l;
+		if (!r->rev) r->qe -= l;
+		else r->qs += l;
+	}
+}
+
+void mm_jump_split(void *km, const mm_idx_t *mi, const mm_mapopt_t *opt, int32_t qlen, const uint8_t *qseq, mm_reg1_t *r, int32_t ts_strand)
+{
+	assert((opt->flag & MM_F_EQX) == 0);
+	mm_jump_split_left(km,  mi, opt, qlen, qseq, r, ts_strand);
+	mm_jump_split_right(km, mi, opt, qlen, qseq, r, ts_strand);
+}

kalloc.c

@@ -40,7 +40,8 @@ void *km_init2(void *km_par, size_t min_core_size)
 	kmem_t *km;
 	km = (kmem_t*)kcalloc(km_par, 1, sizeof(kmem_t));
 	km->par = km_par;
-	km->min_core_size = min_core_size > 0? min_core_size : 0x80000;
+	if (km_par) km->min_core_size = min_core_size > 0? min_core_size : ((kmem_t*)km_par)->min_core_size - 2;
+	else km->min_core_size = min_core_size > 0? min_core_size : 0x80000;
 	return (void*)km;
 }
 
@@ -183,6 +184,16 @@ void *krealloc(void *_km, void *ap, size_t n_bytes) // TODO: this can be made mo
 	return q;
 }
 
+void *krelocate(void *km, void *ap, size_t n_bytes)
+{
+	void *p;
+	if (km == 0 || ap == 0) return ap;
+	p = kmalloc(km, n_bytes);
+	memcpy(p, ap, n_bytes);
+	kfree(km, ap);
+	return p;
+}
+
 void km_stat(const void *_km, km_stat_t *s)
 {
 	kmem_t *km = (kmem_t*)_km;
@@ -203,3 +214,11 @@ void km_stat(const void *_km, km_stat_t *s)
 		s->largest = s->largest > size? s->largest : size;
 	}
 }
+
+void km_stat_print(const void *km)
+{
+	km_stat_t st;
+	km_stat(km, &st);
+	fprintf(stderr, "[km_stat] cap=%ld, avail=%ld, largest=%ld, n_core=%ld, n_block=%ld\n",
+			st.capacity, st.available, st.largest, st.n_blocks, st.n_cores);
+}

kalloc.h

@@ -13,6 +13,7 @@ typedef struct {
 
 void *kmalloc(void *km, size_t size);
 void *krealloc(void *km, void *ptr, size_t size);
+void *krelocate(void *km, void *ap, size_t n_bytes);
 void *kcalloc(void *km, size_t count, size_t size);
 void kfree(void *km, void *ptr);
 
@@ -20,11 +21,29 @@ void *km_init(void);
 void *km_init2(void *km_par, size_t min_core_size);
 void km_destroy(void *km);
 void km_stat(const void *_km, km_stat_t *s);
+void km_stat_print(const void *km);
 
 #ifdef __cplusplus
 }
 #endif
 
+#define Kmalloc(km, type, cnt)       ((type*)kmalloc((km), (cnt) * sizeof(type)))
+#define Kcalloc(km, type, cnt)       ((type*)kcalloc((km), (cnt), sizeof(type)))
+#define Krealloc(km, type, ptr, cnt) ((type*)krealloc((km), (ptr), (cnt) * sizeof(type)))
+
+#define Kgrow(km, type, ptr, __i, __m) do { \
+		if ((__i) >= (__m)) { \
+			(__m) = (__i) + 1; \
+			(__m) += ((__m)>>1) + 16; \
+			(ptr) = Krealloc(km, type, ptr, (__m)); \
+		} \
+	} while (0)
+
+#define Kexpand(km, type, a, m) do { \
+		(m) = (m) >= 4? (m) + ((m)>>1) : 16; \
+		(a) = Krealloc(km, type, (a), (m)); \
+	} while (0)
+
 #define KMALLOC(km, ptr, len) ((ptr) = (__typeof__(ptr))kmalloc((km), (len) * sizeof(*(ptr))))
 #define KCALLOC(km, ptr, len) ((ptr) = (__typeof__(ptr))kcalloc((km), (len), sizeof(*(ptr))))
 #define KREALLOC(km, ptr, len) ((ptr) = (__typeof__(ptr))krealloc((km), (ptr), (len) * sizeof(*(ptr))))
@@ -50,7 +69,7 @@ void km_stat(const void *_km, km_stat_t *s);
 	} kmp_##name##_t; \
 	SCOPE kmp_##name##_t *kmp_init_##name(void *km) { \
 		kmp_##name##_t *mp; \
-		KCALLOC(km, mp, 1); \
+		mp = Kcalloc(km, kmp_##name##_t, 1); \
 		mp->km = km; \
 		return mp; \
 	} \
@@ -66,7 +85,7 @@ void km_stat(const void *_km, km_stat_t *s);
 	} \
 	SCOPE void kmp_free_##name(kmp_##name##_t *mp, kmptype_t *p) { \
 		--mp->cnt; \
-		if (mp->n == mp->max) KEXPAND(mp->km, mp->buf, mp->max); \
+		if (mp->n == mp->max) Kexpand(mp->km, kmptype_t*, mp->buf, mp->max); \
 		mp->buf[mp->n++] = p; \
 	}
 

ksw2.h

@@ -15,6 +15,8 @@
 #define KSW_EZ_SPLICE_FOR  0x100
 #define KSW_EZ_SPLICE_REV  0x200
 #define KSW_EZ_SPLICE_FLANK 0x400
+#define KSW_EZ_SPLICE_CMPLX 0x800  // use the miniprot splice model
+#define KSW_EZ_SPLICE_SCORE 0x1000 // use splice score
 
 // The subset of CIGAR operators used by ksw code.
 // Use MM_CIGAR_* from minimap.h if you need the full list.
@@ -23,6 +25,8 @@
 #define KSW_CIGAR_DEL    2
 #define KSW_CIGAR_N_SKIP 3
 
+#define KSW_SPSC_OFFSET  64
+
 #ifdef __cplusplus
 extern "C" {
 #endif
@@ -68,7 +72,7 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 				   int8_t gapo, int8_t gape, int8_t gapo2, int8_t gape2, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez);
 
 void ksw_exts2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
-				   int8_t gapo, int8_t gape, int8_t gapo2, int8_t noncan, int zdrop, int8_t junc_bonus, int flag, const uint8_t *junc, ksw_extz_t *ez);
+				   int8_t gapo, int8_t gape, int8_t gapo2, int8_t noncan, int zdrop, int end_bonus, int8_t junc_bonus, int8_t junc_pen, int flag, const uint8_t *junc, ksw_extz_t *ez);
 
 void ksw_extf2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t mch, int8_t mis, int8_t e, int w, int xdrop, ksw_extz_t *ez);
 

ksw2_dispatch.c

@@ -80,17 +80,17 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 }
 
 void ksw_exts2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
-				   int8_t q, int8_t e, int8_t q2, int8_t noncan, int zdrop, int8_t junc_bonus, int flag, const uint8_t *junc, ksw_extz_t *ez)
+				   int8_t q, int8_t e, int8_t q2, int8_t noncan, int zdrop, int end_bonus, int8_t junc_bonus, int8_t junc_pen, int flag, const uint8_t *junc, ksw_extz_t *ez)
 {
 	extern void ksw_exts2_sse2(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
-				   int8_t q, int8_t e, int8_t q2, int8_t noncan, int zdrop, int8_t junc_bonus, int flag, const uint8_t *junc, ksw_extz_t *ez);
+				   int8_t q, int8_t e, int8_t q2, int8_t noncan, int zdrop, int end_bonus, int8_t junc_bonus, int8_t junc_pen, int flag, const uint8_t *junc, ksw_extz_t *ez);
 	extern void ksw_exts2_sse41(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
-				   int8_t q, int8_t e, int8_t q2, int8_t noncan, int zdrop, int8_t junc_bonus, int flag, const uint8_t *junc, ksw_extz_t *ez);
+				   int8_t q, int8_t e, int8_t q2, int8_t noncan, int zdrop, int end_bonus, int8_t junc_bonus, int8_t junc_pen, int flag, const uint8_t *junc, ksw_extz_t *ez);
 	if (ksw_simd < 0) ksw_simd = x86_simd();
 	if (ksw_simd & SIMD_SSE4_1)
-		ksw_exts2_sse41(km, qlen, query, tlen, target, m, mat, q, e, q2, noncan, zdrop, junc_bonus, flag, junc, ez);
+		ksw_exts2_sse41(km, qlen, query, tlen, target, m, mat, q, e, q2, noncan, zdrop, end_bonus, junc_bonus, junc_pen, flag, junc, ez);
 	else if (ksw_simd & SIMD_SSE2)
-		ksw_exts2_sse2(km, qlen, query, tlen, target, m, mat, q, e, q2, noncan, zdrop, junc_bonus, flag, junc, ez);
+		ksw_exts2_sse2(km, qlen, query, tlen, target, m, mat, q, e, q2, noncan, zdrop, end_bonus, junc_bonus, junc_pen, flag, junc, ez);
 	else abort();
 }
 #endif

ksw2_extd2_sse.c

@@ -358,7 +358,7 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 			} else H[0] = v8[0] - qe, max_H = H[0], max_t = 0; // special casing r==0
 			// update ez
 			if (en0 == tlen - 1 && H[en0] > ez->mte)
-				ez->mte = H[en0], ez->mte_q = r - en;
+				ez->mte = H[en0], ez->mte_q = r - en0;
 			if (r - st0 == qlen - 1 && H[st0] > ez->mqe)
 				ez->mqe = H[st0], ez->mqe_t = st0;
 			if (ksw_apply_zdrop(ez, 1, max_H, r, max_t, zdrop, e2)) break;

ksw2_exts2_sse.c

@@ -24,14 +24,14 @@
 #ifdef KSW_CPU_DISPATCH
 #ifdef __SSE4_1__
 void ksw_exts2_sse41(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
-				   int8_t q, int8_t e, int8_t q2, int8_t noncan, int zdrop, int8_t junc_bonus, int flag, const uint8_t *junc, ksw_extz_t *ez)
+				   int8_t q, int8_t e, int8_t q2, int8_t noncan, int zdrop, int end_bonus, int8_t junc_bonus, int8_t junc_pen, int flag, const uint8_t *junc, ksw_extz_t *ez)
 #else
 void ksw_exts2_sse2(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
-				   int8_t q, int8_t e, int8_t q2, int8_t noncan, int zdrop, int8_t junc_bonus, int flag, const uint8_t *junc, ksw_extz_t *ez)
+				   int8_t q, int8_t e, int8_t q2, int8_t noncan, int zdrop, int end_bonus, int8_t junc_bonus, int8_t junc_pen, int flag, const uint8_t *junc, ksw_extz_t *ez)
 #endif
 #else
 void ksw_exts2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
-				   int8_t q, int8_t e, int8_t q2, int8_t noncan, int zdrop, int8_t junc_bonus, int flag, const uint8_t *junc, ksw_extz_t *ez)
+				   int8_t q, int8_t e, int8_t q2, int8_t noncan, int zdrop, int end_bonus, int8_t junc_bonus, int8_t junc_pen, int flag, const uint8_t *junc, ksw_extz_t *ez)
 #endif // ~KSW_CPU_DISPATCH
 {
 #define __dp_code_block1 \
@@ -71,6 +71,7 @@ void ksw_exts2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 
 	ksw_reset_extz(ez);
 	if (m <= 1 || qlen <= 0 || tlen <= 0 || q2 <= q + e) return;
+	assert((flag & KSW_EZ_SPLICE_FOR) == 0 || (flag & KSW_EZ_SPLICE_REV) == 0); // can't be both set
 
 	zero_   = _mm_set1_epi8(0);
 	q_      = _mm_set1_epi8(q);
@@ -118,55 +119,100 @@ void ksw_exts2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 
 	// set the donor and acceptor arrays. TODO: this assumes 0/1/2/3 encoding!
 	if (flag & (KSW_EZ_SPLICE_FOR|KSW_EZ_SPLICE_REV)) {
-		int semi_cost = flag&KSW_EZ_SPLICE_FLANK? -noncan/2 : 0; // GTr or yAG is worth 0.5 bit; see PMID:18688272
-		memset(donor, -noncan, tlen_ * 16);
-		memset(acceptor, -noncan, tlen_ * 16);
+		const int sp0[4] = { 8, 15, 21, 30 };
+		int sp[4];
+		if (flag & KSW_EZ_SPLICE_CMPLX) {
+			for (t = 0; t < 4; ++t)
+				sp[t] = (int)((double)sp0[t] / 3. + .499);
+		} else {
+			sp[0] = flag&KSW_EZ_SPLICE_FLANK? noncan / 2 : 0;
+			sp[1] = sp[2] = sp[3] = noncan;
+		}
+		memset(donor,    -sp[3], tlen_ * 16);
+		memset(acceptor, -sp[3], tlen_ * 16);
 		if (!(flag & KSW_EZ_REV_CIGAR)) {
 			for (t = 0; t < tlen - 4; ++t) {
-				int can_type = 0; // type of canonical site: 0=none, 1=GT/AG only, 2=GTr/yAG
-				if ((flag & KSW_EZ_SPLICE_FOR) && target[t+1] == 2 && target[t+2] == 3) can_type = 1; // GTr...
-				if ((flag & KSW_EZ_SPLICE_REV) && target[t+1] == 1 && target[t+2] == 3) can_type = 1; // CTr...
-				if (can_type && (target[t+3] == 0 || target[t+3] == 2)) can_type = 2;
-				if (can_type) ((int8_t*)donor)[t] = can_type == 2? 0 : semi_cost;
+				int z = 3;
+				if (flag & KSW_EZ_SPLICE_FOR) {
+					if (target[t+1] == 2 && target[t+2] == 3)             // |GT.
+						z = target[t+3] == 0 || target[t+3] == 2? -1 : 0; // |GTr or not
+					else if (target[t+1] == 2 && target[t+2] == 1) z = 1; // |GC.
+					else if (target[t+1] == 0 && target[t+2] == 3) z = 2; // |AT.
+				} else if (flag & KSW_EZ_SPLICE_REV) {
+					if (target[t+1] == 1 && target[t+2] == 3)             // |CT. (revcomp of .AG|)
+						z = target[t+3] == 0 || target[t+3] == 2? -1 : 0;
+					else if (target[t+1] == 2 && target[t+2] == 3) z = 2; // |GT. (revcomp of .AC|)
+				}
+				((int8_t*)donor)[t] = z < 0? 0 : -sp[z];
 			}
-			if (junc)
-				for (t = 0; t < tlen - 1; ++t)
-					if (((flag & KSW_EZ_SPLICE_FOR) && (junc[t+1]&1)) || ((flag & KSW_EZ_SPLICE_REV) && (junc[t+1]&8)))
-						((int8_t*)donor)[t] += junc_bonus;
 			for (t = 2; t < tlen; ++t) {
-				int can_type = 0;
-				if ((flag & KSW_EZ_SPLICE_FOR) && target[t-1] == 0 && target[t] == 2) can_type = 1; // ...yAG
-				if ((flag & KSW_EZ_SPLICE_REV) && target[t-1] == 0 && target[t] == 1) can_type = 1; // ...yAC
-				if (can_type && (target[t-2] == 1 || target[t-2] == 3)) can_type = 2;
-				if (can_type) ((int8_t*)acceptor)[t] = can_type == 2? 0 : semi_cost;
+				int z = 3;
+				if (flag & KSW_EZ_SPLICE_FOR) {
+					if (target[t-1] == 0 && target[t] == 2)               // .AG|
+						z = target[t-2] == 1 || target[t-2] == 3? -1 : 0; // yAG| or not
+					else if (target[t-1] == 0 && target[t] == 1) z = 2;   // .AC|
+				} else if (flag & KSW_EZ_SPLICE_REV) {
+					if (target[t-1] == 0 && target[t] == 1)               // .AC| (revcomp of |GT.)
+						z = target[t-2] == 1 || target[t-2] == 3? -1 : 0; // yAC| or not
+					else if (target[t-1] == 2 && target[t] == 1) z = 1;   // .GC| (revcomp of |GC.)
+					else if (target[t-1] == 0 && target[t] == 3) z = 2;   // .AT| (revcomp of |AT.)
+				}
+				((int8_t*)acceptor)[t] = z < 0? 0 : -sp[z];
 			}
-			if (junc)
-				for (t = 0; t < tlen; ++t)
-					if (((flag & KSW_EZ_SPLICE_FOR) && (junc[t]&2)) || ((flag & KSW_EZ_SPLICE_REV) && (junc[t]&4)))
-						((int8_t*)acceptor)[t] += junc_bonus;
 		} else {
 			for (t = 0; t < tlen - 4; ++t) {
-				int can_type = 0; // type of canonical site: 0=none, 1=GT/AG only, 2=GTr/yAG
-				if ((flag & KSW_EZ_SPLICE_FOR) && target[t+1] == 2 && target[t+2] == 0) can_type = 1; // GAy...
-				if ((flag & KSW_EZ_SPLICE_REV) && target[t+1] == 1 && target[t+2] == 0) can_type = 1; // CAy...
-				if (can_type && (target[t+3] == 1 || target[t+3] == 3)) can_type = 2;
-				if (can_type) ((int8_t*)donor)[t] = can_type == 2? 0 : semi_cost;
+				int z = 3;
+				if (flag & KSW_EZ_SPLICE_FOR) {
+					if (target[t+1] == 2 && target[t+2] == 0)             // |GA. (rev of .AG|)
+						z = target[t+3] == 1 || target[t+3] == 3? -1 : 0;
+					else if (target[t+1] == 1 && target[t+2] == 0) z = 2; // |CA. (rev of .AC|)
+				} else if (flag & KSW_EZ_SPLICE_REV) {
+					if (target[t+1] == 1 && target[t+2] == 0)             // |CA. (comp of |GT.)
+						z = target[t+3] == 1 || target[t+3] == 3? -1 : 0;
+					else if (target[t+1] == 1 && target[t+2] == 2) z = 1; // |CG. (comp of |GC.)
+					else if (target[t+1] == 3 && target[t+2] == 0) z = 2; // |TA. (comp of |AT.)
+				}
+				((int8_t*)donor)[t] = z < 0? 0 : -sp[z];
 			}
-			if (junc)
-				for (t = 0; t < tlen - 1; ++t)
-					if (((flag & KSW_EZ_SPLICE_FOR) && (junc[t+1]&2)) || ((flag & KSW_EZ_SPLICE_REV) && (junc[t+1]&4)))
-						((int8_t*)donor)[t] += junc_bonus;
 			for (t = 2; t < tlen; ++t) {
-				int can_type = 0;
-				if ((flag & KSW_EZ_SPLICE_FOR) && target[t-1] == 3 && target[t] == 2) can_type = 1; // ...rTG
-				if ((flag & KSW_EZ_SPLICE_REV) && target[t-1] == 3 && target[t] == 1) can_type = 1; // ...rTC
-				if (can_type && (target[t-2] == 0 || target[t-2] == 2)) can_type = 2;
-				if (can_type) ((int8_t*)acceptor)[t] = can_type == 2? 0 : semi_cost;
+				int z = 3;
+				if (flag & KSW_EZ_SPLICE_FOR) {
+					if (target[t-1] == 3 && target[t] == 2)               // .TG| (rev of |GT.)
+						z = target[t-2] == 0 || target[t-2] == 2? -1 : 0;
+					else if (target[t-1] == 1 && target[t] == 2) z = 1;   // .CG| (rev of |GC.)
+					else if (target[t-1] == 3 && target[t] == 0) z = 2;   // .TA| (rev of |AT.)
+				} else if (flag & KSW_EZ_SPLICE_REV) {
+					if (target[t-1] == 3 && target[t] == 1)               // .TC| (comp of .AG|)
+						z = target[t-2] == 0 || target[t-2] == 2? -1 : 0;
+					else if (target[t-1] == 3 && target[t] == 2) z = 2;   // .TG| (comp of .AC|)
+				}
+				((int8_t*)acceptor)[t] = z < 0? 0 : -sp[z];
 			}
-			if (junc)
-				for (t = 0; t < tlen; ++t)
-					if (((flag & KSW_EZ_SPLICE_FOR) && (junc[t]&1)) || ((flag & KSW_EZ_SPLICE_REV) && (junc[t]&8)))
-						((int8_t*)acceptor)[t] += junc_bonus;
+		}
+	}
+
+	if (junc && (flag & KSW_EZ_SPLICE_SCORE)) { // junc[] keeps the donor score
+		uint8_t donor_val = !!(flag & KSW_EZ_SPLICE_FOR) == !(flag & KSW_EZ_REV_CIGAR)? 0 : 1;
+		for (t = 0; t < tlen - 1; ++t)
+			((int8_t*)donor)[t]    += junc[t+1] == 0xff || (junc[t+1]&1) !=  donor_val? -junc_pen : (int8_t)(junc[t+1]>>1) - (int8_t)KSW_SPSC_OFFSET;
+		for (t = 0; t < tlen - 1; ++t)
+			((int8_t*)acceptor)[t] += junc[t+1] == 0xff || (junc[t+1]&1) != !donor_val? -junc_pen : (int8_t)(junc[t+1]>>1) - (int8_t)KSW_SPSC_OFFSET;
+		//for (t = 0; t < tlen - 1; ++t) if (junc[t+1] != 0xff) fprintf(stderr, "Y2\t%d\t%d\t%c\t%d\n", ((int8_t*)donor)[t], ((int8_t*)acceptor)[t], "DA"[junc[t+1]&1], (int8_t)(junc[t+1]>>1) - (int8_t)KSW_SPSC_OFFSET);
+	} else if (junc) { // junc[] keeps the splice sites
+		if (!(flag & KSW_EZ_REV_CIGAR)) {
+			for (t = 0; t < tlen - 1; ++t)
+				if (((flag & KSW_EZ_SPLICE_FOR) && (junc[t+1]&1)) || ((flag & KSW_EZ_SPLICE_REV) && (junc[t+1]&8)))
+					((int8_t*)donor)[t] += junc_bonus;
+			for (t = 0; t < tlen; ++t)
+				if (((flag & KSW_EZ_SPLICE_FOR) && (junc[t]&2)) || ((flag & KSW_EZ_SPLICE_REV) && (junc[t]&4)))
+					((int8_t*)acceptor)[t] += junc_bonus;
+		} else {
+			for (t = 0; t < tlen - 1; ++t)
+				if (((flag & KSW_EZ_SPLICE_FOR) && (junc[t+1]&2)) || ((flag & KSW_EZ_SPLICE_REV) && (junc[t+1]&4)))
+					((int8_t*)donor)[t] += junc_bonus;
+			for (t = 0; t < tlen; ++t)
+				if (((flag & KSW_EZ_SPLICE_FOR) && (junc[t]&1)) || ((flag & KSW_EZ_SPLICE_REV) && (junc[t]&8)))
+					((int8_t*)acceptor)[t] += junc_bonus;
 		}
 	}
 
@@ -376,7 +422,7 @@ void ksw_exts2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 			} else H[0] = v8[0] - qe, max_H = H[0], max_t = 0; // special casing r==0
 			// update ez
 			if (en0 == tlen - 1 && H[en0] > ez->mte)
-				ez->mte = H[en0], ez->mte_q = r - en;
+				ez->mte = H[en0], ez->mte_q = r - en0;
 			if (r - st0 == qlen - 1 && H[st0] > ez->mqe)
 				ez->mqe = H[st0], ez->mqe_t = st0;
 			if (ksw_apply_zdrop(ez, 1, max_H, r, max_t, zdrop, 0)) break;
@@ -406,10 +452,14 @@ void ksw_exts2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 	if (!approx_max) kfree(km, H);
 	if (with_cigar) { // backtrack
 		int rev_cigar = !!(flag & KSW_EZ_REV_CIGAR);
-		if (!ez->zdropped && !(flag&KSW_EZ_EXTZ_ONLY))
+		if (!ez->zdropped && !(flag&KSW_EZ_EXTZ_ONLY)) {
 			ksw_backtrack(km, 1, rev_cigar, long_thres, (uint8_t*)p, off, off_end, n_col_*16, tlen-1, qlen-1, &ez->m_cigar, &ez->n_cigar, &ez->cigar);
-		else if (ez->max_t >= 0 && ez->max_q >= 0)
+		} else if (!ez->zdropped && (flag&KSW_EZ_EXTZ_ONLY) && ez->mqe + end_bonus > (int)ez->max) {
+			ez->reach_end = 1;
+			ksw_backtrack(km, 1, rev_cigar, long_thres, (uint8_t*)p, off, off_end, n_col_*16, ez->mqe_t, qlen-1, &ez->m_cigar, &ez->n_cigar, &ez->cigar);
+		} else if (ez->max_t >= 0 && ez->max_q >= 0) {
 			ksw_backtrack(km, 1, rev_cigar, long_thres, (uint8_t*)p, off, off_end, n_col_*16, ez->max_t, ez->max_q, &ez->m_cigar, &ez->n_cigar, &ez->cigar);
+		}
 		kfree(km, mem2); kfree(km, off);
 	}
 }

ksw2_extz2_sse.c

@@ -269,7 +269,7 @@ void ksw_extz2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 			} else H[0] = v8[0] - qe - qe, max_H = H[0], max_t = 0; // special casing r==0
 			// update ez
 			if (en0 == tlen - 1 && H[en0] > ez->mte)
-				ez->mte = H[en0], ez->mte_q = r - en;
+				ez->mte = H[en0], ez->mte_q = r - en0;
 			if (r - st0 == qlen - 1 && H[st0] > ez->mqe)
 				ez->mqe = H[st0], ez->mqe_t = st0;
 			if (ksw_apply_zdrop(ez, 1, max_H, r, max_t, zdrop, e)) break;

lchain.c

@@ -6,7 +6,25 @@
 #include "kalloc.h"
 #include "krmq.h"
 
-uint64_t *mg_chain_backtrack(void *km, int64_t n, const int32_t *f, const int64_t *p, int32_t *v, int32_t *t, int32_t min_cnt, int32_t min_sc, int32_t *n_u_, int32_t *n_v_)
+static int64_t mg_chain_bk_end(int32_t max_drop, const mm128_t *z, const int32_t *f, const int64_t *p, int32_t *t, int64_t k)
+{
+	int64_t i = z[k].y, end_i = -1, max_i = i;
+	int32_t max_s = 0;
+	if (i < 0 || t[i] != 0) return i;
+	do {
+		int32_t s;
+		t[i] = 2;
+		end_i = i = p[i];
+		s = i < 0? z[k].x : (int32_t)z[k].x - f[i];
+		if (s > max_s) max_s = s, max_i = i;
+		else if (max_s - s > max_drop) break;
+	} while (i >= 0 && t[i] == 0);
+	for (i = z[k].y; i >= 0 && i != end_i; i = p[i]) // reset modified t[]
+		t[i] = 0;
+	return max_i;
+}
+
+uint64_t *mg_chain_backtrack(void *km, int64_t n, const int32_t *f, const int64_t *p, int32_t *v, int32_t *t, int32_t min_cnt, int32_t min_sc, int32_t max_drop, int32_t *n_u_, int32_t *n_v_)
 {
 	mm128_t *z;
 	uint64_t *u;
@@ -17,33 +35,39 @@ uint64_t *mg_chain_backtrack(void *km, int64_t n, const int32_t *f, const int64_
 	for (i = 0, n_z = 0; i < n; ++i) // precompute n_z
 		if (f[i] >= min_sc) ++n_z;
 	if (n_z == 0) return 0;
-	KMALLOC(km, z, n_z);
+	z = Kmalloc(km, mm128_t, n_z);
 	for (i = 0, k = 0; i < n; ++i) // populate z[]
 		if (f[i] >= min_sc) z[k].x = f[i], z[k++].y = i;
 	radix_sort_128x(z, z + n_z);
 
 	memset(t, 0, n * 4);
 	for (k = n_z - 1, n_v = n_u = 0; k >= 0; --k) { // precompute n_u
-		int64_t n_v0 = n_v;
-		int32_t sc;
-		for (i = z[k].y; i >= 0 && t[i] == 0; i = p[i])
-			++n_v, t[i] = 1;
-		sc = i < 0? z[k].x : (int32_t)z[k].x - f[i];
-		if (sc >= min_sc && n_v > n_v0 && n_v - n_v0 >= min_cnt)
-			++n_u;
-		else n_v = n_v0;
+		if (t[z[k].y] == 0) {
+			int64_t n_v0 = n_v, end_i;
+			int32_t sc;
+			end_i = mg_chain_bk_end(max_drop, z, f, p, t, k);
+			for (i = z[k].y; i != end_i; i = p[i])
+				++n_v, t[i] = 1;
+			sc = i < 0? z[k].x : (int32_t)z[k].x - f[i];
+			if (sc >= min_sc && n_v > n_v0 && n_v - n_v0 >= min_cnt)
+				++n_u;
+			else n_v = n_v0;
+		}
 	}
-	KMALLOC(km, u, n_u);
+	u = Kmalloc(km, uint64_t, n_u);
 	memset(t, 0, n * 4);
 	for (k = n_z - 1, n_v = n_u = 0; k >= 0; --k) { // populate u[]
-		int64_t n_v0 = n_v;
-		int32_t sc;
-		for (i = z[k].y; i >= 0 && t[i] == 0; i = p[i])
-			v[n_v++] = i, t[i] = 1;
-		sc = i < 0? z[k].x : (int32_t)z[k].x - f[i];
-		if (sc >= min_sc && n_v > n_v0 && n_v - n_v0 >= min_cnt)
-			u[n_u++] = (uint64_t)sc << 32 | (n_v - n_v0);
-		else n_v = n_v0;
+		if (t[z[k].y] == 0) {
+			int64_t n_v0 = n_v, end_i;
+			int32_t sc;
+			end_i = mg_chain_bk_end(max_drop, z, f, p, t, k);
+			for (i = z[k].y; i != end_i; i = p[i])
+				v[n_v++] = i, t[i] = 1;
+			sc = i < 0? z[k].x : (int32_t)z[k].x - f[i];
+			if (sc >= min_sc && n_v > n_v0 && n_v - n_v0 >= min_cnt)
+				u[n_u++] = (uint64_t)sc << 32 | (n_v - n_v0);
+			else n_v = n_v0;
+		}
 	}
 	kfree(km, z);
 	assert(n_v < INT32_MAX);
@@ -58,7 +82,7 @@ static mm128_t *compact_a(void *km, int32_t n_u, uint64_t *u, int32_t n_v, int32
 	int64_t i, j, k;
 
 	// write the result to b[]
-	KMALLOC(km, b, n_v);
+	b = Kmalloc(km, mm128_t, n_v);
 	for (i = 0, k = 0; i < n_u; ++i) {
 		int32_t k0 = k, ni = (int32_t)u[i];
 		for (j = 0; j < ni; ++j)
@@ -67,13 +91,13 @@ static mm128_t *compact_a(void *km, int32_t n_u, uint64_t *u, int32_t n_v, int32
 	kfree(km, v);
 
 	// sort u[] and a[] by the target position, such that adjacent chains may be joined
-	KMALLOC(km, w, n_u);
+	w = Kmalloc(km, mm128_t, n_u);
 	for (i = k = 0; i < n_u; ++i) {
 		w[i].x = b[k].x, w[i].y = (uint64_t)k<<32|i;
 		k += (int32_t)u[i];
 	}
 	radix_sort_128x(w, w + n_u);
-	KMALLOC(km, u2, n_u);
+	u2 = Kmalloc(km, uint64_t, n_u);
 	for (i = k = 0; i < n_u; ++i) {
 		int32_t j = (int32_t)w[i].y, n = (int32_t)u[j];
 		u2[i] = u[j];
@@ -114,7 +138,7 @@ static inline int32_t comput_sc(const mm128_t *ai, const mm128_t *aj, int32_t ma
 }
 
 /* Input:
- *   a[].x: tid<<33 | rev<<32 | tpos
+ *   a[].x: rev<<63 | tid<<32 | tpos
  *   a[].y: flags<<40 | q_span<<32 | q_pos
  * Output:
  *   n_u: #chains
@@ -124,8 +148,8 @@ static inline int32_t comput_sc(const mm128_t *ai, const mm128_t *aj, int32_t ma
 mm128_t *mg_lchain_dp(int max_dist_x, int max_dist_y, int bw, int max_skip, int max_iter, int min_cnt, int min_sc, float chn_pen_gap, float chn_pen_skip,
 					  int is_cdna, int n_seg, int64_t n, mm128_t *a, int *n_u_, uint64_t **_u, void *km)
 { // TODO: make sure this works when n has more than 32 bits
-	int32_t *f, *t, *v, n_u, n_v, mmax_f = 0;
-	int64_t *p, i, j, max_ii, st = 0, n_iter = 0;
+	int32_t *f, *t, *v, n_u, n_v, mmax_f = 0, max_drop = bw;
+	int64_t *p, i, j, max_ii, st = 0;
 	uint64_t *u;
 
 	if (_u) *_u = 0, *n_u_ = 0;
@@ -135,10 +159,11 @@ mm128_t *mg_lchain_dp(int max_dist_x, int max_dist_y, int bw, int max_skip, int
 	}
 	if (max_dist_x < bw) max_dist_x = bw;
 	if (max_dist_y < bw && !is_cdna) max_dist_y = bw;
-	KMALLOC(km, p, n);
-	KMALLOC(km, f, n);
-	KMALLOC(km, v, n);
-	KCALLOC(km, t, n);
+	if (is_cdna) max_drop = INT32_MAX;
+	p = Kmalloc(km, int64_t, n);
+	f = Kmalloc(km, int32_t, n);
+	v = Kmalloc(km, int32_t, n);
+	t = Kcalloc(km, int32_t, n);
 
 	// fill the score and backtrack arrays
 	for (i = 0, max_ii = -1; i < n; ++i) {
@@ -149,7 +174,6 @@ mm128_t *mg_lchain_dp(int max_dist_x, int max_dist_y, int bw, int max_skip, int
 		for (j = i - 1; j >= st; --j) {
 			int32_t sc;
 			sc = comput_sc(&a[i], &a[j], max_dist_x, max_dist_y, bw, chn_pen_gap, chn_pen_skip, is_cdna, n_seg);
-			++n_iter;
 			if (sc == INT32_MIN) continue;
 			sc += f[j];
 			if (sc > max_f) {
@@ -179,9 +203,10 @@ mm128_t *mg_lchain_dp(int max_dist_x, int max_dist_y, int bw, int max_skip, int
 		if (max_ii < 0 || (a[i].x - a[max_ii].x <= (int64_t)max_dist_x && f[max_ii] < f[i]))
 			max_ii = i;
 		if (mmax_f < max_f) mmax_f = max_f;
+		//fprintf(stderr, "X1\t%ld\t%ld:%d\t%ld\t%ld:%d\t%ld\t%ld\n", (long)i, (long)(a[i].x>>32), (int32_t)a[i].x, (long)max_j, max_j<0?-1L:(long)(a[max_j].x>>32), max_j<0?-1:(int32_t)a[max_j].x, (long)max_f, (long)v[i]);
 	}
 
-	u = mg_chain_backtrack(km, n, f, p, v, t, min_cnt, min_sc, &n_u, &n_v);
+	u = mg_chain_backtrack(km, n, f, p, v, t, min_cnt, min_sc, max_drop, &n_u, &n_v);
 	*n_u_ = n_u, *_u = u; // NB: note that u[] may not be sorted by score here
 	kfree(km, p); kfree(km, f); kfree(km, t);
 	if (n_u == 0) {
@@ -225,8 +250,8 @@ static inline int32_t comput_sc_simple(const mm128_t *ai, const mm128_t *aj, flo
 mm128_t *mg_lchain_rmq(int max_dist, int max_dist_inner, int bw, int max_chn_skip, int cap_rmq_size, int min_cnt, int min_sc, float chn_pen_gap, float chn_pen_skip,
 					   int64_t n, mm128_t *a, int *n_u_, uint64_t **_u, void *km)
 {
-	int32_t *f,*t, *v, n_u, n_v, mmax_f = 0, max_rmq_size = 0;
-	int64_t *p, i, i0, st = 0, st_inner = 0, n_iter = 0;
+	int32_t *f,*t, *v, n_u, n_v, mmax_f = 0, max_rmq_size = 0, max_drop = bw;
+	int64_t *p, i, i0, st = 0, st_inner = 0;
 	uint64_t *u;
 	lc_elem_t *root = 0, *root_inner = 0;
 	void *mem_mp = 0;
@@ -238,11 +263,12 @@ mm128_t *mg_lchain_rmq(int max_dist, int max_dist_inner, int bw, int max_chn_ski
 		return 0;
 	}
 	if (max_dist < bw) max_dist = bw;
-	if (max_dist_inner <= 0 || max_dist_inner >= max_dist) max_dist_inner = 0;
-	KMALLOC(km, p, n);
-	KMALLOC(km, f, n);
-	KCALLOC(km, t, n);
-	KMALLOC(km, v, n);
+	if (max_dist_inner < 0) max_dist_inner = 0;
+	if (max_dist_inner > max_dist) max_dist_inner = max_dist;
+	p = Kmalloc(km, int64_t, n);
+	f = Kmalloc(km, int32_t, n);
+	t = Kcalloc(km, int32_t, n);
+	v = Kmalloc(km, int32_t, n);
 	mem_mp = km_init2(km, 0x10000);
 	mp = kmp_init_rmq(mem_mp);
 
@@ -300,12 +326,11 @@ mm128_t *mg_lchain_rmq(int max_dist, int max_dist_inner, int bw, int max_chn_ski
 				krmq_interval(lc_elem, root_inner, &s, &lo, &hi);
 				if (lo) {
 					const lc_elem_t *q;
-					int32_t width, n_rmq_iter = 0;
+					int32_t width;
 					krmq_itr_t(lc_elem) itr;
 					krmq_itr_find(lc_elem, root_inner, lo, &itr);
 					while ((q = krmq_at(&itr)) != 0) {
 						if (q->y < (int32_t)a[i].y - max_dist_inner) break;
-						++n_rmq_iter;
 						j = q->i;
 						sc = f[j] + comput_sc_simple(&a[i], &a[j], chn_pen_gap, chn_pen_skip, 0, &width);
 						if (width <= bw) {
@@ -320,7 +345,6 @@ mm128_t *mg_lchain_rmq(int max_dist, int max_dist_inner, int bw, int max_chn_ski
 						}
 						if (!krmq_itr_prev(lc_elem, &itr)) break;
 					}
-					n_iter += n_rmq_iter;
 				}
 			}
 		}
@@ -333,7 +357,7 @@ mm128_t *mg_lchain_rmq(int max_dist, int max_dist_inner, int bw, int max_chn_ski
 	}
 	km_destroy(mem_mp);
 
-	u = mg_chain_backtrack(km, n, f, p, v, t, min_cnt, min_sc, &n_u, &n_v);
+	u = mg_chain_backtrack(km, n, f, p, v, t, min_cnt, min_sc, max_drop, &n_u, &n_v);
 	*n_u_ = n_u, *_u = u; // NB: note that u[] may not be sorted by score here
 	kfree(km, p); kfree(km, f); kfree(km, t);
 	if (n_u == 0) {

main.c

@@ -7,8 +7,6 @@
 #include "mmpriv.h"
 #include "ketopt.h"
 
-#define MM_VERSION "2.23-r1111"
-
 #ifdef __linux__
 #include <sys/resource.h>
 #include <sys/time.h>
@@ -37,12 +35,12 @@ static ko_longopt_t long_options[] = {
 	{ "splice",         ko_no_argument,       310 },
 	{ "cost-non-gt-ag", ko_required_argument, 'C' },
 	{ "no-long-join",   ko_no_argument,       312 },
-	{ "sr",             ko_no_argument,       313 },
+	{ "sr",             ko_optional_argument, 313 },
 	{ "frag",           ko_required_argument, 314 },
 	{ "secondary",      ko_required_argument, 315 },
 	{ "cs",             ko_optional_argument, 316 },
 	{ "end-bonus",      ko_required_argument, 317 },
-	{ "no-pairing",     ko_no_argument,       318 },
+	{ "no-pairing",     ko_no_argument,       318 }, // deprecated but reserved for backward compatibility
 	{ "splice-flank",   ko_required_argument, 319 },
 	{ "idx-no-seq",     ko_no_argument,       320 },
 	{ "end-seed-pen",   ko_required_argument, 321 },
@@ -76,6 +74,18 @@ static ko_longopt_t long_options[] = {
 	{ "cap-kalloc",     ko_required_argument, 349 },
 	{ "q-occ-frac",     ko_required_argument, 350 },
 	{ "chain-skip-scale",ko_required_argument,351 },
+	{ "print-chains",   ko_no_argument,       352 },
+	{ "no-hash-name",   ko_no_argument,       353 },
+	{ "secondary-seq",  ko_no_argument,       354 },
+	{ "ds",             ko_no_argument,       355 },
+	{ "rmq-inner",      ko_required_argument, 356 },
+	{ "spsc",           ko_required_argument, 357 },
+	{ "junc-pen",       ko_required_argument, 358 },
+	{ "pairing",        ko_required_argument, 359 },
+	{ "jump-min-match", ko_required_argument, 360 },
+	{ "write-junc",     ko_no_argument,       361 },
+	{ "pass1",          ko_required_argument, 362 },
+	{ "dbg-seed-occ",   ko_no_argument,       501 },
 	{ "help",           ko_no_argument,       'h' },
 	{ "max-intron-len", ko_required_argument, 'G' },
 	{ "version",        ko_no_argument,       'V' },
@@ -119,12 +129,12 @@ static inline void yes_or_no(mm_mapopt_t *opt, int64_t flag, int long_idx, const
 
 int main(int argc, char *argv[])
 {
-	const char *opt_str = "2aSDw:k:K:t:r:f:Vv:g:G:I:d:XT:s:x:Hcp:M:n:z:A:B:O:E:m:N:Qu:R:hF:LC:yYPo:e:U:";
+	const char *opt_str = "2aSDw:k:K:t:r:f:Vv:g:G:I:d:XT:s:x:Hcp:M:n:z:A:B:b:O:E:m:N:Qu:R:hF:LC:yYPo:e:U:J:j:";
 	ketopt_t o = KETOPT_INIT;
 	mm_mapopt_t opt;
 	mm_idxopt_t ipt;
 	int i, c, n_threads = 3, n_parts, old_best_n = -1;
-	char *fnw = 0, *rg = 0, *junc_bed = 0, *s, *alt_list = 0;
+	char *fnw = 0, *rg = 0, *fn_bed_junc = 0, *fn_bed_jump = 0, *fn_bed_pass1 = 0, *fn_spsc = 0, *s, *alt_list = 0;
 	FILE *fp_help = stderr;
 	mm_idx_reader_t *idx_rdr;
 	mm_idx_t *mi;
@@ -177,6 +187,7 @@ int main(int argc, char *argv[])
 		else if (c == 'm') opt.min_chain_score = atoi(o.arg);
 		else if (c == 'A') opt.a = atoi(o.arg);
 		else if (c == 'B') opt.b = atoi(o.arg);
+		else if (c == 'b') opt.transition = atoi(o.arg);
 		else if (c == 's') opt.min_dp_max = atoi(o.arg);
 		else if (c == 'C') opt.noncan = atoi(o.arg);
 		else if (c == 'I') ipt.batch_size = mm_parse_num(o.arg);
@@ -185,7 +196,13 @@ int main(int argc, char *argv[])
 		else if (c == 'R') rg = o.arg;
 		else if (c == 'h') fp_help = stdout;
 		else if (c == '2') opt.flag |= MM_F_2_IO_THREADS;
-		else if (c == 'o') {
+		else if (c == 'j') fn_bed_jump = o.arg;
+		else if (c == 'J') {
+			int t;
+			t = atoi(o.arg);
+			if (t == 0) opt.flag |= MM_F_SPLICE_OLD;
+			else if (t == 1) opt.flag &= ~MM_F_SPLICE_OLD;
+		} else if (c == 'o') {
 			if (strcmp(o.arg, "-") != 0) {
 				if (freopen(o.arg, "wb", stdout) == NULL) {
 					fprintf(stderr, "[ERROR]\033[1;31m failed to write the output to file '%s'\033[0m: %s\n", o.arg, strerror(errno));
@@ -204,9 +221,8 @@ int main(int argc, char *argv[])
 		else if (c == 309) mm_dbg_flag |= MM_DBG_PRINT_QNAME | MM_DBG_PRINT_ALN_SEQ, n_threads = 1; // --print-aln-seq
 		else if (c == 310) opt.flag |= MM_F_SPLICE; // --splice
 		else if (c == 312) opt.flag |= MM_F_NO_LJOIN; // --no-long-join
-		else if (c == 313) opt.flag |= MM_F_SR; // --sr
 		else if (c == 317) opt.end_bonus = atoi(o.arg); // --end-bonus
-		else if (c == 318) opt.flag |= MM_F_INDEPEND_SEG; // --no-pairing
+		else if (c == 318) opt.flag |= MM_F_INDEPEND_SEG; // --no-pairing (deprecated)
 		else if (c == 320) ipt.flag |= MM_I_NO_SEQ; // --idx-no-seq
 		else if (c == 321) opt.anchor_ext_shift = atoi(o.arg); // --end-seed-pen
 		else if (c == 322) opt.flag |= MM_F_FOR_ONLY; // --for-only
@@ -222,8 +238,9 @@ int main(int argc, char *argv[])
 		else if (c == 336) opt.flag |= MM_F_HARD_MLEVEL; // --hard-mask-level
 		else if (c == 337) opt.max_sw_mat = mm_parse_num(o.arg); // --cap-sw-mat
 		else if (c == 338) opt.max_qlen = mm_parse_num(o.arg); // --max-qlen
-		else if (c == 340) junc_bed = o.arg; // --junc-bed
+		else if (c == 340) fn_bed_junc = o.arg; // --junc-bed
 		else if (c == 341) opt.junc_bonus = atoi(o.arg); // --junc-bonus
+		else if (c == 358) opt.junc_pen = atoi(o.arg); // --junc-pen
 		else if (c == 342) opt.flag |= MM_F_SAM_HIT_ONLY; // --sam-hit-only
 		else if (c == 343) opt.chain_gap_scale = atof(o.arg); // --chain-gap-scale
 		else if (c == 351) opt.chain_skip_scale = atof(o.arg); // --chain-skip-scale
@@ -233,8 +250,29 @@ int main(int argc, char *argv[])
 		else if (c == 348) opt.flag |= MM_F_QSTRAND | MM_F_NO_INV; // --qstrand
 		else if (c == 349) opt.cap_kalloc = mm_parse_num(o.arg); // --cap-kalloc
 		else if (c == 350) opt.q_occ_frac = atof(o.arg); // --q-occ-frac
+		else if (c == 352) mm_dbg_flag |= MM_DBG_PRINT_CHAIN; // --print-chains
+		else if (c == 353) opt.flag |= MM_F_NO_HASH_NAME; // --no-hash-name
+		else if (c == 354) opt.flag |= MM_F_SECONDARY_SEQ; // --secondary-seq
+		else if (c == 355) opt.flag |= MM_F_OUT_DS; // --ds
+		else if (c == 356) opt.rmq_inner_dist = mm_parse_num(o.arg); // --rmq-inner
+		else if (c == 357) fn_spsc = o.arg; // --spsc
+		else if (c == 360) opt.jump_min_match = mm_parse_num(o.arg); // --jump-min-match
+		else if (c == 361) opt.flag |= MM_F_OUT_JUNC | MM_F_CIGAR; // --write-junc
+		else if (c == 362) fn_bed_pass1 = o.arg; // --jump-pass1
+		else if (c == 501) mm_dbg_flag |= MM_DBG_SEED_FREQ; // --dbg-seed-occ
 		else if (c == 330) {
 			fprintf(stderr, "[WARNING] \033[1;31m --lj-min-ratio has been deprecated.\033[0m\n");
+		} else if (c == 313) { // --sr
+			if (o.arg == 0 || strcmp(o.arg, "dna") == 0) {
+				opt.flag |= MM_F_SR;
+			} else if (strcmp(o.arg, "rna") == 0) {
+				opt.flag |= MM_F_SR_RNA;
+			} else if (strcmp(o.arg, "no") == 0) {
+				opt.flag &= ~(uint64_t)(MM_F_SR|MM_F_SR_RNA);
+			} else if (mm_verbose >= 2) {
+				opt.flag |= MM_F_SR;
+				fprintf(stderr, "[WARNING]\033[1;31m --sr only takes 'dna' or 'rna'. Invalid values are assumed to be 'dna'.\033[0m\n");
+			}
 		} else if (c == 314) { // --frag
 			yes_or_no(&opt, MM_F_FRAG_MODE, o.longidx, o.arg, 1);
 		} else if (c == 315) { // --secondary
@@ -257,7 +295,16 @@ int main(int argc, char *argv[])
 		} else if (c == 326) { // --dual
 			yes_or_no(&opt, MM_F_NO_DUAL, o.longidx, o.arg, 0);
 		} else if (c == 347) { // --rmq
-			yes_or_no(&opt, MM_F_RMQ, o.longidx, o.arg, 1);
+			if (o.arg) yes_or_no(&opt, MM_F_RMQ, o.longidx, o.arg, 1);
+			else opt.flag |= MM_F_RMQ;
+		} else if (c == 359) { // --pairing
+			if (strcmp(o.arg, "no") == 0) opt.flag |= MM_F_INDEPEND_SEG;
+			else if (strcmp(o.arg, "weak") == 0) opt.flag |= MM_F_WEAK_PAIRING, opt.flag &= ~(uint64_t)MM_F_INDEPEND_SEG;
+			else {
+				if (strcmp(o.arg, "strong") != 0 && mm_verbose >= 2)
+					fprintf(stderr, "[WARNING]\033[1;31m unrecognized argument for --pairing; assuming 'strong'.\033[0m\n");
+				opt.flag &= ~(uint64_t)(MM_F_INDEPEND_SEG|MM_F_WEAK_PAIRING);
+			}
 		} else if (c == 'S') {
 			opt.flag |= MM_F_OUT_CS | MM_F_CIGAR | MM_F_OUT_CS_LONG;
 			if (mm_verbose >= 2)
@@ -298,10 +345,6 @@ int main(int argc, char *argv[])
 			if (*s == ',') opt.e2 = strtol(s + 1, &s, 10);
 		}
 	}
-	if ((opt.flag & MM_F_SPLICE) && (opt.flag & MM_F_FRAG_MODE)) {
-		fprintf(stderr, "[ERROR]\033[1;31m --splice and --frag should not be specified at the same time.\033[0m\n");
-		return 1;
-	}
 	if (!fnw && !(opt.flag&MM_F_CIGAR))
 		ipt.flag |= MM_I_NO_SEQ;
 	if (mm_check_opt(&ipt, &opt) < 0)
@@ -318,7 +361,7 @@ int main(int argc, char *argv[])
 		fprintf(fp_help, "    -H           use homopolymer-compressed k-mer (preferrable for PacBio)\n");
 		fprintf(fp_help, "    -k INT       k-mer size (no larger than 28) [%d]\n", ipt.k);
 		fprintf(fp_help, "    -w INT       minimizer window size [%d]\n", ipt.w);
-		fprintf(fp_help, "    -I NUM       split index for every ~NUM input bases [4G]\n");
+		fprintf(fp_help, "    -I NUM       split index for every ~NUM input bases [8G]\n");
 		fprintf(fp_help, "    -d FILE      dump index to FILE []\n");
 		fprintf(fp_help, "  Mapping:\n");
 		fprintf(fp_help, "    -f FLOAT     filter out top FLOAT fraction of repetitive minimizers [%g]\n", opt.mid_occ_frac);
@@ -340,6 +383,8 @@ int main(int argc, char *argv[])
 		fprintf(fp_help, "    -z INT[,INT] Z-drop score and inversion Z-drop score [%d,%d]\n", opt.zdrop, opt.zdrop_inv);
 		fprintf(fp_help, "    -s INT       minimal peak DP alignment score [%d]\n", opt.min_dp_max);
 		fprintf(fp_help, "    -u CHAR      how to find GT-AG. f:transcript strand, b:both strands, n:don't match GT-AG [n]\n");
+		fprintf(fp_help, "    -J INT       splice mode. 0: original minimap2 model; 1: miniprot model [1]\n");
+		fprintf(fp_help, "    -j FILE      junctions in BED12 to extend *short* RNA-seq alignment []\n");
 		fprintf(fp_help, "  Input/Output:\n");
 		fprintf(fp_help, "    -a           output in the SAM format (PAF by default)\n");
 		fprintf(fp_help, "    -o FILE      output alignments to FILE [stdout]\n");
@@ -347,21 +392,24 @@ int main(int argc, char *argv[])
 		fprintf(fp_help, "    -R STR       SAM read group line in a format like '@RG\\tID:foo\\tSM:bar' []\n");
 		fprintf(fp_help, "    -c           output CIGAR in PAF\n");
 		fprintf(fp_help, "    --cs[=STR]   output the cs tag; STR is 'short' (if absent) or 'long' [none]\n");
+		fprintf(fp_help, "    --ds         output the ds tag, which is an extension to cs\n");
 		fprintf(fp_help, "    --MD         output the MD tag\n");
 		fprintf(fp_help, "    --eqx        write =/X CIGAR operators\n");
 		fprintf(fp_help, "    -Y           use soft clipping for supplementary alignments\n");
+		fprintf(fp_help, "    -y           copy FASTA/Q comments to output SAM\n");
 		fprintf(fp_help, "    -t INT       number of threads [%d]\n", n_threads);
 		fprintf(fp_help, "    -K NUM       minibatch size for mapping [500M]\n");
 //		fprintf(fp_help, "    -v INT       verbose level [%d]\n", mm_verbose);
 		fprintf(fp_help, "    --version    show version number\n");
 		fprintf(fp_help, "  Preset:\n");
 		fprintf(fp_help, "    -x STR       preset (always applied before other options; see minimap2.1 for details) []\n");
-		fprintf(fp_help, "                 - map-pb/map-ont - PacBio CLR/Nanopore vs reference mapping\n");
-		fprintf(fp_help, "                 - map-hifi - PacBio HiFi reads vs reference mapping\n");
-		fprintf(fp_help, "                 - ava-pb/ava-ont - PacBio/Nanopore read overlap\n");
+		fprintf(fp_help, "                 - lr:hq - accurate long reads (error rate <1%%) against a reference genome\n");
+		fprintf(fp_help, "                 - splice/splice:hq - spliced alignment for long reads/accurate long reads\n");
+		fprintf(fp_help, "                 - splice:sr - spliced alignment for short RNA-seq reads\n");
 		fprintf(fp_help, "                 - asm5/asm10/asm20 - asm-to-ref mapping, for ~0.1/1/5%% sequence divergence\n");
-		fprintf(fp_help, "                 - splice/splice:hq - long-read/Pacbio-CCS spliced alignment\n");
-		fprintf(fp_help, "                 - sr - genomic short-read mapping\n");
+		fprintf(fp_help, "                 - sr - short reads against a reference\n");
+		fprintf(fp_help, "                 - map-pb/map-hifi/map-ont/map-iclr - CLR/HiFi/Nanopore/ICLR vs reference mapping\n");
+		fprintf(fp_help, "                 - ava-pb/ava-ont - PacBio CLR/Nanopore read overlap\n");
 		fprintf(fp_help, "\nSee `man ./minimap2.1' for detailed description of these and other advanced command-line options.\n");
 		return fp_help == stdout? 0 : 1;
 	}
@@ -412,7 +460,26 @@ int main(int argc, char *argv[])
 					__func__, realtime() - mm_realtime0, cputime() / (realtime() - mm_realtime0), mi->n_seq);
 		if (argc != o.ind + 1) mm_mapopt_update(&opt, mi);
 		if (mm_verbose >= 3) mm_idx_stat(mi);
-		if (junc_bed) mm_idx_bed_read(mi, junc_bed, 1);
+		if (fn_bed_junc) {
+			mm_idx_bed_read(mi, fn_bed_junc, 1);
+			if (mi->I == 0 && mm_verbose >= 2)
+				fprintf(stderr, "[WARNING] failed to load the junction BED file\n");
+		}
+		if (fn_bed_jump) {
+			mm_idx_jjump_read(mi, fn_bed_jump, MM_JUNC_ANNO, -1);
+			if (mi->J == 0 && mm_verbose >= 2)
+				fprintf(stderr, "[WARNING] failed to load the jump BED file\n");
+		}
+		if (fn_bed_pass1) {
+			mm_idx_jjump_read(mi, fn_bed_pass1, MM_JUNC_MISC, 5);
+			if (mi->J == 0 && mm_verbose >= 2)
+				fprintf(stderr, "[WARNING] failed to load the pass-1 jump BED file\n");
+		}
+		if (fn_spsc) {
+			mm_idx_spsc_read(mi, fn_spsc, mm_max_spsc_bonus(&opt));
+			if (mi->spsc == 0 && mm_verbose >= 2)
+				fprintf(stderr, "[WARNING] failed to load the splice score file\n");
+		}
 		if (alt_list) mm_idx_alt_read(mi, alt_list);
 		if (argc - (o.ind + 1) == 0) {
 			mm_idx_destroy(mi);

map.c

@@ -10,11 +10,6 @@
 #include "bseq.h"
 #include "khash.h"
 
-struct mm_tbuf_s {
-	void *km;
-	int rep_len, frag_gap;
-};
-
 mm_tbuf_t *mm_tbuf_init(void)
 {
 	mm_tbuf_t *b;
@@ -229,10 +224,10 @@ static mm_reg1_t *align_regs(const mm_mapopt_t *opt, const mm_idx_t *mi, void *k
 	return regs;
 }
 
-void mm_map_frag(const mm_idx_t *mi, int n_segs, const int *qlens, const char **seqs, int *n_regs, mm_reg1_t **regs, mm_tbuf_t *b, const mm_mapopt_t *opt, const char *qname)
+void mm_map_frag_core(const mm_idx_t *mi, int n_segs, const int *qlens, const char **seqs, int *n_regs, mm_reg1_t **regs, mm_tbuf_t *b, const mm_mapopt_t *opt, const char *qname)
 {
 	int i, j, rep_len, qlen_sum, n_regs0, n_mini_pos;
-	int max_chain_gap_qry, max_chain_gap_ref, is_splice = !!(opt->flag & MM_F_SPLICE), is_sr = !!(opt->flag & MM_F_SR);
+	int max_chain_gap_qry, max_chain_gap_ref, is_splice = !!(opt->flag & MM_F_SPLICE), is_sr = !!(opt->flag & MM_F_SR), is_sr_rna = !!(opt->flag & MM_F_SR_RNA);
 	uint32_t hash;
 	int64_t n_a;
 	uint64_t *u, *mini_pos;
@@ -248,7 +243,7 @@ void mm_map_frag(const mm_idx_t *mi, int n_segs, const int *qlens, const char **
 	if (qlen_sum == 0 || n_segs <= 0 || n_segs > MM_MAX_SEG) return;
 	if (opt->max_qlen > 0 && qlen_sum > opt->max_qlen) return;
 
-	hash  = qname? __ac_X31_hash_string(qname) : 0;
+	hash  = qname && !(opt->flag & MM_F_NO_HASH_NAME)? __ac_X31_hash_string(qname) : 0;
 	hash ^= __ac_Wang_hash(qlen_sum) + __ac_Wang_hash(opt->seed);
 	hash  = __ac_Wang_hash(hash);
 
@@ -328,7 +323,7 @@ void mm_map_frag(const mm_idx_t *mi, int n_segs, const int *qlens, const char **
 		mm_hit_sort(b->km, &n_regs0, regs0, opt->alt_drop); // this step can be merged into mm_gen_regs(); will do if this shows up in profile
 	}
 
-	if (mm_dbg_flag & MM_DBG_PRINT_SEED)
+	if (mm_dbg_flag & (MM_DBG_PRINT_SEED|MM_DBG_PRINT_CHAIN))
 		for (j = 0; j < n_regs0; ++j)
 			for (i = regs0[j].as; i < regs0[j].as + regs0[j].cnt; ++i)
 				fprintf(stderr, "CN\t%d\t%s\t%d\t%c\t%d\t%d\t%d\n", j, mi->seq[a[i].x<<1>>33].name, (int32_t)a[i].x, "+-"[a[i].x>>63], (int32_t)a[i].y, (int32_t)(a[i].y>>32&0xff),
@@ -343,7 +338,7 @@ void mm_map_frag(const mm_idx_t *mi, int n_segs, const int *qlens, const char **
 	if (n_segs == 1) { // uni-segment
 		regs0 = align_regs(opt, mi, b->km, qlens[0], seqs[0], &n_regs0, regs0, a);
 		regs0 = (mm_reg1_t*)realloc(regs0, sizeof(*regs0) * n_regs0);
-		mm_set_mapq(b->km, n_regs0, regs0, opt->min_chain_score, opt->a, rep_len, is_sr);
+		mm_set_mapq2(b->km, n_regs0, regs0, opt->min_chain_score, opt->a, rep_len, is_sr || is_sr_rna, is_splice);
 		n_regs[0] = n_regs0, regs[0] = regs0;
 	} else { // multi-segment
 		mm_seg_t *seg;
@@ -352,7 +347,7 @@ void mm_map_frag(const mm_idx_t *mi, int n_segs, const int *qlens, const char **
 		for (i = 0; i < n_segs; ++i) {
 			mm_set_parent(b->km, opt->mask_level, opt->mask_len, n_regs[i], regs[i], opt->a * 2 + opt->b, opt->flag&MM_F_HARD_MLEVEL, opt->alt_drop); // update mm_reg1_t::parent
 			regs[i] = align_regs(opt, mi, b->km, qlens[i], seqs[i], &n_regs[i], regs[i], seg[i].a);
-			mm_set_mapq(b->km, n_regs[i], regs[i], opt->min_chain_score, opt->a, rep_len, is_sr);
+			mm_set_mapq2(b->km, n_regs[i], regs[i], opt->min_chain_score, opt->a, rep_len, is_sr || is_sr_rna, is_splice);
 		}
 		mm_seg_free(b->km, n_segs, seg);
 		if (n_segs == 2 && opt->pe_ori >= 0 && (opt->flag&MM_F_CIGAR))
@@ -364,6 +359,10 @@ void mm_map_frag(const mm_idx_t *mi, int n_segs, const int *qlens, const char **
 	kfree(b->km, u);
 	kfree(b->km, mini_pos);
 
+	if (mi->J && n_segs == 1 && is_splice)
+		for (i = 0; i < n_regs0; ++i)
+			mm_jump_split(b->km, mi, opt, qlens[0], (const uint8_t*)seqs[0], &regs0[i], 0);
+
 	if (b->km) {
 		km_stat(b->km, &kmst);
 		if (mm_dbg_flag & MM_DBG_PRINT_QNAME)
@@ -378,6 +377,18 @@ void mm_map_frag(const mm_idx_t *mi, int n_segs, const int *qlens, const char **
 	}
 }
 
+void mm_map_frag(const mm_idx_t *mi, int n_segs, const int *qlens, const char **seqs, int *n_regs, mm_reg1_t **regs, mm_tbuf_t *b, const mm_mapopt_t *opt, const char *qname)
+{
+	if ((opt->flag & MM_F_WEAK_PAIRING) && n_segs == 2 && opt->pe_ori >= 0 && (opt->flag&MM_F_CIGAR)) {
+		int i;
+		for (i = 0; i < n_segs; ++i)
+			mm_map_frag_core(mi, 1, &qlens[i], &seqs[i], &n_regs[i], &regs[i], b, opt, qname);
+		mm_pair(b->km, opt->max_gap_ref, opt->pe_bonus, opt->a * 2 + opt->b, opt->a, qlens, n_regs, regs);
+	} else {
+		mm_map_frag_core(mi, n_segs, qlens, seqs, n_regs, regs, b, opt, qname);
+	}
+}
+
 mm_reg1_t *mm_map(const mm_idx_t *mi, int qlen, const char *seq, int *n_regs, mm_tbuf_t *b, const mm_mapopt_t *opt, const char *qname)
 {
 	mm_reg1_t *regs;
@@ -452,6 +463,10 @@ static void worker_for(void *_data, long i, int tid) // kt_for() callback
 				r->qs = qlens[j] - r->qe;
 				r->qe = qlens[j] - t;
 				r->rev = !r->rev;
+				if (r->p) {
+					if (r->p->trans_strand == 1) r->p->trans_strand = 2;
+					else if (r->p->trans_strand == 2) r->p->trans_strand = 1;
+				}
 			}
 		}
 	if (mm_dbg_flag & MM_DBG_PRINT_QNAME)
@@ -514,7 +529,7 @@ static void merge_hits(step_t *s)
 				mm_select_sub(km, opt->pri_ratio, s->p->mi->k*2, opt->best_n, 0, opt->max_gap * 0.8, &s->n_reg[k], s->reg[k]);
 				mm_set_sam_pri(s->n_reg[k], s->reg[k]);
 			}
-			mm_set_mapq(km, s->n_reg[k], s->reg[k], opt->min_chain_score, opt->a, rep_len, !!(opt->flag & MM_F_SR));
+			mm_set_mapq2(km, s->n_reg[k], s->reg[k], opt->min_chain_score, opt->a, rep_len, !!(opt->flag & (MM_F_SR|MM_F_SR_RNA)), !!(opt->flag & MM_F_SPLICE));
 		}
 		if (s->n_seg[f] == 2 && opt->pe_ori >= 0 && (opt->flag&MM_F_CIGAR))
 			mm_pair(km, frag_gap_part[0], opt->pe_bonus, opt->a * 2 + opt->b, opt->a, qlens, &s->n_reg[k0], &s->reg[k0]);
@@ -583,23 +598,30 @@ static void *worker_pipeline(void *shared, int step, void *in)
 							mm_err_fwrite(r->p, r->p->capacity, 4, p->fp_split);
 						}
 					}
+				} else if (p->opt->flag & MM_F_OUT_JUNC) { // extra logic for --write-junc
+					for (j = 0; j < s->n_reg[i]; ++j) {
+						const mm_reg1_t *r = &s->reg[i][j];
+						if (r->id != r->parent || r->mapq < 10) continue;
+						mm_write_junc(&p->str, mi, t, r);
+						if (p->str.l > 0) mm_err_puts(p->str.s);
+					}
 				} else if (s->n_reg[i] > 0) { // the query has at least one hit
 					for (j = 0; j < s->n_reg[i]; ++j) {
-						mm_reg1_t *r = &s->reg[i][j];
+						const mm_reg1_t *r = &s->reg[i][j];
 						assert(!r->sam_pri || r->id == r->parent);
 						if ((p->opt->flag & MM_F_NO_PRINT_2ND) && r->id != r->parent)
 							continue;
 						if (p->opt->flag & MM_F_OUT_SAM)
 							mm_write_sam3(&p->str, mi, t, i - seg_st, j, s->n_seg[k], &s->n_reg[seg_st], (const mm_reg1_t*const*)&s->reg[seg_st], km, p->opt->flag, s->rep_len[i]);
 						else
-							mm_write_paf3(&p->str, mi, t, r, km, p->opt->flag, s->rep_len[i]);
+							mm_write_paf4(&p->str, mi, t, r, km, p->opt->flag, s->rep_len[i], s->n_seg[k], i - seg_st);
 						mm_err_puts(p->str.s);
 					}
 				} else if ((p->opt->flag & MM_F_PAF_NO_HIT) || ((p->opt->flag & MM_F_OUT_SAM) && !(p->opt->flag & MM_F_SAM_HIT_ONLY))) { // output an empty hit, if requested
 					if (p->opt->flag & MM_F_OUT_SAM)
 						mm_write_sam3(&p->str, mi, t, i - seg_st, -1, s->n_seg[k], &s->n_reg[seg_st], (const mm_reg1_t*const*)&s->reg[seg_st], km, p->opt->flag, s->rep_len[i]);
 					else
-						mm_write_paf3(&p->str, mi, t, 0, 0, p->opt->flag, s->rep_len[i]);
+						mm_write_paf4(&p->str, mi, t, 0, 0, p->opt->flag, s->rep_len[i], s->n_seg[k], i - seg_st);
 					mm_err_puts(p->str.s);
 				}
 			}

minimap.h

@@ -5,40 +5,49 @@
 #include <stdio.h>
 #include <sys/types.h>
 
-#define MM_F_NO_DIAG       0x001 // no exact diagonal hit
-#define MM_F_NO_DUAL       0x002 // skip pairs where query name is lexicographically larger than target name
-#define MM_F_CIGAR         0x004
-#define MM_F_OUT_SAM       0x008
-#define MM_F_NO_QUAL       0x010
-#define MM_F_OUT_CG        0x020
-#define MM_F_OUT_CS        0x040
-#define MM_F_SPLICE        0x080 // splice mode
-#define MM_F_SPLICE_FOR    0x100 // match GT-AG
-#define MM_F_SPLICE_REV    0x200 // match CT-AC, the reverse complement of GT-AG
-#define MM_F_NO_LJOIN      0x400
-#define MM_F_OUT_CS_LONG   0x800
-#define MM_F_SR            0x1000
-#define MM_F_FRAG_MODE     0x2000
-#define MM_F_NO_PRINT_2ND  0x4000
-#define MM_F_2_IO_THREADS  0x8000
-#define MM_F_LONG_CIGAR    0x10000
-#define MM_F_INDEPEND_SEG  0x20000
-#define MM_F_SPLICE_FLANK  0x40000
-#define MM_F_SOFTCLIP      0x80000
-#define MM_F_FOR_ONLY      0x100000
-#define MM_F_REV_ONLY      0x200000
-#define MM_F_HEAP_SORT     0x400000
-#define MM_F_ALL_CHAINS    0x800000
-#define MM_F_OUT_MD        0x1000000
-#define MM_F_COPY_COMMENT  0x2000000
-#define MM_F_EQX           0x4000000 // use =/X instead of M
-#define MM_F_PAF_NO_HIT    0x8000000 // output unmapped reads to PAF
-#define MM_F_NO_END_FLT    0x10000000
-#define MM_F_HARD_MLEVEL   0x20000000
-#define MM_F_SAM_HIT_ONLY  0x40000000
+#define MM_VERSION "2.29-r1283"
+
+#define MM_F_NO_DIAG       (0x001LL) // no exact diagonal hit
+#define MM_F_NO_DUAL       (0x002LL) // skip pairs where query name is lexicographically larger than target name
+#define MM_F_CIGAR         (0x004LL)
+#define MM_F_OUT_SAM       (0x008LL)
+#define MM_F_NO_QUAL       (0x010LL)
+#define MM_F_OUT_CG        (0x020LL)
+#define MM_F_OUT_CS        (0x040LL)
+#define MM_F_SPLICE        (0x080LL) // splice mode
+#define MM_F_SPLICE_FOR    (0x100LL) // match GT-AG
+#define MM_F_SPLICE_REV    (0x200LL) // match CT-AC, the reverse complement of GT-AG
+#define MM_F_NO_LJOIN      (0x400LL)
+#define MM_F_OUT_CS_LONG   (0x800LL)
+#define MM_F_SR            (0x1000LL)
+#define MM_F_FRAG_MODE     (0x2000LL)
+#define MM_F_NO_PRINT_2ND  (0x4000LL)
+#define MM_F_2_IO_THREADS  (0x8000LL)
+#define MM_F_LONG_CIGAR    (0x10000LL)
+#define MM_F_INDEPEND_SEG  (0x20000LL)
+#define MM_F_SPLICE_FLANK  (0x40000LL)
+#define MM_F_SOFTCLIP      (0x80000LL)
+#define MM_F_FOR_ONLY      (0x100000LL)
+#define MM_F_REV_ONLY      (0x200000LL)
+#define MM_F_HEAP_SORT     (0x400000LL)
+#define MM_F_ALL_CHAINS    (0x800000LL)
+#define MM_F_OUT_MD        (0x1000000LL)
+#define MM_F_COPY_COMMENT  (0x2000000LL)
+#define MM_F_EQX           (0x4000000LL) // use =/X instead of M
+#define MM_F_PAF_NO_HIT    (0x8000000LL) // output unmapped reads to PAF
+#define MM_F_NO_END_FLT    (0x10000000LL)
+#define MM_F_HARD_MLEVEL   (0x20000000LL)
+#define MM_F_SAM_HIT_ONLY  (0x40000000LL)
 #define MM_F_RMQ           (0x80000000LL)
 #define MM_F_QSTRAND       (0x100000000LL)
 #define MM_F_NO_INV        (0x200000000LL)
+#define MM_F_NO_HASH_NAME  (0x400000000LL)
+#define MM_F_SPLICE_OLD    (0x800000000LL)
+#define MM_F_SECONDARY_SEQ (0x1000000000LL)	//output SEQ field for seqondary alignments using hard clipping
+#define MM_F_OUT_DS        (0x2000000000LL)
+#define MM_F_WEAK_PAIRING  (0x4000000000LL)
+#define MM_F_SR_RNA        (0x8000000000LL)
+#define MM_F_OUT_JUNC      (0x10000000000LL)
 
 #define MM_I_HPC          0x1
 #define MM_I_NO_SEQ       0x2
@@ -85,6 +94,8 @@ typedef struct {
 	uint32_t *S;               // 4-bit packed sequence
 	struct mm_idx_bucket_s *B; // index (hidden)
 	struct mm_idx_intv_s *I;   // intervals (hidden)
+	struct mm_idx_spsc_s *spsc;// splice score (hidden)
+	struct mm_idx_jjump_s *J;  // junctions to create jumps (hidden)
 	void *km, *h;
 } mm_idx_t;
 
@@ -92,6 +103,7 @@ typedef struct {
 typedef struct {
 	uint32_t capacity;                  // the capacity of cigar[]
 	int32_t dp_score, dp_max, dp_max2;  // DP score; score of the max-scoring segment; score of the best alternate mappings
+	int32_t dp_max0;                    // DP score before mm_update_dp_max() adjustment
 	uint32_t n_ambi:30, trans_strand:2; // number of ambiguous bases; transcript strand: 0 for unknown, 1 for +, 2 for -
 	uint32_t n_cigar;                   // number of cigar operations in cigar[]
 	uint32_t cigar[];
@@ -108,7 +120,7 @@ typedef struct {
 	int32_t mlen, blen;     // seeded exact match length; seeded alignment block length
 	int32_t n_sub;          // number of suboptimal mappings
 	int32_t score0;         // initial chaining score (before chain merging/spliting)
-	uint32_t mapq:8, split:2, rev:1, inv:1, sam_pri:1, proper_frag:1, pe_thru:1, seg_split:1, seg_id:8, split_inv:1, is_alt:1, strand_retained:1, dummy:5;
+	uint32_t mapq:8, split:2, rev:1, inv:1, sam_pri:1, proper_frag:1, pe_thru:1, seg_split:1, seg_id:8, split_inv:1, is_alt:1, strand_retained:1, is_spliced:1, dummy:4;
 	uint32_t hash;
 	float div;
 	mm_extra_t *p;
@@ -148,9 +160,10 @@ typedef struct {
 	float alt_drop;
 
 	int a, b, q, e, q2, e2; // matching score, mismatch, gap-open and gap-ext penalties
+	int transition; // transition mismatch score (A:G, C:T)
 	int sc_ambi; // score when one or both bases are "N"
 	int noncan;      // cost of non-canonical splicing sites
-	int junc_bonus;
+	int junc_bonus, junc_pen;
 	int zdrop, zdrop_inv;   // break alignment if alignment score drops too fast along the diagonal
 	int end_bonus;
 	int min_dp_max;  // drop an alignment if the score of the max scoring segment is below this threshold
@@ -163,6 +176,8 @@ typedef struct {
 
 	int pe_ori, pe_bonus;
 
+	int32_t jump_min_match;
+
 	float mid_occ_frac;  // only used by mm_mapopt_update(); see below
 	float q_occ_frac;
 	int32_t min_mid_occ, max_mid_occ;
@@ -188,6 +203,11 @@ typedef struct {
 } mm_idx_reader_t;
 
 // memory buffer for thread-local storage during mapping
+struct mm_tbuf_s {
+	void *km;
+	int rep_len, frag_gap;
+};
+
 typedef struct mm_tbuf_s mm_tbuf_t;
 
 // global variables
@@ -398,6 +418,10 @@ int mm_idx_alt_read(mm_idx_t *mi, const char *fn);
 int mm_idx_bed_read(mm_idx_t *mi, const char *fn, int read_junc);
 int mm_idx_bed_junc(const mm_idx_t *mi, int32_t ctg, int32_t st, int32_t en, uint8_t *s);
 
+int mm_max_spsc_bonus(const mm_mapopt_t *mo);
+int32_t mm_idx_spsc_read(mm_idx_t *idx, const char *fn, int32_t max_sc);
+int64_t mm_idx_spsc_get(const mm_idx_t *db, int32_t cid, int64_t st0, int64_t en0, int32_t rev, uint8_t *sc);
+
 // deprecated APIs for backward compatibility
 void mm_mapopt_init(mm_mapopt_t *opt);
 mm_idx_t *mm_idx_build(const char *fn, int w, int k, int flag, int n_threads);

minimap2.1

@@ -1,4 +1,4 @@
-.TH minimap2 1 "18 November 2021" "minimap2-2.23 (r1111)" "Bioinformatics tools"
+.TH minimap2 1 "18 April 2025" "minimap2-2.29 (r1283)" "Bioinformatics tools"
 .SH NAME
 .PP
 minimap2 - mapping and alignment between collections of DNA sequences
@@ -77,7 +77,7 @@ SAM format.
 Minimizer k-mer length [15]
 .TP
 .BI -w \ INT
-Minimizer window size [2/3 of k-mer length]. A minimizer is the smallest k-mer
+Minimizer window size [10]. A minimizer is the smallest k-mer
 in a window of w consecutive k-mers.
 .TP
 .B -H
@@ -88,16 +88,17 @@ on the HPC sequence.
 .BI -I \ NUM
 Load at most
 .I NUM
-target bases into RAM for indexing [4G]. If there are more than
+target bases into RAM for indexing [8G]. If there are more than
 .I NUM
 bases in
 .IR target.fa ,
 minimap2 needs to read
 .I query.fa
-multiple times to map it against each batch of target sequences.
+multiple times to map it against each batch of target sequences. This would create a multi-part index.
 .I NUM
 may be ending with k/K/m/M/g/G. NB: mapping quality is incorrect given a
-multi-part index.
+multi-part index. See also option
+.BR --split-prefix .
 .TP
 .B --idx-no-seq
 Don't store target sequences in the index. It saves disk space and memory but
@@ -254,6 +255,11 @@ or more of the shorter chain [0.5]
 Use the minigraph chaining algorithm [no]. The minigraph algorithm is better
 for aligning contigs through long INDELs.
 .TP
+.BI --rmq-inner \ NUM
+Apply full dynamic programming for anchors within distance
+.I NUM
+[1000].
+.TP
 .B --hard-mask-level
 Honor option
 .B -M
@@ -291,11 +297,13 @@ maximum alignment gap is mostly controlled by
 .B --splice
 Enable the splice alignment mode.
 .TP
-.B --sr
-Enable short-read alignment heuristics. In the short-read mode, minimap2
-applies a second round of chaining with a higher minimizer occurrence threshold
-if no good chain is found. In addition, minimap2 attempts to patch gaps between
-seeds with ungapped alignment.
+.BR --sr [= no | dna | rna ]
+Enable short-read alignment heuristics [no]. If this option is used with no argument,
+.RB ` dna '
+is set. In the DNA short-read mode, minimap2 applies a second round of chaining
+with a higher minimizer occurrence threshold if no good chain is found. In
+addition, minimap2 attempts to patch gaps between seeds with ungapped
+alignment.
 .TP
 .BI --split-prefix \ STR
 Prefix to create temporary files. Typically used for a multi-part index.
@@ -315,9 +323,8 @@ Only map to the reverse complement strand of the reference sequences.
 If yes, sort anchors with heap merge, instead of radix sort. Heap merge is
 faster for short reads, but slower for long reads. [no]
 .TP
-.B --no-pairing
-Treat two reads in a pair as independent reads. The mate related fields in SAM
-are still properly populated.
+.B --no-hash-name
+Produce the same alignment for identical sequences regardless of their sequence names.
 .SS Alignment options
 .TP 10
 .BI -A \ INT
@@ -326,6 +333,10 @@ Matching score [2]
 .BI -B \ INT
 Mismatching penalty [4]
 .TP
+.BI -b \ INT
+Mismatching penalty for transitions [same as
+.BR -B ].
+.TP
 .BI -O \ INT1[,INT2]
 Gap open penalty [4,24]. If
 .I INT2
@@ -339,10 +350,28 @@ costs
 .RI min{ O1 + k * E1 , O2 + k * E2 }.
 In the splice mode, the second gap penalties are not used.
 .TP
+.BI -J \ INT
+Splice model [1]. 0 for the original minimap2 splice model that always penalizes non-GT-AG splicing;
+1 for the miniprot model that considers non-GT-AG. Option
+.B -C
+has no effect with the default
+.BR -J1 .
+.TP
+.BR -j \ FILE
+Junctions used to extend alignment towards ends of reads [].
+.I FILE
+can be gene annotations in the BED12 format (aka 12-column BED), or intron
+positions in 5-column BED with the strand column required. BED12 file can be
+converted from GTF/GFF3 with `paftools.js gff2bed anno.gtf'. This option is
+intended for short RNA-seq reads, while
+.B --junc-bed
+for long noisy RNA-seq reads.
+.TP
 .BI -C \ INT
 Cost for a non-canonical GT-AG splicing (effective with
-.BR --splice )
-[0]
+.B --splice
+.BR -J0 )
+[0].
 .TP
 .BI -z \ INT1[,INT2]
 Truncate an alignment if the running alignment score drops too quickly along
@@ -379,7 +408,16 @@ no attempt to match GT-AG [n]
 Score bonus when alignment extends to the end of the query sequence [0].
 .TP
 .BI --score-N \ INT
-Score of a mismatch involving ambiguous bases [1].
+Penalty of a mismatch involving ambiguous bases [1].
+.TP
+.BR --pairing = strong | weak | no
+How to pair paired-end reads [strong].
+.RB ` no '
+for aligning the two ends in a pair independently with no `properly paired' set.
+.RB ` weak '
+for aligning the two ends independently and then pairing the hits.
+.RB ` strong '
+for jointly aligning and pairing the two ends.
 .TP
 .BR --splice-flank = yes | no
 Assume the next base to a
@@ -398,16 +436,40 @@ on SIRV data, please add
 .B --splice-flank=no
 to the command line.
 .TP
+.BR --spsc \ FILE
+Splice scores []. Each line consists of five fields: 1) contig, 2) offset, 3) `+' or `-', 4) `D' or `A', and 5) score,
+where offset is the number of bases before a splice junction, `D' indicates the
+line corresponds to a donor site and `A' for an acceptor site.
+A positive score suggests the junction is preferred and a negative score
+suggests the junction is not preferred.
+.TP
+.BR --junc-pen \ INT
+Penalty for a position not in FILE specified by
+.B --spsc
+[5]. Effective with
+.B --spsc
+but not
+.BR --junc-bed .
+.TP
 .BR --junc-bed \ FILE
-Gene annotations in the BED12 format (aka 12-column BED), or intron positions
-in 5-column BED. With this option, minimap2 prefers splicing in annotations.
-BED12 file can be converted from GTF/GFF3 with `paftools.js gff2bed anno.gtf'
-[].
+Junctions to prefer during base alignment [].
+Same format as
+.BR -j .
+It is
+.I NOT
+recommended to apply this option to short RNA-seq reads. This would increase
+run time with little improvement to junction accuracy.
 .TP
 .BR --junc-bonus \ INT
-Score bonus for a splice donor or acceptor found in annotation (effective with
-.BR --junc-bed )
-[9].
+Score bonus for a splice donor or acceptor found in annotation [9]. Effective with
+.B --junc-bed
+but not
+.BR --spsc .
+.TP
+.BR --jump-min-match \ INT
+Minimum matching length to create a jump [3]. Equivalent to
+.B STAR
+.BR --alignSJDBoverhangMin .
 .TP
 .BI --end-seed-pen \ INT
 Drop a terminal anchor if
@@ -433,7 +495,7 @@ Set 0 to disable [100m].
 .BI --cap-kalloc \ NUM
 Free thread-local kalloc memory reservoir if after the alignment the size of the reservoir above
 .IR NUM .
-Set 0 to disable [0].
+Set 0 to disable [500m].
 .SS Input/output options
 .TP 10
 .B -a
@@ -465,20 +527,13 @@ Copy input FASTA/Q comments to output.
 .B -c
 Generate CIGAR. In PAF, the CIGAR is written to the `cg' custom tag.
 .TP
-.BI --cs[= STR ]
+.BR --cs [= short | long ]
 Output the
 .B cs
 tag.
-.I STR
-can be either
-.I short
-or
-.IR long .
-If no
-.I STR
-is given,
-.I short
-is assumed. [none]
+If no argument is given,
+.RB ` short '
+is set. [none]
 .TP
 .B --MD
 Output the MD tag (see the SAM spec).
@@ -489,6 +544,29 @@ Output =/X CIGAR operators for sequence match/mismatch.
 .B -Y
 In SAM output, use soft clipping for supplementary alignments.
 .TP
+.B --secondary-seq
+In SAM output, show query sequences for secondary alignments.
+.TP
+.B --write-junc
+Output splice junctions in 6-column BED: contig name, start, end,
+read name, score and strand. Score is the sum of donor and acceptor scores,
+where GT gets 3, GC gets 2 and AT gets 1 at donor sites,
+while AG gets 3 and AC gets 1 at acceptor sites.
+Alignments with mapping quality below 10 are ignored.
+.TP
+.BI --pass1 \ FILE
+Junctions BED file outputted by
+.B --write-junc
+[]. Rows with scores lower than 5 are ignored. When both
+.B -j
+and
+.B --pass1
+are present, junctions in
+.B -j
+are preferred over in
+.BR --pass1
+when there is ambiguity.
+.TP
 .BI --seed \ INT
 Integer seed for randomizing equally best hits. Minimap2 hashes
 .I INT
@@ -549,42 +627,70 @@ are:
 Align noisy long reads of ~10% error rate to a reference genome. This is the
 default mode.
 .TP
+.B lr:hq
+Align accurate long reads (error rate <1%) to a reference genome
+.RB ( -k19
+.B -w19 -U50,500
+.BR -g10k ).
+This was recommended by ONT developers for recent Nanopore reads
+produced with chemistry v14 that can reach ~99% in accuracy.
+It was shown to work better for accurate Nanopore reads
+than
+.BR map-hifi .
+.TP
 .B map-hifi
 Align PacBio high-fidelity (HiFi) reads to a reference genome
-.RB ( -k19
-.B -w19 -U50,500 -g10k -A1 -B4 -O6,26 -E2,1
+.RB ( -xlr:hq
+.B -A1 -B4 -O6,26 -E2,1
 .BR -s200 ).
+It differs from
+.B lr:hq
+only in scoring. It has not been tested whether
+.B lr:hq
+would work better for PacBio HiFi reads.
 .TP
 .B map-pb
 Align older PacBio continuous long (CLR) reads to a reference genome
 .RB ( -Hk19 ).
+Note that this data type is effectively deprecated by HiFi.
+Unless you work on very old data, you probably want to use
+.B map-hifi
+or
+.BR lr:hq .
+.TP
+.B map-iclr
+Align Illumina Complete Long Reads (ICLR) to a reference genome
+.RB ( -k19
+.B -B6 -b4
+.BR -O10,50 ).
+This was recommended by Illumina developers.
 .TP
 .B asm5
 Long assembly to reference mapping
 .RB ( -k19
-.B -w19 -U50,500 --rmq -r100k -g10k -A1 -B19 -O39,81 -E3,1 -s200 -z200
+.B -w19 -U50,500 --rmq -r1k,100k -g10k -A1 -B19 -O39,81 -E3,1 -s200 -z200
 .BR -N50 ).
 Typically, the alignment will not extend to regions with 5% or higher sequence
-divergence. Only use this preset if the average divergence is far below 5%.
+divergence. Use this preset if the average divergence is not much higher than 0.1%.
 .TP
 .B asm10
 Long assembly to reference mapping
 .RB ( -k19
-.B -w19 -U50,500 --rmq -r100k -g10k -A1 -B9 -O16,41 -E2,1 -s200 -z200
+.B -w19 -U50,500 --rmq -r1k,100k -g10k -A1 -B9 -O16,41 -E2,1 -s200 -z200
 .BR -N50 ).
-Up to 10% sequence divergence.
+Use this if the average divergence is around 1%.
 .TP
 .B asm20
 Long assembly to reference mapping
 .RB ( -k19
-.B -w10 -U50,500 --rmq -r100k -g10k -A1 -B4 -O6,26 -E2,1 -s200 -z200
+.B -w10 -U50,500 --rmq -r1k,100k -g10k -A1 -B4 -O6,26 -E2,1 -s200 -z200
 .BR -N50 ).
-Up to 20% sequence divergence.
+Use this if the average divergence is around several percent.
 .TP
 .B splice
 Long-read spliced alignment
 .RB ( -k15
-.B -w5 --splice -g2k -G200k -A1 -B2 -O2,32 -E1,0 -b0 -C9 -z200 -ub --junc-bonus=9 --cap-sw-mem=0
+.B -w5 --splice -g2k -G200k -A1 -B2 -O2,32 -E1,0 -C9 -z200 -ub --junc-bonus=9 --cap-sw-mem=0
 .BR --splice-flank=yes ).
 In the splice mode, 1) long deletions are taken as introns and represented as
 the
@@ -595,15 +701,21 @@ costs are different during chaining; 4) the computation of the
 tag ignores introns to demote hits to pseudogenes.
 .TP
 .B splice:hq
-Long-read splice alignment for PacBio CCS reads
+Spliced alignment for accurate long RNA-seq reads such as PacBio iso-seq
 .RB ( -xsplice
 .B -C5 -O6,24
 .BR -B4 ).
 .TP
+.B splice:sr
+Spliced alignment for short RNA-seq reads
+.RB ( -xsplice:hq
+.B --frag=yes -m25 -s40 -2K100m --heap-sort=yes --pairing=weak --sr=rna --min-dp-len=20
+.BR --secondary=no ).
+.TP
 .B sr
-Short single-end reads without splicing
+Short-read alignment without splicing
 .RB ( -k21
-.B -w11 --sr --frag=yes -A2 -B8 -O12,32 -E2,1 -b0 -r100 -p.5 -N20 -f1000,5000 -n2 -m20
+.B -w11 --sr --frag=yes -A2 -B8 -O12,32 -E2,1 -r100 -p.5 -N20 -f1000,5000 -n2 -m25
 .B -s40 -g100 -2K50m --heap-sort=yes
 .BR --secondary=no ).
 .TP
@@ -676,7 +788,7 @@ s2	i	Chaining score of the best secondary chain
 NM	i	Total number of mismatches and gaps in the alignment
 MD	Z	To generate the ref sequence in the alignment
 AS	i	DP alignment score
-SA	Z	List of other supplementary alignments
+SA	Z	List of other supplementary alignments (with approximate CIGAR strings)
 ms	i	DP score of the max scoring segment in the alignment
 nn	i	Number of ambiguous bases in the alignment
 ts	A	Transcript strand (splice mode only)
@@ -685,6 +797,7 @@ cs	Z	Difference string
 dv	f	Approximate per-base sequence divergence
 de	f	Gap-compressed per-base sequence divergence
 rl	i	Length of query regions harboring repetitive seeds
+zd	i	Alignment broken due to Z-drop; bit 1: left broken; bit 2: right broken
 .TE
 
 .PP

misc/README.md

@@ -16,7 +16,8 @@ minimap2 -c test/MT-human.fa test/MT-orang.fa \
   | paftools.js liftover -l10000 - <(echo -e "MT_orang\t2000\t5000")     # liftOver
 # no test data for the following examples
 paftools.js junceval -e anno.gtf splice.sam > out.txt  # compare splice junctions to annotations
-paftools.js splice2bed anno.gtf > anno.bed             # convert GTF/GFF3 to BED12
+paftools.js splice2bed splice.sam > splice.bed             # convert PAF/SAM to BED12
+paftools.js gff2bed anno.gtf > anno.bed             # convert GTF/GFF3 to BED12
 ```
 
 ## Table of Contents

misc/mmphase.js

@@ -1,335 +0,0 @@
-#!/usr/bin/env k8
-
-var getopt = function(args, ostr) {
-	var oli; // option letter list index
-	if (typeof(getopt.place) == 'undefined')
-		getopt.ind = 0, getopt.arg = null, getopt.place = -1;
-	if (getopt.place == -1) { // update scanning pointer
-		if (getopt.ind >= args.length || args[getopt.ind].charAt(getopt.place = 0) != '-') {
-			getopt.place = -1;
-			return null;
-		}
-		if (getopt.place + 1 < args[getopt.ind].length && args[getopt.ind].charAt(++getopt.place) == '-') { // found "--"
-			++getopt.ind;
-			getopt.place = -1;
-			return null;
-		}
-	}
-	var optopt = args[getopt.ind].charAt(getopt.place++); // character checked for validity
-	if (optopt == ':' || (oli = ostr.indexOf(optopt)) < 0) {
-		if (optopt == '-') return null; //  if the user didn't specify '-' as an option, assume it means null.
-		if (getopt.place < 0) ++getopt.ind;
-		return '?';
-	}
-	if (oli+1 >= ostr.length || ostr.charAt(++oli) != ':') { // don't need argument
-		getopt.arg = null;
-		if (getopt.place < 0 || getopt.place >= args[getopt.ind].length) ++getopt.ind, getopt.place = -1;
-	} else { // need an argument
-		if (getopt.place >= 0 && getopt.place < args[getopt.ind].length)
-			getopt.arg = args[getopt.ind].substr(getopt.place);
-		else if (args.length <= ++getopt.ind) { // no arg
-			getopt.place = -1;
-			if (ostr.length > 0 && ostr.charAt(0) == ':') return ':';
-			return '?';
-		} else getopt.arg = args[getopt.ind]; // white space
-		getopt.place = -1;
-		++getopt.ind;
-	}
-	return optopt;
-}
-
-function read_fastx(file, buf)
-{
-	if (file.readline(buf) < 0) return null;
-	var m, line = buf.toString();
-	if ((m = /^([>@])(\S+)/.exec(line)) == null)
-		throw Error("wrong fastx format");
-	var is_fq = (m[1] == '@');
-	var name = m[2];
-	if (file.readline(buf) < 0)
-		throw Error("missing sequence line");
-	var seq = buf.toString();
-	if (is_fq) { // skip quality
-		file.readline(buf);
-		file.readline(buf);
-	}
-	return [name, seq];
-}
-
-function filter_paf(a, opt)
-{
-	if (a.length == 0) return;
-	var k = 0;
-	for (var i = 0; i < a.length; ++i) {
-		var ai = a[i];
-		if (ai[10] < opt.min_blen) continue;
-		if (ai[9] < ai[10] * opt.min_iden) continue;
-		var clip = [0, 0];
-		if (ai[4] == '+') {
-			clip[0] = ai[2] < ai[7]? ai[2] : ai[7];
-			clip[1] = ai[1] - ai[3] < ai[6] - ai[8]? ai[1] - ai[3] : ai[6] - ai[8];
-		} else {
-			clip[0] = ai[2] < ai[6] - ai[8]? ai[2] : ai[6] - ai[8];
-			clip[1] = ai[1] - ai[3] < ai[7]? ai[1] - ai[3] : ai[7];
-		}
-		if (clip[0] > opt.max_clip_len || clip[1] > opt.max_clip_len) continue;
-		a[k++] = ai;
-	}
-	a.length = k;
-}
-
-function parse_events(t, ev, id, buf)
-{
-	var re = /(:(\d+))|(([\+\-\*])([a-z]+))/g;
-	var m, cs = null;
-	for (var j = 12; j < t.length; ++j) {
-		if ((m = /^cs:Z:(\S+)/.exec(t[j])) != null) {
-			cs = m[1].toLowerCase();
-			break;
-		}
-	}
-	if (cs == null) {
-		warn("Warning: no cs tag for read '" + t[0] + "'");
-		return;
-	}
-	var st = t[2], en = t[3];
-	var x = st;
-	while ((m = re.exec(cs)) != null) {
-		var l;
-		if (m[2] != null) { // an identitcal match ":\d+"
-			l = parseInt(m[2]);
-			// [start, end, type, index, changed_base]
-			ev.push([x, x + l, 0, id]);
-		} else {
-			if (m[4] == '*') {
-				l = 1;
-				ev.push([x, x + 1, 1, id, m[5][0]]);
-			} else if (m[4] == '+') {
-				l = m[5].length;
-				ev.push([x, x + l, 2, id]);
-			} else if (m[4] == '-') {
-				l = 0;
-				ev.push([x, x, -1, id, m[5]]);
-			}
-		}
-		x += l;
-	}
-	if (x != en)
-		throw Error("inconsistent cs for read '" + t[0] + "'");
-}
-
-function find_het_sub(ev, a, opt)
-{
-	var n = a.length, last0_i = -1, h = [], d = [];
-	for (var i = 0; i < n; ++i) h[i] = [], d[i] = [];
-	for (var i = 0; i < ev.length; ++i) {
-		if (ev[i][2] == 0) {
-			if (last0_i < 0 || ev[i][0] != ev[last0_i][0]) last0_i = i;
-			else if (ev[i][1] > ev[last0_i][1])
-				last0_i = i;
-		} else if (ev[i][2] == 1 && last0_i >= 0 && ev[i][0] < ev[last0_i][1]) {
-			if (ev[last0_i][1] - ev[last0_i][0] >= opt.min_mlen) {
-				if (opt.dbg_ev) print("EV", ev[last0_i].join("\t"), "|", ev[i].join("\t"));
-				var e0 = ev[last0_i], hl = h[e0[3]];
-				if (hl.length == 0 || hl[hl.length-1][0] != e0[0])
-					hl.push([e0[0], e0[1]]);
-				d[ev[i][3]].push([ev[i][0], e0[1] - e0[0]]);
-			}
-		}
-	}
-	var b = [];
-	for (var i = 0; i < n; ++i) {
-		var sh = 0, dh = 0;
-		for (var j = 0; j < h[i].length; ++j)
-			sh += h[i][j][1] - h[i][j][0];
-		for (var j = 0; j < d[i].length; ++j)
-			dh += d[i][j][1];
-		// [start, end, index, #consistent, lenConsistent, #conflictive, lenConflictive, identity, mlen]
-		b[i] = [a[i][2], a[i][3], i, h[i].length, sh, d[i].length, dh, a[i][9] / a[i][10], a[i][9]];
-	}
-	return b;
-}
-
-function flt_utg_for_ec(b, opt)
-{
-	var k = 0;
-	for (var i = 0; i < b.length; ++i) {
-		var bi = b[i];
-		if (bi[4] == 0 && bi[6] == 0) b[k++] = bi; // entirely ambiguous
-		else if (bi[6] < (bi[4] + bi[6]) * opt.max_ratio0) b[k++] = bi;
-	}
-	b.length = k;
-	if (b.length == 0) return;
-	// find the longest contiguous segment
-	b.sort(function(x,y) { return x[0]-y[0] });
-	var st = b[0][0], en = b[0][1], max_st = 0, max_en = 0, max_max_en = en;
-	for (var i = 1; i < b.length; ++i) {
-		if (b[i][0] > en) {
-			if (en - st > max_en - max_st)
-				max_st = st, max_en = en;
-			st = b[i][0], en = b[i][1];
-		} else {
-			en = en > b[i][1]? en : b[i][1];
-		}
-		max_max_en = max_max_en > b[i][1]? max_max_en : b[i][1];
-	}
-	if (en - st > max_en - max_st)
-		max_st = st, max_en = en;
-	if (max_max_en != en || st != b[0][0]) {
-		var k = 0;
-		for (var i = 0; i < b.length; ++i)
-			if (b[i][0] < max_en && b[i][1] > max_st)
-				b[k++] = b[i];
-		b.length = k;
-	}
-}
-
-function flt_utg_for_bin(b, opt) // filter out alignments clearly on the wrong phase
-{
-	var k = 0;
-	for (var i = 0; i < b.length; ++i) {
-		var bi = b[i];
-		if (bi[4] + bi[6] == 0 || bi[4] >= (bi[4] + bi[6]) * opt.max_ratio0) b[k++] = bi;
-	}
-	b.length = k;
-}
-
-function ec_core(b, n_a, ev, buf, ecb) // error correction
-{
-	var intv = [];
-	for (var i = 0; i < n_a; ++i)
-		intv[i] = null;
-	intv[b[0][2]] = [b[0][0], b[0][1]];
-	var en = b[0][1];
-	for (var i = 1; i < b.length; ++i) {
-		if (b[i][1] <= en) continue;
-		intv[b[i][2]] = [en, b[i][1]];
-		en = b[i][1];
-	}
-	var k = 0;
-	ecb.capacity = buf.capacity;
-	ecb.length = 0;
-	for (var i = 0; i < ev.length; ++i) {
-		var e = ev[i], I = intv[e[3]];
-		if (I == null) continue;
-		if (e[0] >= I[0] && e[0] < I[1]) { // this is to reduce duplicated events around junctions
-			//print("X", e.join("\t"));
-			if (e[2] == 0) {
-				ecb.length += e[1] - e[0];
-				for (var j = e[0]; j < e[1]; ++j)
-					ecb[k++] = buf[j];
-			} else if (e[2] == 1) {
-				++ecb.length;
-				ecb[k++] = e[4].charCodeAt(0);
-			} else if (e[2] < 0) {
-				ecb.length += e[4].length;
-				for (var j = 0; j < e[4].length; ++j)
-					ecb[k++] = e[4].charCodeAt(j);
-			} // else, skip e[2] == 2
-		}
-	}
-	if (ecb.length != k) throw Error("BUG!");
-}
-
-function process_paf(a, opt, fp_seq, buf, ecb)
-{
-	if (a.length == 0) return;
-	var len = a[0][1], name = a[0][0], seq = null;
-	if (len < opt.min_rlen) return;
-	if (fp_seq) {
-		var ret;
-		while ((ret = read_fastx(fp_seq, buf)) != null)
-			if (ret[0] == a[0][0])
-				break;
-		if (ret == null)
-			throw Error("failed to find sequence for read '" + a[0][0] + "'");
-		name = ret[0], seq = ret[1];
-		if (seq.length != len)
-			throw Error("inconsistent length for read '" + name + "'");
-	}
-	filter_paf(a, opt);
-	if (a.length == 0) return;
-	var ev = [];
-	for (var i = 0; i < a.length; ++i)
-		parse_events(a[i], ev, i, buf);
-	ev.sort(function(x,y) { return x[0]!=y[0]? x[0]-y[0] : x[2]-y[2] });
-	if (seq == null) print("SQ", name, a[0][1], a.length);
-	var b = find_het_sub(ev, a, opt);
-	if (opt.ec) flt_utg_for_ec(b, opt);
-	else flt_utg_for_bin(b, opt);
-	if (seq == null) {
-		for (var i = 0; i < b.length; ++i) {
-			var m, ai = a[b[i][2]], score = 0;
-			for (var j = 10; j < ai.length; ++j)
-				if ((m = /^AS:i:(\d+)/.exec(ai[j])) != null)
-					score = m[1];
-			print("TS", b[i][2], b[i][0], b[i][1], ai.slice(5, 9).join("\t"), b[i].slice(3, 7).join("\t"), score);
-		}
-		print("//");
-	} else { // error correction
-		if (b.length == 0) return;
-		buf.set(seq, 0);
-		ec_core(b, a.length, ev, buf, ecb);
-		print(">" + name);
-		print(ecb);
-	}
-}
-
-function main(args)
-{
-	var c, opt = { min_rlen:5000, min_blen:5000, min_iden:0.8, min_mlen:5, max_clip_len:500, max_ratio0:0.25, dbg_ev:false };
-	while ((c = getopt(args, "l:b:d:m:c:r:E")) != null) {
-		if (c == 'l') opt.min_rlen = parseInt(getopt.arg);
-		else if (c == 'b') opt.min_blen = parseInt(getopt.arg);
-		else if (c == 'd') opt.min_iden = parseFloat(getopt.arg);
-		else if (c == 'm') opt.min_slen = parseInt(getopt.arg);
-		else if (c == 'c') opt.max_clip_len = parseInt(getopt.arg);
-		else if (c == 'r') opt.max_ratio0 = parseFloat(getopt.arg);
-		else if (c == 'E') opt.dbg_ev = true;
-	}
-	if (args.length - getopt.ind < 1) {
-		print("Usage: mmphase.js [options] <map-with-cs.paf> [reads.fa]");
-		print("Options:");
-		print("  -l INT    min read length [" + opt.min_rlen + "]");
-		print("  -b INT    min alignment length [" + opt.min_blen + "]");
-		print("  -d FLOAT  min identity [" + opt.min_iden + "]");
-		print("  -s INT    min match length [" + opt.min_mlen + "]");
-		print("  -c INT    max clip length [" + opt.max_clip_len + "]");
-		print("  -r FLOAT  initial ratio for haplotype filtering [" + opt.max_ratio0 + "]");
-		return 0;
-	}
-
-	opt.ec = args.length - getopt.ind < 2? false : true;
-	if (!opt.ec) {
-		print("CC");
-		print("CC", "SQ  qName  qLen    nHits");
-		print("CC", "TS  index  qStart  qEnd  tName  tLen  tStart  tEnd  nConsistent  lCons  nConflictive  lConf  score");
-		print("CC");
-	}
-
-	var buf = new Bytes(), ecb = new Bytes();
-	var fp_paf = new File(args[getopt.ind]);
-	var fp_seq = args.length - getopt.ind >= 2? new File(args[getopt.ind+1]) : null;
-	var a = [];
-	while (fp_paf.readline(buf) >= 0) {
-		var t = buf.toString().split("\t");
-		if (a.length > 0 && a[0][0] != t[0]) {
-			process_paf(a, opt, fp_seq, buf, ecb);
-			a.length = 0;
-		}
-		for (var i = 1; i <= 3; ++i) t[i] = parseInt(t[i]);
-		if (t[1] < opt.min_rlen) continue;
-		for (var i = 6; i <= 10; ++i) t[i] = parseInt(t[i]);
-		if (t[10] < opt.min_blen) continue;
-		a.push(t);
-	}
-	if (a.length >= 0)
-		process_paf(a, opt, fp_seq, buf, ecb);
-	if (fp_seq) fp_seq.close();
-	fp_paf.close();
-	ecb.destroy();
-	buf.destroy();
-}
-
-var ret = main(arguments)
-exit(ret)

misc/pafcluster.js

@@ -0,0 +1,241 @@
+#!/usr/bin/env k8
+
+"use strict";
+
+Array.prototype.delete_at = function(i) {
+	for (let j = i; j < this.length - 1; ++j)
+		this[j] = this[j + 1];
+	--this.length;
+}
+
+function* getopt(argv, ostr, longopts) {
+	if (argv.length == 0) return;
+	let pos = 0, cur = 0;
+	while (cur < argv.length) {
+		let lopt = "", opt = "?", arg = "";
+		while (cur < argv.length) { // skip non-option arguments
+			if (argv[cur][0] == "-" && argv[cur].length > 1) {
+				if (argv[cur] == "--") cur = argv.length;
+				break;
+			} else ++cur;
+		}
+		if (cur == argv.length) break;
+		let a = argv[cur];
+		if (a[0] == "-" && a[1] == "-") { // a long option
+			pos = -1;
+			let c = 0, k = -1, tmp = "", o;
+			const pos_eq = a.indexOf("=");
+			if (pos_eq > 0) {
+				o = a.substring(2, pos_eq);
+				arg = a.substring(pos_eq + 1);
+			} else o = a.substring(2);
+			for (let i = 0; i < longopts.length; ++i) {
+				let y = longopts[i];
+				if (y[y.length - 1] == "=") y = y.substring(0, y.length - 1);
+				if (o.length <= y.length && o == y.substring(0, o.length)) {
+					k = i, tmp = y;
+					++c; // c is the number of matches
+					if (o == y) { // exact match
+						c = 1;
+						break;
+					}
+				}
+			}
+			if (c == 1) { // find a unique match
+				lopt = tmp;
+				if (pos_eq < 0 && longopts[k][longopts[k].length-1] == "=" && cur + 1 < argv.length) {
+					arg = argv[cur+1];
+					argv.delete_at(cur + 1);
+				}
+			}
+		} else { // a short option
+			if (pos == 0) pos = 1;
+			opt = a[pos++];
+			let k = ostr.indexOf(opt);
+			if (k < 0) {
+				opt = "?";
+			} else if (k + 1 < ostr.length && ostr[k+1] == ":") { // requiring an argument
+				if (pos >= a.length) {
+					arg = argv[cur+1];
+					argv.delete_at(cur + 1);
+				} else arg = a.substring(pos);
+				pos = -1;
+			}
+		}
+		if (pos < 0 || pos >= argv[cur].length) {
+			argv.delete_at(cur);
+			pos = 0;
+		}
+		if (lopt != "") yield { opt: `--${lopt}`, arg: arg };
+		else if (opt != "?") yield { opt: `-${opt}`, arg: arg };
+		else yield { opt: "?", arg: "" };
+	}
+}
+
+function* k8_readline(fn) {
+	let buf = new Bytes();
+	let file = new File(fn);
+	while (file.readline(buf) >= 0) {
+		yield buf.toString();
+	}
+	file.close();
+	buf.destroy();
+}
+
+function merge_hits(b) {
+	if (b.length == 1)
+		return { name1:b[0].name1, name2:b[0].name2, len1:b[0].len1, len2:b[0].len2, min_cov:b[0].min_cov, max_cov:b[0].max_cov, cov1:b[0].cov1, cov2:b[0].cov2, s1:b[0].s1, dv:b[0].dv };
+	b.sort(function(x, y) { return x.st1 - y.st1 });
+	let f = [], bt = [];
+	for (let i = 0; i < b.length; ++i)
+		f[i] = b[i].s1, bt[i] = -1;
+	for (let i = 0; i < b.length; ++i) {
+		for (let j = 0; j < i; ++j) {
+			if (b[j].st2 < b[i].st2) {
+				if (b[j].en1 >= b[i].en1) continue;
+				if (b[j].en2 >= b[i].en2) continue;
+				const ov1 = b[j].en1 <= b[i].st1? 0 : b[i].st1 - b[j].en1;
+				const li1 = b[i].en1 - b[i].st1;
+				const s11 = b[i].s1 / li1 * (li1 - ov1);
+				const ov2 = b[j].en2 <= b[i].st2? 0 : b[i].st2 - b[j].en2;
+				const li2 = b[i].en2 - b[i].st2;
+				const s12 = b[i].s1 / li2 * (li2 - ov2);
+				const s1 = s11 < s12? s11 : s12;
+				if (f[i] < f[j] + s1)
+					f[i] = f[j] + s1, bt[i] = j;
+			}
+		}
+	}
+	let max_i = -1, max_f = 0, d = [];
+	for (let i = 0; i < b.length; ++i)
+		if (max_f < f[i])
+			max_f = f[i], max_i = i;
+	for (let k = max_i; k >= 0; k = bt[k])
+		d.push(k);
+	d = d.reverse();
+	let dv = 0, tot = 0, cov1 = 0, cov2 = 0, st1 = 0, en1 = 0, st2 = 0, en2 = 0;
+	for (let k = 0; k < d.length; ++k) {
+		const i = d[k];
+		tot += b[i].blen;
+		dv += b[i].dv * b[i].blen;
+		if (b[i].st1 > en1) {
+			cov1 += en1 - st1;
+			st1 = b[i].st1, en1 = b[i].en1;
+		} else en1 = en1 > b[i].en1? en1 : b[i].en1;
+		if (b[i].st2 > en2) {
+			cov2 += en2 - st2;
+			st2 = b[i].st2, en2 = b[i].en2;
+		} else en2 = en2 > b[i].en2? en2 : b[i].en2;
+	}
+	dv /= tot;
+	cov1 = (cov1 + (en1 - st1)) / b[0].len1;
+	cov2 = (cov2 + (en2 - st2)) / b[0].len2;
+	const min_cov = cov1 < cov2? cov1 : cov2;
+	const max_cov = cov1 > cov2? cov1 : cov2;
+	//warn(d.length, b[0].name1, b[0].name2, min_cov, max_cov);
+	return { name1:b[0].name1, name2:b[0].name2, len1:b[0].len1, len2:b[0].len2, min_cov:min_cov, max_cov:max_cov, cov1:cov1, cov2:cov2, s1:max_f, dv:dv };
+}
+
+function main(args) {
+	let opt = { min_cov:.9, max_dv:.015, max_diff:20000 };
+	for (const o of getopt(args, "c:d:e:", [])) {
+		if (o.opt == '-c') opt.min_cov = parseFloat(o.arg);
+		else if (o.opt == '-d') opt.max_dv = parseFloat(o.arg);
+		else if (o.opt == '-e') opt.max_diff = parseFloat(o.arg);
+	}
+	if (args.length == 0) {
+		print("Usage: pafcluster.js [options] <ava.paf>");
+		print("Options:");
+		print(`  -c FLOAT     min coverage [${opt.min_cov}]`);
+		print(`  -d FLOAT     max divergence [${opt.max_dv}]`);
+		print(`  -e FLOAT     max difference [${opt.max_diff}]`);
+		return;
+	}
+
+	// read
+	let a = [], len = {}, name2len = {};
+	for (const line of k8_readline(args[0])) {
+		let m, t = line.split("\t");
+		if (t[4] != "+") continue;
+		for (let i = 1; i < 4;  ++i) t[i] = parseInt(t[i]);
+		for (let i = 6; i < 11; ++i) t[i] = parseInt(t[i]);
+		const len1 = t[1], len2 = t[6];
+		let s1 = -1, dv = -1.0;
+		for (let i = 12; i < t.length; ++i) {
+			if ((m = /^(s1|dv):\S:(\S+)/.exec(t[i])) != null) {
+				if (m[1] == "s1") s1 = parseInt(m[2]);
+				else if (m[1] == "dv") dv = parseFloat(m[2]);
+			}
+		}
+		if (s1 < 0 || dv < 0) continue;
+		const cov1 = (parseInt(t[3]) - parseInt(t[2])) / len1;
+		const cov2 = (parseInt(t[8]) - parseInt(t[7])) / len2;
+		const min_cov = cov1 < cov2? cov1 : cov2;
+		const max_cov = cov1 > cov2? cov1 : cov2;
+		name2len[t[0]] = len1;
+		name2len[t[5]] = len2;
+		a.push({ name1:t[0], name2:t[5], len1:len1, len2:len2, min_cov:min_cov, max_cov:max_cov, s1:s1, dv:dv, cov1:cov1, cov2:cov2, st1:t[2], en1:t[3], st2:t[7], en2:t[8], blen:t[10] });
+		len[t[0]] = len1, len[t[5]] = len2;
+	}
+	warn(`Read ${a.length} hits`);
+
+	// merge duplicated hits
+	let h = {};
+	for (let i = 0; i < a.length; ++i) {
+		const key = `${a[i].name1}\t${a[i].name2}`;
+		if (h[key] == null) h[key] = [];
+		h[key].push(a[i]);
+	}
+	a = [];
+	for (const key in h)
+		a.push(merge_hits(h[key]));
+
+	// core loop
+	while (a.length > 1) {
+		// select the sequence with the highest sum of s1
+		let h = {};
+		for (let i = 0; i < a.length; ++i) {
+			if (h[a[i].name1] == null) h[a[i].name1] = 0;
+			h[a[i].name1] += a[i].s1;
+		}
+		let max_s1 = 0, max_name = "";
+		for (const name in h)
+			if (max_s1 < h[name])
+				max_s1 = h[name], max_name = name;
+		// find contigs in the same group
+		h = {};
+		h[max_name] = 1;
+		for (let i = 0; i < a.length; ++i) {
+			if (a[i].name1 != max_name && a[i].name2 != max_name)
+				continue;
+			const diff1 = a[i].len1 * (1.0 - a[i].cov1);
+			const diff2 = a[i].len2 * (1.0 - a[i].cov2);
+			if (a[i].min_cov >= opt.min_cov && a[i].dv <= opt.max_dv && diff1 <= opt.max_diff && diff2 <= opt.max_diff)
+				h[a[i].name1] = h[a[i].name2] = 1;
+		}
+		let n = 0;
+		for (const key in h) {
+			++n;
+			delete name2len[key];
+		}
+		print(`SD\t${max_name}\t${n}`);
+		for (const key in h) print(`CL\t${key}\t${len[key]}`);
+		print("//");
+		// filter out redundant hits
+		let b = [];
+		for (let i = 0; i < a.length; ++i)
+			if (h[a[i].name1] == null && h[a[i].name2] == null)
+				b.push(a[i]);
+		warn(`Reduced the number of hits from ${a.length} to ${b.length}`);
+		a = b;
+	}
+
+	// output remaining singletons
+	for (const key in name2len) {
+		print(`SD\t${key}\t1`);
+		print(`CL\t${key}\t${name2len[key]}`);
+		print(`//`);
+	}
+}
+
+main(arguments);

misc/paftools.js

@@ -1,6 +1,6 @@
 #!/usr/bin/env k8
 
-var paftools_version = '2.23-r1111';
+var paftools_version = '2.29-r1283';
 
 /*****************************
  ***** Library functions *****
@@ -133,26 +133,50 @@ Interval.find_ovlp = function(a, st, en)
 
 function fasta_read(fn)
 {
-	var h = {}, gt = '>'.charCodeAt(0);
+	var h = {}, seqlen = [];
+	var buf = new Bytes();
 	var file = fn == '-'? new File() : new File(fn);
-	var buf = new Bytes(), seq = null, name = null, seqlen = [];
-	while (file.readline(buf) >= 0) {
-		if (buf[0] == gt) {
-			if (seq != null && name != null) {
-				seqlen.push([name, seq.length]);
-				h[name] = seq;
-				name = seq = null;
-			}
-			var m, line = buf.toString();
-			if ((m = /^>(\S+)/.exec(line)) != null) {
-				name = m[1];
-				seq = new Bytes();
-			}
-		} else seq.set(buf);
-	}
-	if (seq != null && name != null) {
-		seqlen.push([name, seq.length]);
-		h[name] = seq;
+	if (typeof k8_version == "undefined") { // for k8-0.x
+		var seq = null, name = null, gt = '>'.charCodeAt(0);
+		while (file.readline(buf) >= 0) {
+			if (buf[0] == gt) {
+				if (seq != null && name != null) {
+					seqlen.push([name, seq.length]);
+					h[name] = seq;
+					name = seq = null;
+				}
+				var m, line = buf.toString();
+				if ((m = /^>(\S+)/.exec(line)) != null) {
+					name = m[1];
+					seq = new Bytes();
+				}
+			} else seq.set(buf);
+		}
+		if (seq != null && name != null) {
+			seqlen.push([name, seq.length]);
+			h[name] = seq;
+		}
+	} else { // for k8-1.x
+		var seq = null, name = null;
+		while (file.readline(buf) >= 0) {
+			var line = buf.toString();
+			if (line[0] == ">") {
+				if (seq != null && name != null) {
+					seqlen.push([name, seq.length]);
+					h[name] = new Uint8Array(seq.buffer);
+					name = seq = null;
+				}
+				var m;
+				if ((m = /^>(\S+)/.exec(line)) != null) {
+					name = m[1];
+					seq = new Bytes();
+				}
+			} else seq.set(line);
+		}
+		if (seq != null && name != null) {
+			seqlen.push([name, seq.length]);
+			h[name] = new Uint8Array(seq.buffer);
+		}
 	}
 	buf.destroy();
 	file.close();
@@ -161,16 +185,27 @@ function fasta_read(fn)
 
 function fasta_free(fa)
 {
-	for (var name in fa)
-		fa[name].destroy();
+	if (typeof k8_version == "undefined")
+		for (var name in fa)
+			fa[name].destroy();
+	// FIXME: for k8-1.0, sequences are not freed. This is ok for now but not general.
 }
 
 Bytes.prototype.reverse = function()
 {
-	for (var i = 0; i < this.length>>1; ++i) {
-		var tmp = this[i];
-		this[i] = this[this.length - i - 1];
-		this[this.length - i - 1] = tmp;
+	if (typeof k8_version === "undefined") { // k8-0.x
+		for (var i = 0; i < this.length>>1; ++i) {
+			var tmp = this[i];
+			this[i] = this[this.length - i - 1];
+			this[this.length - i - 1] = tmp;
+		}
+	} else { // k8-1.x
+		var buf = new Uint8Array(this.buffer);
+		for (var i = 0; i < buf.length>>1; ++i) {
+			var tmp = buf[i];
+			buf[i] = buf[buf.length - i - 1];
+			buf[buf.length - i - 1] = tmp;
+		}
 	}
 }
 
@@ -185,13 +220,24 @@ Bytes.prototype.revcomp = function()
 		for (var i = 0; i < s1.length; ++i)
 			Bytes.rctab[s1.charCodeAt(i)] = s2.charCodeAt(i);
 	}
-	for (var i = 0; i < this.length>>1; ++i) {
-		var tmp = this[this.length - i - 1];
-		this[this.length - i - 1] = Bytes.rctab[this[i]];
-		this[i] = Bytes.rctab[tmp];
+	if (typeof k8_version === "undefined") { // k8-0.x
+		for (var i = 0; i < this.length>>1; ++i) {
+			var tmp = this[this.length - i - 1];
+			this[this.length - i - 1] = Bytes.rctab[this[i]];
+			this[i] = Bytes.rctab[tmp];
+		}
+		if (this.length&1)
+			this[this.length>>1] = Bytes.rctab[this[this.length>>1]];
+	} else { // k8-1.x
+		var buf = new Uint8Array(this.buffer);
+		for (var i = 0; i < buf.length>>1; ++i) {
+			var tmp = buf[buf.length - i - 1];
+			buf[buf.length - i - 1] = Bytes.rctab[buf[i]];
+			buf[i] = Bytes.rctab[tmp];
+		}
+		if (buf.length&1)
+			buf[buf.length>>1] = Bytes.rctab[buf[buf.length>>1]];
 	}
-	if (this.length&1)
-		this[this.length>>1] = Bytes.rctab[this[this.length>>1]];
 }
 
 /********************
@@ -1532,22 +1578,24 @@ function paf_view(args)
 
 function paf_gff2bed(args)
 {
-	var c, fn_ucsc_fai = null, is_short = false, keep_gff = false, print_junc = false, output_gene = false;
-	while ((c = getopt(args, "u:sgjG")) != null) {
+	var c, fn_ucsc_fai = null, is_short = false, keep_gff = false, print_junc = false, output_gene = false, ens_canon_only = false;
+	while ((c = getopt(args, "u:sgjGe")) != null) {
 		if (c == 'u') fn_ucsc_fai = getopt.arg;
 		else if (c == 's') is_short = true;
 		else if (c == 'g') keep_gff = true;
 		else if (c == 'j') print_junc = true;
 		else if (c == 'G') output_gene = true;
+		else if (c == 'e') ens_canon_only = true;
 	}
 
 	if (getopt.ind == args.length) {
 		print("Usage: paftools.js gff2bed [options] <in.gff>");
 		print("Options:");
-		print("  -j       Output junction BED");
-		print("  -s       Print names in the short form");
+		print("  -j       output junction BED");
+		print("  -s       print names in the short form");
 		print("  -u FILE  hg38.fa.fai for chr name conversion");
-		print("  -g       Output GFF (used with -u)");
+		print("  -e       only show transcript tagged with 'Ensembl_canonical'");
+		print("  -g       output GFF (used with -u)");
 		exit(1);
 	}
 
@@ -1606,7 +1654,7 @@ function paf_gff2bed(args)
 		print(a[0][0], st, en, name, 1000, a[0][3], cds_st, cds_en, color, a.length, sizes.join(",") + ",", starts.join(",") + ",");
 	}
 
-	var re_gtf  = /\b(transcript_id|transcript_type|transcript_biotype|gene_name|gene_id|gbkey|transcript_name) "([^"]+)";/g;
+	var re_gtf  = /\b(transcript_id|transcript_type|transcript_biotype|gene_name|gene_id|gbkey|transcript_name|tag) "([^"]+)";/g;
 	var re_gff3 = /\b(transcript_id|transcript_type|transcript_biotype|gene_name|gene_id|gbkey|transcript_name)=([^;]+)/g;
 	var re_gtf_gene  = /\b(gene_id|gene_type|gene_name) "([^;]+)";/g;
 	var re_gff3_gene = /\b(gene_id|gene_type|source_gene|gene_biotype|gene_name)=([^;]+);/g;
@@ -1646,13 +1694,14 @@ function paf_gff2bed(args)
 		if (t[2] != "CDS" && t[2] != "exon") continue;
 		t[3] = parseInt(t[3]) - 1;
 		t[4] = parseInt(t[4]);
-		var id = null, type = "", name = "N/A", biotype = "", m, tname = "N/A";
+		var id = null, type = "", name = "N/A", biotype = "", m, tname = "N/A", ens_canonical = false;
 		while ((m = re_gtf.exec(t[8])) != null) {
 			if (m[1] == "transcript_id") id = m[2];
 			else if (m[1] == "transcript_type") type = m[2];
 			else if (m[1] == "transcript_biotype" || m[1] == "gbkey") biotype = m[2];
 			else if (m[1] == "gene_name" || m[1] == "gene_id") name = m[2];
 			else if (m[1] == "transcript_name") tname = m[2];
+			else if (m[1] == "tag" && m[2] == "Ensembl_canonical") ens_canonical = true;
 		}
 		while ((m = re_gff3.exec(t[8])) != null) {
 			if (m[1] == "transcript_id") id = m[2];
@@ -1661,6 +1710,7 @@ function paf_gff2bed(args)
 			else if (m[1] == "gene_name" || m[1] == "gene_id") name = m[2];
 			else if (m[1] == "transcript_name") tname = m[2];
 		}
+		if (ens_canon_only && !ens_canonical) continue;
 		if (type == "" && biotype != "") type = biotype;
 		if (id == null) throw Error("No transcript_id");
 		if (id != last_id) {
@@ -1690,15 +1740,17 @@ function paf_gff2bed(args)
 
 function paf_sam2paf(args)
 {
-	var c, pri_only = false, long_cs = false;
-	while ((c = getopt(args, "pL")) != null) {
+	var c, pri_only = false, long_cs = false, pri_pri_only = false;
+	while ((c = getopt(args, "pPL")) != null) {
 		if (c == 'p') pri_only = true;
+		else if (c == 'P') pri_pri_only = pri_only = true;
 		else if (c == 'L') long_cs = true;
 	}
 	if (args.length == getopt.ind) {
 		print("Usage: paftools.js sam2paf [options] <in.sam>");
 		print("Options:");
 		print("  -p      convert primary or supplementary alignments only");
+		print("  -P      convert primary alignments only");
 		print("  -L      output the cs tag in the long form");
 		exit(1);
 	}
@@ -1725,6 +1777,7 @@ function paf_sam2paf(args)
 			throw Error("at line " + lineno + ": inconsistent SEQ and QUAL lengths - " + t[9].length + " != " + t[10].length);
 		if (t[2] == '*' || (flag&4) || t[5] == '*') continue;
 		if (pri_only && (flag&0x100)) continue;
+		if (pri_pri_only && (flag&0x900)) continue;
 		var tlen = ctg_len[t[2]];
 		if (tlen == null) throw Error("at line " + lineno + ": can't find the length of contig " + t[2]);
 		// find tags
@@ -1837,7 +1890,10 @@ function paf_sam2paf(args)
 		// optional tags
 		var type = flag&0x100? 'S' : 'P';
 		var tags = ["tp:A:" + type];
-		if (NM != null) tags.push("mm:i:"+mm);
+		if (NM != null) {
+			tags.push("NM:i:"+NM);
+			tags.push("mm:i:"+mm);
+		}
 		tags.push("gn:i:"+(I[1]+D[1]), "go:i:"+(I[0]+D[0]), "cg:Z:" + t[5].replace(/\d+[SH]/g, ''));
 		if (cs_str != null) tags.push("cs:Z:" + cs_str);
 		else if (cs.length > 0) tags.push("cs:Z:" + cs.join(""));
@@ -2047,7 +2103,7 @@ function paf_mapeval(args)
 		warn("Usage: paftools.js mapeval [options] <in.paf>|<in.sam>");
 		warn("Options:");
 		warn("  -r FLOAT   mapping correct if overlap_length/union_length>FLOAT [" + ovlp_ratio + "]");
-		warn("  -Q INT     print wrong mappings with mapQ>INT [don't print]");
+		warn("  -Q INT     print wrong mappings with mapQ>=INT [don't print]");
 		warn("  -m INT     0: eval the longest aln only; 1: first aln only; 2: all primary aln [0]");
 		exit(1);
 	}
@@ -2131,7 +2187,7 @@ function paf_mapeval(args)
 	}
 
 	var lineno = 0, last = null, a = [], n_unmapped = null;
-	var re_cigar = /(\d+)([MIDSHN])/g;
+	var re_cigar = /(\d+)([MIDSHN=X])/g;
 	while (file.readline(buf) >= 0) {
 		var m, line = buf.toString();
 		++lineno;
@@ -2169,7 +2225,7 @@ function paf_mapeval(args)
 				var n_gap = 0, mlen = 0;
 				while ((m = re_cigar.exec(t[5])) != null) {
 					var len = parseInt(m[1]);
-					if (m[2] == 'M') pos_end += len, mlen += len;
+					if (m[2] == 'M' || m[2] == 'X' || m[2] == '=') pos_end += len, mlen += len;
 					else if (m[2] == 'I') n_gap += len;
 					else if (m[2] == 'D') n_gap += len, pos_end += len;
 				}
@@ -2341,12 +2397,15 @@ function paf_pbsim2fq(args)
 
 function paf_junceval(args)
 {
-	var c, l_fuzzy = 0, print_ovlp = false, print_err_only = false, first_only = false, chr_only = false;
-	while ((c = getopt(args, "l:epc")) != null) {
+	var c, l_fuzzy = 0, print_ovlp = false, print_err_only = false, first_only = false, chr_only = false, aa = false, is_bed = false;
+	while ((c = getopt(args, "l:epcab1")) != null) {
 		if (c == 'l') l_fuzzy = parseInt(getopt.arg);
 		else if (c == 'e') print_err_only = print_ovlp = true;
 		else if (c == 'p') print_ovlp = true;
 		else if (c == 'c') chr_only = true;
+		else if (c == 'a') aa = true;
+		else if (c == 'b') is_bed = true;
+		else if (c == '1') first_only = true;
 	}
 
 	if (args.length - getopt.ind < 1) {
@@ -2356,6 +2415,9 @@ function paf_junceval(args)
 		print("  -p        print overlapping introns");
 		print("  -e        print erroreous overlapping introns");
 		print("  -c        only consider alignments to /^(chr)?([0-9]+|X|Y)$/");
+		print("  -a        miniprot PAF as input");
+		print("  -b        BED as input");
+		print("  -1        only process the first alignment of each query");
 		exit(1);
 	}
 
@@ -2409,13 +2471,17 @@ function paf_junceval(args)
 
 	file = getopt.ind+1 >= args.length || args[getopt.ind+1] == '-'? new File() : new File(args[getopt.ind+1]);
 	var last_qname = null;
-	var re_cigar = /(\d+)([MIDNSHP=X])/g;
+	var re_cigar = /(\d+)([MIDNSHP=XFGUV])/g;
 	while (file.readline(buf) >= 0) {
 		var m, t = buf.toString().split("\t");
-		var ctg_name = null, cigar = null, pos = null, qname = t[0];
+		var ctg_name = null, cigar = null, pos = null, qname;
 
 		if (t[0].charAt(0) == '@') continue;
-		if (t[4] == '+' || t[4] == '-' || t[4] == '*') { // PAF
+		if (t[0] == "##PAF") t.shift();
+		qname = t[0];
+		if (is_bed) {
+			ctg_name = t[0], pos = parseInt(t[1]), cigar == null;
+		} else if (t[4] == '+' || t[4] == '-' || t[4] == '*') { // PAF
 			ctg_name = t[5], pos = parseInt(t[7]);
 			var type = 'P';
 			for (i = 12; i < t.length; ++i) {
@@ -2428,6 +2494,10 @@ function paf_junceval(args)
 		} else { // SAM
 			ctg_name = t[2], pos = parseInt(t[3]) - 1, cigar = t[5];
 			var flag = parseInt(t[1]);
+			if (flag & 1) {
+				if (flag & 0x40) qname += '/1';
+				else if (flag & 0x80) qname += '/2';
+			}
 			if (flag&0x100) continue; // secondary
 		}
 
@@ -2445,12 +2515,43 @@ function paf_junceval(args)
 		}
 
 		var intron = [];
-		while ((m = re_cigar.exec(cigar)) != null) {
-			var len = parseInt(m[1]), op = m[2];
-			if (op == 'N') {
-				intron.push([pos, pos + len]);
-				pos += len;
-			} else if (op == 'M' || op == 'X' || op == '=' || op == 'D') pos += len;
+		if (is_bed) {
+			intron.push([pos, parseInt(t[2])]);
+		} else if (aa) {
+			var tmp_junc = [], tmp = 0;
+			while ((m = re_cigar.exec(cigar)) != null) {
+				var len = parseInt(m[1]), op = m[2];
+				if (op == 'N') {
+					tmp_junc.push([tmp, tmp + len]);
+					tmp += len;
+				} else if (op == 'U') {
+					tmp_junc.push([tmp + 1, tmp + len - 2]);
+					tmp += len;
+				} else if (op == 'V') {
+					tmp_junc.push([tmp + 2, tmp + len - 1]);
+					tmp += len;
+				} else if (op == 'M' || op == 'X' || op == '=' || op == 'D') {
+					tmp += len * 3;
+				} else if (op == 'F' || op == 'G') {
+					tmp += len;
+				}
+			}
+			if (t[4] == '+') {
+				for (var i = 0; i < tmp_junc.length; ++i)
+					intron.push([pos + tmp_junc[i][0], pos + tmp_junc[i][1]]);
+			} else if (t[4] == '-') {
+				var glen = parseInt(t[8]) - parseInt(t[7]);
+				for (var i = tmp_junc.length - 1; i >= 0; --i)
+					intron.push([pos + (glen - tmp_junc[i][1]), pos + (glen - tmp_junc[i][0])]);
+			}
+		} else {
+			while ((m = re_cigar.exec(cigar)) != null) {
+				var len = parseInt(m[1]), op = m[2];
+				if (op == 'N') {
+					intron.push([pos, pos + len]);
+					pos += len;
+				} else if (op == 'M' || op == 'X' || op == '=' || op == 'D') pos += len;
+			}
 		}
 		if (intron.length == 0) {
 			++n_sgl;
@@ -2509,6 +2610,276 @@ function paf_junceval(args)
 	}
 }
 
+function paf_exoneval(args) // adapted from paf_junceval()
+{
+	var c, l_fuzzy = 0, print_ovlp = false, print_err_only = false, first_only = false, chr_only = false, aa = false, is_bed = false, use_cds = false, eval_base = false;
+	while ((c = getopt(args, "l:epcab1ds")) != null) {
+		if (c == 'l') l_fuzzy = parseInt(getopt.arg);
+		else if (c == 'e') print_err_only = print_ovlp = true;
+		else if (c == 'p') print_ovlp = true;
+		else if (c == 'c') chr_only = true;
+		else if (c == 'a') aa = true, use_cds = true;
+		else if (c == 'b') is_bed = true;
+		else if (c == '1') first_only = true;
+		else if (c == 'd') use_cds = true;
+		else if (c == 's') eval_base = true;
+	}
+
+	if (args.length - getopt.ind < 1) {
+		print("Usage: paftools.js exoneval [options] <gene.gtf> <aln.sam>");
+		print("Options:");
+		print("  -l INT    tolerance of junction positions (0 for exact) [0]");
+		print("  -d        evaluate coding regions only (exon regions by default)");
+		print("  -a        miniprot PAF as input (force -d)");
+		print("  -p        print overlapping exons");
+		print("  -e        print erroreous overlapping exons");
+		print("  -c        only consider alignments to /^(chr)?([0-9]+|X|Y)$/");
+		print("  -1        only process the first alignment of each query");
+		print("  -b        BED as input");
+		print("  -s        compute base Sn and Sp (more memory)");
+		exit(1);
+	}
+
+	var file, buf = new Bytes();
+
+	warn("Reading reference GTF...");
+	var tr = {};
+	file = args[getopt.ind] == '-'? new File() : new File(args[getopt.ind]);
+	while (file.readline(buf) >= 0) {
+		var m, t = buf.toString().split("\t");
+		if (t[0].charAt(0) == '#') continue;
+		if (use_cds) {
+			if (t[2] != "cds" && t[2] != "CDS") continue;
+		} else {
+			if (t[2] != 'exon') continue;
+		}
+		var st = parseInt(t[3]) - 1;
+		var en = parseInt(t[4]);
+		if ((m = /transcript_id "(\S+)"/.exec(t[8])) == null) continue;
+		var tid = m[1];
+		if (tr[tid] == null) tr[tid] = [t[0], t[6], 0, 0, []];
+		tr[tid][4].push([st, en]); // this keeps transcript
+	}
+	file.close();
+
+	var anno = {};
+	for (var tid in tr) { // traverse each transcript
+		var t = tr[tid];
+		Interval.sort(t[4]);
+		t[2] = t[4][0][0];
+		t[3] = t[4][t[4].length - 1][1];
+		if (anno[t[0]] == null) anno[t[0]] = [];
+		var s = t[4];
+		for (var i = 0; i < s.length; ++i) // traverse each exon
+			anno[t[0]].push([s[i][0], s[i][1]]);
+	}
+	tr = null;
+
+	for (var chr in anno) { // index exons
+		var e = anno[chr];
+		if (e.length == 0) continue;
+		Interval.sort(e);
+		var k = 0;
+		for (var i = 1; i < e.length; ++i) // dedup
+			if (e[i][0] != e[k][0] || e[i][1] != e[k][1])
+				e[++k] = e[i].slice(0);
+		e.length = k + 1;
+		Interval.index_end(e);
+	}
+
+	var n_pri = 0, n_unmapped = 0, n_mapped = 0;
+	var n_exon = 0, n_exon_hit = 0, n_exon_novel = 0;
+
+	file = getopt.ind+1 >= args.length || args[getopt.ind+1] == '-'? new File() : new File(args[getopt.ind+1]);
+	var last_qname = null, qexon = {};
+	var re_cigar = /(\d+)([MIDNSHP=XFGUV])/g;
+
+	warn("Evaluating alignments...");
+	while (file.readline(buf) >= 0) {
+		var m, t = buf.toString().split("\t");
+		var ctg_name = null, cigar = null, pos = null, qname;
+
+		if (t[0].charAt(0) == '@') continue;
+		if (t[0] == "##PAF") t.shift();
+		qname = t[0];
+		if (is_bed) {
+			ctg_name = t[0], pos = parseInt(t[1]), cigar == null;
+		} else if (t[4] == '+' || t[4] == '-' || t[4] == '*') { // PAF
+			ctg_name = t[5], pos = parseInt(t[7]);
+			var type = 'P';
+			for (i = 12; i < t.length; ++i) {
+				if ((m = /^(tp:A|cg:Z):(\S+)/.exec(t[i])) != null) {
+					if (m[1] == 'tp:A') type = m[2];
+					else cigar = m[2];
+				}
+			}
+			if (type == 'S') continue; // secondary
+		} else { // SAM
+			ctg_name = t[2], pos = parseInt(t[3]) - 1, cigar = t[5];
+			var flag = parseInt(t[1]);
+			if (flag&0x100) continue; // secondary
+		}
+
+		if (chr_only && !/^(chr)?([0-9]+|X|Y)$/.test(ctg_name)) continue;
+		if (first_only && last_qname == qname) continue;
+		if (ctg_name == '*') { // unmapped
+			++n_unmapped;
+			continue;
+		} else {
+			++n_pri;
+			if (last_qname != qname) {
+				++n_mapped;
+				last_qname = qname;
+			}
+		}
+
+		var exon = [];
+		if (is_bed) { // BED
+			exon.push([pos, parseInt(t[2])]);
+		} else if (aa) {
+			var tmp_exon = [], tmp = 0, tmp_st = 0;
+			while ((m = re_cigar.exec(cigar)) != null) {
+				var len = parseInt(m[1]), op = m[2];
+				if (op == 'N') {
+					tmp_exon.push([tmp_st, tmp]);
+					tmp_st = tmp + len, tmp += len;
+				} else if (op == 'U') {
+					tmp_exon.push([tmp_st, tmp + 1]);
+					tmp_st = tmp + len - 2, tmp += len;
+				} else if (op == 'V') {
+					tmp_exon.push([tmp_st, tmp + 2]);
+					tmp_st = tmp + len - 1, tmp += len;
+				} else if (op == 'M' || op == 'X' || op == '=' || op == 'D') {
+					tmp += len * 3;
+				} else if (op == 'F' || op == 'G') {
+					tmp += len;
+				}
+			}
+			tmp_exon.push([tmp_st, tmp]);
+			if (t[4] == '+') {
+				for (var i = 0; i < tmp_exon.length; ++i)
+					exon.push([pos + tmp_exon[i][0], pos + tmp_exon[i][1]]);
+			} else if (t[4] == '-') { // For protein-to-genome alignment, the coordinates are on the query strand. Need to flip them.
+				var glen = parseInt(t[8]) - parseInt(t[7]);
+				for (var i = tmp_exon.length - 1; i >= 0; --i)
+					exon.push([pos + (glen - tmp_exon[i][1]), pos + (glen - tmp_exon[i][0])]);
+			}
+		} else {
+			var tmp_st = pos;
+			while ((m = re_cigar.exec(cigar)) != null) {
+				var len = parseInt(m[1]), op = m[2];
+				if (op == 'N') {
+					exon.push([tmp_st, pos]);
+					tmp_st = pos + len, pos += len;
+				} else if (op == 'M' || op == 'X' || op == '=' || op == 'D') pos += len;
+			}
+			exon.push([tmp_st, pos]);
+		}
+		n_exon += exon.length;
+
+		var chr = anno[ctg_name];
+		if (chr != null) {
+			for (var i = 0; i < exon.length; ++i) {
+				if (eval_base) {
+					if (qexon[ctg_name] == null) qexon[ctg_name] = [];
+					qexon[ctg_name].push([exon[i][0], exon[i][1]]);
+				}
+				var o = Interval.find_ovlp(chr, exon[i][0], exon[i][1]);
+				if (o.length > 0) {
+					var hit = false;
+					for (var j = 0; j < o.length; ++j) {
+						var st_diff = exon[i][0] - o[j][0];
+						var en_diff = exon[i][1] - o[j][1];
+						if (st_diff < 0) st_diff = -st_diff;
+						if (en_diff < 0) en_diff = -en_diff;
+						if (st_diff <= l_fuzzy && en_diff <= l_fuzzy)
+							++n_exon_hit, hit = true;
+						if (hit) break;
+					}
+					if (print_ovlp) {
+						var type = hit? 'C' : 'P';
+						if (hit && print_err_only) continue;
+						var x = '[';
+						for (var j = 0; j < o.length; ++j) {
+							if (j) x += ', ';
+							x += '(' + o[j][0] + "," + o[j][1] + ')';
+						}
+						x += ']';
+						print(type, qname, i+1, ctg_name, exon[i][0], exon[i][1], x);
+					}
+				} else {
+					++n_exon_novel;
+					if (print_ovlp)
+						print('N',  qname, i+1, ctg_name, exon[i][0], exon[i][1]);
+				}
+			}
+		} else {
+			n_exon_novel += exon.length;
+		}
+	}
+	file.close();
+
+	buf.destroy();
+
+	if (!print_ovlp) {
+		print("# unmapped reads: " + n_unmapped);
+		print("# mapped reads: " + n_mapped);
+		print("# primary alignments: " + n_pri);
+		print("# predicted exons: " + n_exon);
+		print("# non-overlapping exons: " + n_exon_novel);
+		print("# correct exons: " + n_exon_hit + " (" + (n_exon_hit / n_exon * 100).toFixed(2) + "%)");
+	}
+
+	function merge_and_index(ex) {
+		for (var chr in ex) {
+			var a = [];
+			e = ex[chr];
+			Interval.sort(e);
+			var st = e[0][0], en = e[0][1];
+			for (var i = 1; i < e.length; ++i) { // merge
+				if (e[i][0] > en) {
+					a.push([st, en]);
+					st = e[i][0], en = e[i][1];
+				} else {
+					en = en > e[i][1]? en : e[i][1];
+				}
+			}
+			a.push([st, en]);
+			Interval.index_end(a);
+			ex[chr] = a;
+		}
+	}
+
+	function cal_sn(a0, a1) {
+		var tot = 0, cov = 0;
+		for (var chr in a1) {
+			var e0 = a0[chr], e1 = a1[chr];
+			for (var i = 0; i < e1.length; ++i)
+				tot += e1[i][1] - e1[i][0];
+			if (e0 == null) continue;
+			for (var i = 0; i < e1.length; ++i) {
+				var o = Interval.find_ovlp(e0, e1[i][0], e1[i][1]);
+				for (var j = 0; j < o.length; ++j) { // this only works when there are no overlaps between intervals
+					var st = e1[i][0] > o[j][0]? e1[i][0] : o[j][0];
+					var en = e1[i][1] < o[j][1]? e1[i][1] : o[j][1];
+					cov += en - st;
+				}
+			}
+		}
+		return [tot, cov];
+	}
+
+	if (eval_base) {
+		warn("Computing base Sn and Sp...");
+		merge_and_index(qexon);
+		merge_and_index(anno);
+		var sn = cal_sn(qexon, anno);
+		var sp = cal_sn(anno, qexon);
+		print("Base Sn: " + sn[1] + " / " + sn[0] + " = " + (sn[1] / sn[0] * 100).toFixed(2) + "%");
+		print("Base Sp: " + sp[1] + " / " + sp[0] + " = " + (sp[1] / sp[0] * 100).toFixed(2) + "%");
+	}
+}
+
 // evaluate overlap sensitivity
 function paf_ov_eval(args)
 {
@@ -2704,6 +3075,23 @@ function paf_misjoin(args)
 		return len < (en - st) * cen_ratio? false : true;
 	}
 
+	function test_cen_point(cen, chr, x) {
+		var b = cen[chr];
+		if (b == null) return false;
+		for (var j = 0; j < b.length; ++j)
+			if (x >= b[j][0] && x < b[j][1])
+				return true;
+		return false;
+	}
+
+	if (show_err || show_long) {
+		print("C\tJ  inter-chromosomal misjoin");
+		print("C\tj  inter-chromosomal misjoin with both breakpoints ending in centromeres");
+		print("C\tG  long gap on the reference genome");
+		print("C\tg  long gap on the reference genome with both breakpoints ending in centromeres");
+		print("C\tM  closed inversion");
+		print("C");
+	}
 	function process(a) {
 		var k = 0;
 		for (var i = 0; i < a.length; ++i) {
@@ -2716,14 +3104,17 @@ function paf_misjoin(args)
 		a = a.sort(function(x,y){return x[2]-y[2]});
 		if (show_long) for (var i = 0; i < a.length; ++i) print(a[i].join("\t"));
 		for (var i = 1; i < a.length; ++i) {
-			var ov = [false, false];
+			var ov = [false, false], end_cen = [false, false];
 			ov[0] = test_cen(cen, a[i-1][5], a[i-1][7], a[i-1][8]);
 			ov[1] = test_cen(cen, a[i][5], a[i][7], a[i][8]);
+			end_cen[0] = test_cen_point(cen, a[i-1][5], a[i-1][4] == '+'? a[i-1][8] : a[i-1][7]);
+			end_cen[1] = test_cen_point(cen, a[i][5],   a[i][4] == '+'?   a[i][7]   : a[i][8]);
 			if (a[i-1][5] != a[i][5]) { // different chr
 				if (ov[0] || ov[1]) ++n_diff[1];
 				else if (show_err) {
-					print("J", a[i-1].slice(0, 12).join("\t"));
-					print("J", a[i].slice(0, 12).join("\t"));
+					var label = end_cen[0] && end_cen[1]? 'j' : 'J';
+					print(label, a[i-1].slice(0, 12).join("\t"));
+					print(label, a[i].slice(0, 12).join("\t"));
 				}
 				++n_diff[0];
 			} else if (a[i-1][4] == a[i][4]) { // a gap
@@ -2733,8 +3124,9 @@ function paf_misjoin(args)
 				if (gap > max_gap) {
 					if (ov[0] || ov[1]) ++n_gap[1];
 					else if (show_err) {
-						print("G", a[i-1].slice(0, 12).join("\t"));
-						print("G", a[i].slice(0, 12).join("\t"));
+						var label = end_cen[0] && end_cen[1]? 'g' : 'G';
+						print(label, a[i-1].slice(0, 12).join("\t"));
+						print(label, a[i].slice(0, 12).join("\t"));
 					}
 					++n_gap[0];
 				}
@@ -2852,6 +3244,7 @@ function paf_sveval(args)
 			if (bed != null && bed[t[0]] == null) continue;
 			if (t[4] == '<INV>' || t[4] == '<INVDUP>') continue; // no inversion
 			if (/[\[\]]/.test(t[4])) continue; // no break points
+			if (t[6] != "." && t[6] != "PASS") continue;
 			var st = parseInt(t[1]) - 1, en = st + t[3].length;
 			// parse svlen
 			var b = _paf_get_alen(t), svlen = b[0];
@@ -3084,6 +3477,183 @@ function paf_pafcmp(args)
 	buf.destroy();
 }
 
+function paf_longcs2seq(args) {
+	var c, opt = { query:false };
+	while ((c = getopt(args, "q")) != null)
+		if (c == 'q') opt.query = true;
+	if (args.length == getopt.ind) {
+		print("Usage: paftools.js longcs2seq [-q] <long-cs.paf>");
+		return;
+	}
+	var re_cs = /([:=*+-])(\d+|[A-Za-z]+)/g
+	var buf = new Bytes();
+	var file = args[getopt.ind] == "-"? new File() : new File(args[getopt.ind]);
+	while (file.readline(buf) >= 0) {
+		var m, cs = null, t = buf.toString().split("\t");
+		for (var i = 12; i < t.length; ++i)
+			if ((m = /^cs:Z:(\S+)/.exec(t[i])) != null) {
+				cs = m[1];
+				break;
+			}
+		if (cs == null) continue;
+		var ts = "", qs = "";
+		while ((m = re_cs.exec(cs)) != null) {
+			if (m[1] == "=") ts += m[2], qs += m[2];
+			else if (m[1] == "+") qs += m[2].toUpperCase();
+			else if (m[1] == "-") ts += m[2].toUpperCase();
+			else if (m[1] == "*") ts += m[2][0].toUpperCase(), qs += m[2][1].toUpperCase();
+			else if (m[1] == ":") throw Error("Long cs is required");
+		}
+		if (opt.query) {
+			print(">" + t[0] + "_" + t[2] + "_" + t[3]);
+			print(qs);
+		} else {
+			print(">" + t[5] + "_" + t[7] + "_" + t[8]);
+			print(ts);
+		}
+	}
+	file.close();
+	buf.destroy();
+}
+
+function paf_paf2gff(args) {
+	var c, opt = { aa:false };
+	var re_cigar = /(\d+)([A-Z=])/g;
+	while ((c = getopt(args, "a")) != null) {
+		if (c == 'a') opt.aa = true;
+	}
+	if (args.length == getopt.ind) {
+		print("Usage: paftools.js paf2gff [-a] <in.paf>");
+		return;
+	}
+	var buf = new Bytes();
+	var file = args[getopt.ind] == '-'? new File() : new File(args[getopt.ind]);
+	var hid = 1, last_name = null;
+	while (file.readline(buf) >= 0) {
+		var m, t = buf.toString().split("\t");
+		if (t[5] == '*') continue; // skip unmapped lines
+
+		if (t[0] != last_name) last_name = t[0], hid = 1;
+		else ++hid;
+		for (var i = 1; i <= 3; ++i) t[i] = parseInt(t[i]);
+		for (var i = 6; i <= 11; ++i) t[i] = parseInt(t[i]);
+		var cigar = null, score = null, np = null, dist_stop = null, dist_start = null;
+		for (var i = 12; i < t.length; ++i) {
+			if ((m = /^(cg:Z|AS:i|np:i|da:i|do:i):(\S+)/.exec(t[i])) != null) {
+				if (m[1] == 'cg:Z') cigar = m[2];
+				else if (m[1] == 'AS:i') score = parseInt(m[2]);
+				else if (m[1] == 'np:i') np = parseInt(m[2]);
+				else if (m[1] == 'do:i') dist_stop = parseInt(m[2]);
+				else if (m[1] == 'da:i') dist_start = parseInt(m[2]);
+			}
+		}
+		if (cigar == null) throw Error("failed to find the cg:Z tag");
+		if (score == null) throw Error("failed to find the AS:i tag");
+
+		var st = 0, en = 0, phase = 0, pseudo = false, fs = 0, a = [];
+		if (dist_start != null && dist_start == 0)
+			a.push([t[5], 'paf2gff', 'start_codon', 0, 3, 0, t[4], '.', 0]);
+		while ((m = re_cigar.exec(cigar)) != null) {
+			var len = parseInt(m[1]);
+			if (m[2] == 'M' || m[2] == 'D') {
+				en += opt.aa? len * 3 : len;
+			} else if (m[2] == 'F' || m[2] == 'G' || m[2] == 'R') {
+				en += len, pseudo = true, fs = 1;
+			} else if (m[2] == 'N') {
+				a.push([t[5], 'paf2gff', 'exon', st, en, 0, t[4], phase, fs]);
+				st = en + len, en += len, phase = 0, fs = 0;
+			} else if (m[2] == 'U') { // ...xGT...AGxx...
+				a.push([t[5], 'paf2gff', 'exon', st, en + 1, 0, t[4], phase, fs]);
+				st = en + len - 2, en += len, phase = 2, fs = 0;
+			} else if (m[2] == 'V') { // ...xxGT...AGx...
+				a.push([t[5], 'paf2gff', 'exon', st, en + 2, 0, t[4], phase, fs]);
+				st = en + len - 1, en += len, phase = 1, fs = 0;
+			}
+		}
+		a.push([t[5], 'paf2gff', 'exon', st, en, 0, t[4], phase, fs]);
+		if (en != t[8] - t[7]) throw Error("inconsistent cigar");
+		if (dist_stop != null && dist_stop == 0)
+			a.push([t[5], 'paf2gff', 'stop_codon', en, en + 3, 0, t[4], '.', 0]);
+		var type = pseudo? 'pseudogene' : 'protein_coding';
+		var attr = ['transcript_id=' + t[0] + '#' + hid, 'transcript_type=' + type].join(";");
+		var trans_attr = 'identity=' + (t[9] / t[10]).toFixed(4);
+		if (np != null) trans_attr += ';positive=' + (np * 3 / t[10]).toFixed(4);
+		trans_attr += ';aa_start=' + t[2];
+		trans_attr += ';aa_end=' + (t[1] - t[3]);
+		if (dist_start != null && dist_start >= 0) trans_attr += ';dist_start_codon=' + dist_start;
+		if (dist_stop != null && dist_stop >= 0) trans_attr += ';dist_stop_codon=' + dist_stop;
+		var trans_st = t[7], trans_en = t[8];
+		if (dist_stop != null && dist_stop == 0) {
+			if (t[4] == '-') trans_st -= 3;
+			else trans_en += 3;
+		}
+		print([t[5], 'paf2gff', 'transcript', trans_st + 1, trans_en, score, t[4], '.', attr + ';' + trans_attr].join("\t"));
+		if (opt.aa && t[4] == '-') {
+			var b = [], len = t[8] - t[7];
+			for (var i = a.length - 1; i >= 0; --i) {
+				var x = len - a[i][3];
+				a[i][3] = len - a[i][4];
+				a[i][4] = x;
+				//a[i][7] = a[i][7] == 0? 0 : 3 - a[i][7]; // not sure if this line is needed
+				b.push(a[i]);
+			}
+			a = b;
+		}
+		for (var i = 0; i < a.length; ++i) {
+			if (!pseudo && a[i][2] == "exon") a[i][2] = "CDS";
+			a[i][3] += t[7] + 1;
+			a[i][4] += t[7];
+			a[i][8] = attr + ";frameshift=" + a[i][8];
+			print(a[i].join("\t"));
+		}
+	}
+	file.close();
+	buf.destroy();
+}
+
+function paf_gff2junc(args) {
+	var c, feat = "CDS";
+	while ((c = getopt(args, "f:")) != null) {
+		if (c == 'f') feat = getopt.arg;
+	}
+	if (getopt.ind == args.length) {
+		print("Usage: paftools.js gff2junc [-f feature] <in.gff3>");
+		return;
+	}
+	var buf = new Bytes();
+	var file = args[getopt.ind] == "-"? new File() : new File(args[getopt.ind]);
+
+	function process_a(a) {
+		if (a.length < 2) return;
+		a = a.sort(function(x, y) { return x[4] - y[4] });
+		for (var i = 1; i < a.length; ++i)
+			print([a[i][1], a[i-1][5], a[i][4], a[i][0], 0, a[i][7]].join("\t"));
+	}
+
+	var a = [];
+	while (file.readline(buf) >= 0) {
+		var m, t = buf.toString().split("\t");
+		if (t[0][0] == '#') continue;
+		if (t[2].toLowerCase() != feat.toLowerCase()) continue;
+		//print(t.join("\t"));
+		if ((m = /\bParent=([^;]+)/.exec(t[8])) == null) {
+			warn("Can't find Parent");
+			continue;
+		}
+		t[3] = parseInt(t[3]) - 1;
+		t[4] = parseInt(t[4]);
+		t.unshift(m[1]);
+		if (a.length > 0 && a[0][0] != m[1]) {
+			process_a(a);
+			a.length = 0;
+			a.push(t);
+		} else a.push(t);
+	}
+	process_a(a);
+	file.close();
+	buf.destroy();
+}
+
 /*************************
  ***** main function *****
  *************************/
@@ -3098,6 +3668,9 @@ function main(args)
 		print("  sam2paf    convert SAM to PAF");
 		print("  delta2paf  convert MUMmer's delta to PAF");
 		print("  gff2bed    convert GTF/GFF3 to BED12");
+		print("  gff2junc   convert GFF3 to junction BED");
+		print("  longcs2seq convert long-cs PAF to sequences");
+//		print("  paf2gff    convert PAF to GFF3 (tested for miniprot only)");
 		print("");
 		print("  stat       collect basic mapping information in PAF/SAM");
 		print("  asmstat    collect basic assembly information");
@@ -3115,6 +3688,7 @@ function main(args)
 		print("  mason2fq   convert mason2-simulated SAM to FASTQ");
 		print("  pbsim2fq   convert PBSIM-simulated MAF to FASTQ");
 		print("  junceval   evaluate splice junction consistency with known annotations");
+		print("  exoneval   evaluate exon-level consistency with known annotations");
 		print("  ov-eval    evaluate read overlap sensitivity using read-to-ref mapping");
 		exit(1);
 	}
@@ -3125,6 +3699,7 @@ function main(args)
 	else if (cmd == 'delta2paf') paf_delta2paf(args);
 	else if (cmd == 'splice2bed') paf_splice2bed(args);
 	else if (cmd == 'gff2bed') paf_gff2bed(args);
+	else if (cmd == 'gff2junc') paf_gff2junc(args);
 	else if (cmd == 'stat') paf_stat(args);
 	else if (cmd == 'asmstat') paf_asmstat(args);
 	else if (cmd == 'asmgene') paf_asmgene(args);
@@ -3138,10 +3713,13 @@ function main(args)
 	else if (cmd == 'mason2fq') paf_mason2fq(args);
 	else if (cmd == 'pbsim2fq') paf_pbsim2fq(args);
 	else if (cmd == 'junceval') paf_junceval(args);
+	else if (cmd == 'exoneval') paf_exoneval(args);
 	else if (cmd == 'ov-eval') paf_ov_eval(args);
 	else if (cmd == 'vcfstat') paf_vcfstat(args);
 	else if (cmd == 'sveval') paf_sveval(args);
 	else if (cmd == 'vcfsel') paf_vcfsel(args);
+	else if (cmd == 'longcs2seq') paf_longcs2seq(args);
+	else if (cmd == 'paf2gff') paf_paf2gff(args);
 	else if (cmd == 'version') print(paftools_version);
 	else throw Error("unrecognized command: " + cmd);
 }

mmpriv.h

@@ -13,6 +13,8 @@
 #define MM_DBG_PRINT_QNAME   0x2
 #define MM_DBG_PRINT_SEED    0x4
 #define MM_DBG_PRINT_ALN_SEQ 0x8
+#define MM_DBG_PRINT_CHAIN   0x10
+#define MM_DBG_SEED_FREQ     0x20
 
 #define MM_SEED_LONG_JOIN  (1ULL<<40)
 #define MM_SEED_IGNORE     (1ULL<<41)
@@ -22,6 +24,9 @@
 #define MM_SEED_SEG_SHIFT  48
 #define MM_SEED_SEG_MASK   (0xffULL<<(MM_SEED_SEG_SHIFT))
 
+#define MM_JUNC_ANNO 0x1
+#define MM_JUNC_MISC 0x2
+
 #ifndef kroundup32
 #define kroundup32(x) (--(x), (x)|=(x)>>1, (x)|=(x)>>2, (x)|=(x)>>4, (x)|=(x)>>8, (x)|=(x)>>16, ++(x))
 #endif
@@ -31,6 +36,7 @@
 
 #define MALLOC(type, len) ((type*)malloc((len) * sizeof(type)))
 #define CALLOC(type, len) ((type*)calloc((len), sizeof(type)))
+#define REALLOC(type, ptr, cnt) ((type*)realloc((ptr), (cnt) * sizeof(type)))
 
 #ifdef __cplusplus
 extern "C" {
@@ -50,6 +56,12 @@ typedef struct {
 	mm128_t *a;
 } mm_seg_t;
 
+typedef struct {
+	int32_t off, off2, cnt;
+	int16_t strand;
+	uint16_t flag;
+} mm_idx_jjump1_t;
+
 double cputime(void);
 double realtime(void);
 long peakrss(void);
@@ -67,19 +79,23 @@ double mm_event_identity(const mm_reg1_t *r);
 int mm_write_sam_hdr(const mm_idx_t *mi, const char *rg, const char *ver, int argc, char *argv[]);
 void mm_write_paf(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, void *km, int64_t opt_flag);
 void mm_write_paf3(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, void *km, int64_t opt_flag, int rep_len);
+void mm_write_paf4(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, void *km, int64_t opt_flag, int rep_len, int n_seg, int seg_idx);
 void mm_write_sam(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, int n_regs, const mm_reg1_t *regs);
 void mm_write_sam2(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, int seg_idx, int reg_idx, int n_seg, const int *n_regs, const mm_reg1_t *const* regs, void *km, int64_t opt_flag);
 void mm_write_sam3(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, int seg_idx, int reg_idx, int n_seg, const int *n_regss, const mm_reg1_t *const* regss, void *km, int64_t opt_flag, int rep_len);
+void mm_write_junc(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r);
 
+// indexing related in index.c
 void mm_idxopt_init(mm_idxopt_t *opt);
 const uint64_t *mm_idx_get(const mm_idx_t *mi, uint64_t minier, int *n);
 int32_t mm_idx_cal_max_occ(const mm_idx_t *mi, float f);
 int mm_idx_getseq2(const mm_idx_t *mi, int is_rev, uint32_t rid, uint32_t st, uint32_t en, uint8_t *seq);
-mm_reg1_t *mm_align_skeleton(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int qlen, const char *qstr, int *n_regs_, mm_reg1_t *regs, mm128_t *a);
 mm_reg1_t *mm_gen_regs(void *km, uint32_t hash, int qlen, int n_u, uint64_t *u, mm128_t *a, int is_qstrand);
+int mm_idx_bed_read(mm_idx_t *mi, const char *fn, int read_junc);
+int mm_idx_jjump_read(mm_idx_t *mi, const char *fn, int flag, int min_sc);
+const mm_idx_jjump1_t *mm_idx_jump_get(const mm_idx_t *db, int32_t cid, int32_t st, int32_t en, int32_t *n);
 
-mm128_t *mm_chain_dp(int max_dist_x, int max_dist_y, int bw, int max_skip, int max_iter, int min_cnt, int min_sc, float gap_scale,
-					 int is_cdna, int n_segs, int64_t n, mm128_t *a, int *n_u_, uint64_t **_u, void *km);
+// chaining in lchain.c
 mm128_t *mg_lchain_dp(int max_dist_x, int max_dist_y, int bw, int max_skip, int max_iter, int min_cnt, int min_sc, float chn_pen_gap, float chn_pen_skip,
 					  int is_cdna, int n_segs, int64_t n, mm128_t *a, int *n_u_, uint64_t **_u, void *km);
 mm128_t *mg_lchain_rmq(int max_dist, int max_dist_inner, int bw, int max_chn_skip, int cap_rmq_size, int min_cnt, int min_sc, float chn_pen_gap, float chn_pen_skip,
@@ -96,8 +112,12 @@ void mm_select_sub_multi(void *km, float pri_ratio, float pri1, float pri2, int
 int mm_filter_strand_retained(int n_regs, mm_reg1_t *r);
 void mm_filter_regs(const mm_mapopt_t *opt, int qlen, int *n_regs, mm_reg1_t *regs);
 void mm_hit_sort(void *km, int *n_regs, mm_reg1_t *r, float alt_diff_frac);
-void mm_set_mapq(void *km, int n_regs, mm_reg1_t *regs, int min_chain_sc, int match_sc, int rep_len, int is_sr);
+void mm_set_mapq2(void *km, int n_regs, mm_reg1_t *regs, int min_chain_sc, int match_sc, int rep_len, int is_sr, int is_splice);
 void mm_update_dp_max(int qlen, int n_regs, mm_reg1_t *regs, float frac, int a, int b);
+void mm_jump_split(void *km, const mm_idx_t *mi, const mm_mapopt_t *opt, int32_t qlen, const uint8_t *qseq, mm_reg1_t *r, int32_t ts_strand);
+
+mm_reg1_t *mm_align_skeleton(void *km, const mm_mapopt_t *opt, const mm_idx_t *mi, int qlen, const char *qstr, int *n_regs_, mm_reg1_t *regs, mm128_t *a);
+void mm_enlarge_cigar(mm_reg1_t *r, uint32_t n_cigar);
 
 void mm_est_err(const mm_idx_t *mi, int qlen, int n_regs, mm_reg1_t *regs, const mm128_t *a, int32_t n, const uint64_t *mini_pos);
 
@@ -105,6 +125,8 @@ mm_seg_t *mm_seg_gen(void *km, uint32_t hash, int n_segs, const int *qlens, int
 void mm_seg_free(void *km, int n_segs, mm_seg_t *segs);
 void mm_pair(void *km, int max_gap_ref, int dp_bonus, int sub_diff, int match_sc, const int *qlens, int *n_regs, mm_reg1_t **regs);
 
+void mm_jump_split(void *km, const mm_idx_t *mi, const mm_mapopt_t *opt, int32_t qlen, const uint8_t *qseq, mm_reg1_t *r, int32_t ts_strand);
+
 FILE *mm_split_init(const char *prefix, const mm_idx_t *mi);
 mm_idx_t *mm_split_merge_prep(const char *prefix, int n_splits, FILE **fp, uint32_t *n_seq_part);
 int mm_split_merge(int n_segs, const char **fn, const mm_mapopt_t *opt, int n_split_idx);

options.c

@@ -8,7 +8,7 @@ void mm_idxopt_init(mm_idxopt_t *opt)
 	opt->k = 15, opt->w = 10, opt->flag = 0;
 	opt->bucket_bits = 14;
 	opt->mini_batch_size = 50000000;
-	opt->batch_size = 4000000000ULL;
+	opt->batch_size = 8000000000ULL;
 }
 
 void mm_mapopt_init(mm_mapopt_t *opt)
@@ -45,6 +45,7 @@ void mm_mapopt_init(mm_mapopt_t *opt)
 	opt->alt_drop = 0.15f;
 
 	opt->a = 2, opt->b = 4, opt->q = 4, opt->e = 2, opt->q2 = 24, opt->e2 = 1;
+	opt->transition = 0;
 	opt->sc_ambi = 1;
 	opt->zdrop = 400, opt->zdrop_inv = 200;
 	opt->end_bonus = -1;
@@ -54,13 +55,15 @@ void mm_mapopt_init(mm_mapopt_t *opt)
 	opt->max_clip_ratio = 1.0f;
 	opt->mini_batch_size = 500000000;
 	opt->max_sw_mat = 100000000;
-	opt->cap_kalloc = 1000000000;
+	opt->cap_kalloc = 500000000;
 
 	opt->rank_min_len = 500;
 	opt->rank_frac = 0.9f;
 
 	opt->pe_ori = 0; // FF
 	opt->pe_bonus = 33;
+
+	opt->jump_min_match = 3;
 }
 
 void mm_mapopt_update(mm_mapopt_t *opt, const mm_idx_t *mi)
@@ -74,6 +77,7 @@ void mm_mapopt_update(mm_mapopt_t *opt, const mm_idx_t *mi)
 		if (opt->max_mid_occ > opt->min_mid_occ && opt->mid_occ > opt->max_mid_occ)
 			opt->mid_occ = opt->max_mid_occ;
 	}
+	if (opt->bw_long < opt->bw) opt->bw_long = opt->bw;
 	if (mm_verbose >= 3)
 		fprintf(stderr, "[M::%s::%.3f*%.2f] mid_occ = %d\n", __func__, realtime() - mm_realtime0, cputime() / (realtime() - mm_realtime0), opt->mid_occ);
 }
@@ -89,7 +93,7 @@ int mm_set_opt(const char *preset, mm_idxopt_t *io, mm_mapopt_t *mo)
 	if (preset == 0) {
 		mm_idxopt_init(io);
 		mm_mapopt_init(mo);
-	} else if (strcmp(preset, "map-ont") == 0) { // this is the same as the default
+	} else if (strcmp(preset, "lr") == 0 || strcmp(preset, "map-ont") == 0) { // this is the same as the default
 	} else if (strcmp(preset, "ava-ont") == 0) {
 		io->flag = 0, io->k = 15, io->w = 5;
 		mo->flag |= MM_F_ALL_CHAINS | MM_F_NO_DIAG | MM_F_NO_DUAL | MM_F_NO_LJOIN;
@@ -104,16 +108,33 @@ int mm_set_opt(const char *preset, mm_idxopt_t *io, mm_mapopt_t *mo)
 		mo->min_chain_score = 100, mo->pri_ratio = 0.0f, mo->max_chain_skip = 25;
 		mo->bw_long = mo->bw;
 		mo->occ_dist = 0;
-	} else if (strcmp(preset, "map-hifi") == 0 || strcmp(preset, "map-ccs") == 0) {
+	} else if (strcmp(preset, "lr:hq") == 0 || strcmp(preset, "map-hifi") == 0 || strcmp(preset, "map-ccs") == 0) {
 		io->flag = 0, io->k = 19, io->w = 19;
 		mo->max_gap = 10000;
-		mo->a = 1, mo->b = 4, mo->q = 6, mo->q2 = 26, mo->e = 2, mo->e2 = 1;
-		mo->occ_dist = 500;
 		mo->min_mid_occ = 50, mo->max_mid_occ = 500;
-		mo->min_dp_max = 200;
+		if (strcmp(preset, "map-hifi") == 0 || strcmp(preset, "map-ccs") == 0) {
+			mo->a = 1, mo->b = 4, mo->q = 6, mo->q2 = 26, mo->e = 2, mo->e2 = 1;
+			mo->min_dp_max = 200;
+		}
+	} else if (strcmp(preset, "lr:hqae") == 0) { // high-quality assembly evaluation
+		io->flag = 0, io->k = 25, io->w = 51;
+		mo->flag |= MM_F_RMQ;
+		mo->min_mid_occ = 50, mo->max_mid_occ = 500;
+		mo->rmq_inner_dist = 5000;
+		mo->occ_dist = 200;
+		mo->best_n = 100;
+		mo->chain_gap_scale = 5.0f;
+	} else if (strcmp(preset, "map-iclr-prerender") == 0) {
+		io->flag = 0, io->k = 15;
+		mo->b = 6, mo->transition = 1;
+		mo->q = 10, mo->q2 = 50;
+	} else if (strcmp(preset, "map-iclr") == 0) {
+		io->flag = 0, io->k = 19;
+		mo->b = 6, mo->transition = 4;
+		mo->q = 10, mo->q2 = 50;
 	} else if (strncmp(preset, "asm", 3) == 0) {
 		io->flag = 0, io->k = 19, io->w = 19;
-		mo->bw = mo->bw_long = 100000;
+		mo->bw = 1000, mo->bw_long = 100000;
 		mo->max_gap = 10000;
 		mo->flag |= MM_F_RMQ;
 		mo->min_mid_occ = 50, mo->max_mid_occ = 500;
@@ -145,7 +166,7 @@ int mm_set_opt(const char *preset, mm_idxopt_t *io, mm_mapopt_t *mo)
 		mo->mid_occ = 1000;
 		mo->max_occ = 5000;
 		mo->mini_batch_size = 50000000;
-	} else if (strncmp(preset, "splice", 6) == 0 || strcmp(preset, "cdna") == 0) {
+	} else if (strcmp(preset, "splice") == 0 || strcmp(preset, "splice:hq") == 0 || strcmp(preset, "splice:sr") == 0 || strcmp(preset, "cdna") == 0) {
 		io->flag = 0, io->k = 15, io->w = 5;
 		mo->flag |= MM_F_SPLICE | MM_F_SPLICE_FOR | MM_F_SPLICE_REV | MM_F_SPLICE_FLANK;
 		mo->max_sw_mat = 0;
@@ -153,13 +174,31 @@ int mm_set_opt(const char *preset, mm_idxopt_t *io, mm_mapopt_t *mo)
 		mo->a = 1, mo->b = 2, mo->q = 2, mo->e = 1, mo->q2 = 32, mo->e2 = 0;
 		mo->noncan = 9;
 		mo->junc_bonus = 9;
+		mo->junc_pen = 5;
 		mo->zdrop = 200, mo->zdrop_inv = 100; // because mo->a is halved
-		if (strcmp(preset, "splice:hq") == 0)
-			mo->junc_bonus = 5, mo->b = 4, mo->q = 6, mo->q2 = 24;
+		if (strcmp(preset, "splice:hq") == 0) {
+			mo->noncan = 5, mo->b = 4, mo->q = 6, mo->q2 = 24;
+		} else if (strcmp(preset, "splice:sr") == 0) {
+			mo->flag |= MM_F_NO_PRINT_2ND | MM_F_2_IO_THREADS | MM_F_HEAP_SORT | MM_F_FRAG_MODE | MM_F_WEAK_PAIRING | MM_F_SR_RNA;
+			mo->noncan = 5, mo->b = 4, mo->q = 6, mo->q2 = 24;
+			mo->min_chain_score = 25;
+			mo->min_dp_max = 40;
+			mo->min_ksw_len = 20;
+			mo->pe_ori = 0<<1|1; // FR
+			mo->best_n = 10;
+			mo->mini_batch_size = 100000000;
+		}
 	} else return -1;
 	return 0;
 }
 
+int mm_max_spsc_bonus(const mm_mapopt_t *mo)
+{
+	int max_sc = (mo->q2 + 1) / 2 - 1;
+	max_sc = max_sc > mo->q2 - mo->q? max_sc : mo->q2 - mo->q;
+	return max_sc;
+}
+
 int mm_check_opt(const mm_idxopt_t *io, const mm_mapopt_t *mo)
 {
 	if (mo->bw > mo->bw_long) {
@@ -214,6 +253,11 @@ int mm_check_opt(const mm_idxopt_t *io, const mm_mapopt_t *mo)
 			fprintf(stderr, "[ERROR]\033[1;31m scoring system violating ({-O}+{-E})+({-O2}+{-E2}) <= 127\033[0m\n");
 		return -1;
 	}
+	if (mo->sc_ambi < 0 || mo->sc_ambi >= mo->b) {
+		if (mm_verbose >= 1)
+			fprintf(stderr, "[ERROR]\033[1;31m --score-N should be within [0,{-B})\033[0m\n");
+		return -1;
+	}
 	if (mo->zdrop < mo->zdrop_inv) {
 		if (mm_verbose >= 1)
 			fprintf(stderr, "[ERROR]\033[1;31m Z-drop should not be less than inversion-Z-drop\033[0m\n");

pyproject.toml

@@ -0,0 +1,2 @@
+[build-system]
+requires = ["setuptools", "wheel", "Cython"]

python/README.rst

@@ -77,7 +77,9 @@ This constructor accepts the following arguments:
 
 * **min_chain_score**: minimum chaing score
 
-* **bw**: chaining and alignment band width
+* **bw**: chaining and alignment band width (initial chaining and extension)
+
+* **bw_long**: chaining and alignment band width (RMQ-based rechaining and closing gaps)
 
 * **best_n**: max number of alignments to return
 

python/cmappy.h

@@ -71,13 +71,13 @@ static inline void mm_reset_timer(void)
 }
 
 extern unsigned char seq_comp_table[256];
-static inline mm_reg1_t *mm_map_aux(const mm_idx_t *mi, const char *seq1, const char *seq2, int *n_regs, mm_tbuf_t *b, const mm_mapopt_t *opt)
+static inline mm_reg1_t *mm_map_aux(const mm_idx_t *mi, const char* seqname, const char *seq1, const char *seq2, int *n_regs, mm_tbuf_t *b, const mm_mapopt_t *opt)
 {
 	mm_reg1_t *r;
 
 	Py_BEGIN_ALLOW_THREADS
 	if (seq2 == 0) {
-		r = mm_map(mi, strlen(seq1), seq1, n_regs, b, opt, NULL);
+		r = mm_map(mi, strlen(seq1), seq1, n_regs, b, opt, seqname);
 	} else {
 		int _n_regs[2];
 		mm_reg1_t *regs[2];
@@ -94,7 +94,7 @@ static inline mm_reg1_t *mm_map_aux(const mm_idx_t *mi, const char *seq1, const
 			seq[1][i] = seq_comp_table[t];
 		}
 		if (len[1]&1) seq[1][len[1]>>1] = seq_comp_table[(uint8_t)seq[1][len[1]>>1]];
-		mm_map_frag(mi, 2, len, (const char**)seq, _n_regs, regs, b, opt, NULL);
+		mm_map_frag(mi, 2, len, (const char**)seq, _n_regs, regs, b, opt, seqname);
 		for (i = 0; i < _n_regs[1]; ++i)
 			regs[1][i].rev = !regs[1][i].rev;
 		*n_regs = _n_regs[0] + _n_regs[1];

python/cmappy.pxd

@@ -36,9 +36,10 @@ cdef extern from "minimap.h":
 		float alt_drop
 
 		int a, b, q, e, q2, e2
+		int transition
 		int sc_ambi
 		int noncan
-		int junc_bonus
+		int junc_bonus, junc_pen
 		int zdrop, zdrop_inv
 		int end_bonus
 		int min_dp_max
@@ -51,6 +52,8 @@ cdef extern from "minimap.h":
 
 		int pe_ori, pe_bonus
 
+		int jump_min_match;
+
 		float mid_occ_frac
 		float q_occ_frac
 		int32_t min_mid_occ
@@ -128,7 +131,7 @@ cdef extern from "cmappy.h":
 
 	void mm_reg2hitpy(const mm_idx_t *mi, mm_reg1_t *r, mm_hitpy_t *h)
 	void mm_free_reg1(mm_reg1_t *r)
-	mm_reg1_t *mm_map_aux(const mm_idx_t *mi, const char *seq1, const char *seq2, int *n_regs, mm_tbuf_t *b, const mm_mapopt_t *opt)
+	mm_reg1_t *mm_map_aux(const mm_idx_t *mi, const char* seqname, const char *seq1, const char *seq2, int *n_regs, mm_tbuf_t *b, const mm_mapopt_t *opt)
 	char *mappy_fetch_seq(const mm_idx_t *mi, const char *name, int st, int en, int *l)
 	mm_idx_t *mappy_idx_seq(int w, int k, int is_hpc, int bucket_bits, const char *seq, int l)
 

python/mappy.pyx

@@ -3,7 +3,7 @@ from libc.stdlib cimport free
 cimport cmappy
 import sys
 
-__version__ = '2.23'
+__version__ = '2.29'
 
 cmappy.mm_reset_timer()
 
@@ -96,6 +96,7 @@ cdef class Alignment:
 		a = [str(self._q_st), str(self._q_en), strand, self._ctg, str(self._ctg_len), str(self._r_st), str(self._r_en),
 			str(self._mlen), str(self._blen), str(self._mapq), tp, ts, "cg:Z:" + self.cigar_str]
 		if self._cs != "": a.append("cs:Z:" + self._cs)
+		if self._MD != "": a.append("MD:Z:" + self._MD)
 		return "\t".join(a)
 
 cdef class ThreadBuffer:
@@ -112,7 +113,7 @@ cdef class Aligner:
 	cdef cmappy.mm_idxopt_t idx_opt
 	cdef cmappy.mm_mapopt_t map_opt
 
-	def __cinit__(self, fn_idx_in=None, preset=None, k=None, w=None, min_cnt=None, min_chain_score=None, min_dp_score=None, bw=None, best_n=None, n_threads=3, fn_idx_out=None, max_frag_len=None, extra_flags=None, seq=None, scoring=None):
+	def __cinit__(self, fn_idx_in=None, preset=None, k=None, w=None, min_cnt=None, min_chain_score=None, min_dp_score=None, bw=None, bw_long=None, best_n=None, n_threads=3, fn_idx_out=None, max_frag_len=None, extra_flags=None, seq=None, scoring=None, sc_ambi=None, max_chain_skip=None):
 		self._idx = NULL
 		cmappy.mm_set_opt(NULL, &self.idx_opt, &self.map_opt) # set the default options
 		if preset is not None:
@@ -125,6 +126,7 @@ cdef class Aligner:
 		if min_chain_score is not None: self.map_opt.min_chain_score = min_chain_score
 		if min_dp_score is not None: self.map_opt.min_dp_max = min_dp_score
 		if bw is not None: self.map_opt.bw = bw
+		if bw_long is not None: self.map_opt.bw_long = bw_long
 		if best_n is not None: self.map_opt.best_n = best_n
 		if max_frag_len is not None: self.map_opt.max_frag_len = max_frag_len
 		if extra_flags is not None: self.map_opt.flag |= extra_flags
@@ -136,6 +138,8 @@ cdef class Aligner:
 				self.map_opt.q2, self.map_opt.e2 = scoring[4], scoring[5]
 				if len(scoring) >= 7:
 					self.map_opt.sc_ambi = scoring[6]
+		if sc_ambi is not None: self.map_opt.sc_ambi = sc_ambi
+		if max_chain_skip is not None: self.map_opt.max_chain_skip = max_chain_skip
 
 		cdef cmappy.mm_idx_reader_t *r;
 
@@ -161,7 +165,7 @@ cdef class Aligner:
 	def __bool__(self):
 		return (self._idx != NULL)
 
-	def map(self, seq, seq2=None, buf=None, cs=False, MD=False, max_frag_len=None, extra_flags=None):
+	def map(self, seq, seq2=None, name=None, buf=None, cs=False, MD=False, max_frag_len=None, extra_flags=None):
 		cdef cmappy.mm_reg1_t *regs
 		cdef cmappy.mm_hitpy_t h
 		cdef ThreadBuffer b
@@ -172,6 +176,7 @@ cdef class Aligner:
 		cdef cmappy.mm_mapopt_t map_opt
 
 		if self._idx == NULL: return
+		if ((self.map_opt.flag & 4) and (self._idx.flag & 2)): return
 		map_opt = self.map_opt
 		if max_frag_len is not None: map_opt.max_frag_len = max_frag_len
 		if extra_flags is not None: map_opt.flag |= extra_flags
@@ -182,11 +187,20 @@ cdef class Aligner:
 		km = cmappy.mm_tbuf_get_km(b._b)
 
 		_seq = seq if isinstance(seq, bytes) else seq.encode()
+		if name is not None:
+			_name = name if isinstance(name, bytes) else name.encode()
+
 		if seq2 is None:
-			regs = cmappy.mm_map_aux(self._idx, _seq, NULL,  &n_regs, b._b, &map_opt)
+			if name is None:
+				regs = cmappy.mm_map_aux(self._idx, NULL, _seq, NULL,  &n_regs, b._b, &map_opt)
+			else:
+				regs = cmappy.mm_map_aux(self._idx, _name, _seq, NULL,  &n_regs, b._b, &map_opt)
 		else:
 			_seq2 = seq2 if isinstance(seq2, bytes) else seq2.encode()
-			regs = cmappy.mm_map_aux(self._idx, _seq, _seq2, &n_regs, b._b, &map_opt)
+			if name is None:
+				regs = cmappy.mm_map_aux(self._idx, NULL, _seq, _seq2, &n_regs, b._b, &map_opt)
+			else:
+				regs = cmappy.mm_map_aux(self._idx, _name, _seq, _seq2, &n_regs, b._b, &map_opt)
 
 		try:
 			i = 0
@@ -197,11 +211,12 @@ cdef class Aligner:
 					c = h.cigar32[k]
 					cigar.append([c>>4, c&0xf])
 				if cs or MD: # generate the cs and/or the MD tag, if requested
+					_cur_seq = _seq2 if h.seg_id > 0 and seq2 is not None else _seq
 					if cs:
-						l_cs_str = cmappy.mm_gen_cs(km, &cs_str, &m_cs_str, self._idx, &regs[i], _seq, 1)
+						l_cs_str = cmappy.mm_gen_cs(km, &cs_str, &m_cs_str, self._idx, &regs[i], _cur_seq, 1)
 						_cs = cs_str[:l_cs_str] if isinstance(cs_str, str) else cs_str[:l_cs_str].decode()
 					if MD:
-						l_cs_str = cmappy.mm_gen_MD(km, &cs_str, &m_cs_str, self._idx, &regs[i], _seq)
+						l_cs_str = cmappy.mm_gen_MD(km, &cs_str, &m_cs_str, self._idx, &regs[i], _cur_seq)
 						_MD = cs_str[:l_cs_str] if isinstance(cs_str, str) else cs_str[:l_cs_str].decode()
 				yield Alignment(h.ctg, h.ctg_len, h.ctg_start, h.ctg_end, h.strand, h.qry_start, h.qry_end, h.mapq, cigar, h.is_primary, h.mlen, h.blen, h.NM, h.trans_strand, h.seg_id, _cs, _MD)
 				cmappy.mm_free_reg1(&regs[i])
@@ -217,6 +232,7 @@ cdef class Aligner:
 		cdef int l
 		cdef char *s
 		if self._idx == NULL: return
+		if ((self.map_opt.flag & 4) and (self._idx.flag & 2)): return
 		s = cmappy.mappy_fetch_seq(self._idx, name.encode(), start, end, &l)
 		if l == 0: return None
 		r = s[:l] if isinstance(s, str) else s[:l].decode()

python/minimap2.py

@@ -5,7 +5,7 @@ import getopt
 import mappy as mp
 
 def main(argv):
-	opts, args = getopt.getopt(argv[1:], "x:n:m:k:w:r:c")
+	opts, args = getopt.getopt(argv[1:], "x:n:m:k:w:r:cM")
 	if len(args) < 2:
 		print("Usage: minimap2.py [options] <ref.fa>|<ref.mmi> <query.fq>")
 		print("Options:")
@@ -16,10 +16,11 @@ def main(argv):
 		print("  -w INT      minimizer window length")
 		print("  -r INT      band width")
 		print("  -c          output the cs tag")
+		print("  -M          output the MD tag")
 		sys.exit(1)
 
 	preset = min_cnt = min_sc = k = w = bw = None
-	out_cs = False
+	out_cs = out_MD = False
 	for opt, arg in opts:
 		if opt == '-x': preset = arg
 		elif opt == '-n': min_cnt = int(arg)
@@ -28,11 +29,12 @@ def main(argv):
 		elif opt == '-k': k = int(arg)
 		elif opt == '-w': w = int(arg)
 		elif opt == '-c': out_cs = True
+		elif opt == '-M': out_MD = True
 
 	a = mp.Aligner(args[0], preset=preset, min_cnt=min_cnt, min_chain_score=min_sc, k=k, w=w, bw=bw)
 	if not a: raise Exception("ERROR: failed to load/build index file '{}'".format(args[0]))
 	for name, seq, qual in mp.fastx_read(args[1]): # read one sequence
-		for h in a.map(seq, cs=out_cs): # traverse hits
+		for h in a.map(seq, cs=out_cs, MD=out_MD): # traverse hits
 			print('{}\t{}\t{}'.format(name, len(seq), h))
 
 if __name__ == "__main__":

seed.c

@@ -7,7 +7,7 @@ void mm_seed_mz_flt(void *km, mm128_v *mv, int32_t q_occ_max, float q_occ_frac)
 	mm128_t *a;
 	size_t i, j, st;
 	if (mv->n <= q_occ_max || q_occ_frac <= 0.0f || q_occ_max <= 0) return;
-	KMALLOC(km, a, mv->n);
+	a = Kmalloc(km, mm128_t, mv->n);
 	for (i = 0; i < mv->n; ++i)
 		a[i].x = mv->a[i].x, a[i].y = i;
 	radix_sort_128x(a, a + mv->n);
@@ -112,7 +112,8 @@ mm_seed_t *mm_collect_matches(void *km, int *_n_m, int qlen, int max_occ, int ma
 	}
 	for (i = 0, n_m = 0, *rep_len = 0, *n_a = 0; i < n_m0; ++i) {
 		mm_seed_t *q = &m[i];
-		//fprintf(stderr, "X\t%d\t%d\t%d\n", q->q_pos>>1, q->n, q->flt);
+		if (mm_dbg_flag & MM_DBG_SEED_FREQ)
+			fprintf(stderr, "SF\t%d\t%d\t%d\n", q->q_pos>>1, q->n, q->flt);
 		if (q->flt) {
 			int en = (q->q_pos >> 1) + 1, st = en - q->q_span;
 			if (st > rep_en) {

setup.py

@@ -23,7 +23,7 @@ def readme():
 
 setup(
 	name = 'mappy',
-	version = '2.23',
+	version = '2.29',
 	url = 'https://github.com/lh3/minimap2',
 	description = 'Minimap2 python binding',
 	long_description = readme(),
@@ -33,7 +33,7 @@ setup(
 	keywords = 'sequence-alignment',
 	scripts = ['python/minimap2.py'],
 	ext_modules = [Extension('mappy',
-		sources = ['python/mappy.pyx', 'align.c', 'bseq.c', 'lchain.c', 'seed.c', 'format.c', 'hit.c', 'index.c', 'pe.c', 'options.c',
+		sources = ['python/mappy.pyx', 'align.c', 'bseq.c', 'lchain.c', 'seed.c', 'format.c', 'hit.c', 'index.c', 'pe.c', 'jump.c', 'options.c',
 				   'ksw2_extd2_sse.c', 'ksw2_exts2_sse.c', 'ksw2_extz2_sse.c', 'ksw2_ll_sse.c',
 				   'kalloc.c', 'kthread.c', 'map.c', 'misc.c', 'sdust.c', 'sketch.c', 'esterr.c', 'splitidx.c'],
 		depends = ['minimap.h', 'bseq.h', 'kalloc.h', 'kdq.h', 'khash.h', 'kseq.h', 'ksort.h',