xxay/hifiasm
1
0
Fork 0
mirror of https://github.com/chhylp123/hifiasm.git synced 2026-10-08 02:08:12 +08:00
Code Issues Packages Projects Releases Wiki Activity

more cutting methods

Browse Source
This commit is contained in:
Haoyu Cheng
2019-11-20 15:43:44 -05:00
parent e8ff0c1dbc
commit 3425bd020b
6 changed files with 660 additions and 118 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Assembly.cpp
+25 -16
View File
@@ -1180,6 +1180,7 @@ void push_overlaps(ma_hit_t_alloc* paf, overlap_region_alloc* overlap_list, int
tmp.bl = R_INF.read_length[overlap_list->list[i].y_id]; tmp.bl = R_INF.read_length[overlap_list->list[i].y_id];
tmp.ml = overlap_list->list[i].strong; tmp.ml = overlap_list->list[i].strong;
tmp.no_l_indel = overlap_list->list[i].without_large_indel;
add_ma_hit_t_alloc(paf, &tmp); add_ma_hit_t_alloc(paf, &tmp);
} }
@@ -1281,7 +1282,7 @@ overlap_region_alloc* overlap_list, UC_Read* x_read, UC_Read* y_read)
tmp.bl = R_INF.read_length[overlap_list->list[i].y_id]; tmp.bl = R_INF.read_length[overlap_list->list[i].y_id];
tmp.ml = overlap_list->list[i].strong; tmp.ml = overlap_list->list[i].strong;
tmp.no_l_indel = overlap_list->list[i].without_large_indel;
@@ -1842,7 +1843,7 @@ void* Overlap_calculate_heap_merge(void* arg)
long long num_correct_base = 0; long long num_correct_base = 0;
long long num_second_correct_base = 0; long long num_second_correct_base = 0;
long long j; long long j;
int fully_cov; int fully_cov, abnormal;
int thr_ID = *((int*)arg); int thr_ID = *((int*)arg);
uint64_t POA_i; uint64_t POA_i;
@@ -1919,7 +1920,7 @@ void* Overlap_calculate_heap_merge(void* arg)
correct_overlap(&overlap_list, &R_INF, &g_read, &correct, &overlap_read, &POA_Graph, &DAGCon, correct_overlap(&overlap_list, &R_INF, &g_read, &correct, &overlap_read, &POA_Graph, &DAGCon,
&matched_overlap_0, &matched_overlap_1, &potiental_matched_overlap_0, &potiental_matched_overlap_1, &matched_overlap_0, &matched_overlap_1, &potiental_matched_overlap_0, &potiental_matched_overlap_1,
&current_cigar, &hap, &second_round, 0, 1, &fully_cov); &current_cigar, &hap, &second_round, 0, 1, &fully_cov, &abnormal);
num_read_base = num_read_base + g_read.length; num_read_base = num_read_base + g_read.length;
num_correct_base = num_correct_base + correct.corrected_base; num_correct_base = num_correct_base + correct.corrected_base;
@@ -1933,18 +1934,6 @@ void* Overlap_calculate_heap_merge(void* arg)
R_INF.paf[i].is_fully_corrected = 0; R_INF.paf[i].is_fully_corrected = 0;
if(fully_cov) if(fully_cov)
{ {
/**
if(current_cigar.length == 1 && second_round.cigar.length == 1)
{
if(Get_Cigar_Type(current_cigar.record[0]) == 0 &&
Get_Cigar_Length(current_cigar.record[0]) == g_read.length &&
Get_Cigar_Type(second_round.cigar.record[0]) == 0 &&
Get_Cigar_Length(second_round.cigar.record[0]) == g_read.length)
{
R_INF.paf[i].is_fully_corrected = 1;
}
}
**/
if( if(
get_cigar_errors(&current_cigar) == 0 get_cigar_errors(&current_cigar) == 0
&& &&
@@ -1953,6 +1942,10 @@ void* Overlap_calculate_heap_merge(void* arg)
R_INF.paf[i].is_fully_corrected = 1; R_INF.paf[i].is_fully_corrected = 1;
} }
} }
R_INF.paf[i].is_abnormal = abnormal;
push_overlaps(&(R_INF.paf[i]), &overlap_list, 1); push_overlaps(&(R_INF.paf[i]), &overlap_list, 1);
@@ -2828,8 +2821,8 @@ void* Final_overlap_calculate_heap_merge(void* arg)
long long matched_overlap_1 = 0; long long matched_overlap_1 = 0;
long long potiental_matched_overlap_0 = 0; long long potiental_matched_overlap_0 = 0;
long long potiental_matched_overlap_1 = 0; long long potiental_matched_overlap_1 = 0;
long long num_read_base = 0;
long long num_correct_base = 0; long long num_correct_base = 0;
long long num_read_base = 0;
long long num_second_correct_base = 0; long long num_second_correct_base = 0;
long long j, inner_j; long long j, inner_j;
@@ -2937,6 +2930,7 @@ void* Final_overlap_calculate_heap_merge(void* arg)
overlap_list.list[j].is_match = 1; overlap_list.list[j].is_match = 1;
overlap_list.list[j].strong = R_INF.paf[i].buffer[inner_j].ml; overlap_list.list[j].strong = R_INF.paf[i].buffer[inner_j].ml;
overlap_list.list[j].without_large_indel = R_INF.paf[i].buffer[inner_j].no_l_indel;
overlap_list.mapped_overlaps_length++; overlap_list.mapped_overlaps_length++;
if(overlap_list.list[j].strong == 1) if(overlap_list.list[j].strong == 1)
@@ -2951,6 +2945,15 @@ void* Final_overlap_calculate_heap_merge(void* arg)
{ {
fprintf(stderr, "error\n"); fprintf(stderr, "error\n");
} }
if(overlap_list.list[j].without_large_indel == 0)
{
num_second_correct_base++;
}
} }
j++; j++;
inner_j++; inner_j++;
@@ -2958,6 +2961,7 @@ void* Final_overlap_calculate_heap_merge(void* arg)
} }
///recover missing exact overlaps
reverse_complement(g_read.seq, g_read.length); reverse_complement(g_read.seq, g_read.length);
for (j = 0; j < overlap_list.length; j++) for (j = 0; j < overlap_list.length; j++)
{ {
@@ -2978,6 +2982,7 @@ void* Final_overlap_calculate_heap_merge(void* arg)
{ {
overlap_list.list[j].is_match = 1; overlap_list.list[j].is_match = 1;
overlap_list.list[j].strong = 0; overlap_list.list[j].strong = 0;
overlap_list.list[j].without_large_indel = 1;
overlap_list.mapped_overlaps_length++; overlap_list.mapped_overlaps_length++;
potiental_matched_overlap_0++; potiental_matched_overlap_0++;
} }
@@ -3024,11 +3029,15 @@ void* Final_overlap_calculate_heap_merge(void* arg)
total_potiental_matched_overlap_0 += potiental_matched_overlap_0; total_potiental_matched_overlap_0 += potiental_matched_overlap_0;
total_potiental_matched_overlap_1 += potiental_matched_overlap_1; total_potiental_matched_overlap_1 += potiental_matched_overlap_1;
total_num_correct_base += num_correct_base; total_num_correct_base += num_correct_base;
total_second_num_correct_base += num_second_correct_base;
complete_threads++; complete_threads++;
if(complete_threads == thread_num) if(complete_threads == thread_num)
{ {
fprintf(stderr, "overlaps with large indels: %llu\n", total_second_num_correct_base);
fprintf(stderr, "weak overlaps: %llu\n", total_matched_overlap_0); fprintf(stderr, "weak overlaps: %llu\n", total_matched_overlap_0);
fprintf(stderr, "strong overlaps: %llu\n", total_matched_overlap_1); fprintf(stderr, "strong overlaps: %llu\n", total_matched_overlap_1);
fprintf(stderr, "recover weak overlaps: %llu\n", total_potiental_matched_overlap_0); fprintf(stderr, "recover weak overlaps: %llu\n", total_potiental_matched_overlap_0);
Correct.cpp
+160 -12
View File
@@ -6695,7 +6695,7 @@ void generate_consensus(overlap_region_alloc* overlap_list, All_reads* R_INF,
inline int get_available_fully_covered_interval(long long window_start, long long window_end, inline int get_available_fully_covered_interval(long long window_start, long long window_end,
overlap_region_alloc* overlap_list, Correct_dumy* dumy, long long* real_length) overlap_region_alloc* overlap_list, Correct_dumy* dumy, long long* real_length, long long* real_length_100)
{ {
long long i; long long i;
int flag = 0; int flag = 0;
@@ -6745,9 +6745,13 @@ overlap_region_alloc* overlap_list, Correct_dumy* dumy, long long* real_length)
if (overlap_length == Len && overlap_list->list[i].is_match == 1) if (overlap_length == Len && overlap_list->list[i].is_match == 1)
{ {
//dumy->length++;
(*real_length)++; (*real_length)++;
} }
if (overlap_length == Len && overlap_list->list[i].is_match == 100)
{
(*real_length_100)++;
}
} }
if(overlap_list->list[i].x_pos_s > window_end) if(overlap_list->list[i].x_pos_s > window_end)
@@ -6768,17 +6772,19 @@ overlap_region_alloc* overlap_list, Correct_dumy* dumy, long long* real_length)
} }
int check_if_fully_covered(overlap_region_alloc* overlap_list, int check_if_fully_covered(overlap_region_alloc* overlap_list,
All_reads* R_INF, UC_Read* g_read, Correct_dumy* dumy, Graph* g) All_reads* R_INF, UC_Read* g_read, Correct_dumy* dumy, Graph* g, int* abnormal)
{ {
long long overlap_length; long long overlap_length;
long long window_start, window_end; long long window_start, window_end;
int return_flag = 1;
(*abnormal) = 0;
Window_Pool w_inf; Window_Pool w_inf;
init_Window_Pool(&w_inf, g_read->length, WINDOW, TAIL_LENGTH); init_Window_Pool(&w_inf, g_read->length, WINDOW, TAIL_LENGTH);
int flag = 0; int flag = 0;
long long realLen; long long realLen, tmpLen;
while(get_Window(&w_inf, &window_start, &window_end) && flag != -2) while(get_Window(&w_inf, &window_start, &window_end) && flag != -2)
{ {
dumy->length = 0; dumy->length = 0;
@@ -6788,7 +6794,7 @@ All_reads* R_INF, UC_Read* g_read, Correct_dumy* dumy, Graph* g)
///dumy->lengthNT返回的是有效不完全重叠的数量 ///dumy->lengthNT返回的是有效不完全重叠的数量
///return overlaps that is overlaped with [window_start, window_end] ///return overlaps that is overlaped with [window_start, window_end]
flag = get_available_fully_covered_interval(window_start, window_end, flag = get_available_fully_covered_interval(window_start, window_end,
overlap_list, dumy, &realLen); overlap_list, dumy, &realLen, &tmpLen);
switch (flag) switch (flag)
@@ -6803,12 +6809,26 @@ All_reads* R_INF, UC_Read* g_read, Correct_dumy* dumy, Graph* g)
if(realLen < MIN_COVERAGE_THRESHOLD * 2) if(realLen < MIN_COVERAGE_THRESHOLD * 2)
{ {
return 0; return_flag = 0;
//return 0;
}
if(realLen == 0)
{
///that means this window is a middle window
if(window_start != 0 && window_end != g_read->length - 1)
{
(*abnormal) = 1;
}
else if((*abnormal)==0)
{
(*abnormal) = 2;
}
} }
} }
return return_flag;
return 1; //return 1;
} }
@@ -9430,6 +9450,132 @@ int force_repeat)
}
void lable_large_indels(overlap_region_alloc* overlap_list, All_reads* R_INF, long long read_length,
Correct_dumy* dumy)
{
long long i, j;
long long cigar_i, operation, operationLen;
int is_delete = 0;
CIGAR* cigar;
for (i = 0; i < overlap_list->length; i++)
{
///should has at least 3 windows for this overlap
if (overlap_list->list[i].is_match == 1 && overlap_list->list[i].w_list_length >= 3)
{
///here w_list_length >= 3
///skip the first and last window
for (j = 1; j < overlap_list->list[i].w_list_length - 1; j++)
{
///this window is not matched, it seems to have large difference
if(overlap_list->list[i].w_list[j].y_end == -1)
{
overlap_list->list[i].is_match = 100;
is_delete = 1;
goto end_rem;
}
cigar = &(overlap_list->list[i].w_list[j].cigar);
///if there are <=2 cigar elements, skip it
if(cigar->length < 3)
{
continue;
}
///skip the first and last cigar elements
for (cigar_i = 1; cigar_i < cigar->length - 1; cigar_i++)
{
operation = cigar->C_C[cigar_i];
operationLen = cigar->C_L[cigar_i];
if(operationLen <= 5)
{
continue;
}
///>=6 bp deletion or insertion
if(operation == 2 || operation == 3)
{
overlap_list->list[i].is_match = 100;
is_delete = 1;
goto end_rem;
}
}
}
}
end_rem:
overlap_list->list[i].w_list_length >= 3;
}
if(is_delete == 1)
{
long long overlap_length;
long long window_start, window_end;
Window_Pool w_inf;
init_Window_Pool(&w_inf, read_length, WINDOW, TAIL_LENGTH);
int flag = 0;
long long realLen, realLen_100;
int to_recover = 0;
while(get_Window(&w_inf, &window_start, &window_end) && flag != -2)
{
dumy->length = 0;
dumy->lengthNT = 0;
///flag返回的是重叠数量
///dumy->length返回的是有效完全重叠的数量
///dumy->lengthNT返回的是有效不完全重叠的数量
///return overlaps that is overlaped with [window_start, window_end]
flag = get_available_fully_covered_interval(window_start, window_end,
overlap_list, dumy, &realLen, &realLen_100);
switch (flag)
{
case 1: ///找到匹配
break;
case 0: ///没找到匹配
break;
case -2: ///下一个window也不会存在匹配, 直接跳出
break;
}
///it seems there is a long indel at the reference read itself
if(realLen == 0 && realLen_100 > 0)
{
to_recover = 1;
break;
}
}
if(to_recover == 1)
{
for (i = 0; i < overlap_list->length; i++)
{
if (overlap_list->list[i].is_match == 100)
{
overlap_list->list[i].is_match = 1;
}
}
}
}
for (i = 0; i < overlap_list->length; i++)
{
if (overlap_list->list[i].is_match == 1)
{
overlap_list->list[i].without_large_indel = 1;
}
if (overlap_list->list[i].is_match == 100)
{
overlap_list->list[i].is_match = 1;
overlap_list->list[i].without_large_indel = 0;
}
}
} }
@@ -10285,6 +10431,8 @@ void partition_overlaps(overlap_region_alloc* overlap_list, All_reads* R_INF,
generate_haplotypes_DP(hap, overlap_list, R_INF, g_read->length, force_repeat); generate_haplotypes_DP(hap, overlap_list, R_INF, g_read->length, force_repeat);
///generate_haplotypes_naive(hap, overlap_list, R_INF, g_read->length, force_repeat); ///generate_haplotypes_naive(hap, overlap_list, R_INF, g_read->length, force_repeat);
lable_large_indels(overlap_list, R_INF, g_read->length, dumy);
///debug_snp_matrix(hap); ///debug_snp_matrix(hap);
} }
@@ -10542,7 +10690,7 @@ void correct_overlap(overlap_region_alloc* overlap_list, All_reads* R_INF,
long long* potiental_matched_overlap_0, long long* potiental_matched_overlap_1, long long* potiental_matched_overlap_0, long long* potiental_matched_overlap_1,
Cigar_record* current_cigar, haplotype_evdience_alloc* hap, Cigar_record* current_cigar, haplotype_evdience_alloc* hap,
Round2_alignment* second_round, int force_repeat, int is_consensus, Round2_alignment* second_round, int force_repeat, int is_consensus,
int* fully_cov) int* fully_cov, int* abnormal)
{ {
reverse_complement(g_read->seq, g_read->length); reverse_complement(g_read->seq, g_read->length);
@@ -10594,8 +10742,8 @@ void correct_overlap(overlap_region_alloc* overlap_list, All_reads* R_INF,
partition_overlaps(overlap_list, R_INF, g_read, dumy, hap, force_repeat); partition_overlaps(overlap_list, R_INF, g_read, dumy, hap, force_repeat);
print_overlap("m64016_190918_162737/53545052/ccs", // print_overlap("m64016_190918_162737/53545052/ccs",
overlap_list->list[0].x_id, overlap_list, R_INF, 1); // overlap_list->list[0].x_id, overlap_list, R_INF, 1);
@@ -10606,7 +10754,7 @@ void correct_overlap(overlap_region_alloc* overlap_list, All_reads* R_INF,
} }
(*fully_cov) = check_if_fully_covered(overlap_list, R_INF, g_read, dumy, g); (*fully_cov) = check_if_fully_covered(overlap_list, R_INF, g_read, dumy, g, abnormal);
} }
Correct.h
+1 -1
View File
@@ -1230,7 +1230,7 @@ void correct_overlap(overlap_region_alloc* overlap_list, All_reads* R_INF,
long long* potiental_matched_overlap_0, long long* potiental_matched_overlap_1, long long* potiental_matched_overlap_0, long long* potiental_matched_overlap_1,
Cigar_record* current_cigar, haplotype_evdience_alloc* hap, Cigar_record* current_cigar, haplotype_evdience_alloc* hap,
Round2_alignment* second_round, int force_repeat, int is_consensus, Round2_alignment* second_round, int force_repeat, int is_consensus,
int* fully_cov); int* fully_cov, int* abnormal);
void init_Correct_dumy(Correct_dumy* list); void init_Correct_dumy(Correct_dumy* list);
void destory_Correct_dumy(Correct_dumy* list); void destory_Correct_dumy(Correct_dumy* list);
void clear_Correct_dumy(Correct_dumy* list, overlap_region_alloc* overlap_list); void clear_Correct_dumy(Correct_dumy* list, overlap_region_alloc* overlap_list);
Hash_Table.h
+1
View File
@@ -124,6 +124,7 @@ typedef struct
uint64_t align_length; uint64_t align_length;
///uint64_t total_errors; ///uint64_t total_errors;
uint8_t is_match; uint8_t is_match;
uint8_t without_large_indel;
uint64_t non_homopolymer_errors; uint64_t non_homopolymer_errors;
window_list* w_list; window_list* w_list;
Overlaps.cpp
+468 -89
View File
@@ -354,6 +354,17 @@ inline void set_reverse_overlap(ma_hit_t* dest, ma_hit_t* source)
{ {
dest->ml = source->ml = 1; dest->ml = source->ml = 1;
} }
if(dest->no_l_indel == 0 || source->no_l_indel == 0)
{
dest->no_l_indel = source->no_l_indel = 0;
}
else
{
dest->no_l_indel = source->no_l_indel = 1;
}
/****************************may have bugs********************************/ /****************************may have bugs********************************/
dest->bl = Get_qe(*dest) - Get_qs(*dest); dest->bl = Get_qe(*dest) - Get_qs(*dest);
} }
@@ -405,6 +416,7 @@ void normalize_ma_hit_t(ma_hit_t_alloc* sources, long long num_sources)
{ {
///must have this line ///must have this line
new_element.ml = 1; new_element.ml = 1;
new_element.no_l_indel = 1;
set_reverse_overlap(&new_element, &(sources[i].buffer[j])); set_reverse_overlap(&new_element, &(sources[i].buffer[j]));
add_ma_hit_t_alloc(&(sources[tn]), &new_element); add_ma_hit_t_alloc(&(sources[tn]), &new_element);
si_overlaps++; si_overlaps++;
@@ -2966,6 +2978,8 @@ uint32_t startNode, uint32_t endNode, int max_dist, buf_t* bub)
} }
int find_single_link(asg_t *g, uint32_t link_beg, int linkLen, uint32_t* link_end) int find_single_link(asg_t *g, uint32_t link_beg, int linkLen, uint32_t* link_end)
{ {
uint32_t v, w; uint32_t v, w;
@@ -3595,6 +3609,47 @@ int asg_arc_del_triangular_advance(asg_t *g, long long max_dist)
} }
int asg_arc_del_triangular_directly(asg_t *g, long long max_dist)
{
double startTime = Get_T();
///the reason is that each read has two direction (query->target, target->query)
uint32_t v, w, n_vtx = g->n_seq * 2, n_reduced = 0;
int flag0, flag1, node;
for (v = 0; v < n_vtx; ++v)
{
uint32_t nv = asg_arc_n(g, v);
asg_arc_t *av = asg_arc_a(g, v);
if (g->seq[v>>1].del)
{
continue;
}
if(nv < 2)
{
continue;
}
//test_triangular_directly();
}
if (n_reduced) {
asg_cleanup(g);
asg_symm(g);
}
fprintf(stderr, "[M::%s] removed %d triangular overlaps\n",
__func__, n_reduced);
fprintf(stderr, "[M::%s] takes %0.2f s\n\n", __func__, Get_T()-startTime);
return n_reduced;
}
int asg_arc_del_triangular_advance_debug(asg_t *g, long long max_dist) int asg_arc_del_triangular_advance_debug(asg_t *g, long long max_dist)
{ {
double startTime = Get_T(); double startTime = Get_T();
@@ -4288,7 +4343,7 @@ uint32_t startNode, uint32_t endNode)
} }
int test_single_node_bubble_directly(asg_t *g, uint32_t v, long long longLen_thres) int test_single_node_bubble_directly(asg_t *g, uint32_t v, long long longLen_thres, ma_hit_t_alloc* sources)
{ {
uint32_t w, vEnd; uint32_t w, vEnd;
@@ -4323,7 +4378,8 @@ int test_single_node_bubble_directly(asg_t *g, uint32_t v, long long longLen_thr
longLen = Len[1]; longLen = Len[1];
/****************************may have bugs********************************/ /****************************may have bugs********************************/
if(asg_arc_a(g, w)[0].el == 0 || asg_arc_a(g, w^1)[0].el == 0) ///if(asg_arc_a(g, w)[0].el == 0 || asg_arc_a(g, w^1)[0].el == 0)
if(asg_arc_a(g, w)[0].el == 0 || asg_arc_a(g, w^1)[0].el == 0 || sources[w>>1].is_abnormal == 1)
{ {
asg_seq_del(g, w>>1); asg_seq_del(g, w>>1);
n_reduced++; n_reduced++;
@@ -4348,7 +4404,8 @@ int test_single_node_bubble_directly(asg_t *g, uint32_t v, long long longLen_thr
longLen = Len[0]; longLen = Len[0];
/****************************may have bugs********************************/ /****************************may have bugs********************************/
if(asg_arc_a(g, w)[0].el == 0 || asg_arc_a(g, w^1)[0].el == 0) ///if(asg_arc_a(g, w)[0].el == 0 || asg_arc_a(g, w^1)[0].el == 0)
if(asg_arc_a(g, w)[0].el == 0 || asg_arc_a(g, w^1)[0].el == 0 || sources[w>>1].is_abnormal == 1)
{ {
asg_seq_del(g, w>>1); asg_seq_del(g, w>>1);
n_reduced++; n_reduced++;
@@ -4367,41 +4424,54 @@ int test_single_node_bubble_directly(asg_t *g, uint32_t v, long long longLen_thr
} }
else if(Len[0] == 1 && Len[1] == 1) else if(Len[0] == 1 && Len[1] == 1)
{ {
w = av[0].v; flag0 = sources[av[0].v>>1].is_abnormal;
flag0 = asg_arc_a(g, w)[0].el + asg_arc_a(g, w^1)[0].el; flag1 = sources[av[1].v>>1].is_abnormal;
w = av[1].v;
flag1 = asg_arc_a(g, w)[0].el + asg_arc_a(g, w^1)[0].el; if(flag0 == 1 && flag1 == 0)
///>=2 means this is an exact overlap {
if(flag0 < 2 && flag1 >= 2) asg_seq_del(g, av[0].v>>1);
n_reduced++;
}
else if(flag0 == 0 && flag1 == 1)
{
asg_seq_del(g, av[1].v>>1);
n_reduced++;
}
else
{ {
w = av[0].v; w = av[0].v;
flag0 = asg_arc_a(g, w)[0].el + asg_arc_a(g, w^1)[0].el;
/****************************may have bugs********************************/
if(asg_arc_a(g, w)[0].el == 0 || asg_arc_a(g, w^1)[0].el == 0)
{
asg_seq_del(g, w>>1);
n_reduced++;
}
/****************************may have bugs********************************/
}
if(flag0 >= 2 && flag1 < 2)
{
w = av[1].v; w = av[1].v;
flag1 = asg_arc_a(g, w)[0].el + asg_arc_a(g, w^1)[0].el;
/****************************may have bugs********************************/ ///>=2 means this is an exact overlap
if(asg_arc_a(g, w)[0].el == 0 || asg_arc_a(g, w^1)[0].el == 0) if(flag0 < 2 && flag1 >= 2)
{ {
asg_seq_del(g, w>>1); w = av[0].v;
n_reduced++;
/****************************may have bugs********************************/
if(asg_arc_a(g, w)[0].el == 0 || asg_arc_a(g, w^1)[0].el == 0)
{
asg_seq_del(g, w>>1);
n_reduced++;
}
/****************************may have bugs********************************/
}
if(flag0 >= 2 && flag1 < 2)
{
w = av[1].v;
/****************************may have bugs********************************/
if(asg_arc_a(g, w)[0].el == 0 || asg_arc_a(g, w^1)[0].el == 0)
{
asg_seq_del(g, w>>1);
n_reduced++;
}
/****************************may have bugs********************************/
} }
/****************************may have bugs********************************/
} }
} }
else
{
fprintf(stderr, "error\n");
}
} }
return n_reduced; return n_reduced;
} }
@@ -4451,7 +4521,7 @@ int asg_arc_del_single_node_bubble(asg_t *g, long long max_dist)
return n_reduced; return n_reduced;
} }
int asg_arc_del_single_node_directly(asg_t *g, long long longLen_thres) int asg_arc_del_single_node_directly(asg_t *g, long long longLen_thres, ma_hit_t_alloc* sources)
{ {
double startTime = Get_T(); double startTime = Get_T();
///the reason is that each read has two direction (query->target, target->query) ///the reason is that each read has two direction (query->target, target->query)
@@ -4470,7 +4540,7 @@ int asg_arc_del_single_node_directly(asg_t *g, long long longLen_thres)
continue; continue;
} }
n_reduced += test_single_node_bubble_directly(g, v, longLen_thres); n_reduced += test_single_node_bubble_directly(g, v, longLen_thres, sources);
} }
@@ -5056,18 +5126,19 @@ long long weakID, uint32_t w_qs, uint32_t w_qe)
} }
inline int check_weak_ma_hit_reverse(ma_hit_t_alloc* aim_paf, ma_hit_t_alloc* reverse_paf_list, inline int check_weak_ma_hit_reverse(ma_hit_t_alloc* r_paf, ma_hit_t_alloc* r_paf_source,
long long weakID) long long weakID)
{ {
long long i = 0; long long i = 0;
long long strongID, index; long long strongID, index;
///all overlaps coming from another haplotye are strong ///all overlaps coming from another haplotye are strong
for (i = 0; i < aim_paf->length; i++) for (i = 0; i < r_paf->length; i++)
{ {
strongID = Get_tn(aim_paf->buffer[i]); strongID = Get_tn(r_paf->buffer[i]);
index = get_specific_overlap index = get_specific_overlap
(&(reverse_paf_list[strongID]), strongID, weakID); (&(r_paf_source[strongID]), strongID, weakID);
if(index != -1) ///must be a strong overlap
if(index != -1 && r_paf_source[strongID].buffer[index].ml == 1)
{ {
return 0; return 0;
} }
@@ -5253,6 +5324,41 @@ int asg_arc_del_short_diploid_unclean(asg_t *g, float drop_ratio, ma_hit_t_alloc
return n_short; return n_short;
} }
int asg_arc_del_too_short_overlaps(asg_t *g, long long dropLen)
{
double startTime = Get_T();
uint32_t v, n_vtx = g->n_seq * 2, n_short = 0;
for (v = 0; v < n_vtx; ++v)
{
if (g->seq[v>>1].del) continue;
if (g->seq_vis[v] != 0) continue;
asg_arc_t *av = asg_arc_a(g, v);
uint32_t i, thres, nv = asg_arc_n(g, v);
///if there is just one overlap, do nothing
if (nv < 2) continue;
//av[0] has the most overlap length
///remove short overlaps
if(av[0].ol < dropLen) continue;
for (i = nv - 1; i >= 1 && av[i].ol < dropLen; --i);
for (i = i + 1; i < nv; ++i)
av[i].del = 1, ++n_short;
}
asg_cleanup(g);
asg_symm(g);
fprintf(stderr, "[M::%s] removed %d short overlaps\n", __func__, n_short);
fprintf(stderr, "[M::%s] takes %0.2f s\n\n", __func__, Get_T()-startTime);
return n_short;
}
int asg_arc_del_short_diploid_unclean_exact(asg_t *g, float drop_ratio, ma_hit_t_alloc* sources) int asg_arc_del_short_diploid_unclean_exact(asg_t *g, float drop_ratio, ma_hit_t_alloc* sources)
{ {
@@ -6460,7 +6566,10 @@ int asg_arc_del_short_diploid_by_exact(asg_t *g, int max_ext, ma_hit_t_alloc* so
sources[v>>1].is_fully_corrected == 1 && sources[v>>1].is_fully_corrected == 1 &&
sources[w>>1].is_fully_corrected == 0) sources[w>>1].is_fully_corrected == 0)
{ {
if(av[ov_max_i].el == 1 && sources[av[ov_max_i].v>>1].is_fully_corrected) /****************************may have bugs********************************/
///if(av[ov_max_i].el == 1 && sources[av[ov_max_i].v>>1].is_fully_corrected)
/****************************may have bugs********************************/
if(av[ov_max_i].el == 1 && sources[av[ov_max_i].v>>1].is_fully_corrected == 1)
{ {
if (kv > 1 && kw > 1) { if (kv > 1 && kw > 1) {
to_del = 1; to_del = 1;
@@ -6491,6 +6600,104 @@ int asg_arc_del_short_diploid_by_exact(asg_t *g, int max_ext, ma_hit_t_alloc* so
} }
int asg_arc_del_short_diploi_by_suspect_edge(asg_t *g, int max_ext, ma_hit_t_alloc* sources)
{
double startTime = Get_T();
kvec_t(uint64_t) b;
memset(&b, 0, sizeof(b));
uint32_t v, n_vtx = g->n_seq * 2;
long long n_cut = 0;
for (v = 0; v < n_vtx; ++v)
{
///if(g->seq_vis[v] == 0)
{
asg_arc_t *av = asg_arc_a(g, v);
uint32_t nv = asg_arc_n(g, v);
if (nv < 2) continue;
long long i;
for (i = 0; i < nv; ++i)
{
///means there is a large indel at this edge
if(av[i].no_l_indel == 0)
{
kv_push(uint64_t, b, (uint64_t)(av[i].ol << 32 | (av - g->arc + i)));
}
}
}
}
fprintf(stderr, "[M::%s] %lld unsorted pending overlaps\n", __func__, b.n);
radix_sort_arch64(b.a, b.a + b.n);
fprintf(stderr, "[M::%s] %lld sorted pending overlaps\n", __func__, b.n);
long long k;
for (k = 0; k < b.n; k++)
{
asg_arc_t *a = &g->arc[(uint32_t)b.a[k]];
///v is self id, w is the id of another end
uint32_t i, iv, iw, v = (a->ul)>>32, w = a->v^1, to_del = 0;
uint32_t nv = asg_arc_n(g, v), nw = asg_arc_n(g, w), kv, kw;
asg_arc_t *av, *aw;
///nv must be >= 2
if (nv == 1 && nw == 1) continue;
av = asg_arc_a(g, v);
aw = asg_arc_a(g, w);
///calculate the longest edge for v and w
for (i = 0, kv = 0; i < nv; ++i)
{
if (av[i].del) continue;
++kv;
}
for (i = 0, kw = 0; i < nw; ++i)
{
if (aw[i].del) continue;
++kw;
}
if (kv == 1 && kw == 1) continue;
///to see which one is the current edge (from v and w)
for (iv = 0; iv < nv; ++iv)
if (av[iv].v == (w^1)) break;
for (iw = 0; iw < nw; ++iw)
if (aw[iw].v == (v^1)) break;
///if one edge has been deleted, it should be deleted in both direction
if (av[iv].del && aw[iw].del) continue;
if (kv > 1 && kw > 1) {
to_del = 1;
} else if (kw == 1) {
if (asg_topocut_aux(g, w^1, max_ext) < max_ext) to_del = 1;
} else if (kv == 1) {
if (asg_topocut_aux(g, v^1, max_ext) < max_ext) to_del = 1;
}
if (to_del)
av[iv].del = aw[iw].del = 1, ++n_cut;
}
free(b.a);
if (n_cut)
{
asg_cleanup(g);
asg_symm(g);
}
fprintf(stderr, "[M::%s] removed %d suspect overlaps\n", __func__, n_cut);
fprintf(stderr, "[M::%s] takes %0.2f s\n\n", __func__, Get_T()-startTime);
return n_cut;
}
int asg_arc_del_false_node(asg_t *g, int max_ext) int asg_arc_del_false_node(asg_t *g, int max_ext)
{ {
double startTime = Get_T(); double startTime = Get_T();
@@ -7296,6 +7503,7 @@ void read_ma(ma_hit_t* x, FILE* fp)
fread(&(x->ts), sizeof(x->ts), 1, fp); fread(&(x->ts), sizeof(x->ts), 1, fp);
fread(&(x->te), sizeof(x->te), 1, fp); fread(&(x->te), sizeof(x->te), 1, fp);
fread(&(x->el), sizeof(x->el), 1, fp); fread(&(x->el), sizeof(x->el), 1, fp);
fread(&(x->no_l_indel), sizeof(x->no_l_indel), 1, fp);
uint32_t t; uint32_t t;
fread(&(t), sizeof(t), 1, fp); fread(&(t), sizeof(t), 1, fp);
@@ -7333,6 +7541,7 @@ int load_ma_hit_ts(ma_hit_t_alloc** x, char* read_file_name)
for (i = 0; i < n_read; i++) for (i = 0; i < n_read; i++)
{ {
fread(&((*x)[i].is_fully_corrected), sizeof((*x)[i].is_fully_corrected), 1, fp); fread(&((*x)[i].is_fully_corrected), sizeof((*x)[i].is_fully_corrected), 1, fp);
fread(&((*x)[i].is_abnormal), sizeof((*x)[i].is_abnormal), 1, fp);
fread(&((*x)[i].length), sizeof((*x)[i].length), 1, fp); fread(&((*x)[i].length), sizeof((*x)[i].length), 1, fp);
(*x)[i].buffer = (ma_hit_t*)malloc(sizeof(ma_hit_t)*(*x)[i].length); (*x)[i].buffer = (ma_hit_t*)malloc(sizeof(ma_hit_t)*(*x)[i].length);
@@ -7357,6 +7566,7 @@ void write_ma(ma_hit_t* x, FILE* fp)
fwrite(&(x->ts), sizeof(x->ts), 1, fp); fwrite(&(x->ts), sizeof(x->ts), 1, fp);
fwrite(&(x->te), sizeof(x->te), 1, fp); fwrite(&(x->te), sizeof(x->te), 1, fp);
fwrite(&(x->el), sizeof(x->el), 1, fp); fwrite(&(x->el), sizeof(x->el), 1, fp);
fwrite(&(x->no_l_indel), sizeof(x->no_l_indel), 1, fp);
uint32_t t = x->ml; uint32_t t = x->ml;
fwrite(&(t), sizeof(t), 1, fp); fwrite(&(t), sizeof(t), 1, fp);
@@ -7383,6 +7593,7 @@ void write_ma_hit_ts(ma_hit_t_alloc* x, long long n_read, char* read_file_name)
for (i = 0; i < n_read; i++) for (i = 0; i < n_read; i++)
{ {
fwrite(&(x[i].is_fully_corrected), sizeof(x[i].is_fully_corrected), 1, fp); fwrite(&(x[i].is_fully_corrected), sizeof(x[i].is_fully_corrected), 1, fp);
fwrite(&(x[i].is_abnormal), sizeof(x[i].is_abnormal), 1, fp);
fwrite(&(x[i].length), sizeof(x[i].length), 1, fp); fwrite(&(x[i].length), sizeof(x[i].length), 1, fp);
for (k = 0; k < x[i].length; k++) for (k = 0; k < x[i].length; k++)
{ {
@@ -7440,6 +7651,192 @@ char* output_file_name)
} }
// in a resolved bubble, mark unused vertices and arcs as "reduced"
static void asg_bub_backtrack(asg_t *g, uint32_t v0, buf_t *b)
{
uint32_t i, v;
///assert(b->S.n == 1);
///first remove all nodes in this bubble
for (i = 0; i < b->b.n; ++i)
g->seq[b->b.a[i]>>1].del = 1;
///second remove all edges (self/reverse for each edge) in this bubble
for (i = 0; i < b->e.n; ++i) {
asg_arc_t *a = &g->arc[b->e.a[i]];
///remove this edge self
a->del = 1;
///remove the reverse direction
asg_arc_del(g, a->v^1, a->ul>>32^1, 1);
}
///v is the sink of this bubble
v = b->S.a[0];
///recover node
do {
uint32_t u = b->a[v].p; // u->v
g->seq[v>>1].del = 0;
asg_arc_del(g, u, v, 0);
asg_arc_del(g, v^1, u^1, 0);
v = u;
} while (v != v0);
}
// count the number of outgoing arcs, excluding reduced arcs
static inline int count_out(const asg_t *g, uint32_t v)
{
uint32_t i, n, nv = asg_arc_n(g, v);
const asg_arc_t *av = asg_arc_a(g, v);
for (i = n = 0; i < nv; ++i)
if (!av[i].del) ++n;
return n;
}
// pop bubbles from vertex v0; the graph MJUST BE symmetric: if u->v present, v'->u' must be present as well
static uint64_t asg_bub_pop1(asg_t *g, uint32_t v0, int max_dist, buf_t *b)
{
uint32_t i, n_pending = 0;
uint64_t n_pop = 0;
///if this node has been deleted
if (g->seq[v0>>1].del) return 0; // already deleted
///asg_arc_n(n0)
if ((uint32_t)g->idx[v0] < 2) return 0; // no bubbles
///S saves nodes with all incoming edges visited
b->S.n = b->T.n = b->b.n = b->e.n = 0;
///for each node, b->a saves all related information
b->a[v0].c = b->a[v0].d = 0;
///b->S is the nodes with all incoming edges visited
kv_push(uint32_t, b->S, v0);
do {
///v is a node that all incoming edges have been visited
///d is the distance from v0 to v
uint32_t v = kv_pop(b->S), d = b->a[v].d, c = b->a[v].c;
uint32_t nv = asg_arc_n(g, v);
asg_arc_t *av = asg_arc_a(g, v);
///why we have this assert?
///assert(nv > 0);
///all out-edges of v
for (i = 0; i < nv; ++i) { // loop through v's neighbors
/**
p->ul: |____________31__________|__________1___________|______________32_____________|
qn direction of overlap length of this node (not overlap length)
(in the view of query)
p->v : |___________31___________|__________1___________|
tn reverse direction of overlap
(in the view of target)
p->ol: overlap length
**/
uint32_t w = av[i].v, l = (uint32_t)av[i].ul; // v->w with length l
binfo_t *t = &b->a[w];
///that means there is a circle, directly terminate the whole bubble poping
if (w == v0) goto pop_reset;
///if this edge has been deleted
if (av[i].del) continue;
///push the edge
kv_push(uint32_t, b->e, (g->idx[v]>>32) + i);
///find a too far path? directly terminate the whole bubble poping
if (d + l > max_dist) break; // too far
if (t->s == 0) { // this vertex has never been visited
kv_push(uint32_t, b->b, w); // save it for revert
///t->p means the in-node of w is v
///t->s = 1 means w has been visited
///d is len(v0->v), l is len(v->w), so t->d is len(v0->w)
t->p = v, t->s = 1, t->d = d + l;
///incoming edges of w
t->r = count_out(g, w^1);
++n_pending;
} else { // visited before
///c seems the max weight of node
if (c + 1 > t->c || (c + 1 == t->c && d + l > t->d)) t->p = v;
if (c + 1 > t->c) t->c = c + 1;
///update len(v0->w)
if (d + l < t->d) t->d = d + l; // update dist
}
///assert(t->r > 0);
//if all incoming edges of w have visited
//push it to b->S
if (--(t->r) == 0) {
uint32_t x = asg_arc_n(g, w);
if (x) kv_push(uint32_t, b->S, w);
else kv_push(uint32_t, b->T, w); // a tip
--n_pending;
}
}
///if i < nv, that means (d + l > max_dist)
if (i < nv || b->S.n == 0) goto pop_reset;
} while (b->S.n > 1 || n_pending);
asg_bub_backtrack(g, v0, b);
n_pop = 1 | (uint64_t)b->T.n<<32;
pop_reset:
for (i = 0; i < b->b.n; ++i) { // clear the states of visited vertices
binfo_t *t = &b->a[b->b.a[i]];
t->s = t->c = t->d = 0;
}
return n_pop;
}
// pop bubbles
int asg_pop_bubble(asg_t *g, int max_dist)
{
uint32_t v, n_vtx = g->n_seq * 2;
uint64_t n_pop = 0;
buf_t b;
if (!g->is_symm) asg_symm(g);
memset(&b, 0, sizeof(buf_t));
///set information for each node
b.a = (binfo_t*)calloc(n_vtx, sizeof(binfo_t));
//traverse all node with two directions
for (v = 0; v < n_vtx; ++v) {
uint32_t i, n_arc = 0, nv = asg_arc_n(g, v);
asg_arc_t *av = asg_arc_a(g, v);
///some node could be deleted
if (nv < 2 || g->seq[v>>1].del) continue;
///some edges could be deleted
for (i = 0; i < nv; ++i) // asg_bub_pop1() may delete some edges/arcs
if (!av[i].del) ++n_arc;
if (n_arc > 1)
n_pop += asg_bub_pop1(g, v, max_dist, &b);
}
free(b.a); free(b.S.a); free(b.T.a); free(b.b.a); free(b.e.a);
if (n_pop) asg_cleanup(g);
fprintf(stderr, "[M::%s] popped %d bubbles and trimmed %d tips\n", __func__, (uint32_t)n_pop, (uint32_t)(n_pop>>32));
return n_pop;
}
void output_contig_graph(asg_t *sg, ma_sub_t* coverage_cut, char* output_file_name, long long n_read, long long bubble_dist)
{
asg_pop_bubble(sg, bubble_dist);
ma_ug_t *ug = NULL;
ug = ma_ug_gen(sg);
ma_ug_seq(ug, &R_INF, coverage_cut, n_read);
fprintf(stdout, "Writing unitig GFA to disk ...... \n");
char* gfa_name = (char*)malloc(strlen(output_file_name)+35);
sprintf(gfa_name, "%s.contig.gfa", output_file_name);
FILE* output_file = fopen(gfa_name, "w");
ma_ug_print(ug, &R_INF, coverage_cut, output_file);
fclose(output_file);
sprintf(gfa_name, "%s.simple.contig.gfa", output_file_name);
output_file = fopen(gfa_name, "w");
ma_ug_print_simple(ug, &R_INF, coverage_cut, output_file);
fclose(output_file);
free(gfa_name);
ma_ug_destroy(ug);
}
void build_string_graph_without_clean( void build_string_graph_without_clean(
int min_dp, ma_hit_t_alloc* sources, ma_hit_t_alloc* reverse_sources, int min_dp, ma_hit_t_alloc* sources, ma_hit_t_alloc* reverse_sources,
@@ -7459,8 +7856,8 @@ char* output_file_name, long long bubble_dist, int read_graph, int write)
// debug_info_of_specfic_read("m64016_190918_162737/92668450/ccs", // debug_info_of_specfic_read("m64016_190918_162737/92668450/ccs",
// sources, reverse_sources, -1, "init"); // sources, reverse_sources, -1, "init");
debug_info_of_specfic_read("m64016_190918_162737/53545052/ccs", // debug_info_of_specfic_read("m64016_190918_162737/53545052/ccs",
sources, reverse_sources, -1, "init"); // sources, reverse_sources, -1, "init");
@@ -7472,8 +7869,8 @@ char* output_file_name, long long bubble_dist, int read_graph, int write)
// debug_info_of_specfic_read("m64016_190918_162737/92668450/ccs", // debug_info_of_specfic_read("m64016_190918_162737/92668450/ccs",
// sources, reverse_sources, -1, "normalize"); // sources, reverse_sources, -1, "normalize");
debug_info_of_specfic_read("m64016_190918_162737/53545052/ccs", // debug_info_of_specfic_read("m64016_190918_162737/53545052/ccs",
sources, reverse_sources, -1, "normalize"); // sources, reverse_sources, -1, "normalize");
@@ -7513,9 +7910,6 @@ char* output_file_name, long long bubble_dist, int read_graph, int write)
asg_t *sg = NULL; asg_t *sg = NULL;
sg = ma_sg_gen(sources, n_read, coverage_cut, max_hang_length, mini_overlap_length); sg = ma_sg_gen(sources, n_read, coverage_cut, max_hang_length, mini_overlap_length);
@@ -7523,11 +7917,6 @@ char* output_file_name, long long bubble_dist, int read_graph, int write)
// debug_info_of_specfic_node("m64016_190918_162737/72220752/ccs", sg, "sg_gen"); // debug_info_of_specfic_node("m64016_190918_162737/72220752/ccs", sg, "sg_gen");
asg_arc_del_trans(sg, GAP_FUZZ); asg_arc_del_trans(sg, GAP_FUZZ);
// debug_info_of_specfic_node("m64016_190918_162737/72220752/ccs", sg, "del_trans"); // debug_info_of_specfic_node("m64016_190918_162737/72220752/ccs", sg, "del_trans");
@@ -7543,7 +7932,7 @@ char* output_file_name, long long bubble_dist, int read_graph, int write)
asg_arc_del_short_diploid_unclean(sg, corase_ovlp_drop_ratio, sources, reverse_sources); ///asg_arc_del_short_diploid_unclean(sg, corase_ovlp_drop_ratio, sources, reverse_sources);
@@ -7584,36 +7973,25 @@ char* output_file_name, long long bubble_dist, int read_graph, int write)
fprintf(stderr, "\n\n**********%d-th round drop: drop_ratio = %f**********\n", fprintf(stderr, "\n\n**********%d-th round drop: drop_ratio = %f**********\n",
i, drop_ratio); i, drop_ratio);
// debug_info_of_specfic_node("m64016_190918_162737/72220752/ccs", sg, "step");
while(1) while(1)
{ {
int tri_flag = 0; int tri_flag = 0;
tri_flag += asg_arc_del_self_circle_contig(sg); tri_flag += asg_arc_del_self_circle_contig(sg);
// fprintf(stderr, "tri_flag: %d\n", tri_flag);
// fflush(stderr);
///asg_arc_del_single_node_bubble(sg, bubble_dist); ///asg_arc_del_single_node_bubble(sg, bubble_dist);
tri_flag += asg_arc_del_single_node_directly(sg, MAX_SHORT_TIPS); tri_flag += asg_arc_del_single_node_directly(sg, MAX_SHORT_TIPS, sources);
// fprintf(stderr, "tri_flag: %d\n", tri_flag);
// fflush(stderr);
tri_flag += asg_arc_del_triangular_advance(sg, bubble_dist); tri_flag += asg_arc_del_triangular_advance(sg, bubble_dist);
// fprintf(stderr, "tri_flag: %d\n", tri_flag);
// fflush(stderr);
tri_flag += asg_arc_del_cross_bubble(sg, bubble_dist); tri_flag += asg_arc_del_cross_bubble(sg, bubble_dist);
// fprintf(stderr, "tri_flag: %d\n", tri_flag);
// fflush(stderr);
///asg_arc_del_single_node_bubble(sg, bubble_dist); ///asg_arc_del_single_node_bubble(sg, bubble_dist);
tri_flag += asg_arc_del_single_node_directly(sg, MAX_SHORT_TIPS); tri_flag += asg_arc_del_single_node_directly(sg, MAX_SHORT_TIPS, sources);
// fprintf(stderr, "tri_flag: %d\n", tri_flag);
// fflush(stderr);
if(tri_flag == 0) if(tri_flag == 0)
{ {
break; break;
} }
} }
/****************************may have bugs********************************/ /****************************may have bugs********************************/
@@ -7624,37 +8002,21 @@ char* output_file_name, long long bubble_dist, int read_graph, int write)
/****************************may have bugs********************************/ /****************************may have bugs********************************/
/****************************may have bugs********************************/ /****************************may have bugs********************************/
//asg_arc_identify_simple_bubbles(sg); ///asg_arc_identify_simple_bubbles_multi(sg, 1);
asg_arc_identify_simple_bubbles_multi(sg, 1); asg_arc_identify_simple_bubbles_multi(sg, 0);
///asg_arc_del_short_diploid_unclean_exact(sg, drop_ratio, sources); ///asg_arc_del_short_diploid_unclean_exact(sg, drop_ratio, sources);
asg_arc_del_short_diploid_by_exact(sg, MAX_SHORT_TIPS, sources); asg_arc_del_short_diploid_by_exact(sg, MAX_SHORT_TIPS, sources);
asg_cut_tip(sg, MAX_SHORT_TIPS); asg_cut_tip(sg, MAX_SHORT_TIPS);
/****************************may have bugs********************************/ /****************************may have bugs********************************/
// fprintf(stderr, "bugs\n");
// fflush(stderr);
//asg_arc_identify_simple_bubbles(sg); //asg_arc_identify_simple_bubbles(sg);
asg_arc_identify_simple_bubbles_multi(sg, 1); asg_arc_identify_simple_bubbles_multi(sg, 1);
// fprintf(stderr, "asg_arc_identify_simple_bubbles_multi\n");
// fflush(stderr);
asg_arc_del_short_diploid_by_length(sg, drop_ratio, MAX_SHORT_TIPS); asg_arc_del_short_diploid_by_length(sg, drop_ratio, MAX_SHORT_TIPS);
asg_cut_tip(sg, MAX_SHORT_TIPS); asg_cut_tip(sg, MAX_SHORT_TIPS);
// fprintf(stderr, "asg_arc_del_short_diploid_by_length\n");
// fflush(stderr);
// debug_info_of_specfic_node("m64016_190918_162737/141297762/ccs", sg);
asg_arc_identify_simple_bubbles_multi(sg, 0); asg_arc_identify_simple_bubbles_multi(sg, 0);
asg_arc_del_short_false_link(sg, 0.6, bubble_dist); asg_arc_del_short_false_link(sg, 0.6, bubble_dist);
@@ -7677,7 +8039,7 @@ char* output_file_name, long long bubble_dist, int read_graph, int write)
// fprintf(stderr, "tri_flag: %d\n", tri_flag); // fprintf(stderr, "tri_flag: %d\n", tri_flag);
// fflush(stderr); // fflush(stderr);
///asg_arc_del_single_node_bubble(sg, bubble_dist); ///asg_arc_del_single_node_bubble(sg, bubble_dist);
tri_flag += asg_arc_del_single_node_directly(sg, MAX_SHORT_TIPS); tri_flag += asg_arc_del_single_node_directly(sg, MAX_SHORT_TIPS, sources);
// fprintf(stderr, "tri_flag: %d\n", tri_flag); // fprintf(stderr, "tri_flag: %d\n", tri_flag);
// fflush(stderr); // fflush(stderr);
tri_flag += asg_arc_del_triangular_advance(sg, bubble_dist); tri_flag += asg_arc_del_triangular_advance(sg, bubble_dist);
@@ -7689,7 +8051,7 @@ char* output_file_name, long long bubble_dist, int read_graph, int write)
// fprintf(stderr, "tri_flag: %d\n", tri_flag); // fprintf(stderr, "tri_flag: %d\n", tri_flag);
// fflush(stderr); // fflush(stderr);
///asg_arc_del_single_node_bubble(sg, bubble_dist); ///asg_arc_del_single_node_bubble(sg, bubble_dist);
tri_flag += asg_arc_del_single_node_directly(sg, MAX_SHORT_TIPS); tri_flag += asg_arc_del_single_node_directly(sg, MAX_SHORT_TIPS, sources);
// fprintf(stderr, "tri_flag: %d\n", tri_flag); // fprintf(stderr, "tri_flag: %d\n", tri_flag);
// fflush(stderr); // fflush(stderr);
@@ -7700,6 +8062,17 @@ char* output_file_name, long long bubble_dist, int read_graph, int write)
} }
asg_arc_del_short_diploi_by_suspect_edge(sg, MAX_SHORT_TIPS, sources);
asg_cut_tip(sg, MAX_SHORT_TIPS);
// asg_arc_identify_simple_bubbles_multi(sg, 0);
// asg_arc_del_too_short_overlaps(sg, 1000);
// asg_cut_tip(sg, MAX_SHORT_TIPS);
///asg_arc_del_triangular_advance_debug(sg, bubble_dist); ///asg_arc_del_triangular_advance_debug(sg, bubble_dist);
/** /**
@@ -7740,14 +8113,20 @@ char* output_file_name, long long bubble_dist, int read_graph, int write)
/****************************may have bugs********************************/ /****************************may have bugs********************************/
/**
memset(sg->seq_vis, 0, sg->n_seq*2*sizeof(uint8_t));
asg_arc_del_short_diploid_by_exact(sg, MAX_SHORT_TIPS, sources);
asg_cut_tip(sg, MAX_SHORT_TIPS);
**/
// debug_info_of_specfic_node("m64016_190918_162737/141297762/ccs", sg); // debug_info_of_specfic_node("m64016_190918_162737/141297762/ccs", sg);
out: out:
output_unitig_graph(sg, coverage_cut, output_file_name, n_read); output_unitig_graph(sg, coverage_cut, output_file_name, n_read);
output_read_graph(sg, coverage_cut, output_file_name, n_read); output_read_graph(sg, coverage_cut, output_file_name, n_read);
output_contig_graph(sg, coverage_cut, output_file_name, n_read, 50000);
asg_destroy(sg); asg_destroy(sg);
free(coverage_cut); free(coverage_cut);
Overlaps.h
+5
View File
@@ -7,6 +7,7 @@
///#define MIN_OVERLAP_LEN 2000 ///#define MIN_OVERLAP_LEN 2000
///#define MIN_OVERLAP_LEN 500 ///#define MIN_OVERLAP_LEN 500
///#define MIN_OVERLAP_LEN 50
#define MIN_OVERLAP_LEN 50 #define MIN_OVERLAP_LEN 50
#define MIN_OVERLAP_COVERAGE 1 #define MIN_OVERLAP_COVERAGE 1
#define MAX_HANG_LEN 1000 #define MAX_HANG_LEN 1000
@@ -42,6 +43,7 @@ typedef struct {
uint32_t ml:31, rev:1; uint32_t ml:31, rev:1;
uint32_t bl:31, del:1; uint32_t bl:31, del:1;
uint8_t el; uint8_t el;
uint8_t no_l_indel;
} ma_hit_t; } ma_hit_t;
@@ -50,6 +52,7 @@ typedef struct {
uint32_t size; uint32_t size;
uint32_t length; uint32_t length;
uint8_t is_fully_corrected; uint8_t is_fully_corrected;
uint8_t is_abnormal;
} ma_hit_t_alloc; } ma_hit_t_alloc;
@@ -96,6 +99,7 @@ typedef struct {
uint32_t ol:31, del:1; uint32_t ol:31, del:1;
uint8_t strong; uint8_t strong;
uint8_t el; uint8_t el;
uint8_t no_l_indel;
} asg_arc_t; } asg_arc_t;
typedef struct { typedef struct {
@@ -194,6 +198,7 @@ static inline int ma_hit2arc(const ma_hit_t *h, int ql, int tl, int max_hang, fl
p->strong = h->ml; p->strong = h->ml;
p->el = h->el; p->el = h->el;
p->no_l_indel = h->no_l_indel;
return l; return l;
} }