xxay/hifiasm
1
0
Fork 0
mirror of https://github.com/chhylp123/hifiasm.git synced 2026-10-03 10:38:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

update for r350

Browse Source
This commit is contained in:
chhylp123
2021-07-26 00:16:00 -04:00
parent b14f894160
commit 321acb2d8a
12 changed files with 241 additions and 130 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/source/faq.rst
+22 -8
View File
@@ -34,19 +34,19 @@ Are diploid genomes supported?
Are polyploid genomes supported?
-------------------------------------
The ``*r_utg.gfa`` and ``*p_utg.gfa`` are lossless so that they also work for polyploid genomes. However, currently the contig-generation modules of hifiasm are designed for diploid samples, which means both the partially phased assembly and the fully-phased assembly does not directly support polyploid genomes. Please use primary assembly for polyploid samples and run multiple rounds of purging steps using third-party tools such as purge_dups.
The ``*r_utg.gfa`` and ``*p_utg.gfa`` are lossless so that they also work for polyploid genomes. However, currently the contig-generation modules of hifiasm are designed for diploid samples, which means both the partially phased assembly and the fully-phased assembly does not directly support polyploid genomes. If it is set to >2, the quality of primary assembly for polyploid genomes might be improved. Please use primary assembly for polyploid samples and run multiple rounds of purging steps using third-party tools such as purge_dups.
Why one Hi-C integrated assembly is larger than another one?
------------------------------------------------------------
For some samples like human male, the paternal haplotype should be larger than the maternal haplotype. However, if one assembly is much larger than another one, it should be the issues of hifiasm. To fix it, please set smaller value for ``-s`` (default: 0.55).
Another possibility is that hifiasm misidentifies coverage threshold for homozygous reads. For instance, hifiasm prints the following information during assembly without setting the option ``--purge-cov``:
Another possibility is that hifiasm misidentifies coverage threshold for homozygous reads. For instance, hifiasm prints the following information during assembly:
::
[M::purge_dups] purge duplication coverage threshold: 46
[M::purge_dups] purge duplication coverage threshold: 36
In this example, hifiasm identifies the coverage threshold for homozygous reads as ``46/1.25 = 36``. If it is significantly smaller than the homozygous coverage peak, hifiasm will generate two unbalanced assemblies. In this case, please set ``--purge-cov`` to homozygous coverage peak. Please note that tuning ``--purge-cov`` may affect ``*p_utg*gfa`` so that ``*hic*.bin`` should be deleted. Since v0.15.5, hifiasm can detect such changes and renew Hi-C bin files automatically.
In this example, hifiasm identifies the coverage threshold for homozygous reads as ``36``. If it is significantly smaller than the homozygous coverage peak, hifiasm will generate two unbalanced assemblies. In this case, please set ``--hom-cov`` to homozygous coverage peak. Please note that tuning ``--hom-cov`` may affect ``*p_utg*gfa`` so that ``*hic*.bin`` should be deleted. Since v0.15.5, hifiasm can detect such changes and renew Hi-C bin files automatically.
@@ -57,12 +57,12 @@ For Hi-C integrated assembly, why the assembly size of both haplotypes are much
[M::stat] # heterozygous bases: 645155110; # homozygous bases: 1495396634
If most bases of a diploid sample are homozygous, the coverage threshold is wrongly determined by hifiasm. For instance, hifiasm prints the following information during assembly without setting the option ``--purge-cov``:
If most bases of a diploid sample are homozygous, the coverage threshold is wrongly determined by hifiasm. For instance, hifiasm prints the following information during assembly:
::
[M::purge_dups] purge duplication coverage threshold: 46
[M::purge_dups] purge duplication coverage threshold: 36
In this example, hifiasm identifies the coverage threshold for homozygous reads as ``46/1.25 = 36``. If it is much smaller than homozygous coverage peak, hifiasm thinks most reads are homozygous and assign them to both assemblies, making both of them much larger than the estimated haploid genome size. In this case, please set ``--purge-cov`` to homozygous coverage peak. Please note that tuning ``--purge-cov`` may affect ``*p_utg*gfa`` so that ``*hic*.bin`` should be deleted. Since v0.15.5, hifiasm can detect such changes and renew Hi-C bin files automatically.
In this example, hifiasm identifies the coverage threshold for homozygous reads as ``36``. If it is much smaller than homozygous coverage peak, hifiasm thinks most reads are homozygous and assign them to both assemblies, making both of them much larger than the estimated haploid genome size. In this case, please set ``--hom-cov`` to homozygous coverage peak. Please note that tuning ``--hom-cov`` may affect ``*p_utg*gfa`` so that ``*hic*.bin`` should be deleted. Since v0.15.5, hifiasm can detect such changes and renew Hi-C bin files automatically.
.. _hic-iss:
@@ -80,10 +80,14 @@ Why the size of primary assembly or partially phased assembly is much larger tha
It could be because the estimated genome size is incorrect. Another possibility is that hifiasm does not perform enough purging. Setting smaller value for ``-s`` (default: 0.55) or turning ``--purge-cov`` should be helpful. See :ref:`loginter` for more details.
.. _p-hamming:
Why the hamming error rate or the swith error rate of trio-binning assembly is very high?
---------------------------------------------------------------------------------------------------------------
In rare cases, a potential issue is that a few contigs may misjoin two haplotypes. For example, half of a contig come from mother while another half come from father. Such misjoined contigs can be fixed by manually breaking. The coordinates of problematic regions can be found by A-lines in GFA file or ``yak trioeval -e`` (see `issue 37 <https://github.com/chhylp123/hifiasm/issues/37>`_ for more details). However, if there are many misjoined contigs or the switch/hamming error rate reported by ``yak trioeval`` is very high, users should check if the parental data is correct (see `issue 130 <https://github.com/chhylp123/hifiasm/issues/130#issuecomment-862347943>`_ for more details).
Another possibility is that there are some unitigs in unitig graph misjoining two haplotypes. Such problematic unitigs might be ignored by the graph-binning strategy. Set smaller value for ``--t-occ`` forcedly remove unitig including unexpected haplotype-specific reads.
Why does hifiasm stuck or crash?
-------------------------------------
In most cases, it is caused by the low quality HiFi reads. A good HiFi dataset should have a k-mer plot like `issue10 <https://github.com/chhylp123/hifiasm/issues/10#issuecomment-616213684>`_ or `issue49 <https://github.com/chhylp123/hifiasm/issues/49#issue-729106823>`_. In contrast, low quality HiFi data often lead to weird k-mer plot like `issue93 <https://github.com/chhylp123/hifiasm/issues/93#issue-852259042>`_. Such weird k-mer plots usually indicate insufficient coverage or presence of contaminants. See :ref:`loginter` for more details. If the HiFi data look fine, please raise an issue at the `issue page <https://github.com/chhylp123/hifiasm/issues>`_.
@@ -98,7 +102,7 @@ Can I generate HiFi-only assembly first, and then add Hi-C or trio data later?
What is the minimum read coverage required for hifiasm?
-------------------------------------------------------
Usually >=14x HiFi reads per haplotype. Higher coverage might be able to improve the contiguity of assembly.
Usually >=13x HiFi reads per haplotype. Higher coverage might be able to improve the contiguity of assembly.
Why the primary assembly is more contiguous than the fully-phased assemblies and the partially phased assemblies (i.e. ``*.hap*.p_ctg.gfa``)?
----------------------------------------------------------------------------------------------------------------------------------------------------
@@ -107,3 +111,13 @@ Why the primary assembly is more contiguous than the fully-phased assemblies and
When producing fully-phased assemblies and partially phased assemblies, hifiasm is designed to keep both haplotypes contiguous. It is important for many downstream applications like SV calling.
My assembly is fragmented or not contiguous enough, how do I improve it?
--------------------------------------------------------------------------
Raising ``-D`` or ``-N`` may improve the resolution of repetitive regions but takes longer time. These two options affect all types of assemblies and usually do not have a negative impact on the assembly quality. In contrast, ``--purge-max`` only affects primary assembly. Setting larger value for ``--purge-max`` makes primary assembly more contiguous but may collapse repeats or segmental duplications.
If the assembly is too fragmented, users should check if HiFi data is good enough. See `Why does hifiasm stuck or crash?`_ for details.
How do I avoid misassemblies?
--------------------------------------------------------------------------
Set smaller value for ``--purge-max``, ``-s`` and ``-O``, or use the ``-u`` option.