Short Read Aligners/ Snp Calling Comparison

Question

[Pierre Lindenbaum]

Dear lazyweb,

my question is somehow related to some previous questions asked here: ( http://biostar.stackexchange.com/questions/788 , http://biostar.stackexchange.com/questions/31, http://biostar.stackexchange.com/questions/613 , etc...).

I'm looking for a table that would summarize the pro's and the con's of each tool for NGS (mappers and/or snp callers), the algorithm used, if the quality of the fastq, is taken into account, how the gaps & the indels are handled, a publication, etc... ... e.g.: something that would look like the following (almost empty) page on wikipedia: http://en.wikipedia.org/wiki/List_of_sequence_alignment_software#Short-Read_Sequence_Alignment

Thanks,

Pierre

UPDATE: @Julien_f suggested this link: http://lh3lh3.users.sourceforge.net/NGSalign.shtml

Update2: another cool resource cited in Marc's paper: http://seqanswers.com/wiki/Software/list

[Jorge Amigo]

that would definitely be a valuable resource.

[Michael]

Maybe we should write a paper on this ;)

[lh3]

In general, I am against a long list. It tells me nothing about the speed, memory, friendliness and accuracy. After reading that, we still do not know what we should choose. It is more important to know how often these are used in practice.

[lh3]

In addition, you should go to seqanswers.com. Your answers will be better replied there.

Answer 1

[lh3]

Although there are over 30 aligners, I guess 90% of researchers only choose from the list below:

• Illumina: Bowtie, BWA, Mosaik and Novoalign (Stampy is also a good one)
• SOLiD: Bfast, Bowtie and BWA (perhaps also NovoalignCS?)
• 454: SSAHA2, LAST, BLAT and Mosaik (I would recommend BWA-SW actually)

Carefully studying their features would not take much time. In addition, to "feel" a tool, we'd better use it rather than just look at the feature table.

As to SNP callers, I guess 90% who do not write in-house tools are using GATK and SAMtools/MAQ. SNVMix is also a good one, especially for cancer sequencing. As there are only a few of them, it is easy to try each.

In addition, usual SNP calling on deep resequencing data is not hard. Many groups use their in-house tools and being highly tuned for specific data, in-house tools may have better performance.

At last, as I am promoting everywhere, apply BAQ implemented in samtools. Doing this will improve the accuracy for all SNP callers so far.

[Erik Garrison]

I should note that I've observed a marked decrease in sensitivity when using BAQ. Is it still suitable for current illumina data.

The sensitivity loss is observed in simulation and also in large-scale resequencing runs.

[lh3]

With better realignment, BAQ is optional. I agree that the original BAQ hurts sensitivity to MNPs, but extended BAQ is better. If I am right, BCM, UM and Sanger are still using extended BAQ at different steps. I do not remember they have obviously lower sensitivity than BC. I have not looked into 1000g calls for quite some time, so could be wrong.

Answer 2

[Frédéric Bigey]

May I suggest the reading of:

Nielsen R, Paul JS, Albrechtsen A, Song YS. Genotype and SNP calling from next-generation sequencing data. Nat Rev Genet. 2011 Jun;12(6):443-51. Review.

[lh3]

+1. A very good review.

[Giovanni M Dall'Olio]

What happened to this link?

Answer 3

[Marc Perry]

Did you already check out this review from last year?

Nat Methods. 2009 Nov;6(11 Suppl):S6-S12. Sense from sequence reads: methods for alignment and assembly. by Flicek P, Birney E.

[Pierre Lindenbaum]

I accepted your anwser as I found some cool resources in this paper.

[Pierre Lindenbaum]

I accepted your answer as I found some cool resources in this paper.

Answer 4

[Pablo Marin-Garcia]

Have you seen this paper? http://genomemedicine.com/content/5/3/28 Low concordance of multiple variant-calling pipelines: practical implications for exome and genome sequencing Jason O'Rawe1,2, Tao Jiang3, Guangqing Sun3, Yiyang Wu1,2, Wei Wang4, Jingchu Hu3, Paul Bodily5, Lifeng Tian6, Hakon Hakonarson6, W Evan Johnson7, Zhi Wei4, Kai Wang8,9* and Gholson J Lyon1,2,9*

Not a table per se and software used is a bit outdated (bwa is 0.5.9) but give the list of commands used and could be useful as template for doing the comparison by yourself.

Answer 5

[Malachi Griffith]

Here is a very handy list of short read aligners/mappers:

HTS Mappers

Tools for mapping high-throughput sequencing data (Pubmed)