Variant Caller Of Choice?

Question

[richardc.gsc]

Hi all, Currently we use mpileup and Strelka as our variant callers in our illumina read cancer genome pipeline.

We have been relatively happy with the results from both but we like to keep on top of things. Has anyone had positive experiences with other variant callers that you would recommend testing out?

I am interested in both individual human SNP calling as well as somatic calling (ie. what is in the tumour and not in the matched normal).

I appreciate any feedback. thanks!

[Ashutosh Pandey]

You can also try GATK Unified Genotyper.

[Jorge Amigo]

from GATK's best practice:

Choosing HaplotypeCaller or UnifiedGenotyper

• We believe the best possible caller in the GATK is the HaplotypeCaller, which combines a local de novo assembler with a more advanced HMM likelihood function than the UnifiedGenotyper. It should produce excellent SNP, MNP, indel, and short SV calls. It should be the go-to calling algorithm for most projects. It is, for example, how we make our Phase II call set for 1000 Genomes.
• Currently the HaplotypeCaller only supports diploid calling. If you want to call non-diploid samples you'll need to use the UnifiedGenotyper.
• At the moment the HaplotypeCaller does not support multithreading. For now you should indeed stick with the UG if you wish to use the -nt option. However you can use Queue to parallelize execution of HaplotypeCaller.
• If for some reason you cannot use the HaplotyperCaller do fall back to the UnifiedGenotyper protocol below. Otherwise try out the HaplotypeCaller and let us know about your experiences here on the forum!

See more at: http://gatkforums.broadinstitute.org/discussion/1186/best-practice-variant-detection-with-the-gatk-v4-for-release-2-0#sthash.1CMxyQIK.dpuf

[User 59]

I think you will find GATK's UnifiedGenotpyer is pretty unsuitable for somatic mutations: http://gatkforums.broadinstitute.org/discussion/2405/can-i-use-unifiedgenotyper-to-call-somatic-variants-in-tumorcontrol-tissues

[Ashutosh Pandey]

You are right. I should have been more specific in answering. I only use it for SNP calling in mice strains.

Answer 1

[Jordan]

We had had good luck with MuTect from Broad Institute. And Varscan is another tools which takes in both Tumor and Normal pairs.

Apart from that, UnifiedGenotyper from Broad is something which people normally use for individual somatic calling in humans.

[User 59]

The more I use VarScan the more I like it. SomaticSniper is another choice for paired/normal calling

Answer 2

[Carlos Borroto]

I'm currently testing out gkno.me. I love the simplicity, the brilliant use of GNU Make to avoid rerunning unnecessary steps with unchanged input or options and the flexibility to accommodate future progress in the field.

It comes with already predefined pipelines with good documentation, also has documentation on how to build your own pipeline and allows you to integrate new tools.

Answer 3

[Sangwoo Kim]

Oh. This is a good chance to advertise our new somatic mutation caller Virmid. http://sourceforge.net/p/virmid/wiki/Home/

It considers sample impurity (like proportion of stromal cells in the tumor sample) to improve sensitivity. We compared performance with MuTect and Strelka. They are both good tools. But you can try ours if you suspect some sample impurity. We expect the paper to be published soon.

Answer 4

[Malachi Griffith]

This post:

Best Software For Detection Of Somatic Mutations From Matched Tumor:Normal Ngs Data

describes quite a few of them.