No Somatic SNV Calling with GATK?

Question

[sichan]

Hi,

I see that GATK should not be used to call somatic variants from tumor/normal pairs:

http://gatkforums.broadinstitute.org/discussion/2405/can-i-use-unifiedgenotyper-to-call-somatic-variants-in-tumor-control-tissues

Is this still true today? I ask because someone who is no longer in our group generated 'somatic calls' by

• calling SNPs on the normal
• calling SNPs on the matched tumor
• subtracting the normal SNPs from the tumor SNPs

I'm trying to determine how useful those calls are (probably not much).

This post from ~2 years ago also suggests that GATK should not be used for somatic calls: Difference Between Somatic And Germline Variant Calling?

Thank you

[sichan]

Thanks everyone! Yes, I forgot to mention that I had already started running MuTect, which is also from Broad.

[User 59]

Try to refrain from posting thanks as an answer. An upvote or accepted answer will do just fine!

Answer 1

[Sean Davis]

You should definitely abandon those calls and use a tool designed for somatic variant calling. There are numerous tools to choose from.

[vdauwera]

You can use MuTect, which is based on GATK and is available for download from the GATK website.

[Sean Davis]

Yes, MuTect is one of many and a popular one. On the downside, it is pretty slow unless one takes the time to run it in parallel.

[User 59]

I could't agree more. This naive subtractive approach is in no way comparable to say VarScan2.