varscan2: filters

[mrfox]

Hi all and hopefully Dan,

I am still not very sure about how to use the Varscan2 filters for SNVs and Indels obtained. Below are my questions, any advice on the clarification of using the filters will be appreciated.

1) For a single sample,

Step 1. Use pileup2snp and pileup2indel call SNVs and Indels.

Step2. Run varscan2 filter on SNVs using the Indels as an input. Then filter indels using java -jar varscan2 filter output.indel >filtered.indel.

Aternatively,
Step2. Another option is to filter indel first and use the filtered indels as input to filter SNVs.

Which one is better for step 2?

2) For a Normal/Tumor sample,
From the context of the Varscan2 instruction, the "somaticFilter" is designed only for filtering SNVs, not for indels. So there is actually no filter for Indels. The Indel file obtained from calling Varscan2 somatic will be the final reported Indel. Is that correct? Is somaticFilter applicable to Indels?

Of course, we can change any somatic indel with read2>0 into Germline.

Thank you all!

[dkoboldt]

Hao,

Thank you for the question. First of all, I would use *unfiltered* indels to filter SNPs (as long as your variant calling criteria are reasonable), because your goal is to remove variants that are due to alignment issues, and even unfiltered indel calls are a guide to these. But use your judgment: if the unfiltered indels remove a significant portion of dbSNPs, then there's a problem.

The somaticFilter works on somatic indel as well as SNV files from VarScan somatic. I do recommend that you use it, but you're correct to think of applying your own filters to those (especially to catch germline variants). Any time we see "somatic" indels with evidence in the normal sample, or that match dbSNP variants, we remove them.