This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Varscan Output Strandness

Hello, I am carefully performing an SNP analysis, with BWA, SamTools, Picard, Bcftools and VarScan, primarily. The one thing I am not sure I have under control is the strand information in my data. My reads are 75bp paired-end from a cDNA library not strand-specific. My question is, how does varscan handle this? Will the input have info about this? I know that some A-to-G calls will be lost in the absence of strand information. Other than that, will the final vcf file contain this data? whether the SNP is on the plus or minus strand? If not? how would I best tackle this problem? Thanks, G.

snp strand
• 3,730 views • link
updated by Sean Davis Sean Davis • written by GPR GPR

1 answer

SNVs are always on both strands since an SNV on the forward strand implies the complement base on the reverse. I believe VarScan reports SNVs on the forward strand.

• 0 views • link
written by Sean Davis Sean Davis

OK, thanks. I guess my confusion stems from the fact that in VarScan.jar filter, there is the option (--min-strands2) to be set as 1 or 2. What's the recommended value here? Should the variant be observed in the two strands? or is 1 enough?

• 0 views • link
written by GPR GPR

Only you can answer this question. The "strand" in this case has to do with the strandedness of the READ. If the variant is only on one strand, that may represent a false positive finding. I would recommend using VarScan liberally (let everything through) and then filter after-the-fact.

• 0 views • link
written by Sean Davis Sean Davis

Yep, Sean nailed it. SNVs that only appear on one strand or the other are often false-positives. That said, if sequence coverage is low at that site, you may only see reads from one strand just by chance. It's up to you to decide what acceptable thresholds are.

• 0 views • link
written by Chris Miller Chris Miller

Thanks All!. Will keep this in mind.

• 0 views • link
written by GPR GPR

Log in to answer this question.

More posts like this