about the use of varscan, somaticsnipper and samtools

Question

[Bogdan]

Dear all,

please could you advise : before the use of eitther varscan, somaticsnipper or samtools to call the somatic mutations, would we need to :

a, mark the duplicates b. realign the data around indels (as we would do typically with GATK, before running Mutect) ?

many thanks !

bogdan

Answer 1

[Chris Miller]

a) yes, mark duplicates in your bam files

b) no, local indel realignment is not necessary

c) I wouldn't rely solely on somatic sniper or varscan for detecting indels. They'll provide some useful information, but Pindel-C and GATK generally perform better on indel calling.

[Len Trigg]

Hi Chris, Can you elaborate on why you say that local indel realignment is not necessary? Is it that you see no benefit, or that some callers do not need the local realignment preprocessing in order to deal with non-canonical alignments? Our (RTG) somatic caller internally does something similar to local realignment to candidate haplotypes when calling complex regions such as indels or dense snps, so realignment is not needed as a separate preprocessing step, and I imagine that callers incorporating local assembly also do not need realignment preprocessing.

[Chris Miller]

Definitely the latter. I can't comment on whether there is any benefit, but the specific tools he mentioned do not require it.

Answer 2

[Bogdan]

Thanks Chris ! Talking about somatic variant calling, which one of these 4 would you recommend : STRELKA, MUTECT, VARSCAN or SOMATICSNIPPER ? Or perhaps some other somatic variant caller ?

what about the SV detection or CNV ? Any recommendation is welcome ! many thanks !

[Chris Miller]

• We use all four of those and combine them in various ways to up our sensitivity.
• For SVs, I like Manta or Lumpy. For CNV, I'm partial to my own package - copyCat (https://github.com/chrisamiller/copycat)

Answer 3

[Bogdan]

Chris, thank you very much !