Bug Bulletin: we have identified a bug that affects indexing when producing gzipped VCFs. This will be fixed in the upcoming 3.2 release; in the meantime you need to reindex gzipped VCFs using Tabix.

Multi-sample calling vs single sample calling

SharonCoxSharonCox Posts: 4Member

Dear GATK team, I have used the UG following the best practice GATK workflow to call snps and Indels from exomeseq data of 24 human samples. First I called snps and Indels separately for each bam file, and I obtained separate vcfs. Then I decided to try to call the snps and Indels all in one go. I noticed that the output was quite different and the number of inser/delitions was higher when I called variants in contemporary (starting from separate bam files: -I sample1.bam -I sample2.bam...ETC). I also noticed that the called indels mostly were adjacent to tricky areas such as repetitive elements (ATATATATATATAT) or next to polyAAAAA. These snps and Indels weren't called by the UG when I called the variants separately. Is it more error prone to call variants in contemporary?

Answers

  • Geraldine_VdAuweraGeraldine_VdAuwera Posts: 5,235Administrator, GSA Member admin

    Hi Sharon,

    Ah, I understand your question better now -- I thought you were calling indels and SNPs together, then separately, and seeing differences in the totals of variants called (there shouldn't be).

    If your samples belong to a coherent cohort (as opposed to being all a bunch of individuals lumped together at random) then it is better to call them together. What you're seeing here is probably that you have some variants which don't look very good by themselves (the program is not confident that they are real -- possibly because the context is tricky as you describe), but occur in many of your samples. When the program sees that they are present in many samples, even at low confidence, this increases the overall confidence that those sites are variant. This is the strength of multisample calling.

    Geraldine Van der Auwera, PhD

Sign In or Register to comment.