Bug Bulletin: The recent 3.2 release fixes many issues. If you run into a problem, please try the latest version before posting a bug report, as your problem may already have been solved.

Joint/batch variant calling when too few variants per sample

nishnish Posts: 1Member

Hi there,

We are sequencing a set of regions that covers about 1.5 megabases in total. We're running into problems with VQSR -- VariantRecalibrator says there are too few variants to do recalibration. To give a sense of numbers, in one sample we have about 3000 SNVs and 600 indels.

We seem to have too few indels to do VQSR on them and have a couple of questions:

  1. Can we combine multiple samples to increase the number of variants, or does VariantRecalibrator need to work on each sample individually?

  2. If we do not use VQSR for indels, should we also avoid VQSR for the SNPs?

  3. The other question is whether joint or batch variant calling across several samples would help us in this case?

Thanks in advance!

Answers

  • Geraldine_VdAuweraGeraldine_VdAuwera Posts: 5,840Administrator, GATK Developer admin

    VQSR definitely works better on multisample cohorts than on single samples. For exomes we recommend processing at least 30 samples together. If you do not have enough samples in your cohort, you can get additional exomes from the 1000 Genomes project, call them together with your samples and then recalibrate them all together.

    If you're still having trouble with the indels, it's OK to do VQSR on the SNPs but hard-filter the indels, as long as you analyze the results separately.

    Geraldine Van der Auwera, PhD

Sign In or Register to comment.