Entering edit mode
6.9 years ago
bitpir
▴
250
I have exome sequences from 6 families (~38 samples) of European descent with similar hereditary diseases (e.g. epilepsy). Each family is supposedly unrelated. Some samples are from healthy family members, others are infected. What's the best way to do joint genotyping for this type of analysis? Should I (i) do joint genotyping on all 38 samples, then split them into individual vcf afterwards (for variant analysis), or (ii) do joint genotyping only for samples in the same family (so 6 separate joint-genotypings)? Also, let's say my samples comprise of different ethnicities, is it still better to do joint-genotyping or it won't really improve the variant calling in this case? Thanks for your help!
All of them together. However, beware that you might loose singletons when using joint genotyping. R packages like VariantAnnotation can handle multi-sample-vcfs - depends on the tool though.
great! Thanks for the tip!
I'd join genotype every sample together, and retain them as a single vcf unless retaining them in one file doesn't make sense. Given that you don't need to separate them to analyze them separately, I don't see why they should be in separate VCFs.
Hi Ram, I don't need to analyze each of them separately, but was planning to analyze on per family basis as some genes could be dominant/recessive/compound hets, etc. Thanks for the advice!
No matter how samples are analyzed, it doesn't matter how many VCFs they are stored in. Calling heavily unrelated samples together can confound your results, but you're the best judge of how phenotypically alike you deem your samples.