I am looking for some advice on hard filtering by allele balance.
I have 94 exomes that have been analyzed using the GATK pipeline per the current “Best Practices”. I did VQSR and filtered for “pass” variants. I also am only considering variants with a depth of at least 20 for downstream analysis. After doing association analysis in PLINK and sanger sequencing the variants with high scores, it had become clear that most of the significant SNPs are errors in genotyping. The sample is being called as 0/1 but it should be 0/0. I’ve looked back at the VCF files for these sites and I can tell that the genotypes “look wrong”. There will be sites with allele frequencies like 150/5 or even 45/0 and they are being called as heterozygotes. I am trying to find a way to filter out these genotypes before association testing. I would go cross-eyed if I just had to eyeball the VCF for every site. Someone suggested filtering by allele balance or BAF. I could get GATK to output allele balance across all samples for ABhet and ABhom and I also can have it calculate the allele balance on a per sample per site basis. Now that I have done this, it isn’t clear to me what thresholds I should use. I can tell I am going to lose a lot of genotypes but I guess no significant associations are better than sequencing error.
I’d appreciate any feedback from the community on this issue.
I have 94 exomes that have been analyzed using the GATK pipeline per the current “Best Practices”. I did VQSR and filtered for “pass” variants. I also am only considering variants with a depth of at least 20 for downstream analysis. After doing association analysis in PLINK and sanger sequencing the variants with high scores, it had become clear that most of the significant SNPs are errors in genotyping. The sample is being called as 0/1 but it should be 0/0. I’ve looked back at the VCF files for these sites and I can tell that the genotypes “look wrong”. There will be sites with allele frequencies like 150/5 or even 45/0 and they are being called as heterozygotes. I am trying to find a way to filter out these genotypes before association testing. I would go cross-eyed if I just had to eyeball the VCF for every site. Someone suggested filtering by allele balance or BAF. I could get GATK to output allele balance across all samples for ABhet and ABhom and I also can have it calculate the allele balance on a per sample per site basis. Now that I have done this, it isn’t clear to me what thresholds I should use. I can tell I am going to lose a lot of genotypes but I guess no significant associations are better than sequencing error.
I’d appreciate any feedback from the community on this issue.