ExAC DNPs

[alexpenson]

DNPs composed of two ExAC SNPs are not annotated.

Solution: add these DNPs (and TNPs?) to the vcf.

[alexpenson]

http://exac.broadinstitute.org/region/6-41903782-41903782

[ckandoth]

Sample VCF where 1000g, NHLBI EVS, and ExAC allele freqs are only provided for the SNPs, but not for the DNP:

##fileformat=VCFv4.2
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
##FORMAT=<ID=AD,Number=G,Type=Integer,Description="Allelic Depths of REF and ALT(s) in the order listed">
##FORMAT=<ID=DP,Number=1,Type=Integer,Description="Read Depth">
#CHROM  POS ID  REF ALT QUAL    FILTER  INFO    FORMAT  TUMOR   NORMAL
6   41903782    .   A   C   .   .   .   GT:AD:DP    0/1:10,10:20    0/0:22,0:22
6   41903783    .   G   A   .   .   .   GT:AD:DP    0/1:11,10:21    0/0:21,1:22
6   41903782    .   AG  CA  .   .   .   GT:AD:DP    0/1:10,10:20    0/0:21,0:21

[ckandoth]

I figure I could make this part of this gist for VEP installation, where I already require the user to edit the ExAC VCF with bcftools. bcftools concat has a feature to --ligate phased VCFs by matching overlapping haplotypes... which would have been perfect iff everyone reported phased VCFs. 🙄

Really, we just need a simple script that merges adjacent ExAC SNPs with similar allele-counts. The ExAC VCF is already sorted by position. So a one-liner is possible.

[ckandoth]

I just figured out VEP's internal solution when reporting allele freqs from 1000g, ESP, or ExAC. They apparently report the freq of all non-reference alleles at the given POS, without trying to match the variant allele to the input. So in the above example, the DNP at position 41903782 would get assigned the allele freq of SNPs and/or indels at the same position. This would include the allele freq of the SNP at position 41903782.

[ckandoth]

Just hit a bottleneck. 6-41903782-A-C has AC=65104, and 6-41903783-G-A has AC=5264. And they are found in phase only in 2432 individuals. This phased ACs are only reported in the ExAC web pages, and is not in their downloadable VCF. Ideally, a future release will list multiallelic loci with separate allele counts for SNPs and MNPs. But right now, we just don't have the correct ACs for DNPs or MNPs.

So as a workaround, I will report the sum of ACs of all non-reference alleles at a given position, even if it doesn't match the input allele. Same as explained in my last comment here. So for an input of 6-41903782-AG-CA, the allele count reported will be the sum of ACs of 6-41903782-A-C and 6-41903782-A-T

[ckandoth]

This is now handled in 86e58e3