vcf2vcf genotyping using bam input

[fpbarthel]

One of the really nice features of vcf2vcf is the genotyping feature. I don't know any other software that really has this capability. I've tried using freebayes force-calling which is very fast but it doesn't work as expected (skips a handful of variants), and bcftools/samtools don't seem to offer this possibility without additional post-processing (as vcf2vcf essentially does)

However, I've found vcf2vcf to be too slow when genotyping a large number of variants. I wonder if this could be made more efficient by using samtools mpileup on a bed file covering all of the features in the VCF file and post-processing the results, rather than calling samtools mpileup separately for each variant in the VCF file?

Unrelated, but wondering why the DP tag is chosen over DP4 to output read depth?

[ckandoth]

Thanks. I agree - vcf2vcf's genotyping feature should be sped up with a BED file. I use a similar strategy to speed up samtools faidx to pull flanking bps. But that was easy since samtools faidx can take many regions in command-line.

Speeding up vcf2vcf genotyping will need to remain in my backlog. It will be a while till I can get to it. I'll leave this issue open. In the meantime, look at GetBaseCountsMultiSample. It accepts either VCF or MAF as input, and produces a MAF-like output file.

In vcf2vcf output, DP is for total depth. In other VCF specs, DP4 lists 4 values for fwd/rev read counts of REF/ALT alleles, but it does not work for multi-allelic ALTs. So mpileup uses ADF and ADR instead to represent fwd/rev read counts for all alleles.