Hi:
I'm running HaplotypeCaller on whole-genome data. I'm interested in only several thousand specific positions across the genome, all of which are in dbSNP. Some of these are indels, which makes me more inclined to use HaplotypeCaller. However, I want to know the quality/likelihood of the bases called at each position in downstream analysis. So, my goal is to obtain a VCF file containing a call at every dbSNP position, whether or not it's variant with respect to the reference. I need each associated with a likelihood from VQSR, with VQSLOD having the same meaning whether or not the site is variant.
It seems that I could do this with UnifiedGenotyper --output_mode EMIT_ALL_CONFIDENT_SITES. How can I replicate this with HaplotypeCaller? Second, is there a better way to achieve this? Should I not be using HaplotypeCaller at all?
So far, I ran HaplotypeCaller with --emitRefConfidence BP_RESOLUTION --dbsnp dbsnp.vcf, followed by GenotypeGVCFs --includeNonVariantSites --dbsnp dbsnp.vcf. However, the rsIDs aren't used to populate the ID column in the resulting VCF (is this a bug?), so I can't then remove non-dbSNP positions. Secondly, annotations aren't added to positions matching the reference.
Thanks in advance. Douglas