Calling heterozygosity in NGS data

[andy11]

In genotyping from NGS data, what is the best way to account for bias (sequencing, mapping, etc) when identifying a locus as homozygous vs heterozygous?

Thanks for any help!

[Heisman]

Not quite sure what you mean by locus. Do you mean specific SNP calls? If so there are many programs out there to call SNPs and will attempt to correctly determine if they are homozygous or heterozygous.

As for bias, it's good to make sure to remove any PCR duplicates.

[andy11]

Yes, exactly. I was wondering what those programs do to determine if a locus is homozygous or heterozygous.

[Heisman]

SAMtools mpileup is pretty good. It's the only common one I'm really familiar with so I don't want to try to compare it to others in any way, but I know it's used fairly commonly.

[sdvie]

Yes, exactly. I was wondering what those programs do to determine if a locus is homozygous or heterozygous.

The programs that I know (samtools mpileup + bcf tools and GATK unified genotyper), basically count the number of reads in your experiment supporting one or another allele at a SNP position, and then they estimate whether the ratio of 1st allele to 2nd allele is 50 : 50 or 100 : 0 (usually one of the alleles corresponds to the reference sequence, and the other to the alternative sequence).

The field "AF" (allele frequency) in the vcf-format file gives you this information.

Hope that helps.

Cheers,
Sophia