Can I say variant called from GIAB confidence region from any sample is real?

Question

[Jian]

Hi all,

I have a question about GIAB confidence region! I know it is normally used for benchmarking. But when people do variant calling from their sample. Is it ok to say that variant called from the confidence region is real?

Thanks!

Answer 1

[Jeremy Leipzig]

Those are very specific to individual samples in GIAB. There are literally different confidence region files for HG002, HG003, and HG004 for instance.

[Jian]

Thanks for your reply.

what if a set of variants (GATK calls) from sample A can also be found from the HG002 high-confidence call set within the high-confidence region (V3). May I say that I am confident that these variants are real in sample A?

[Jeremy Leipzig]

The high-confidence regions are really geared toward defining the areas that were fair game for the benchmarking studies. They aren't supposed to be used for evaluating individual variant calls.

You'll know if you have low coverage (DP). You'll know if you have trouble mapping (MAPQ). These are all considered when computing the genotype quality (GQ).

Use your genotype quality score when talking about called variants for sample A.