little discrepancies gnomAD DB vs. website

Question

[cocchi.e89]

I found that the latest DB (downloadable here) have little discrepancies respect the website data displayed.

As example (exome data):

16-90162186-T-C DB allele frequency: 0.371754 vs website (https://gnomad.broadinstitute.org/variant/16-90162186-T-C) AF: 0.3775

12-133435716-T-G DB AF: 0.22511 vs website (https://gnomad.broadinstitute.org/variant/12-133435716-T-G) AF: 0.2241

and the same results for population and GENOMES data. Does anybody has got a clue?

Thanks a lot in advance for any help!

[Ram]

I think this is a known discrepancy. Maybe someone more familiar with gnomAD can explain exactly why this happens.

[finswimmer]

What do you mean by "DB"? The vcf files one can download?

Please notice that in gnomAD they divide the frequency in exomes and genomes. The data you are showing for "DB" looks like the exome data, where else the "website" data looks like the total (exome+genome) data.

fin swimmer

[cocchi.e89]

Yes VCF files downloaded, I’m well aware of it, I’m speaking about both (in example of exomes data) but is the same for genomes.

Answer 1

[tpoterba]

Looking at your 16-90162186-T-C example, the AF in the VCF/HT is 0.37175. The allele frequency listed under "exomes" on the browser at your link is 0.3718. These match.

The 0.3775 number is the total AF -- a weighted average cross the genomes and exomes.