Difference in allele count between ExAC and gnomad exomes

Question

[dolevrahat]

Hello

I have recently queried the allele count for some variants using both gnomAD and ExAC. I found that for some variants, the allele count was much greater in ExAC then in gnomAD genomes. For example, for the variant chr13:25670851:A:G (hg19 coordinates) ExAC reports 194 occurrences, while gnomad reports 18 occurrences in exomes.

I understand from the gnomaAD release notes that this difference may be due to differences in the variant calling algorithm between ExAC and gnomAD.

Is anyone aware of any other possible causes for this difference? For example is it possible that some exomes are included in ExAC but not in gnomAD? The release notes are not too explicit about this, although they do mention that some exomes belonging to closely related individuals were excluded from gnomAD.

Any further information will be most appreciated.

Thanks in advance!

Answer 1

[Kevin Blighe]

Given the large numbers of people being sequenced by these consortia and the way in which they tackle biases, there should not be huge differences in reported allele frequencies that are not related to technical artifacts. 'Technical artifacts', here, can relate to differences in analysis methods and / or problems during sequencing.

For the particular example that you've mentioned, it has the following warning on gnomAD:

Warning! This variant is found in a segmental duplication region (SEGDUP), which may indicate a low-quality site.

...on ExAc, it states that it has the following VCF FILTER tag:

VQSRTrancheSNP99.80to99.90

So, it's one to be questioned and not trusted.

Kevin