gnomAD AF filters - when people use sub populations

[davmlaw]

Research use any populations other than Finnish/Ashkenazi Jewish - which means we can't use our gnomAD cohort genotype filters and get the speed boost.

We could make our gnomad AF division based on ANY of the sub populations, not sure how much bigger that would make it.

It doesn't appear to make much difference, maybe stops a few extras going in but they are almost always filtered out by eg damaging filters below - maybe we could get them to switch to using AF or popmax?

[davmlaw]

• Record what % of test VCF is rare/common
• Modify generate_gnomad_upload_preprocess to test whether any of the AFs are over 5
• Re-run import, compare data with previous runs

[davmlaw]

We want a "common" list of variants that has no rare variants in any sample. We need to decrease the "common" list from where it is with gnomad AF. Basically common should only be where ALL populations are > 5% - perhaps not including the tiny pops

Investigate how big this file will be and what the rare percent is for a sample and whether it gets too high. For HSS2008 trio it was 3.22%