Is joint (combined genome+exome) FAF unreliable if either genome or exome fails filters?

I’m annotating my variants of interest using the published gnomad vcf’s. I intended to simply use the “fafmax_faf95_max_joint” INFO column subject to the FILTER column.

However, if the joint value includes both genome and exome data even if one fails filters, then I would want to look at only the “fafmax_faf95_max” column from the PASS data. Is my understanding correct?

@pgarg Thanks for posting this question, when creating the release data, we only filtered out chrM, AS_lowqual sites and AC_raw == 0 variants, but the FILTER column from variant QC (“PASS” or not) need to be interpreted with precaution by users, we would have some variants that didn’t PASS in both datasets due to different coverage.