We are overlapping gnomAD annotations with genebass variants. I noted that among the missense variants, ~17% of gnomAD variants have missing REVEL_MAX annotations, ~5% have missing SIFT / polyphen annotations.
I wonder if the genomAD team have ideas on reasons behind these missingness. For example, how many missing annotations are due to filtering, and how many can be addressed by re-running these in-silico tools.
Hello,
For gnomAD v4 release, we applied a different to handle these in silico predictors, you can find the explanations on our Help page and if you want to see the detailed process for them, please refer to our code.
Specifically for REVEL, when the score for a variant is from an unmatched transcript (different version of Ensembl in REVEL and in gnomAD v4), we newly added supplementary files for these variants. You can find the README and resources here on our downloads page.
Hope it helps!