Earlier this year a paper by Shah et al. in AJHG suggested that our allele frequency filtering approach “may be prone to removing pathogenic variants”. In this response, we show that this is incorrect, and demonstrate that our approach is robust biorxiv.org/content/early/… 1/5
Shah et al. reported 15 ClinVar variants that are ‘inappropriately filtered’ by our approach. We show that these are either (a) incorrect use of our approach e.g. filtering founder variants or singletons, (b) not high-confidence pathogenic alleles, or (c) very low penetrance. 2/5
The penetrance estimates for all variants with sufficient evidence to be classed as (Likely) Pathogenic, are well below the 50% threshold used to define the max population AF for each disease. Allowing for lower penetrance when defining max AF would retain these variants. 3/5
This analysis also shows how using strict frequency thresholds allows investigation of the genetic architecture of disease. We can identify these low penetrance alleles, flagging them for more nuanced reporting and tailored counselling. 4/5
Our original approach, for reference, was published in @GIMJournal last year nature.com/articles/gim20… @drjamesware @dgmacarthur @cureffi 5/5
