The PM5 rule in the 2015 ACMG/AMP variant classification guidelines applies to missense variants at an amino acid position where a different missense variant is pathogenic. Experimental support for this moderate-level pathogenicity criterion is lacking. In this study, we investigated whether Multiplexed Assays of Variant Effect (MAVEs)-derived functional effect predictions for missense variants at the residue level support the PM5 rule used in clinical diagnostics. We extracted data from MAVE studies pertaining to 39 unique genes. We determined the probability that a missense variant at a given residue is classified as loss-of-function (LOF), when at least one other missense variant at the same residue is classified as LOF. We computed likelihood ratios of pathogenicity at each residue and assessed the degree to which variant characteristics (Grantham scores, in silico predictions) mediate the correlation of intra-residue MAVE functional effects scores. Across a total of 13,351 residues with at least one predicted LOF missense variant in the MAVE, 10,437 (78.2%) had at least one other missense variant predicted LOF and only 10.1% had all possible missense variants predicted LOF. Likelihood ratios would be high enough to qualify as supporting- or moderate-level evidence in the ACMG/AMP framework once the proportion of loss-of-function (LOF) missense variants at the corresponding residue position exceeded 67.5% and 81.2%, respectively. These cut-offs applied to only 30.8% and 21.0% of the residues across these 39 genes, respectively. Variant- and gene-level factors had modest impacts on these correlations. Proposed refinements to the 2015 variant classification rules for missense variant showed improved alignment with MAVE outputs. Data from MAVE studies do not support the widespread use of the PM5 pathogenicity criterion as moderate-level evidence for clinically classifying missense variants.
Dai et al. (Thu,) studied this question.