Combining MAVEs and computational predictors improves variant classification across ancestries in hereditary cancer genes

ACMG/AMP-calibrated combination of MAVE evidence and population-free variant effect predictors

Many variant effect predictors show ancestry-related bias because they are trained on clinical or population datasets that under-represent global diversity. The authors compare classification based on multiplexed assays of variant effect (MAVEs) and on three population-free predictors (GEMME, EVE, CPT-1), calibrated to ACMG/AMP guidelines, across five cancer-associated genes with high-quality MAVE datasets. MAVEs are more conservative and decisive (fewer variants classified as pathogenic, fewer indeterminate classifications) and more sensitive in functionally assayed regions, whereas predictors identify a broader range of pathogenic variants. By combining the evidence strengths of both approaches, over 90% of variants of uncertain significance in the SG10K_Health and Mexico City Prospective Study reference datasets are reclassified as at least likely benign or likely pathogenic. Variants with discordant classifications are isolated to highlight method-specific limitations.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

The value is twofold: evidence that does not depend on the patient's ancestry, and a quantified complementarity between experimental data and prediction. The over-90% reclassification rate, however, applies to population reference datasets rather than patients, and the abstract neither names the five genes nor gives the number of variants involved. The approach remains limited to genes already covered by a high-quality MAVE.

Analysis by Dr Thibaut Benquey