Empirically calibrated allele frequency thresholds for ACMG BA1, BS1 and PM2 evidence criteria

Empirical threshold calibration by gene-normalized kernel density estimation, stratified by inheritance mode and gene-level missense constraint

The ACMG/AMP criteria based on allele frequency (BA1, BS1, PM2) are still applied at fixed defaults while computational predictors have been systematically recalibrated. The authors stratified ClinVar missense variants annotated against gnomAD v4.1.1 by inheritance mode and gene-level missense constraint, then fitted gene-normalized kernel density estimates to pathogenic and benign variants in a sliding window along the constraint axis, placing thresholds where the likelihood ratio crossed ACMG/AMP evidence strengths at a prior of 0.0441. The derived thresholds varied systematically with constraint, differed between inheritance modes, and departed from the fixed defaults in both directions. On held-out genes, stratified cutoffs reached 96.7% accuracy versus 90.1% unstratified; restricted to the 73 ClinGen expert panel genes with autosomal dominant or recessive inheritance, they reached 91.0% accuracy at 69.5% variant coverage against 88.8% accuracy at 86.2% coverage for the panel-specified cutoffs.

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

This is the logical extension of in silico predictor recalibration: nothing justifies a single frequency threshold serving both a constrained dominant gene and a tolerant recessive one. The practical strength is the format — a lookup table deployable across thousands of genes no expert panel covers — but the accuracy gain is paid for in coverage (69.5% of variants versus 86.2%), mechanically increasing the number of variants left with no frequency evidence at all. A preprint: not to be wired into a classification workflow before peer review and independent validation.

Analysis by Dr Thibaut Benquey