Beyond a single threshold: calibration of CADD v1.7 and AlphaGenome for ACMG/AMP computational evidence across variant classes, model versions, and reference data

Calibrated CADD v1.7 and AlphaGenome AVI on rare ClinVar missense variants yield moderate pathogenic and strong benign evidence, with AVI giving final-direction evidence for 72.5% of resolved VUS.

• Why it matters: Computational ACMG/AMP evidence depends on properly calibrated score cutoffs, yet genome-wide predictors are often used with one threshold and AVI lacked calibration.
• What they did: Using a prespecified plan and the ClinGen calibration method, the team calibrated CADD v1.7 and AVI on rare ClinVar missense variants, then tested whether thresholds held for later classifications.
• The result: Thresholds held for variants classified after calibration; AVI outperformed CADD in each comparison and, for 1.2 million current missense VUS, agreed on pathogenic/benign evidence for four in ten and strong pathogenic evidence for 186 VUS.

AI summary of the abstract · may contain errors · not medical advice · read the paper.