Predictor concordance for HUMSAVAR variants classified as uncertain significance (US). Each variant is evaluated independently by AlphaMissense (AM), VESM-3B, and VESPA. Scores are empirically calibrated pathogenicity probabilities (see Documentation).
The numerical score is the empirically calibrated pathogenicity probability for the variant. The associated call is the binary classification obtained by applying the decision threshold of 0.50 to that score. Red indicates P (high-effect) and green indicates B (low-effect). Score colours follow the corresponding P/B call; they do not represent generic numerical score ranges.
Calibrated: AM ≥ 0.50 → P · VESM-3B ≥ 0.50 → P · VESPA ≥ 0.50 → P
Raw-score equivalents of 0.50: AM 0.29 · VESM-3B −8.04 · VESPA 0.45
Consensus records the pattern of calls across the three predictors: 3P means all three independently give a high-effect call; 2P · 1B and 1P · 2B indicate predictor disagreement; and 3B means all three give a low-effect call. The calibration was derived from HUMSAVAR LP/P and LB/B variants and is descriptive of that dataset rather than an independent performance estimate. These predictor calls are computational annotations and should not be interpreted as clinical reclassification of the HUMSAVAR variant.