The ACMG/AMP 2015 guidelines classify a genetic variant into one of five tiers by combining independent evidence criteria. Tavtigian et al. (2018) showed this rule set is equivalent to a Bayesian point system: each criterion carries a strength-of-evidence weight, and the weights sum to a total that maps onto a posterior probability of pathogenicity.
points = 8·PVS1 + 4·ΣPS − 4·ΣBS + 2·ΣPM
OddsPath = 350^(points / 8)
P(path) = OddsPath·Prior / (OddsPath·Prior + (1 − Prior))
Verified numerically against the published Tavtigian et al. (2018) worked cases: PVS1 alone (8 pts, prior 0.10) → P=0.975, inside the paper's "Likely Pathogenic" band; adding one strong criterion (12 pts) → P>0.99, "Pathogenic". The formula in this simulator (and its 3D sibling) reproduces those reference values exactly — no correction was needed.
- PVS1 — a "very strong" criterion (e.g. a null variant in a gene where loss-of-function is a known disease mechanism), worth 8 points.
- PS1–PS4 — "strong" pathogenic criteria (e.g. the same amino-acid change seen before as pathogenic), 4 points each.
- PM1–PM6 — "moderate" pathogenic criteria (e.g. a mutational hotspot, absent from population databases), 2 points each.
- BS1–BS4 — "strong" benign criteria (e.g. allele frequency too high for the disorder), −4 points each, pulling the balance toward benign.
- Prior P(pathogenic) — the pre-test probability that any variant in this gene is disease-causing before any evidence is applied; the 3D sibling fixes this at 10% (the ACMG-recommended default), this 2D version lets you drag it from 1%–50% to see how a rarer or more penetrant gene shifts every tier boundary.
The resulting posterior probability sorts variants into five clinical tiers: Pathogenic (P>0.99), Likely Pathogenic (0.90–0.99), Uncertain Significance / VUS (0.10–0.90), Likely Benign (0.001–0.10) and Benign (P<0.001) — the same boundaries a clinical laboratory reports to a physician. Drag inside the top panel to rotate the scale's viewing angle.