Table 5 vs Bayesian Points — where the two ACMG combining systems agree, and where they don't

Every one of the classifier's 29 curated point-mutation fixtures, scored independently by both the original 2015 categorical rule table (Richards et al.) and the 2020 Bayesian points reformulation (Tavtigian et al.). Circles mark the Table 5 result, diamonds mark the Bayesian result — a line connects the two only where they disagree. Hover any marker for the variant ID, points, and reasoning.

Pathogenic Likely Pathogenic Uncertain Significance Likely Benign Benign
Four permanent, regression-tested divergences — CAPN3_c.1939G>T, DMD_SYNTH_PATHOGENIC_01, DMD_c.2302C>T, and DMD_c.8944C>T — all share the same root cause: one Very Strong criterion plus one Supporting/Moderate criterion clears the Bayesian point threshold, but that exact combination was never one of Table 5's hand-enumerated rules. A fifth instance of the same shape appeared at CAPN3_c.550del (batch 25) when real functional-assay evidence pushed it to 9 Bayesian points against a Table 5 VUS — and resolved itself three batches later (batch 28) when two real published homozygous observations gave it PM3 at Strong strength, satisfying Table 5's own rule directly. Both systems now agree: Pathogenic, 13 points — matching the variant's real ClinVar classification.