Evaluating ACMG Classification Software

Manually applying the ACMG/AMP 2015 evidence framework to every variant in a WES or WGS callset does not scale - a single exome can carry tens of thousands of candidate variants. Automated classification tools exist to apply the same evidence criteria consistently and show their reasoning, not to replace expert judgment.

• Evidence coverage: does the tool evaluate the full criteria set - PVS1, PS1-PS4, PM1-PM6, PP1-PP5, BA1, BS1-BS4 - or only a subset?
• Transparency: can you see exactly which criteria were applied to a given variant and why, or only the final P/LP/VUS/LB/B label?
• Independent cross-check: does the tool support comparing its output against a second, independently-implemented classification engine for discordant calls?
• Manual override: can a reviewer change a classification and record the reason, or is the automated output final?
• Update cadence: how often are underlying population frequency, clinical significance, and literature sources refreshed?

• No way to audit which evidence criteria drove a classification - a black-box verdict is hard to defend in a clinical report
• No manual override path - automated classification should support expert review, not bypass it
• Marketing language implying the tool makes a diagnosis rather than classifying a variant under ACMG/AMP evidence rules
• No distinction between research-use and clinical/diagnostic-grade validation status