The familiar five-tier classification of sequence variants provides a shared language: pathogenic, likely pathogenic, uncertain significance, likely benign and benign. These categories summarize evidence about a variant in a defined disease context. They do not, on their own, establish whether a particular person has the condition or specify that person’s clinical management.
Start with the right context
The 2015 ACMG/AMP consensus framework organizes population, computational, functional, familial and other evidence. Its application requires a relevant gene-disease relationship and an appropriate mechanism of disease. A predicted damaging change in a gene is not necessarily disease-causing: the consequence must fit how variation in that gene produces the condition under consideration. Read the foundational ACMG/AMP framework.
In practice, curators need to document both the evidence used and the evidence considered but rejected. A variant of uncertain significance reflects unresolved evidence, not a partial diagnosis. Clinical action should not be justified by treating that uncertainty as if it were a confirmed pathogenic finding.
Evidence strength has assumptions
Tavtigian and colleagues modeled the ACMG/AMP framework in Bayesian terms, clarifying how evidence of different strengths can change the odds of pathogenicity. This perspective helps explain why multiple weak observations are not interchangeable with one well-validated observation. It also highlights the importance of avoiding correlated evidence and understanding the assumptions behind numerical outputs. Read the Bayesian modeling study.
For example, several computational predictors may depend on overlapping training data or similar biological features. Agreement among them should not automatically be treated as several independent experiments. Likewise, an assay and a prediction concerning the same molecular event may require careful handling to avoid counting the same underlying evidence twice.
Make the conclusion auditable
- Specify the condition, inheritance pattern and relevant transcript.
- Record the evidence sources and the date they were assessed.
- State any applicable gene-specific interpretation rules.
- Explain unresolved contradictions rather than hiding them within a score.
Classification can change when new population, family or functional evidence appears. A good report makes that possibility understandable without suggesting that every result is unstable. The framework is most useful when it disciplines expert judgment: a transparent account of why the available evidence supports a category, with a separate clinical assessment of whether that category explains the case.