ACMG Variant Classification

The American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) published standardized guidelines in 2015 for the interpretation of sequence variants. This 5-tier classification system provides a consistent framework for categorizing variants based on the strength and type of available evidence, directly informing clinical management.

• 5-tier classification system:

  • Pathogenic (Class 5): Strong evidence that the variant causes disease. Actionable: can be used for clinical decision-making, predictive testing, and prenatal diagnosis.
  • Likely pathogenic (Class 4): Greater than 90% certainty of pathogenicity. Clinically actionable: treated similarly to pathogenic variants in most contexts.
  • Variant of uncertain significance (VUS, Class 3): Insufficient evidence to classify as benign or pathogenic. Should NOT be used for clinical decision-making. Requires follow-up and potential reclassification over time.
  • Likely benign (Class 2): Greater than 90% certainty that the variant does not cause disease. Generally not reported or acted upon clinically.
  • Benign (Class 1): Strong evidence that the variant is not disease-causing. Common polymorphisms typically fall here.

• Evidence categories (each with varying strength levels: supporting, moderate, strong, very strong):

  • Population data (PM2, BS1, BA1): Variant frequency in population databases (e.g., gnomAD). Rare variants get supporting evidence for pathogenicity; common variants (above disease-specific thresholds) are evidence for benign classification. BA1 = allele frequency >5% in any population is stand-alone benign.
  • Computational/predictive data (PP3, BP4): In silico tools (REVEL, CADD, SpliceAI) predict functional impact based on conservation, protein structure, and splicing effects.
  • Functional data (PS3, BS3): Well-established functional assays demonstrating a damaging or neutral effect on the gene/protein.
  • Segregation data (PP1, BS4): Co-segregation of the variant with disease in affected family members (or lack thereof).
  • De novo data (PS2, PM6): Confirmed de novo occurrence in a patient with the phenotype (with confirmed maternity and paternity) is strong evidence for pathogenicity.
  • Allelic/genotype data (PM3, BP2): Variant found in trans with a known pathogenic variant in a recessive disorder supports pathogenicity.
  • Prevalence in affected vs controls (PS4): Statistically significant enrichment in affected individuals compared to controls.

• Combining criteria: Variants are classified by tallying the number and strength of criteria met. For example, pathogenic requires: (1) 1 very strong + 1 strong, OR (2) 1 very strong + 2 moderate, OR (3) 2 strong, OR (4) 1 strong + 3 moderate, among other combinations.

• ClinGen Sequence Variant Interpretation (SVI) working group: Has published updated recommendations refining original ACMG/AMP criteria, including the Bayesian point-based system for more quantitative classification.

• Pathogenic/Likely pathogenic: Report to ordering clinician. Inform genetic counseling, cascade testing in family members, and management changes.
• VUS: Report but clearly state it should not be used to make clinical decisions. Encourage periodic follow-up for reclassification. Options include: segregation studies in family, functional studies, or waiting for additional case reports.
• Likely benign/Benign: Often not included in clinical reports, or mentioned only in supplementary data.
• Reclassification: Variants are reclassified as new evidence emerges. Laboratories have a responsibility to update classifications, and clinicians should recontact laboratories periodically for VUS results.
• Gene-specific guidelines: ClinGen expert panels develop modified ACMG criteria for individual genes (e.g., PTEN, RASopathy genes, hearing loss genes) with calibrated thresholds that improve classification accuracy.

• VUS is not "negative": A VUS does not mean the variant is benign. Patients and clinicians may misinterpret VUS results.
• Subjectivity remains: Despite standardization, different laboratories may classify the same variant differently. ClinVar discrepancies highlight this issue.
• Not designed for somatic variants: ACMG/AMP 2015 guidelines are for germline variants. Somatic variant interpretation uses the separate AMP/ASCO/CAP guidelines (Tier I–IV system).
• Does not address structural variants or CNVs directly: Separate ACMG guidelines exist for CNV interpretation (2020).
• Evidence may be limited for rare genes: Ultra-rare conditions may lack functional data, population frequency data, and published case reports.