Skip to main content
In Development
🚀 Coming Soon! Seeking Testers

We're seeking organizational testers for early access to our AI-powered genomics platform.

Part 2 of 6
Variant Classification Series

The 28 ACMG/AMP Criteria: A Deep Dive into the Rules of Evidence

November 23, 2025
25 min read
RW

Ryan Wentzel

Founder & CEO, Humanome.AI

1. Introduction: The Periodic Table of Interpretation

If variant classification is a language, the 28 ACMG/AMP criteria are its alphabet. These codes—ranging from PVS1 to BP7—represent distinct lines of biological, clinical, and statistical evidence. Mastering them requires understanding not just their definitions, but their limitations, exceptions, and the specific contexts in which they apply.

Figure 1: The Evidence Landscape

PVS1
PS1
PS2
PS3
PS4
PM1
PM2
PM3
PM4
PM5
PM6
PP1
PP2
PP3
PP4
PP5
BA1
BS1
BS2
BS3
BS4
BP1
BP2
BP3
BP4
BP5
BP6
BP7

2. PVS1: The Heavy Hitter (Null Variants)

Definition: Null variant (nonsense, frameshift, canonical +/- 1 or 2 splice sites, initiation codon, single or multi-exon deletion) in a gene where LoF is a known mechanism of disease.

Critical Caveats (Tayoun et al., 2018)

  • Nonsense Mediated Decay (NMD): Does the variant occur in the last exon or the last 50bp of the penultimate exon? If so, it may escape NMD and result in a truncated protein rather than no protein. PVS1 may need to be downgraded to Strong or Moderate.
  • Splice Rescue: Does the splice site variant allow for the use of a cryptic splice site that preserves the reading frame?
  • Gene Mechanism: Is the gene actually haploinsufficient? Some genes cause disease only via dominant negative mechanisms (e.g., GJB2 dominant deafness vs recessive).

3. Strong Evidence (PS1-PS4)

PS1: Same Amino Acid Change

Same amino acid change as a previously established pathogenic variant regardless of nucleotide change. Example: c.123G>A (p.Val41Ile) is new, but c.123G>T (p.Val41Ile) is known pathogenic.

PS2: De Novo (Confirmed)

De novo (both maternity and paternity confirmed) in a patient with the disease and no family history. If paternity is not confirmed, this is downgraded to PM6.

PS3: Functional Studies

Well-established in vitro or in vivo functional studies supportive of a damaging effect on the gene or gene product. Note: Not all assays are created equal. A Western blot showing protein absence is strong; a yeast-2-hybrid assay might only be supporting.

PS4: Prevalence in Cases

The prevalence of the variant in affected individuals is significantly increased compared with the prevalence in controls. This is the "Case-Control" criterion.

4. Moderate Evidence (PM1-PM6)

These criteria carry less weight but are often the workhorses of classification.

  • PM1 (Hotspot): Located in a mutational hot spot and/or critical and well-established functional domain (e.g., the DNA binding domain of TP53).
  • PM2 (Absent in Controls): Absent from controls (or at extremely low frequency if recessive) in Exome Sequencing Project, 1000 Genomes Project, or ExAC/gnomAD. Update: As discussed in Part 1, ClinGen has recommended downgrading this to Supporting (PM2_Supporting) due to the expansion of databases.
  • PM3 (In Trans): For recessive disorders, detected in trans with a pathogenic variant.

5. Supporting Evidence (PP1-PP5)

Supporting evidence is the weakest tier. Multiple supporting criteria are usually needed to move the needle.

PP3: In-Silico Predictions

Multiple lines of computational evidence support a deleterious effect on the gene or gene product (conservation, evolutionary, splicing impact, etc.).

Tools: REVEL, CADD, SpliceAI, PolyPhen-2, SIFT.

Note: You cannot count PP3 multiple times for different algorithms. It is one piece of evidence, regardless of how many tools agree.

6. Benign Criteria

The benign criteria mirror the pathogenic ones but provide evidence against disease causality.

  • BA1 (Stand-alone): Allele frequency > 5%. (See our Population Databases post).
  • BS1 (Strong): Allele frequency is greater than expected for disorder.
  • BS2 (Observed in Healthy): Observed in a healthy adult individual for a recessive (homozygous), dominant (heterozygous), or X-linked (hemizygous) disorder, with full penetrance expected at an early age.
  • BP4 (In-Silico): Prediction algorithms suggest no impact.

7. Combining Rules: The Matrix of Truth

How do we turn these codes into a classification? The 2015 guidelines provide explicit logic:

ClassificationRequired Evidence Combinations
Pathogenic
  • 1 Very Strong (PVS1) + ≥1 Strong (PS) OR ≥2 Moderate (PM)
  • 1 Strong (PS) + ≥3 Moderate (PM)
Likely Pathogenic
  • 1 Very Strong (PVS1) + 1 Moderate (PM)
  • 1 Strong (PS) + 1-2 Moderate (PM)
  • ≥3 Moderate (PM)
Benign
  • 1 Stand-alone (BA1)
  • ≥2 Strong (BS)

Tags

ACMG Criteria
PVS1
Clinical Genetics
Education
Humanome.AI - Genomic Variant Intelligence Assistant