1. Introduction
The 2015 ACMG/AMP guidelines provided a necessary common language, but their "rule-based" logic (e.g., "1 Strong + 2 Moderate") has significant limitations. It treats evidence as categorical rather than continuous, leading to "cliff-edge" effects where a variant falls just short of a classification despite overwhelming evidence.
2. Limitations of the 2015 Rules
Consider a scenario where you have 6 pieces of "Supporting" evidence. Under the 2015 rules, this might not sum to "Pathogenic" because the rules explicitly ask for Strong or Moderate evidence. Yet, statistically, 6 independent lines of supporting evidence yield a posterior probability of pathogenicity >99%.
This rigidity led the ClinGen Sequence Variant Interpretation (SVI) working group to develop a more flexible, mathematically grounded approach.
3. The Tavtigian Framework (2018)
In 2018, Tavtigian et al. published a Bayesian framework that calibrated the ACMG criteria. They assigned an "Odds of Pathogenicity" (OddsPath) to each strength level:
- Very Strong: OddsPath = 350:1
- Strong: OddsPath = 18.7:1
- Moderate: OddsPath = 4.33:1
- Supporting: OddsPath = 2.08:1
By assuming a prior probability of 0.1 (10%), one can calculate the posterior probability using these odds.
4. The Point System
To make this usable in a clinical lab without complex calculators, the OddsPath values were converted into a simple additive point system.
ACMG Points
ClinGen SVI CompatibleBenign Points (Negative)
5. Classification Thresholds
Once you sum the points, the classification is determined by these thresholds:
| Classification | Point Range | Posterior Probability |
|---|---|---|
| Pathogenic | ≥ 10 | > 99% |
| Likely Pathogenic | 6 - 9 | 90% - 99% |
| VUS | 0 - 5 | 10% - 90% |
| Likely Benign | -1 to -6 | < 10% |
| Benign | ≤ -7 | < 0.1% |
6. Resolving Conflicts
What happens if you have PVS1 (+8 points) and BS1 (-4 points)?
The Conflict Rule
If a variant has both Pathogenic and Benign criteria applied, it defaults to VUS unless the conflict can be resolved.
Example: A frameshift variant (PVS1) is found at 1% frequency in gnomAD (BS1).
Resolution: The high frequency suggests the frameshift is not disease-causing (perhaps it's in a pseudogene or an exon not expressed in the relevant tissue). The PVS1 should likely be removed or the variant classified as Benign.
7. Case Study: BRCA1 c.68_69del
Let's apply the point system to the famous 185delAG founder mutation.
- PVS1 (+8): Frameshift leading to NMD.
- PS4 (+4): Significantly enriched in breast cancer cases vs controls.
- PP1 (+1): Cosegregates with disease in multiple families.
Total Score: 8 + 4 + 1 = 13 Points
Result: Pathogenic (>99%)