ClinGen SVI / VCEP Points — Bayesian Variant Classifier

Why use it

Gene-specific ClinGen VCEP specifications, the calibrated PP3/BP4 strengths from Pejaver et al., and the SVI recommendations on PVS1 and PM2 all express evidence at strengths the 2015 combining rules cannot represent. This calculator takes the strength as an input rather than assuming it, so a VCEP specification can be applied as written. It also makes the distance from a threshold visible — knowing a variant sits at 5 points rather than 9 tells you how much new evidence would change the answer.

About

A point-based implementation of ACMG/AMP variant classification in which each piece of evidence contributes a number of points determined by the strength at which it was applied, and the sum maps onto the five-tier classification. Pathogenic evidence scores 8 points at very strong, 4 at strong, 2 at moderate and 1 at supporting; benign evidence scores −4 at strong, −2 at moderate and −1 at supporting, with BA1 acting as a stand-alone benign filter. Tavtigian et al. derived these weights by modelling the 2015 combining rules as a naive Bayesian classifier and showed that the point thresholds reproduce the original rules almost exactly — while, unlike those rules, accommodating evidence applied at a modulated strength. That is what makes this the arithmetic ClinGen VCEPs actually work in: a gene-specific specification that says "apply PM2 at supporting" or "PS3 at moderate for this assay" has a natural home here and none in the 2015 tables.

When to use

• Applying a ClinGen Variant Curation Expert Panel specification that modulates criterion strengths for a specific gene
• Classifying a variant where PP3/BP4 strength comes from a calibrated predictor score (Pejaver et al. 2022) rather than a fixed supporting weight
• Applying the SVI recommendations on PVS1 (Abou Tayoun et al.) where the decision tree returns PVS1_Strong, _Moderate or _Supporting
• Cross-checking a rule-based ACMG/AMP 2015 classification, and understanding any disagreement between the two
• Judging how far a VUS sits from the Likely pathogenic or Likely benign threshold when prioritising variants for re-review
• Laboratory variant review, where the point total is a compact and auditable summary of the evidence

Formula

Pearls & pitfalls

• The identity of a criterion does not affect the arithmetic — only its strength does. Two moderate criteria score the same 4 points whether they are PM2 and PM5 or PS3_Moderate and PM1.
• BA1 is a stand-alone benign filter, not a −8 point item. When it applies the variant is benign regardless of what else was collected, which is why it is a separate switch here.
• ClinGen SVI recommends applying PM2 at SUPPORTING strength (1 point) rather than moderate for most genes. Entering it as moderate is the single most common way to inflate a total.
• PP3 and BP4 have calibrated strengths (Pejaver et al. 2022): depending on the predictor score, PP3 can reach moderate or strong, and BP4 can reach strong. Do not apply them at a flat supporting weight without checking the calibration.
• The PVS1 decision tree (Abou Tayoun et al. 2018) frequently returns something weaker than very strong — a null variant in the last exon, or in a transcript of uncertain relevance, is not automatically 8 points.
• Point thresholds reproduce the 2015 combining rules closely but not perfectly. Where the two disagree, the disagreement is informative: it usually means the evidence is sitting exactly on a boundary and deserves discussion rather than an automatic answer.
• A 0-point result is a genuine VUS with no evidence, and a 5-point result is a VUS one supporting criterion away from Likely pathogenic. Both are "VUS"; they are not the same clinical situation, which is why the total is reported alongside the tier.
• The system permits combinations the 2015 rules forbid — for example, opposing pathogenic and benign evidence that cancels to a middling total. A near-zero total built from strong evidence on both sides is a conflict to resolve by expert review, not a real "uncertain" answer.

Next steps

Pathogenic / Likely pathogenic: use for clinical decision-making and cascade testing, per gene-specific management guidelines. VUS: do NOT use for clinical decision-making and do NOT offer predictive testing to relatives; the point total shows how much additional evidence would move it. Likely benign / Benign: not contributory. Always record which criteria were applied and at which strength — the total is only reproducible if the inputs are documented, and a VCEP specification is only followed if the deviations from default strength are written down.

Evidence

Tavtigian SV, Greenblatt MS, Harrison SM, et al. Modeling the ACMG/AMP variant classification guidelines as a Bayesian classification framework. Genet Med. 2018;20(9):1054-1060. Point thresholds: Pathogenic ≥10, Likely pathogenic 6-9, VUS 0-5, Likely benign −1 to −6, Benign ≤−7. The framework is implemented in the ClinGen Pathogenicity Calculator (Patel RY et al., Genome Med. 2017;9:3) and is the arithmetic used by ClinGen Variant Curation Expert Panels.

References

1. Tavtigian SV et al. Modeling the ACMG/AMP variant classification guidelines as a Bayesian classification framework. Genet Med. 2018;20(9):1054-1060. — link
2. Patel RY et al. ClinGen Pathogenicity Calculator: a configurable system for assessing pathogenicity of genetic variants. Genome Med. 2017;9:3. — link
3. Richards S et al. Standards and guidelines for the interpretation of sequence variants. Genet Med. 2015;17(5):405-424. — link
4. Pejaver V et al. Calibration of computational tools for missense variant pathogenicity classification and ClinGen recommendations for PP3/BP4 criteria. Am J Hum Genet. 2022;109(12):2163-2177. — link
5. Abou Tayoun AN et al. Recommendations for interpreting the loss of function PVS1 ACMG/AMP variant criterion. Hum Mutat. 2018;39(11):1517-1524. — link