ACMG Variant Classification Interpreter
ACMG Variant Classification Interpreter
Turn counts of ACMG/AMP criteria into Pathogenic, Likely pathogenic, Uncertain significance, Likely benign or Benign. The headline is the Tavtigian point score that ClinGen expert panels now use; the 2015 Table 5 verdict is printed beside every result, because that is what the report in front of you was written against.
ACMG/AMP criteria to a class
Criterion counts → 5-tier classA nonsense variant in a gene where loss of function is the established disease mechanism, absent from population databases, and segregating with disease in one family: one very strong criterion (PVS1), one moderate and one supporting
The point scale, and where it comes from
− 8 × (BA1) − 4 × (benign strong) − 2 × (benign moderate) − 1 × (benign supporting)
- 1, 2, 4, 8
- supporting, moderate, strong and very strong. The doubling is not arbitrary: it comes from fitting an exponential to the odds of pathogenicity implied by the 2015 rules, so that one very strong is worth the same as two strong, four moderate or eight supporting
- OddsPath
- the odds of pathogenicity contributed per point, 350 to the power of one eighth, which is 2.0797. Very strong evidence is defined as an OddsPath of 350, and every other strength is a power of the same base
- posterior
- OddsPath × 0.10 ÷ (OddsPath × 0.10 + 0.90), starting from the framework’s prior probability of pathogenicity of 0.10. At 10 points that is 0.9941, at 9 points 0.9878, at 6 points 0.8999, at minus 6 points 0.00137 and at minus 7 points 0.00066 — which reproduces every published class boundary
- what the page cannot do
- decide whether a criterion applies. It counts what you tell it. PM2 at moderate rather than supporting, an uncalibrated functional assay counted at strong, or PP5 counted at all will each move the answer, and the page has no way of knowing
Worked example
A nonsense variant in a gene where loss of function is the established disease mechanism, absent from population databases, and segregating with disease in one family: one very strong criterion (PVS1), one moderate and one supporting
Very strong: 1 × 8 = 8 points
Moderate: 1 × 2 = 2 points
Supporting: 1 × 1 = 1 point
No benign evidence, so nothing is subtracted
8 + 2 + 1 = 11 points, which is 10 or more → Pathogenic
Checking the 2015 rules by hand: Table 5 Pathogenic (i)(c) is one very strong criterion with one moderate and one supporting. Met. The two methods agree, which they do for the great majority of combinations
Now take away the supporting criterion. Points fall to 10, still Pathogenic, and Table 5 (i)(b) needs two moderate criteria, which are not present — so the 2015 rules drop to Likely pathogenic via (i) of the likely pathogenic list while the points hold at Pathogenic. That one-step disagreement is exactly what the second line of the result is for
Points to class — and the national disagreement on the benign side
| Points | ClinGen / Tavtigian | ACGS (United Kingdom) | Posterior probability of pathogenicity |
|---|---|---|---|
| 10 or more | Pathogenic | Pathogenic | 0.99 or above |
| 6 to 9 | Likely pathogenic | Likely pathogenic | 0.90 to below 0.99 |
| 0 to 5 | Uncertain significance | Uncertain significance | 0.10 to below 0.90 |
| −1 to −5 | Likely benign | Likely benign | above 0.001, below 0.10 |
| −6 | Likely benign | Benign | 0.00137 — above the 0.001 line |
| −7 or fewer | Benign | Benign | 0.001 or below |
The 2015 combining rules, in full
| Class | Any one of these combinations |
|---|---|
| Pathogenic | 1 very strong AND (at least 1 strong, OR at least 2 moderate, OR 1 moderate and 1 supporting, OR at least 2 supporting) · OR at least 2 strong · OR 1 strong AND (at least 3 moderate, OR 2 moderate and at least 2 supporting, OR 1 moderate and at least 4 supporting) |
| Likely pathogenic | 1 very strong and 1 moderate · 1 strong and 1 to 2 moderate · 1 strong and at least 2 supporting · at least 3 moderate · 2 moderate and at least 2 supporting · 1 moderate and at least 4 supporting |
| Benign | 1 stand-alone (BA1) · OR at least 2 strong |
| Likely benign | 1 strong and 1 supporting · OR at least 2 supporting |
| Uncertain significance | None of the above is met, OR the benign and pathogenic criteria are contradictory |
Which criteria the ClinGen SVI has changed since 2015
| Criterion | What the SVI recommended | Effect here |
|---|---|---|
| PVS1 | A decision tree by predicted consequence — a nonsense variant in the last exon, or an in-frame deletion of a non-critical region, is not very strong | Count PVS1 at the weight the tree gives it, not automatically at very strong |
| PM2 | Downgrade to supporting for most genes | 2 points becomes 1 |
| PP3 / BP4 | Use calibrated computational thresholds, which can support supporting, moderate or strong weight in either direction | A prediction tool score is no longer automatically one supporting point |
| PP5 / BP6 | Withdrawn | Do not count them at all |
| PP1 / BS4 / PP4 | Segregation and phenotype specificity guidance, with likelihood-ratio thresholds for the weight | Count at the weight the likelihood ratio supports |
| PS2 / PM6 | Points for confirmed and assumed de novo occurrences, by phenotype specificity | Merged into one code, OBS_DNV, in the draft v4 standard |
Which framework is current, and why this page prints two answers
The 2015 ACMG/AMP guideline is still the published standard, and it is still what almost every clinical report is written against. But it has not stood still. The ClinGen Sequence Variant Interpretation working group has amended it one criterion at a time — PVS1 became a decision tree, PM2 was downgraded to supporting for most genes, PP3 and BP4 were recalibrated against real data, and PP5 and BP6 were withdrawn altogether — and gene-specific Variant Curation Expert Panels have layered their own specifications on top. A 2015 classification made strictly by the letter of the original paper would today be out of date in several specific ways.
The successor is real and is not yet published. It is called SVC v4.0 and it is a joint product of ACMG, AMP, CAP and ClinGen. At the time of writing it is in pilot: a multi-site pilot of the forthcoming standards was presented at the 2026 ACMG meeting and the document itself has not been issued. So the honest position, which this page states rather than hides, is that there is no formally adopted successor to classify against yet. What there is, and what v4 is built on, is the point system of Tavtigian and colleagues: a demonstration that the 2015 combining rules are very nearly a Bayesian classifier in disguise, and that giving supporting, moderate, strong and very strong the values 1, 2, 4 and 8 reproduces them. That point system is not a proposal any more. ClinGen Variant Curation Expert Panels have adopted it in their published criteria specifications, which is why the headline on this page is a point score.
The 2015 verdict is printed beside it on every result for a practical reason. A reader holding a report that says “Likely pathogenic (PM2, PP3, PP1)” needs to be able to check that reasoning on its own terms, and the criterion codes in v4 will not be the same codes — PS2 and PM6 are already merged into OBS_DNV in the draft. Two answers, both labelled, is more useful than one answer with the working hidden.
Where the two disagree, the disagreement is usually informative rather than a defect. Three patterns account for most of it. A single very strong criterion scores 8 points, which is Likely pathogenic, but matches no row of Table 5 at all, so the 2015 rules return Uncertain significance; several national guidelines including ACGS formalise this by requiring at least two criteria for any classification other than BA1. Two strong criteria score 8 points, which is Likely pathogenic, but Table 5 makes them Pathogenic outright. And when pathogenic and benign evidence are both present, Table 5 declares a contradiction and returns Uncertain significance whatever the weights, where the point scale simply subtracts. None of these is a rounding artefact; each is a real difference in how the two methods handle sparse or conflicting evidence, and the page shows you both so you can see which one you are relying on.
One thing the page deliberately does not do is decide whether a criterion applies. That is the whole of the difficulty and none of the arithmetic. If you count PM2 at moderate when your gene’s expert panel has downgraded it, or count an uncalibrated functional assay at strong, the total will be wrong and nothing here will notice. Related pages: the variant allele frequency calculator for reading an allele fraction that does not look germline, the NGS variant confirmation interpreter for whether a call needs orthogonal confirmation before it is reported at all, and the Hardy-Weinberg carrier frequency calculator and residual carrier risk calculator for the population-frequency side of BA1 and BS1.
Frequently asked questions
Has the 2015 ACMG/AMP variant classification guideline been replaced?
Not yet. A successor exists in draft — SVC v4.0, developed jointly by ACMG, AMP, CAP and ClinGen — and it is in pilot rather than published, with a multi-site pilot reported at the 2026 ACMG annual meeting. The 2015 guideline therefore remains the published standard, but it has been amended substantially by ClinGen Sequence Variant Interpretation working group recommendations on individual criteria (PVS1, PM2, PP3/BP4, PP1/BS4, PS2/PM6) and by gene-specific Variant Curation Expert Panel specifications. In practice, classifying strictly by the 2015 paper alone is no longer current, and classifying by v4 is not yet possible.
What is the ACMG points system and who uses it?
It assigns 1 point to supporting evidence, 2 to moderate, 4 to strong and 8 to very strong, with the same values negative on the benign side, and then reads the total against five bands: 10 or more Pathogenic, 6 to 9 Likely pathogenic, 0 to 5 Uncertain significance, minus 1 to minus 6 Likely benign, minus 7 or fewer Benign. It comes from Tavtigian and colleagues, who showed that these values reproduce the 2015 combining rules within a Bayesian framework with a prior probability of pathogenicity of 0.10. It is not a proposal: ClinGen Variant Curation Expert Panels have adopted it in their published criteria specifications, and the forthcoming v4 standard is built on it.
Why does the points answer sometimes differ from the 2015 rules?
Three situations account for nearly all of it. A single criterion, however strong, matches no row of Table 5, so one very strong criterion is 8 points — Likely pathogenic — but Uncertain significance under the 2015 rules; several national guidelines including ACGS require at least two criteria before any classification other than BA1. Two strong criteria are 8 points, again Likely pathogenic, but the 2015 rules make them Pathogenic. And when pathogenic and benign criteria are both met, the 2015 rules call that contradictory and return Uncertain significance whatever the weights, where the point total simply nets them off. This page prints both verdicts so the difference is visible rather than silent.
Is a variant of uncertain significance a weak positive?
No. It is a statement that the available evidence does not reach either threshold, and it should not drive clinical management — no surgery, no screening change, no cascade testing on the basis of a VUS alone. Most variants are VUS and most of them will eventually be reclassified as benign, because benign evidence accrues faster than pathogenic evidence once population databases grow. The draft v4 standard splits the VUS band into low, mid and high, which is likely to be the change readers notice most, and it does not change what a VUS means: not enough evidence.
How many points is BA1 worth?
Minus 8, which reaches the Benign band by itself. That matches its status in the 2015 rules, where BA1 is the only stand-alone criterion: an allele frequency too high to be consistent with the disease classifies the variant as Benign regardless of any other evidence. This page shows both, and adds a line to the result whenever BA1 has been counted, because it is the one criterion whose effect is not a matter of arithmetic.
Can I use this for copy number variants?
No. Copy number variants are classified by a separate joint ACMG/ClinGen technical standard with its own scoring, in which the score runs on a different scale entirely and the thresholds are 0.99 for pathogenic and minus 0.99 for benign rather than 10 and minus 7. Sequence variant points and copy number variant points are not interchangeable and must not be added together. For copy number work start from the copy number from log2 ratio calculator instead.
Related calculators
References
- Richards S, Aziz N, Bale S, Bick D, Das S, Gastier-Foster J, Grody WW, Hegde M, Lyon E, Spector E, Voelkerding K, Rehm HL. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet Med. 2015;17(5):405–424. doi:10.1038/gim.2015.30. Table 5, the rules for combining criteria, is the second verdict printed on every result here.
- Li Q, Wang K. InterVar: Clinical Interpretation of Genetic Variants by the 2015 ACMG-AMP Guidelines. Am J Hum Genet. 2017;100(2):267–280. doi:10.1016/j.ajhg.2017.01.004. The reference implementation of Table 5; its
classfy()function is what the 2015 expressions on this page were transcribed from, condition by condition, because the ACMG PDF of the original table could not be retrieved. - Tavtigian SV, et al. Fitting a naturally scaled point system to the ACMG/AMP variant classification guidelines. Hum Mutat. 2020;41(10):1734–1737. doi:10.1002/humu.24088 (PMID 32720330). Point values 1, 2, 4 and 8 and the five class bands.
- Tavtigian SV, Greenblatt MS, Harrison SM, et al. Modeling the ACMG/AMP variant classification guidelines as a Bayesian classification framework. Genet Med. 2018. The prior probability of pathogenicity of 0.10, the OddsPath of 350 for very strong evidence, and the posterior thresholds of 0.99, 0.90, 0.10 and 0.001 from which the point bands on this page were re-derived.
- ClinGen Sequence Variant Interpretation working group. Criteria Specification Registry, specification 1570648177 version 1.0.0: pathogenic codes 1, 2, 4 and 8 points and benign codes minus 1, minus 2, minus 4 and minus 8, with classification at 10 or more, 6 to 9, 0 to 5, minus 1 to minus 6 and minus 7 or fewer, the panel having “adopted the point-based combining scoring in Table 2 and Table 3 of PMID: 32720330”, which “replaces the combining rules listed below”.
- Association for Clinical Genomic Science. ACGS Best Practice Guidelines for Variant Classification in Rare Disease, 2020 (v4.01) and 2024. Same point values, but Likely benign minus 1 to minus 5 and Benign minus 6 or fewer; and “With the exception of BA1 stand-alone, a minimum of two criteria are required to classify a variant as (likely) benign or (likely) pathogenic”. The one row where this page and ACGS differ.
- Biesecker LG, et al. Piloting the forthcoming ACMG/AMP/CAP/ClinGen standards for sequence variant classification. Genet Med Open. 2026; abstract P593, ACMG Annual Clinical Genetics Meeting. The evidence that v4.0 is in pilot and not yet published.
- Abou Tayoun AN, et al.; ClinGen Sequence Variant Interpretation Working Group. Recommendations for interpreting the loss of function PVS1 ACMG/AMP variant criterion. Hum Mutat. 2018. Why PVS1 is not automatically very strong.
Medical Disclaimer: The tools and content provided here are for educational and reference purposes only. They are not intended to substitute for professional medical advice, diagnosis, or treatment. Clinical decisions should always be based on the comprehensive assessment of a qualified healthcare professional.