TL;DR

If you are searching for the pathogenicity calculator for a variant, enter your variant definition into the input box at the top of the page or visit the main page to view examples, for example https://genebe.net/variant/hg38/chr6:160585140-T%3EG. Below is a draft of the implementation description.

Genebe ACMG Implementation

ACMG classification (ACMG Standards and Guidelines (2015)) is a widely used framework for interpreting and reporting the clinical significance of genetic variants. The framework includes criteria that help classify variants into five categories: Pathogenic (P), Likely Pathogenic (LP), Uncertain Significance (VUS), Likely Benign (LB), and Benign (B).
Genebe's ACMG Implementation is a reliable tool for clinical genetics research. It provides a comprehensive list of ACMG criteria for gene variant pathogenicity assessment.

Usage

The calculator is available on every Genebe variant page. To view the variant pathogenicity assignment, simply enter the variant description in the search box located at the top of the page. You can easily refine the verdict at a later time on the same page by setting ACMG entries on/off. Check variant description examples on a GeneBe homepage.
ACMG entries that could be guessed based on existing databases were precomputed. The implementation uses a variety of databases, including GnomAD, BayesDel, MetaRNN, and many more.
Every ACMG entry in Genebe calculator is editable, allowing for customization based on specific research needs. The overall score is computed dynamically, providing a real-time assessment of the pathogenicity of a given gene variant.

Implementation details

If a variant affects multiple transcripts, by default the one with the most disruptive effect is chosen. Then we choose the MANE transcript, canonical transcript and longest transcript. You can change transcript later using the UI.
Not every ACMG rule can be assigned automatically. For example, our calculator has no information about a variant's de novo status. In such cases, rules are not evaluated. However, users are encouraged to provide the information manually and trigger the rules. The computed score and overall pathogenicity of the variant will be updated accordingly. Similarly, the strength of every rule can be refined manually.
ACMG entries that were implemented include:

Pathogenic

PVS1: Null variant (nonsense, frameshift, canonical ± 1 or 2 splice sites, initiation codon, single exon deletions) in a gene where LOF is a known mechanism of disease.

Reported for variants with a High impact effect, according to VEP, with the exception of splice_donor_5th_base_variant, which is also considered High impact. If a variant affects a gene with an unknown Loss of Function mechanism for disease, the strength is limited to Strong. If a variant is located in the last exon and its distance to the end of the CDS is smaller than 2% of the CDS length, the strength is limited to Strong. If a variant is a splicing variant and splicing pathogenicity scores suggest a Benign effect, the strength is limited to Strong. Otherwise, a strength of VeryStrong is reported.
The decision on whether LOF is a known mechanism of disease in the considered gene is based on (1) the value of OeLOF statistics from the GnomAD project and (2) the number of ClinVar submissions with a Pathogenic assignment.

PS1: Same amino acid change as a previously established pathogenic variant regardless of nucleotide change.

The amino acid change caused by the variant is compared to the ClinVar database. If the same aminoacid and Pathogenic change is found and the rule is triggered, the strength is determined. For ClinVar entries with 2 or more stars, the strength is set to Strong. For 1 star, it is set to Moderate. In case of no stars, the strength is set to Supporting.

PS2: De novo (both maternity and paternity confirmed) in a patient with the disease and no family history.

Not evaluated. Genebe has no access to such information. Please set up manually.

PS3: Well-established in vitro or in vivo functional studies supportive of a damaging effect on the gene or gene product.

Not evaluated. Please set up manually. If you know publications that connects in vitro or in vivo functional studies supportive of a damaging effect on the gene or gene product please let us know.

PS4: Patient's phenotype or family history is highly specific for a disease with a single genetic etiology.

Not evaluated. Please set up manually. We don't yet support phenotypes in Genebe.

PM1: Located in a mutational hot spot and/or critical and well-established functional domain (e.g., active site of an enzyme) without benign variation.

Neighborhood analysis: Regions of the CDS within a distance of 15 amino acids are analyzed. If the number of rare missense or in-frame pathogenic variants in ClinVar is significantly higher than the number of benign variants, the rule is triggered.
Uniprot domains are also analyzed. If the number of rare pathogenic variants affecting the domain is significantly higher than the number of benign variants, the rule is triggered.

PM2: Absent from controls (or at extremely low frequency if recessive) in Exome Sequencing Project, 1000 Genomes Project, or Exome Aggregation Consortium.

For variants with high gnomAD coverage, the gnomAD Genomes and gnomAD Exomes allele frequency (AF) and allele count (AC) are analyzed. Genes with autosomal dominance, X-linked, and autosomal recessive inheritance patterns are analyzed with adjusted thresholds for AF and AC values. The AC of variants on X chromosomes is additionally analyzed.

PM3: For recessive disorders, detected in trans with a pathogenic variant.

Not evaluated. Please set up manually.

PM4: Protein length changes as a result of in-frame deletions/insertions in a nonrepeat region or stop-loss variants.

The RepeatMasker data is utilized to identify repeat regions, and the rule is applied to nonframeshift and stoploss variants.

PM5: Novel missense change at an amino acid residue where a different missense change determined to be pathogenic has been seen before.

The amino acid change caused by the variant is compared to the ClinVar database. If different, Pathogenic aminoacid change in the same amminoacid is found and the rule is triggered.

PM6: Assumed de novo, but without confirmation of paternity and maternity.

Not evaluated. Please set up manually.

PP1: Cosegregation with disease in multiple affected family members in a gene definitively known to cause the disease.

Not evaluated. Please set up manually.

PP2: Missense variant in a gene that has a low rate of benign missense variation and in which missense variants are a common mechanism of disease.

Value of misZ score from gnomAD is used for determination of genes in which missense variants are a common mechanism of disease.

PP3: Multiple lines of computational evidence support a deleterious effect on the gene or gene product (conservation, evolutionary, splicing impact, etc.).

According to the suggestions in https://doi.org/10.1186/s13073-022-01120-z we have chosen single scorer, MetaRNN. If MetaRNN annotation is not present, we use BayesDel. For splice site variants RF and ADA scores are used.

PP4: Patient's phenotype or family history is highly specific for a disease with a single genetic etiology, but the genetic etiology has not been established definitively.

Not evaluated. Please set up manually.

PP5: Reputable source recently reports variant as pathogenic, but the evidence is not available to the laboratory to perform an independent evaluation.

ClinVar database is searched for the analyzed variant. If found, strength is applied according to the number of stars.

Benign

BA1: Allele frequency is >5% in Exome Sequencing Project, 1000 Genomes Project, or Exome Aggregation Consortium.

Allele frequencies from projects gnomAD Genomes and gnomadExomes are taken into account. According to recomentations included in https://clinicalgenome.org/site/assets/files/3460/ba1exceptionlist07302018.pdf we don't call BA1 for these well known, common variants.

BS1: Allele frequency is greater than expected for disorder.

For genes with a lot of Pathogenic variants present in ClinVar we try to estimate the minimum AF level of pathogenicity detection. Otherwise the threshold of 0.015 is used. Only gnomAD Genomes and gnomadExomes allele frequences are considered.

BS2: 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

GnomAD Genomes and Gnomade Exomes Allele Counts and Homozygote Counts are used. For chromosome X the number of AC in male is used.

BS3: Well-established in vitro or in vivo functional studies show no damaging effect on protein function or splicing.

Not evaluated. Please set up manually.

BS4: Lack of segregation in affected members of a family.

Not evaluated. Please set up manually.

BP1: Missense variant in a gene for which primarily truncating variants are known to cause disease.

Based on the ClinVar data we compute fraction of benign and missense benign variants. If it's higher than the threshold for the gene, BP1 rule is called.

BP2: Observed in trans with a pathogenic variant for a fully penetrant dominant gene/disorder or observed in cis with a pathogenic variant in any inheritance pattern.

Not evaluated. Please set up manually.

BP3: In-frame deletions/insertions in a repetitive region without a known function.

Called for non frameshift variant in a repetitive region, defined by Repeat Masker.

BP4: Multiple lines of computational evidence suggest no impact on gene or gene product (conservation, evolutionary, splicing impact, etc.).

According to the suggestions in https://www.biorxiv.org/content/10.1101/2022.03.17.484479v1 we have chosen single scorer, MetaRNN. If MetaRNN annotationi is not present, we use BayesDel. For splice site variants RF and ADA scores are used.

BP5: Variant found in a case with an alternate molecular basis for disease.

Not evaluated. Please set up manually.

BP6: Reputable source recently reports variant as benign, but the evidence is not available to the laboratory to perform an independent evaluation.

We search for the variant in ClinVar. Rule is called if reported as Benign. Strength is set according to the number of stars in ClinVar: VeryStrong for more than 2 stars, Strong for 2, Moderate for 1 and Supporting value for 0 stars.

BP7: A synonymous (silent) variant for which splicing prediction algorithms predict no impact to the splice consensus sequence nor the creation of a new splice site AND the nucleotide is not highly conserved.

Nucleotide conservation is determined using the PhyloP100 score.