ACMG Methodology
Complete documentation of how Helena processes, annotates, and classifies genetic variants. Every threshold, database version, and classification rule used in production is documented on this page. This documentation is intended for clinical geneticists, laboratory directors, and accreditation auditors.
Variant classification follows the ACMG/AMP 2015 framework (Richards et al., Genetics in Medicine, 2015) implemented through the Bayesian point-based system (Tavtigian et al., Hum Mutat. 2018;39(11):1485-1492. PMID: 30311386) with BayesDel ClinGen SVI calibrated thresholds (Pejaver et al., Am J Hum Genet. 2022;109(12):2163-2177. PMID: 36413997) and SpliceAI integration aligned to ClinGen SVI 2023 recommendations (Walker et al., Am J Hum Genet. 2023;110(7):1046-1067. PMID: 37352859). Optional ClinGen VCEP gene-specific specification overlay is available for approximately 50-60 genes. Classification is strictly evidence-based. No machine learning model determines variant pathogenicity.
Contents
Pipeline Overview
Six-stage processing pipeline transforms a raw VCF file into fully classified variants. Total execution time under 15 minutes for a whole genome (~4M variants) on dedicated hardware.
VCF Parsing
~60sStandard VCF file parsed into columnar in-memory database. Multi-allelic handling, genome build detection (GRCh38 required).
Quality Filtering
~5sConfigurable quality, depth, and genotype quality thresholds. ClinVar-listed pathogenic variants are protected from filtering.
VEP Annotation
~3-4 minEnsembl Variant Effect Predictor for consequence, impact, transcript, and protein annotations. Parallel processing across chromosomes.
Reference DB Annotation
~5-10sPopulation frequencies, clinical significance, functional predictions, gene constraint, phenotype associations, and dosage sensitivity loaded from 16 reference databases.
ACMG Classification
<1sSQL-based ACMG/AMP 2015 classification using Bayesian point framework (Tavtigian et al. 2018). 19 automated criteria evaluated with calibrated evidence strength, point-based classification thresholds applied, continuous confidence scores assigned. Optional VCEP gene-specific overlay for ~50-60 genes. Homozygous reference genotypes (hom_ref) from multi-allelic sites excluded from classification.
Export
~5sGene-level summaries exported for streaming. Classified variants persisted to analytical database for downstream services.
Maximum Sensitivity Approach
Helena classifies all variants that pass quality filtering. There is no frequency-based or impact-based pre-filtering at any stage. A common variant (e.g., gnomAD allele frequency 40%) is still classified -- it will receive a Benign classification via BA1, but it is not silently discarded before classification. This design ensures that no variant is excluded from clinical review by an automated filter. The geneticist decides clinical relevance based on the complete classification and annotation data.
Reference Databases
All reference data is stored locally on EU-based infrastructure. No variant data is sent to external APIs during processing. Database versions are fixed per deployment and documented here.
gnomAD
Population allele frequencies (global and population-specific), allele counts, homozygote counts
Used by: BA1 (allele frequency > 5%), BS1 (elevated frequency), PM2 (absent in controls), BS2 (homozygote count)
Source: gnomad.broadinstitute.org
ClinVar
Clinical significance assertions, review star levels, disease associations, submitter information
Used by: PS1 (known pathogenic), PP5 (reputable source pathogenic), BP6 (reputable source benign), ClinVar override logic, quality filter rescue
Source: ncbi.nlm.nih.gov/clinvar
dbNSFP
Functional impact predictions from multiple algorithms, conservation scores
Used by: PP3/BP4 primary tool: BayesDel_noAF with ClinGen SVI calibrated thresholds (Pejaver et al. 2022). Display predictors: SIFT, AlphaMissense, MetaSVM, DANN, PhyloP, GERP (available for clinical review, not used in classification logic)
SpliceAI
Splice impact predictions (4 delta scores: acceptor gain, acceptor loss, donor gain, donor loss)
Used by: PP3_splice (max score >= 0.2), BP4 guard (max score < 0.1), BP7 (synonymous splice check)
Source: Illumina / Ensembl
gnomAD Constraint
Gene-level constraint metrics indicating tolerance to loss-of-function and missense variation
Used by: PVS1 (LoF intolerance), PP2 (missense constraint), BP1 (LoF tolerance)
Source: gnomad.broadinstitute.org
MANE Select Transcripts
Canonical transcript definitions with CDS start/end coordinates for each protein-coding gene. MANE Select is the single representative transcript per gene agreed upon by NCBI and Ensembl. MANE Plus Clinical covers additional clinically relevant transcripts for 66 genes.
Used by: PVS1 MANE positional guard (v3.22.0, HELIX-CR-2026-055): variants outside MANE Select CDS do not receive PVS1 or PM4. Exon-level pext aggregation (CR-056): pext scores computed per MANE Select CDS exon.
gnomAD pext (Proportion Expressed)
Per-exon expression proportion across 49 GTEx v10 tissues. For each base in the coding sequence, pext measures what proportion of the gene total expression passes through that position. Aggregated to exon level as mean_pext (arithmetic mean across tissues) and max_tissue_pext (maximum across 49 tissues). Max tissue pext used for classification to avoid false negatives for tissue-specific genes.
Used by: PVS1 expression-aware guard (v3.23.0, HELIX-CR-2026-056): max_tissue_pext >= 0.9 = PVS1 Very Strong, 0.1-0.9 = PVS1 downgraded to Strong, < 0.1 = PVS1 blocked. PM4 binary guard (blocked at < 0.1). Scientific basis: Cummings 2020 PMID:32461655.
Source: gnomAD / Broad Institute
HPO (Multi-Source Enriched)
Gene-to-phenotype associations aggregated from 6 curated sources: HPO Consortium (5,173 genes), Orphanet disease-to-HPO (3,176 genes with frequency data), DECIPHER G2P 7 clinical panels (2,125 genes), Monarch Initiative (4,791 genes), ClinVar-MedGen P/LP chain mapping (5,258 genes), and manual clinical curation. Source priority ordering with confidence-based filtering.
Used by: PP4 (patient phenotype matching), Phenotype Matching Service (semantic similarity scoring), Screening Service (gene clinical breadth proxy)
Source: HPO Consortium + Orphanet + G2P + Monarch + ClinVar-MedGen
ClinGen
Dosage sensitivity scores (haploinsufficiency, triplosensitivity)
Used by: PVS1 (haploinsufficiency_score = 3 as constraint gate fallback for X-linked genes, v3.11.5), BS1 (inheritance-aware frequency threshold proxy), BP2 (trans with pathogenic in recessive)
Source: clinicalgenome.org
Orphanet / Orphadata
Gene-disease-inheritance mode annotations for rare diseases (autosomal dominant, autosomal recessive, X-linked dominant, X-linked recessive)
Used by: BS1 (inheritance-aware frequency threshold: Orphanet AD/XLD -> 0.1%, Orphanet AR -> 5%), PVS1 disease association gate (gene must have Orphanet entry for PVS1 to apply)
Source: orphadata.com (INSERM, France)
ClinGen VCEP Specifications
Gene-specific ACMG criteria thresholds from ClinGen Variant Curation Expert Panels
Used by: BA1, BS1, PM2 (gene-specific frequency thresholds), PVS1 (applicability gate for gain-of-function genes), PVS1 disease association gate
Source: cspec.genome.network
DECIPHER G2P
Gene-disease associations from 7 curated clinical panels: DD (developmental disorders), Eye, Cardiac, Skin, Skeletal, Cancer, Ear. Each entry includes allelic requirement, molecular mechanism (gain of function, loss of function, dominant negative), and confidence level (definitive, strong, moderate). Molecular mechanism data used for PVS1 GoF/DN guard (v3.17.0).
Used by: HPO enrichment pipeline (source 3 of 6, v3.12.0). PVS1 GoF/DN guard: gof_genes_unified view provides 348 pure GoF/DN/GoE monoallelic genes where PVS1 is blocked (v3.18.0, HELIX-REF-003). Sources: G2P (199 genes), GoFCards (179 genes), manual curation (27 genes). Dual-mechanism genes excluded from view to preserve PVS1 eligibility for LoF phenotypes.
Monarch Initiative
Gene-phenotype associations aggregated from HPO Consortium redistribution via Monarch Knowledge Graph. Broad coverage of gene-disease relationships from multiple biomedical ontologies.
Used by: HPO enrichment pipeline (source 4 of 6). Large-scale gene coverage complements primary HPO Consortium data.
ClinVar-MedGen HPO Mapping
Gene-phenotype associations derived from ClinVar P/LP variant submissions mapped through MedGen CUI to HPO terms. Chain: ClinVar P/LP variant -> gene -> disease (MedGen CUI) -> HPO terms. Largest single contributor of new genes to the HPO enrichment pipeline.
Used by: HPO enrichment pipeline (source 5 of 6). Covers 5,258 genes including many not present in primary HPO Consortium or Orphanet datasets.
Source: NCBI ClinVar + MedGen
InterPro / Pfam
Comprehensive protein domain classification catalog from Pfam. Full Pfam entry set loaded into reference database with domain accession, type (domain/family/repeat/motif), and functional annotation. Critical functional domains marked for PM1 criterion (ACMG 2015: variant in mutational hotspot or well-established functional domain).
Used by: PM1 (critical functional domain detection via is_critical flag, v3.15.0). 49 domains marked as critical across 14 categories: kinase, DNA binding, ion channel, GTPase, protease, signaling, tumor suppressor, calcium binding, structural, receptor, hearing, nuclear hormone receptor, ubiquitin, RNA binding.
Source: EMBL-EBI InterPro
ClinGen Gene-Disease Validity
Gene-disease validity classifications (Definitive, Strong, Moderate, Limited) with mode of inheritance for autosomal dominant, autosomal recessive, and X-linked disorders. Curated by ClinGen Gene Curation Expert Panels.
Used by: PVS1 constraint gate (AD Definitive/Strong genes bypass pLI/LOEUF, v3.16.6). PVS1 AR LoF gene bypass (AR Definitive/Strong with verified LoF mechanism, v3.16.0). Disease association gate for PVS1, PP3_Strong, LP4/LP5/LP6, and ClinVar override (v3.16.0).
Source: clinicalgenome.org
GoFCards
Curated gain-of-function variant database with experimental evidence (animal model, cell model), pathways, disorder associations, and proprietary evidence score (Pscore). Each variant entry includes PMID, functional description, and experimental validation status.
Used by: PVS1 GoF guard via gof_genes_unified view (v3.18.0, HELIX-REF-003). Gene-level aggregation: genes with 3+ curated GoF variants, average Pscore >= 2.0, and at least one variant with animal or cell model evidence qualify as GoF with confidence "strong". 217 genes passed threshold and are integrated into gene_disease_mechanism table.
ClinGen Dosage Sensitivity (Full)
Full dosage sensitivity curation including haploinsufficiency score (0-3, 30, 40), triplosensitivity score, up to 6 PMIDs per gene, date last evaluated, and disease IDs. Replaces the legacy 3-column clingen table (v3.18.0). HI scores: 0 = no evidence, 1 = little evidence, 2 = emerging evidence, 3 = sufficient evidence, 30 = AR phenotype, 40 = dosage sensitivity unlikely.
Used by: PVS1 constraint gate (HI=3 as fallback for X-linked genes, v3.11.5). Disease association gate (HI IN 2, 3, 30 per HELIX-CR-2026-025, v3.18.1). BS1 inheritance proxy. BP2 (HI=30 for AR phenotype).
Source: clinicalgenome.org
Ensembl VEP
Variant consequence prediction, protein impact, transcript annotation, functional domain mapping
Used by: PVS1 (consequence type), PM1 (Pfam domains), PM4 (in-frame indels), BP3 (non-critical regions), BP7 (synonymous), all impact-based criteria
Source: ensembl.org
ACMG/AMP Classification
Variant classification follows the 2015 ACMG/AMP guidelines with 28 evidence criteria evaluated systematically. 19 criteria are fully automated; 9 require manual curation by the reviewing geneticist.
Classification Priority Order
Classification logic is applied in strict priority order. Higher-priority rules are evaluated first, and the first matching rule determines the final classification:
BA1 Stand-alone
Allele frequency > 5% is always classified Benign. BA1 is the only stand-alone criterion in the ACMG framework and cannot be overridden by any other evidence, including ClinVar assertions.
Conflicting Evidence
If a variant has pathogenic evidence at moderate strength or above (PVS, PS, or PM criteria triggered) AND strong benign evidence (BS criteria triggered), the variant is classified as VUS regardless of the individual evidence strength. This is a conservative approach that prioritizes clinical safety.
ClinVar Override
ClinVar classification is applied only when no conflicting computational evidence exists. Requires minimum review star level (default: 1 star). ClinVar VUS does not override computational classification.
Bayesian Point System
Each triggered criterion contributes points based on its evidence strength: Very Strong (+8), Strong (+4), Moderate (+2), Supporting (+1) for pathogenic; Strong (-4), Supporting (-1) for benign. Total points determine classification: >= 10 Pathogenic, 6-9 Likely Pathogenic, 0-5 VUS, -1 to -5 Likely Benign, <= -6 Benign. This system is mathematically equivalent to the original 18 ACMG combining rules while filling gaps for evidence combinations not explicitly covered.
Default
Variants that do not meet any of the above criteria are classified as Uncertain Significance (VUS).
Classification Output
Each variant receives one of five standard ACMG classifications, a list of all criteria that were triggered with evidence strength levels (e.g., "PVS1,PM2,PP3_Strong"), a Bayesian point total, and a continuous confidence score derived from the distance between the point total and the nearest classification boundary:
Pathogenic
>= 10 pts
confidence: 0.80-0.99
Likely Pathogenic
6-9 pts
confidence: 0.70-0.90
VUS
0-5 pts
confidence: 0.30-0.60
Likely Benign
-1 to -5 pts
confidence: 0.70-0.90
Benign
<= -6 pts
confidence: 0.80-0.99
Automated Criteria (19 of 28)
These criteria are evaluated automatically for every quality-passing variant. Exact conditions and thresholds are documented below. Each criterion lists the databases it depends on and known limitations.
Pathogenic Evidence
Null variant in gene where loss-of-function is a known disease mechanism
Conditions
Impact = HIGH
Consequence: frameshift, stop_gained, splice_acceptor, or splice_donor variant
Gene constraint: pLI > 0.9 OR LOEUF < 0.35 OR gene is in the curated autosomal recessive LoF gene list (~150 genes with Definitive/Strong evidence for biallelic LoF disease mechanism) OR ClinGen haploinsufficiency_score = 3 OR gene is in the curated autosomal dominant ClinVar LOF gene list (AD genes with ClinVar P/LP LOF variants at 2+ review stars but uninformative gnomAD constraint, v3.13.0)
Last-exon NMD downgrade (v3.9.0): Truncating variants in the last exon are downgraded from PVS1 Very Strong (+8 points) to PVS1_Strong (+4 points). Last-exon variants escape nonsense-mediated mRNA decay (NMD) because there is no downstream exon-exon junction. The resulting truncated protein may retain partial function or exert a dominant-negative effect. Detection uses the VEP exon_number column (format X/Y): last exon when X == Y. Single-exon genes (1/1) correctly receive PVS1_Strong. NULL or missing exon_number defaults to Very Strong (conservative). Reference: Abou Tayoun et al. 2018, Figure 1, Node 3.
MANE Select positional guard (v3.22.0, HELIX-CR-2026-055): Variants outside the MANE Select/Plus Clinical CDS region do not receive PVS1 or PM4. VEP may annotate variants against non-canonical transcripts that extend beyond the canonical coding sequence, producing false loss-of-function predictions. 19,354 MANE transcripts loaded with CDS coordinates. NULL MANE data = PVS1 applied (conservative fallback). Reference: Abou Tayoun et al. 2018, PMID:30192042.
Expression-aware guard (v3.23.0, HELIX-CR-2026-056): Per-exon pext scores from gnomAD v4.1 (GTEx v10, 49 tissues) modulate PVS1. Three tiers based on max_tissue_pext: >= 0.9 = PVS1 Very Strong (unchanged), 0.1-0.9 = PVS1_Strong[pext] (downgraded to Strong), < 0.1 = PVS1 blocked (exon not expressed in any tissue). NULL pext = Very Strong (conservative fallback). PM4 blocked at pext < 0.1. Scientific basis: Cummings 2020 (PMID:32461655) -- 22.8% false pLoF filtered, < 4% true pathogenic lost.
Exclusions
NMD-escaping transcripts (consequence contains NMD_escaping_variant). Note: NMD_transcript_variant is NOT excluded -- it indicates the transcript undergoes NMD, which is evidence FOR loss-of-function (v3.11.3).
Non-canonical splice site variants: splice_donor_5th_base_variant (position +5), splice_donor_region_variant (positions +3 to +8), splice_acceptor_5th_base_variant (position -5). ClinGen PVS1 Decision Tree specifies canonical +/-1,2 only. Extended splice variants retain PP3_splice access (v3.11.4).
Gain-of-function / dominant-negative / gain-of-expression genes: multi-source guard (v3.18.0, HELIX-REF-003). Primary: gof_genes_unified view (348 pure GoF/DN/GoE monoallelic genes from G2P + GoFCards + manual curation, confidence definitive/strong/moderate). Fallback: GOF_AD_GENES curated list (29 genes including ANKRD26, C1QTNF5, CALM2, MYH3, PRKAG2, SKI, TUBB3 and 22 others with PMID-documented non-LoF mechanism). Dual-mechanism genes (e.g., SCN5A, LMNA, KCNH2, MYH7) excluded from view and retain PVS1 eligibility. ClinGen PVS1 Decision Tree (Abou Tayoun 2018): "Is LoF a known mechanism of disease?" -- if NO, PVS1 not applicable.
Gain-of-function gene guard history: v3.16.8 (HELIX-CR-2026-022) introduced GOF_AD_GENES (36 curated genes). v3.17.0 (HELIX-CR-2026-023) added G2P as primary source, reduced GOF_AD_GENES to 14 fallback genes, and fixed GNAS dual-mechanism error.
Stop-retained and stop-lost variants
HLA gene family (HLA-A, HLA-B, HLA-C, HLA-DRA, HLA-DRB1, HLA-DRB5, HLA-DQA1, HLA-DQB1, HLA-DPA1, HLA-DPB1, HLA-E, HLA-F, HLA-G, HLA-DMA, HLA-DMB, HLA-DOA, HLA-DOB)
Homozygous reference genotypes (hom_ref) from multi-allelic VCF sites
Databases: VEP (consequence, impact), gnomAD Constraint (pLI, LOEUF), curated AR LoF gene list (ClinGen Gene-Disease Validity), curated AD ClinVar LOF gene list (v3.13.0), Orphanet (gene-disease association), ClinVar (gene-level P/LP check), ClinGen (haploinsufficiency_score), VCEP (gene coverage), gof_genes_unified view (348 GoF/DN/GoE genes from G2P + GoFCards + manual, v3.20.1), GOF_AD_GENES fallback (29 genes, v3.20.1), gene_disease_mechanism table (disease association gate, v3.18.2), MANE Select transcripts v1.4 (positional guard, v3.22.0), gnomAD pext v4.1 GTEx v10 (expression-aware guard, v3.23.0)
Known limitations
Does not evaluate reading frame rescue via downstream in-frame reinitiation
Tissue-specific expression is assessed via gnomAD pext (proportion expressed across transcripts, GTEx v10) for PVS1 and PM4 modulation (v3.23.0, HELIX-CR-2026-056). Exons with max tissue pext < 0.1 block PVS1; 0.1-0.9 downgrade PVS1 to Strong. Alternative transcript usage addressed via MANE Select positional guard (v3.22.0, HELIX-CR-2026-055): variants outside MANE Select CDS do not receive PVS1 or PM4.
VCEP gene-specific PVS1 applicability gate available for ~50-60 genes (e.g., PVS1 disabled for gain-of-function genes like MYOC). Generic GoF/DN/GoE guard via gof_genes_unified (347 genes from G2P + GoFCards + manual curation) + GOF_AD_GENES fallback (28 genes) covers all other known non-LoF mechanism genes (v3.18.0, HELIX-REF-003). Generic thresholds used for all other genes.
Autosomal recessive gene bypass uses a curated list of ~150 genes with Definitive/Strong ClinGen validity or equivalent published evidence for biallelic LoF disease mechanism. VCEP AR genes (e.g., CDH23, GJB2, MYO7A, PAH, SLC26A4, USH2A) are handled via the VCEP pvs1_applicable gate and are not duplicated in the curated list. Categories covered: neurodegeneration, metabolic (amino acid, fatty acid oxidation, glycogen storage, lysosomal, peroxisomal), ciliopathies, sensory (hearing, vision), immune/hematologic, cardiac, neuromuscular, connective tissue, kidney, endocrine, liver, and respiratory.
Autosomal dominant ClinVar LOF gene bypass (v3.13.0) uses a curated list of AD genes with ClinVar P/LP LOF variants (2+ review stars) but uninformative gnomAD constraint (pLI < 0.9, LOEUF >= 0.35, HI != 3). Targets small genes where gnomAD constraint is statistically underpowered. New genes added only with clinical trigger and verification. Reference: Abou Tayoun 2018 PMID:30192042, HELIX-CR-2026-010.
Same amino acid change as an established pathogenic variant
Conditions
ClinVar clinical significance: Pathogenic, Pathogenic/Likely_pathogenic, or Likely_pathogenic
ClinVar review stars >= 2
Databases: ClinVar (clinical_significance, review_stars)
Known limitations
Matches on exact variant position and allele, not amino acid position (PM5 amino acid-level matching is disabled)
ClinVar assertions may lag behind current evidence for recently reclassified variants
Located in a mutational hot spot or well-established functional domain
Conditions
Consequence: missense_variant (v3.20.1, HELIX-CR-2026-035). PM1 is defined for missense variants in critical functional domains (Richards 2015 Table 3, ClinGen SVI Walker 2023 Section 3.3). Synonymous, frameshift, splice, and other non-missense variants do not receive PM1.
Variant overlaps a critical functional domain from interpro_pfam_domains reference table (~49 domains marked is_critical=true, sourced from InterPro Pfam catalog, v3.15.0)
Databases: VEP (domains)
Known limitations
Only curated critical domains trigger PM1: kinase catalytic, DNA-binding, ion channel pores, enzyme active sites, and other VCEP-documented domains
Generic structural domains (e.g., Caveolin, coiled-coil, DUF) excluded from PM1
Non-Pfam functional domains (e.g., UniProt annotated regions) not considered
Absent from controls or at extremely low frequency in population databases
Conditions
gnomAD global allele frequency < 0.0001 (0.01%) OR variant absent from gnomAD (NULL frequency = never observed in ~800K individuals)
NULL gnomAD allele frequency correctly treated as "absent from controls" per Richards et al. 2015 Table 3
Databases: gnomAD v4.1 (global_af)
Known limitations
Does not apply population-specific frequency adjustments
ClinGen SVI PM2_Supporting downgrade implemented for VCEP genes (v3.20.0, HELIX-CR-2026-034). 46 VCEP genes use PM2 at Supporting strength; 11 genes retain Moderate (RASopathy, ENIGMA BRCA1/2, HBOP). Non-VCEP genes use full Moderate strength.
When VCEP gene-specific specifications are enabled, PM2 threshold may differ from the generic 0.01% (e.g., 0% for RASopathy genes where any population frequency argues against pathogenicity)
Detected in trans with a pathogenic variant for recessive disorders
Conditions
Variant flagged as compound heterozygote candidate (compound_het_candidate = true)
Databases: Pipeline-internal (compound heterozygote detection)
Known limitations
Compound heterozygote status inferred from genotype data without formal phasing
Does not confirm that the trans variant is pathogenic
Trio data or long-read phasing would provide definitive confirmation
Protein length change in a non-repetitive region
Conditions
Consequence: in-frame insertion or in-frame deletion
Located within a Pfam functional domain
Not in a repetitive or low-complexity region (domains not containing "tandem", "repeat", "lowcomplexity", or "Seg")
Exclusions
HLA gene family (same exclusion list as PVS1)
Databases: VEP (consequence, domains)
Known limitations
Repetitive region detection based on VEP domain annotations only
Does not evaluate whether the in-frame change disrupts a critical functional residue
Missense variant in a gene with low rate of benign missense variation
Conditions
Consequence: missense variant
Gene constraint: pLI > 0.5 AND mis_z > 2.0 (missense constraint)
Disease association gate (v3.18.5): gene must have established disease association via same 7 sources as PVS1 gate. Prevents PP2 in genes without known Mendelian disease mechanism.
MPC regional constraint guard (v3.18.6, HELIX-CR-2026-032): mpc_score >= 1.0 required when MPC data is available. Regions with mpc_score < 1.0 are unconstrained for missense -- PP2 blocked. NULL MPC data falls back to gene-level constraint (bypass). Reference: Samocha et al. 2014.
Databases: VEP (consequence), gnomAD Constraint (pLI, mis_z), MPC regional constraint (mpc_score), disease association (7 sources)
Known limitations
pLI measures loss-of-function constraint; mis_z measures missense constraint
Both metrics required: pLI > 0.5 ensures LoF intolerance, mis_z > 2.0 ensures missense variants are a disease mechanism (Samocha et al. 2014)
MPC regional constraint data available for ~18K genes. NULL MPC falls back to gene-level constraint only.
Computational evidence supports a deleterious effect (two independent paths)
Conditions
Path A (Missense -- BayesDel): Requires consequence = missense_variant (v3.12.1, HELIX-CR-2026-009). BayesDel_noAF is calibrated exclusively for missense variants (Pejaver et al. 2022); non-missense variants (stop_gained, frameshift) receive spurious scores from dbNSFP positional overlap and are excluded. Three strength levels: PP3_Strong (>= 0.518, +4 points), PP3_Moderate (0.290-0.517, +2 points), PP3_Supporting (0.130-0.289, +1 point). Scores below 0.130 are indeterminate. Requires missense relevance guard v2.0 (v3.19.0, HELIX-CR-2026-033): 5 gates determine if missense is a disease mechanism. Gate 1: mis_z > 2.0 (direct constraint). Gate 2: mis_z > 1.5 AND pLI < 0.5 (AR genes). Gate 3: mis_z > 0.5 AND pLI > 0.9 (AD LoF-intolerant). Gate 4a: gene in refdb.clinvar_missense_genes (2,136 genes with ClinVar P/LP missense evidence). Gate 4b: pLI > 0.7 AND mis_z > 0.5 AND gene has disease association (gray-zone genes). NULL fallback: pLI > 0.5 without mis_z data. PP3_Strong additionally requires established disease association (v3.8.1 gate, same 7 sources as PVS1 gate).
PM1 + PP3 double-counting guard: When PM1 (functional domain) applies alongside PP3_Strong, PP3 is downgraded to PP3_Moderate. Combined PM1 + PP3 capped at Strong equivalent (4 points) per ClinGen SVI.
Path B (Splice evidence -- 3 strength levels, v3.15.1): SpliceAI max_score evaluated with evidence strength modulation. PP3_splice_Strong (>= 0.8, +4 points), PP3_splice_Moderate (0.5-0.799, +2 points), PP3_splice (0.2-0.499, +1 point Supporting). PP3_splice_Strong requires disease association gate (same as PP3_Strong missense). All levels excluded when PVS1 applies (ClinGen SVI 2023 double-counting guard).
Paths A and B are independent. A variant can trigger both if it has a BayesDel score and a splice prediction.
Exclusions
PP3_splice not applied when PVS1 is triggered at any strength level (Very Strong or Strong/last-exon). Prevents double-counting loss-of-function and splice evidence per ClinGen SVI 2023.
BayesDel indeterminate range (-0.180 to 0.129): no PP3 or BP4 applied
Databases: dbNSFP 4.9c (bayesdel_noaf_score), SpliceAI (max_score)
Known limitations
BayesDel_noAF used to avoid circular reasoning with PM2/BA1/BS1 frequency criteria (Pejaver et al. 2022)
PM1 + PP3 cap assumes PM1 = Moderate (2 points). Extensible if future VCEP upgrades PM1 strength.
BayesDel does not reach PP3_Very_Strong per Pejaver calibration data
Patient phenotype is highly specific for a disease with a single genetic etiology
Conditions
Requires patient HPO terms to be provided for the analysis session
Trigger condition A: >= 3 patient HPO terms match the gene HPO profile
Trigger condition B: >= 2 patient HPO terms match AND gene has <= 5 total HPO associations (highly specific gene-phenotype relationship)
Exclusions
Not evaluated when no patient HPO terms are provided
Databases: HPO (gene-phenotype associations)
Known limitations
HPO matching is exact term overlap, not semantic similarity (ontology hierarchy not used at this stage)
Semantic similarity-based matching is performed by the downstream Phenotype Matching Service
Reputable source reports variant as pathogenic
Conditions
ClinVar clinical significance: Pathogenic, Pathogenic/Likely_pathogenic, or Likely_pathogenic
ClinVar review stars >= 1 AND < 2 (lower confidence than PS1)
Exclusions
Does not apply when PS1 already applies (prevents double-counting ClinVar evidence at different strength levels)
Databases: ClinVar (clinical_significance, review_stars)
Known limitations
ClinGen SVI has recommended retiring PP5 as a standalone criterion; retained for maximum sensitivity
Benign Evidence
Allele frequency is above 5% in population databases
Conditions
gnomAD global allele frequency > 0.05 (5%)
Databases: gnomAD v4.1 (global_af)
Known limitations
Uses global allele frequency only; population-specific BA1 thresholds not implemented
BA1 is the only stand-alone ACMG criterion. A variant meeting BA1 is classified as Benign regardless of any other evidence, including ClinVar assertions. When VCEP gene-specific specifications are enabled, the BA1 threshold may be lower than 5% for specific genes (e.g., 0.1% for Cardiomyopathy genes).
Allele frequency is greater than expected for the disorder
Conditions
Inheritance-aware frequency threshold determined by a 5-level cascade (v3.7.0):
Priority 1 -- VCEP gene-specific: BS1 threshold from published ClinGen VCEP specification (~100 genes)
Priority 2 -- ClinGen HI score = 3: allele frequency >= 0.001 (0.1%) (~400 genes)
Priority 3 -- Orphanet AD inheritance: allele frequency >= 0.001 (0.1%) for genes with autosomal dominant disease association (~1,694 genes)
Priority 3b -- Orphanet XLD inheritance: allele frequency >= 0.001 (0.1%) for genes with X-linked dominant disease association (~90 genes). XLR genes intentionally excluded (fall to AR default).
Priority 4 -- HPO AD fallback: allele frequency >= 0.001 (0.1%) for genes annotated with HP:0000006 (autosomal dominant inheritance) (~200-400 additional genes)
Default -- AR threshold: allele frequency >= 0.05 (5%) for all remaining genes
Upper bound: allele frequency <= BA1 threshold (BA1 takes precedence)
Exclusions
Does not apply when BA1 applies (BA1 takes precedence)
XLR (X-linked recessive) genes do not receive the AD threshold -- carrier females are unaffected and higher carrier frequencies are expected
Databases: gnomAD v4.1 (global_af), ClinGen (haploinsufficiency_score), Orphanet/Orphadata (orphanet_gene_inheritance: has_ad, has_ar, has_xld), HPO (hpo_ids for HP:0000006), gnomAD Constraint (pLI, oe_lof_upper for constraint-implied AD fallback), AR_LOF_GENES curated list
Known limitations
Orphanet covers rare diseases only; some common AD conditions may not be catalogued (covered by HPO AD fallback or AR default)
Dual AD+AR inheritance genes (1,027 genes, e.g. KCNJ11) receive the AD threshold (0.1%) as conservative default -- AR pathogenic variants are typically well below 0.1% AF
Orphanet data updated twice per year (June + December releases). Coverage expands with each release.
When VCEP gene-specific specifications are enabled, VCEP thresholds take highest priority and override all other cascade levels
Criteria string includes inheritance source annotation: BS1[ClinGen-HI], BS1[Orphanet-AD], BS1[Orphanet-XLD], BS1[HPO-AD]
Data source: Orphadata (INSERM, France). License: CC-BY-4.0. Attribution: Orphadata: Free access data from Orphanet. INSERM 1999. Available on https://www.orphadata.com.
Observed in a healthy adult individual for a fully penetrant early-onset disorder
Conditions
gnomAD homozygote count > 15
Databases: gnomAD v4.1 (global_hom)
Known limitations
Does not distinguish between early-onset and late-onset conditions
Fixed threshold; not adjusted for disease penetrance or frequency
Missense variant in a gene for which primarily truncating variants are known to cause disease
Conditions
Consequence: missense variant
Impact: MODERATE
Gene constraint: pLI < 0.1 (gene is tolerant to loss-of-function)
pLI value must be present (non-NULL)
Missense constraint guard (v3.6.8): mis_z must be < 2.0 or absent. Genes with high mis_z have missense as primary disease mechanism.
ClinVar pathogenic missense guard (v3.8.2): BP1 does not apply when ClinVar has a Pathogenic missense variant in the same gene. If ClinVar confirms pathogenic missense, the premise of BP1 is falsified.
Exclusions
Genes with mis_z >= 2.0 (missense is a known disease mechanism)
Genes with ClinVar Pathogenic/Likely Pathogenic missense variants (session-local check)
Databases: VEP (consequence, impact), gnomAD Constraint (pLI, mis_z), ClinVar (clinical_significance, consequence)
Known limitations
Low pLI used as proxy for "primarily truncating variants cause disease"
mis_z measures heterozygous missense constraint. AR enzyme genes (e.g. MCCC2) may have low mis_z despite pathogenic homozygous missense -- the ClinVar missense guard (v3.8.2) addresses this gap.
Observed in trans with a pathogenic variant for a fully penetrant dominant disorder
Conditions
Compound heterozygote candidate (compound_het_candidate = true)
ClinGen haploinsufficiency score = 30 (dosage sensitivity unlikely)
Databases: Pipeline-internal (compound heterozygote detection), ClinGen (haploinsufficiency_score)
Known limitations
Trans observation inferred without formal phasing
Haploinsufficiency score of 30 is a specific ClinGen code for "dosage sensitivity unlikely"
In-frame insertion or deletion in a repetitive region without a known function
Conditions
Consequence: in-frame insertion or in-frame deletion
Located in a repetitive/low-complexity region (domains contain "tandem", "repeat", "lowcomplexity", or "Seg"), OR not in any Pfam domain, OR no domain annotation available
Databases: VEP (consequence, domains)
Known limitations
Complementary to PM4 (PM4 requires Pfam domain; BP3 requires absence of critical domain)
Computational evidence suggests no impact on gene or gene product
Conditions
Requires consequence = missense_variant (v3.12.1, HELIX-CR-2026-009). BayesDel_noAF score evaluated with ClinGen SVI calibrated thresholds (Pejaver et al. 2022). Two strength levels: BP4_Moderate (<= -0.361, -2 points), BP4_Supporting (-0.360 to -0.181, -1 point).
SpliceAI max_score must be < 0.1 or absent (no predicted splice impact)
Exclusions
BayesDel indeterminate range (-0.180 to 0.129): no BP4 applied
Databases: dbNSFP 4.9c (bayesdel_noaf_score), SpliceAI (max_score)
Known limitations
BayesDel_noAF does not reach BP4_Strong per Pejaver calibration data
SpliceAI guard prevents BP4 for variants with any predicted splice impact
Reputable source reports variant as benign
Conditions
ClinVar clinical significance: Benign, Benign/Likely_benign, or Likely_benign
ClinVar review stars >= 1
Databases: ClinVar (clinical_significance, review_stars)
Known limitations
ClinGen SVI has recommended retiring BP6 as a standalone criterion; retained for maximum sensitivity
Synonymous variant with no predicted impact on splicing
Conditions
Consequence: synonymous variant
Not in a splice region (consequence does not contain "splice_region")
SpliceAI max_score <= 0.1 or absent
Databases: VEP (consequence), SpliceAI (max_score)
Known limitations
Conservation filter intentionally omitted per Walker et al. 2023 Table S13 recommendation ("no improvement in negative predictive value" with conservation filter)
Aligned with ClinGen SVI 2023 (Walker et al.) Figure 4 decision tree for synonymous variant classification.
Computational Predictors (PP3 / BP4)
PP3 and BP4 use BayesDel_noAF as the primary classification tool with ClinGen SVI calibrated thresholds (Pejaver et al. 2022). BayesDel is a Bayesian framework that integrates deleteriousness scores from multiple underlying predictors into a single calibrated score. The noAF variant (without allele frequency) is used to avoid circular reasoning with PM2, BA1, and BS1 frequency criteria.
| BayesDel_noAF Score | Evidence Strength | ACMG Code | Bayesian Points |
|---|---|---|---|
| >= 0.518 | Strong pathogenic | PP3_Strong | +4 |
| 0.290 to 0.517 | Moderate pathogenic | PP3_Moderate | +2 |
| 0.130 to 0.289 | Supporting pathogenic | PP3_Supporting | +1 |
| -0.180 to 0.129 | Indeterminate | None | 0 |
| -0.360 to -0.181 | Supporting benign | BP4_Supporting | -1 |
| <= -0.361 | Moderate benign | BP4_Moderate | -2 |
PM1 + PP3 Double-Counting Guard
When PM1 (functional domain, Moderate = 2 points) applies alongside PP3_Strong (4 points), the combined evidence would exceed the ClinGen SVI recommended cap of Strong equivalent (4 points). In this case, PP3_Strong is downgraded to PP3_Moderate (2 points), yielding a combined 2 + 2 = 4 points. PP3_Moderate and PP3_Supporting are not affected by this cap.
Why BayesDel
The ClinGen SVI Working Group calibrated four computational tools (BayesDel, MutPred2, REVEL, VEST4) and demonstrated that a single calibrated tool with evidence strength modulation provides more accurate classification than a fixed-threshold multi-predictor consensus. BayesDel_noAF was selected because it explicitly excludes allele frequency from its model (avoiding circular reasoning with PM2/BA1/BS1), is precomputed in dbNSFP 4.9c, and reaches both PP3_Strong and BP4_Moderate in the Pejaver calibration -- providing the widest evidence strength range among available tools.
Display Predictors (Not Used in Classification)
The following predictors are available alongside every variant for the reviewing geneticist to inspect, but they are not used in the PP3/BP4 classification logic. They were used in the weighted consensus approach prior to v3.4 and are retained for clinical reference:
SIFT
dbNSFP 4.9c
AlphaMissense
dbNSFP 4.9c
MetaSVM
dbNSFP 4.9c
DANN
dbNSFP 4.9c
PhyloP (100-way)
dbNSFP 4.9c
GERP++
dbNSFP 4.9c
SpliceAI Integration
Splice impact predictions are integrated following ClinGen Sequence Variant Interpretation (SVI) Working Group recommendations (Walker et al., 2023).
SpliceAI predicts the impact of each variant on mRNA splicing through four delta scores: acceptor gain (DS_AG), acceptor loss (DS_AL), donor gain (DS_DG), and donor loss (DS_DL). The maximum of these four scores is used for classification thresholds.
Scores are sourced from Ensembl precomputed MANE transcript predictions, not computed at runtime. This ensures reproducibility and avoids runtime dependencies on external services.
Thresholds
Strong evidence for spliceogenicity (+4 points). Requires disease association gate (v3.15.1).
Moderate evidence for spliceogenicity (+2 points).
Supporting evidence for spliceogenicity (+1 point).
Required for BP4 to apply. Prevents benign classification when splice impact is predicted.
Required for synonymous variant BP7 classification. Confirms no splice impact.
PVS1 Double-Counting Guard
When PVS1 (loss-of-function) is triggered for a variant, PP3_splice is not applied. This prevents double-counting the same biological mechanism (splice disruption leading to loss of function) as both PVS1 and PP3 evidence, per ClinGen SVI recommendation.
Reference: Walker LC et al. Am J Hum Genet. 2023;110(7):1046-1067. PMID: 37352859
Classification Logic
Classification uses the Bayesian point-based framework (Tavtigian et al. 2018, 2020) which is mathematically equivalent to the original 18 ACMG combining rules while providing proper classifications for evidence combinations not explicitly covered by the 2015 guidelines.
Bayesian Point System (Primary)
Each evidence criterion contributes points based on its strength level. The total determines classification:
Pathogenic Evidence Points
Benign Evidence Points
Classification Thresholds
Pathogenic
>= 10 pts
Likely Path.
6 to 9 pts
VUS
0 to 5 pts
Likely Benign
-1 to -5 pts
Benign
<= -6 pts
High-Confidence Conflict Safety Check
When pathogenic evidence at Strong or Very Strong level conflicts with Strong benign evidence (BS), the variant is flagged for manual review regardless of the point total. This prevents automated resolution of genuinely conflicting high-quality evidence.
ACMG 2015 Combining Rules (Reference)
The original 18 ACMG 2015 combining rules are a special case of the Bayesian point system -- every rule produces the same classification under both approaches. They are retained here as a reference.
Pathogenic (8 rules)
Likely Pathogenic (7 rules)
Benign (2 rules)
Likely Benign (2 rules)
Conflicting Evidence Handling
The Bayesian point system naturally handles most conflicting evidence through point summation. For example, PM2 (+2) and BS1 (-4) yield a net of -2 = Likely Benign. Under the previous v3.3 system, this combination would have been flagged as conflicting and defaulted to VUS. The point-based approach is more nuanced and clinically appropriate. However, when pathogenic evidence at Strong or Very Strong level directly conflicts with Strong benign evidence, the variant is flagged for manual review regardless of point total (see High-Confidence Conflict Safety Check above). BA1 remains a stand-alone override handled at a higher priority level.
VCEP Gene-Specific Specifications
ClinGen Variant Curation Expert Panels (VCEPs) adapt generic ACMG/AMP 2015 criteria to specific genes or diseases. Helena implements approved VCEP specifications as an optional overlay on top of the standard classification.
How It Works
The standard classification pipeline runs first, producing a generic ACMG classification for all variants. For variants in genes with available VCEP specifications, gene-specific thresholds are applied via a lightweight overlay that modifies frequency cutoffs (BA1, BS1, PM2) and criterion applicability (PVS1) based on the published VCEP specification. The Bayesian point total is then recalculated with the modified criteria.
What VCEPs Modify
Gene-specific allele frequency threshold for stand-alone benign (e.g., 0.1% for Cardiomyopathy genes instead of generic 5%)
Gene-specific elevated frequency threshold, already mode-of-inheritance-aware
Gene-specific absent-in-controls threshold (e.g., 0% for RASopathy genes)
Gene-specific applicability gate. Set to FALSE for gain-of-function genes (e.g., MYOC in glaucoma)
Toggle Behavior
VCEP overlay is enabled by default and can be disabled per case in case settings. When enabled, variants in VCEP-covered genes display an audit trail marker (e.g., "[VCEP:Hearing Loss v1.0]") in the criteria string.
Coverage
Approved VCEP specifications are sourced from the ClinGen Criteria Specification Registry (CSpec). Approximately 50-60 genes have published specifications, including panels for Hearing Loss, Cardiomyopathy (MYH7, MYBPC3), RASopathy (PTPN11, BRAF, SOS1), PTEN, CDH1, TP53, PAH, and BRCA1/BRCA2 (ENIGMA). For all other genes, generic ACMG 2015 thresholds are used.
ClinVar Override Logic
ClinVar clinical significance assertions are used as classification evidence, but only under specific conditions that prevent overriding computational evidence when conflicts exist.
When ClinVar override IS applied
ClinVar has a Pathogenic, Likely Pathogenic, Benign, or Likely Benign assertion with at least 1 review star, AND no conflicting computational evidence exists (no BA1, no conflicting pathogenic+benign criteria at moderate+ strength). For P/LP assertions with only 1 review star (single submitter), the gene must also have an established disease association via at least one of five sources: ClinVar P/LP (gene-level, requires 2+ review stars to prevent circular self-validation), Orphanet, ClinGen haploinsufficiency score, curated AR LoF gene list, or VCEP coverage (v3.10.0 disease association gate, v3.10.1 CTE fix). P/LP assertions with 2+ review stars bypass this gate. Benign/Likely Benign assertions are not gated.
When ClinVar override is NOT applied
BA1 applies (frequency > 5% always overrides ClinVar). OR conflicting evidence exists (pathogenic + benign criteria both triggered). OR ClinVar asserts VUS (VUS does not override computational classification). OR ClinVar review stars are below the minimum threshold. OR ClinVar P/LP with 1 review star in a gene without established disease association (v3.10.0 gate, v3.10.1 circular reference fix -- the gene-level ClinVar disease check requires 2+ stars to prevent a 1-star submission from self-validating its own gate).
When ClinVar classification is used, the criteria string includes "ClinVar" as the first element (e.g., "ClinVar,PM2,PP3") to make the evidence source explicit. When a ClinVar P/LP override is blocked by the disease association gate (v3.10.0), the criteria string shows "ClinVar_gated" as the first element to indicate that the ClinVar assertion was present but not applied. ClinVar review star level is available alongside every variant for the reviewing geneticist to assess assertion quality.
Default minimum review stars for override: 1 (configurable per deployment)
Manual Review Criteria (9 of 28)
These criteria require information that cannot be determined from a single-sample VCF file -- family segregation data, functional study results, confirmed de novo status, or case-level clinical context. They must be evaluated by the reviewing geneticist.
De novo variant (confirmed paternity and maternity)
Requires trio sequencing data and confirmed parental relationships. Cannot be determined from single-sample VCF analysis.
Well-established in vitro or in vivo functional studies show a deleterious effect
Requires curation of published functional assay data. Automated literature extraction of functional evidence is not yet implemented.
Prevalence of the variant in affected individuals is significantly increased compared with controls
Requires case-control study data or odds ratios not available in standard annotation databases.
Novel missense change at an amino acid residue where a different pathogenic missense has been observed
Currently disabled. Requires normalized HGVSp matching against ClinVar at the amino acid position level. Will be enabled when ClinVar preprocessing provides standardized protein-level coordinates.
Disabled (pending implementation)Assumed de novo without confirmation of paternity and maternity
Requires family structure information not available in single-sample analysis.
Cosegregation with disease in multiple affected family members
Requires multi-generational pedigree data and segregation analysis.
Well-established in vitro or in vivo functional studies show no deleterious effect
Requires curation of published functional assay data (benign counterpart of PS3).
Lack of segregation in affected members of a family
Requires family segregation data not available in single-sample analysis.
Variant found in a case with an alternate molecular basis for disease
Requires clinical case-level information about alternative diagnoses.
Quality Filtering
Three configurable quality presets control the stringency of variant filtering. Quality filtering occurs before annotation and classification.
| Preset | Quality (QUAL) | Depth (DP) | Genotype Quality (GQ) | Recommended Use |
|---|---|---|---|---|
| Strict | >= 30 | >= 20 | >= 30 | High-confidence clinical reporting |
| Balanced | >= 20 | >= 15 | >= 20 | Standard clinical analysis (default) |
| Permissive | >= 10 | >= 10 | >= 10 | Maximum sensitivity / research |
ClinVar Rescue Mechanism
Variants with documented clinical significance in ClinVar (Pathogenic or Likely Pathogenic) that fail quality thresholds are not discarded. They are flagged as rescued variants and proceed through the classification pipeline. This prevents clinically significant findings from being silently excluded due to sequencing quality in low-coverage regions -- a deliberate design decision for clinical safety.
Limitations and Disclaimers
Helena is a clinical decision support tool, not a diagnostic device. All classifications require review and confirmation by a qualified clinical geneticist.
9 of 28 ACMG criteria require information not available from single-sample VCF analysis (segregation, functional studies, de novo confirmation). These criteria must be evaluated manually by the reviewing geneticist.
SpliceAI predictions are computational. RNA splicing studies remain the gold standard for confirming splice-altering effects.
Population frequency data from gnomAD may underrepresent certain ethnic groups and geographic populations. Allele frequency thresholds should be interpreted in the context of the patient's ancestry.
ClinVar assertions vary in quality and currency. Review star levels are displayed alongside all ClinVar-derived evidence to enable informed interpretation.
Structural variants (SVs), copy number variants (CNVs), and repeat expansions are not currently classified by this pipeline.
Mitochondrial variants are processed through the same pipeline using nuclear ACMG rules as an approximation. Dedicated mitochondrial classification guidelines (e.g., MitoMap criteria) are not yet implemented.
VCEP gene-specific specifications are implemented as a threshold overlay for BA1, BS1, PM2, and PVS1 applicability. The overlay does not implement VCEP-specific functional assay interpretation (PS3/BS3) or gene-specific segregation logic (PP1/BS4), which require manual curation. Coverage is limited to approximately 50-60 genes with published ClinGen CSpec specifications; all other genes use generic ACMG 2015 thresholds.
PM5 (novel missense at known pathogenic amino acid position) is currently disabled pending standardized protein-level coordinate matching in the ClinVar preprocessing pipeline.
Compound heterozygote detection is inferred from genotype data without long-read phasing or trio analysis. Formal phasing should be performed for clinical confirmation.
The curated autosomal recessive LoF gene list (~150 genes) for PVS1 bypass covers genes with Definitive/Strong evidence for biallelic LoF disease mechanism. AR genes outside this list that lack constraint (low pLI, high LOEUF) will not trigger PVS1 unless they have VCEP gene-specific specifications.
Homozygous reference genotypes (hom_ref) from multi-allelic VCF sites are excluded from classification. These represent alleles not present in the patient.
Results should always be interpreted in the context of the patient's clinical presentation, family history, and other available clinical information.
Version History
Every methodology change is versioned and documented. The version number corresponds to the classification engine version in production.
PVS1: Expression-aware guard using gnomAD v4.1 pext data (GTEx v10, GRCh38, 49 tissues). Per-exon max_tissue_pext modulates PVS1 application (HELIX-CR-2026-056, Cummings et al. 2020 PMID:32461655). Three tiers: max_tissue_pext >= 0.9 = PVS1 Very Strong (unchanged), 0.1-0.9 = PVS1_Strong[pext] (downgraded from Very Strong to Strong), < 0.1 = PVS1 blocked (exon not expressed). NULL pext = Very Strong (conservative fallback). PM4 receives binary pext guard (blocked at < 0.1).
PVS1: 189,856 exon records for 18,923 MANE Select genes loaded from gnomAD v4.1 base-level pext TSV. Distribution: 56.3% full (>= 0.9), 42.5% downgrade (0.1-0.9), 1.2% block (< 0.1). Uses max_tissue_pext (maximum across 49 GTEx v10 tissues) to avoid false negatives for tissue-specific genes.
PVS1: Human-readable annotations in acmg_criteria ("[PVS1 blocked: low exon expression (pext X.XX)]", "[PVS1 downgraded: partial exon expression (pext X.XX)]") and processing_notes (full sentence with gene name, threshold, source).
Reference databases: gnomAD pext (v4.1, GTEx v10) and MANE Select transcripts (v1.4) added as reference data sources.
Scientific basis: Cummings 2020 demonstrated that expression-based annotation selectively filters 22.8% of falsely annotated pLoF variants while removing less than 4% of true pathogenic variants.
PVS1: MANE Select positional guard (HELIX-CR-2026-055, Abou Tayoun 2018 PMID:30192042). Variants outside the MANE Select/Plus Clinical CDS region do not receive PVS1 or PM4. Addresses false PVS1 application on variants annotated by VEP against non-canonical transcripts that extend beyond the canonical coding sequence.
PVS1: 19,354 MANE transcripts (19,288 Select + 66 Plus Clinical) with CDS coordinates loaded from MANE GFF3 v1.4. LEFT JOIN with GROUP BY prevents row duplication for 66 genes with both Select and Plus Clinical transcripts.
PVS1: Inline NULL-safe guards on 6 sites (pvs_count, ps_count/pvs1_last_exon, pm_count/pm4, has_pvs1_full, has_pvs1_last_exon, has_pm4). NULL MANE data = PVS1 applied (conservative fallback for genes not yet in MANE).
Criteria string: [PVS1_BLOCKED:outside_MANE_CDS] marker appended for HIGH impact variants outside MANE CDS. Processing notes: PVS1_BLOCKED:outside_MANE_CDS(gene,mane_gene) audit trail.
Clinical trigger: HG2022_057 FANCB chrX:14797196 p.Gln864Ter classified LP (PVS1+PM2) against non-canonical transcript ENST00000696353. Position is 46 kb upstream of MANE Select CDS (14,843,567-14,865,510). Reclassified to VUS (PM2 only).
CLCN1 removed from AR_LOF_GENES (HELIX-CR-2026-053). Myotonia congenita has both AD (Thomsen) and AR (Becker) forms. Het LoF can cause AD Thomsen disease. CLCN1 receives AD/AR via dual HPO check.
Annotation Phase 7.5: dual HPO inheritance detection (HELIX-CR-2026-051). Genes with both HP:0000006 (AD) and HP:0000007 (AR), confirmed by gene_disease_mechanism biallelic, now receive AD/AR instead of AD. 312 genes annotation improved.
CYP11A1 added to AR_LOF_GENES. Congenital adrenal insufficiency, AR. OMIM #613743.
PM2 frequency field alignment for VCEP FAF thresholds (HELIX-CR-2026-047). VCEP genes with pm2_frequency_field=af_grpmax now compare PM2 against af_grpmax instead of global_af. K1 bottlenecked NULL guard: af_grpmax NULL + global_af NOT NULL = PM2 not satisfied.
PM2 criteria string audit trail: PM2[af_grpmax] and PM2_Supporting[af_grpmax] markers.
Show older versions (44 more)
BP1: reference-based missense mechanism guard (HELIX-CR-2026-044). BP1 blocked when gene is in refdb.clinvar_missense_genes (>= 2 ClinVar P/LP missense at 2+ stars). Extends session-local ClinVar guard with reference-level check.
PP3_splice missense double-counting guard (HELIX-CR-2026-043). PP3_splice blocked when variant already receives PP3 from BayesDel missense path. ClinGen SVI: PP3 is a single criterion.
BP1: GoF/DN gene guard (HELIX-CR-2026-039v2). BP1 blocked in GoF/DN genes where truncating variants do not cause disease via LoF. PLIN1 added to GOF_AD_GENES (DN mechanism, FPLD4).
PVS1/PM3: Dual-mechanism gene guard (HELIX-CR-2026-037). gof_genes_exclusive view excludes 55 dual-mechanism genes from GoF guard, restoring PVS1/PM3 for AR LoF pathway.
PM3: GoF/DN gene exclusion (HELIX-CR-2026-036). Compound het in pure GoF/DN genes no longer satisfies PM3. PM3 presupposes AR mechanism.
Disease association gate: gene_disease_mechanism table added as 7th source in pvs1_disease_gate and gene_has_disease_association (HELIX-CR-2026-026). 256 genes in gene_disease_mechanism (G2P 225, GoFCards 35, manual 2) were not covered by existing 6 disease gate sources. Confidence filter: definitive/strong/moderate only (predicted excluded per CTO directive). Clinical trigger: SNRPN c.-391+1_-391+2insGA false VUS after v3.18.1 HI fix.
Disease association gate: haploinsufficiency_score IS NOT NULL replaced with IN (2, 3, 30) in pvs1_disease_gate and gene_has_disease_association (HELIX-CR-2026-025). HI=0 (no evidence), HI=1 (little evidence), HI=40 (dosage sensitivity unlikely) no longer satisfy disease gate. 475 genes lost false disease association, 117 with no other disease source. Clinical trigger: SLC22A18 c.-133+2T>C false LP in gene without Mendelian disease.
PVS1 GoF/DN guard: gof_genes_g2p view (198 genes) replaced by gof_genes_unified view (347 genes) as primary PVS1 GoF exclusion source (HELIX-REF-003 Phase 2A). gof_genes_unified draws from 3 sources: G2P (199 genes), GoFCards (179 genes), manual curation (27 genes). Dual-mechanism genes correctly excluded. GOF_AD_GENES Python constant retained as fallback.
Reference database: gene_disease_mechanism table created in reference_db (2,617 records). Unified schema for molecular mechanism data from G2P, GoFCards, and manual curation. Columns: gene_symbol, disease_id, disease_name, mechanism (LoF/GoF/DN/GoE), confidence, source, allelic_requirement, pmid, curation_date, curator, last_reviewed, notes.
Reference database: GoFCards Release 1.0 integrated (579 genes, 3,161 curated GoF variants). Gene-level aggregation threshold: 3+ variants, Pscore >= 2.0, at least 1 variant with animal or cell model evidence = confidence "strong". 217 genes qualified.
Reference database: ClinGen Dosage Sensitivity full TSV integrated (1,626 genes, 23 columns). Replaces legacy 3-column clingen table. Backward-compatible clingen table recreated from full data.
Reference: GoFCards PMID:39578693, ClinGen clinicalgenome.org, HELIX-REF-003.
PVS1: 12 GoF/DN/GoE AD genes added to GOF_AD_GENES (HELIX-REF-003 Phase 2B). Systemic gap analysis: 28 AD genes depend ONLY on ClinGen AD Definitive path for PVS1, without mechanism check. 13 identified as HIGH risk (established non-LoF mechanism). TRAF7 deferred (insufficient germline functional evidence).
Genes added: CALM2 (DN calmodulinopathy), CALM3 (DN calmodulinopathy), DNAJB6 (DN myofibrillar myopathy), GJB6 (DN deafness, ClinGen AR=Refuted), GREM1 (regulatory GoE), HSPB8 (toxic GoF CMT2L), INF2 (DN formin FSGS5), NEFL (DN neurofilament CMT), PRKAG2 (GoF AMPK activation), SKI (DN Shprintzen-Goldberg), TNNC1 (GoF Ca2+ sensitivity HCM), TUBB3 (DN microtubule CFEOM3A).
Cross-referenced with LoGoFunc (variant-level GoF/LoF predictions), GoFCards (curated GoF variants), ClinGen Dosage (HI=0 signal), and G2P mechanism data.
PVS1: ANKRD26 and C1QTNF5 added to GOF_AD_GENES (HELIX-CR-2026-024). ANKRD26: Thrombocytopenia 2, gain-of-expression mechanism -- 5UTR mutations disrupt RUNX1/FLI1 silencing. ClinGen Dosage explicitly states "gain-of-function rather than haploinsufficiency". C1QTNF5: Late-onset retinal degeneration, dominant-negative mechanism -- mutant destabilizes wildtype in gC1q domain co-expression.
Clinical trigger: PD2025_095 ANKRD26 chr10:27044191 classified P [PVS1,PM2,PM3], should be VUS. C1QTNF5 chr11:119339479 received PVS1_Strong for frameshift despite all known pathogenic being missense.
PVS1: G2P molecular mechanism integration (HELIX-CR-2026-023). DECIPHER Gene2Phenotype (G2P) database now provides primary GoF/DN guard for PVS1. gof_genes_g2p view contains 198 pure GoF/DN monoallelic genes where PVS1 is blocked. G2P is queried at classification time via reference_db ATTACH.
PVS1: GOF_AD_GENES reduced from 36 to 14 fallback genes (genes not in G2P: 6 neurodegeneration/toxic aggregation, 3 somatic/neomorphic, 5 absent from G2P 2026-02-28). PRSS1 added to fallback (GoF trypsinogen autoactivation, clinical trigger PD2025_090).
PVS1: GNAS removed from GOF_AD_GENES (dual-mechanism: McCune-Albright = GoF, pseudohypoparathyroidism 1A = LoF). G2P view correctly excludes GNAS. GNAS LoF variants now receive PVS1.
PVS1: 49 dual-mechanism genes (SCN5A, LMNA, KCNH2, KCNQ1, FGFR1, etc.) excluded from GoF view, preserving PVS1 for legitimate LoF phenotypes.
PVS1: G2P confidence filter: only definitive, strong, moderate included. Limited confidence excluded to prevent blocking PVS1 on weak GoF evidence.
PVS1: Guard order: G2P view (primary) -> GOF_AD_GENES (fallback). Identical pattern to refdb.ar_lof_genes_clingen and refdb.disease_associated_genes_clingen.
DECIPHER G2P: Molecular mechanism data (g2p_gene_mechanism table, 2,372 records) added to reference_db alongside existing HPO enrichment data (v3.12.0).
Loader script: scripts/load_g2p_mechanism.py with dry-run, force flags, and 9-point validation (total records, mechanism distribution, view count, dual-mechanism exclusion, spot checks).
Subtractive change: can only remove PVS1 from GoF/DN genes. Cannot create false P/LP.
Clinical trigger: PD2025_090 PRSS1 p.Gly177Ter (stop_gained). GoF gene, PVS1 scientifically incorrect. BA1 protected in this case (AF 41%), but rare PRSS1 truncating variants without BA1 could reach LP[PVS1,PM2] incorrectly.
Reference: Abou Tayoun 2018 PMID:30192042, Josephs 2023 PMID:37884986, HELIX-CR-2026-023.
PVS1: GoF AD gene exclusion (HELIX-CR-2026-022). GOF_AD_GENES curated list (36 genes) blocks PVS1 for autosomal dominant genes where the disease mechanism is gain-of-function, dominant-negative, or toxic aggregation. ClinGen PVS1 Decision Tree Node 1: "Is LoF a known mechanism of disease?" -- if NO, PVS1 not applicable.
PVS1: GOF_AD_GENES categories: RAS/MAPK pathway (12 genes), receptor tyrosine kinases (10), PI3K/AKT/mTOR (2), GTPases (3), metabolic enzymes (2), kinases (2), neurodegeneration (6). Each gene documented with PMID for GoF mechanism.
PVS1: ClinGen AD Definitive/Strong genes retain PVS1 access even if in GOF_AD_GENES -- VCEP and ClinGen disease validity take precedence.
Clinical trigger: PD2025_082 RAC1 p.Val14SerfsTer4 (frameshift in GoF gene). SOD1 discordance in same session.
Reference: Abou Tayoun 2018 PMID:30192042, HELIX-CR-2026-022.
LP classification: Disease association gate for LP rules LP4, LP5, LP6 (HELIX-CR-2026-021). LP rules that rely solely on moderate and supporting criteria now require established disease association via at least one of five sources (same as PVS1 and PP3_Strong gates).
LP4 (>=3 PM), LP5 (2 PM + >=2 PP), LP6 (1 PM + >=4 PP): these rules can be satisfied by PM1+PM2+PM4 or similar combinations without any gene-disease evidence. Gate prevents LP classification in genes without known Mendelian disease mechanism.
LP1, LP2, LP3 not gated: they require PVS or PS criteria, which already have their own disease association gates.
Subtractive change: can only downgrade LP to VUS, cannot upgrade.
PVS1: ClinGen AD Definitive/Strong genes bypass pLI/LOEUF constraint gate. Genes with ClinGen Gene-Disease Validity rating of Definitive or Strong for autosomal dominant inheritance now qualify for PVS1 regardless of population constraint metrics.
PVS1: This addresses genes like TUBB1 (pLI near 0, LOEUF=1.586) with ClinGen AD Definitive that were excluded from PVS1 by the population constraint gate.
ClinGen Gene-Disease Validity integration: ar_lof_genes_clingen and disease_associated_genes_clingen reference tables added to reference_db. Replaces in-memory Python constants for AR LoF gene list and disease association checks.
PVS1 AR gene bypass now queries refdb.ar_lof_genes_clingen at classification time instead of Python AR_LOF_GENES constant.
Disease association check now queries refdb.disease_associated_genes_clingen for PVS1, PP3_Strong, LP disease gates, and ClinVar override gate.
Reference database count: 14 (previously 13).
PM1: Critical functional domains migrated from Python constant (CRITICAL_PFAM_DOMAINS) to reference database table (refdb.interpro_pfam_domains). HELIX-CR-2026-012.
PM1: Full InterPro Pfam catalog (27,481 entries) loaded into reference_db. 49 domains marked as critical (is_critical=true) across 14 categories.
PM1: Pre-build approach -- critical accessions loaded from refdb at classification time, OR LIKE clauses generated in Python. Zero performance regression vs v3.14.0.
PM1: Adding a new critical domain is now a database UPDATE, not a code change + rebuild + deploy.
InterPro / Pfam added as 13th reference database.
Loader script: scripts/load_interpro_pfam.py with InterPro API fetch, JSON cache for offline fallback, and hard-fail validation.
Reference: HELIX-CR-2026-012.
PM1: CRITICAL_PFAM_DOMAINS converted from Pfam short names to InterPro-verified Pfam accession numbers (HELIX-CR-2026-011). VEP annotation records domain data as Pfam:PFxxxxx accession numbers, but PM1 SQL searched for short names (RyR, Pkinase, Ion_trans). Result: 0 legitimate PM1 applications since curated list introduction. Fix: all 49 domains now use verified accession numbers.
PM1: Removed Rho (covered by PF00071 Ras superfamily, VEP confirms RAC1 annotated with PF00071) and Peptidase_S1 (covered by PF00089 Trypsin, VEP confirms F10 annotated with PF00089).
PM1: Corrected accessions -- Dynamin_N: PF01031 (Dynamin_M) corrected to PF00350 (Dynamin_N). PH: PF00168 (C2 domain) corrected to PF00169 (PH domain). FBN_EGF: mapped to PF00008 (EGF-like domain).
PM1: All 49 accessions verified against InterPro REST API (ebi.ac.uk/interpro) on 2026-03-16. Domain coverage validated against patient VCF data (RYR1 PF01365, KRAS PF00071, KCNH2 PF00520, CDH23 PF00028, MYO7A PF00063, COL1A1 PF01391, CFTR PF00005, EGFR PF07714).
PM1: Post-fix validation on HG2023_206: PM1 count increased from 2 (false positive) to 1,139 (legitimate). RYR1 c.6635T>C correctly receives PM1. CRYGD false positives eliminated. Zero B/LB classification changes.
Clinical trigger: HG2023_206 RYR1 c.6635T>C (p.Val2212Ala) in RyR domain (Pfam:PF01365) classified VUS [PM2,PP3_Supporting] without PM1. CG classified LP. PM1 now applied: VUS [PM1,PM2,PP3_Supporting].
Reference: HELIX-CR-2026-011, Richards et al. 2015 PMID:25741868, ClinGen SVI PM1 guidance.
PVS1: ClinVar LOF evidence added as fifth constraint gate path (HELIX-CR-2026-010). Small autosomal dominant genes with ClinVar P/LP LOF variants (2+ review stars) but uninformative gnomAD constraint (pLI < 0.9, LOEUF >= 0.35, HI != 3) now qualify for PVS1 via CLINVAR_LOF_AD_GENES curated list.
gene_has_disease_association: CLINVAR_LOF_AD_GENES added as sixth source in the disease association check.
Clinical trigger: GCM2 p.Arg131Ter (stop_gained) in PD2025_098. GCM2 is a 5-exon AD gene for Familial hypoparathyroidism with ClinVar Pathogenic p.Tyr136Ter (2 stars) but pLI=0.0, LOEUF=1.205, HI=NULL. Now LP [PVS1,PM2].
Reference: Abou Tayoun 2018 PMID:30192042, HELIX-CR-2026-010.
PP3/BP4: Missense consequence guard (HELIX-CR-2026-009). PP3 BayesDel path (Strong/Moderate/Supporting) and BP4 BayesDel path (Moderate/Supporting) now require consequence = missense_variant. BayesDel_noAF is calibrated exclusively for missense variants (Pejaver et al. 2022).
Clinical trigger: GCM2 p.Arg131Ter (stop_gained) in PD2025_098 received PP3_Strong (BayesDel=0.66) + PM2 = LP. Correct: PM2 only = VUS. 2 non-missense variants with false PP3_Strong eliminated.
Subtractive change: can only downgrade (LP to VUS), cannot upgrade. PP3_splice (SpliceAI) unaffected.
Reference: Pejaver 2022 PMID:36413997, Walker 2023 PMID:37352859, HELIX-CR-2026-009.
HELIX-REF-001: Multi-Source HPO Enrichment Pipeline. HPO annotation expanded from single-source (HPO Consortium, 5,173 genes, 320K records) to multi-source enriched view (6 sources, 5,688 genes, 927K records, +10% gene coverage).
HPO sources: HPO Consortium (primary), Orphanet disease-to-HPO via Orphadata XML (3,176 genes with frequency data), DECIPHER G2P from 7 clinical panels (2,125 genes), Monarch Initiative (4,791 genes), ClinVar-MedGen chain mapping from P/LP variants (5,258 genes), and manual clinical curation for common-disease genes.
Enrichment depth: 3,292 existing genes received additional HPO terms. 515 net new genes added. 1,505 genes covered by all 5 automated sources.
Source priority ordering: Orphanet > G2P > HPO Consortium > Monarch > ClinVar-MedGen. Low-confidence entries filtered from clinical view.
PP4 criterion evaluates against enriched HPO set. No ACMG criteria logic changes.
Data integrity validated: 0 NULLs, 0 invalid HPO IDs, 0 duplicate PKs across 926,930 records.
PVS1: Non-canonical splice site variants excluded from PVS1 application. VEP consequence terms splice_donor_5th_base_variant (position +5), splice_donor_region_variant (positions +3 to +8), and splice_acceptor_5th_base_variant (acceptor position -5) now block PVS1. ClinGen PVS1 Decision Tree (Abou Tayoun et al. 2018) specifies canonical +/-1,2 splice sites only. Extended splice region variants retain access to PP3_splice (SpliceAI >= 0.2) at Supporting evidence level.
Clinical trigger: PDE6B c.1401+4_1401+48del in PD2025_061 -- VUS vs ClinVar Benign (2 stars). VEP annotates as splice_donor_5th_base_variant. PVS1 created false conflicting evidence (PVS1 vs BS1+BS2+BP6 = VUS).
Impact: 2 of 18 splice_donor PVS1 variants affected in PD2025_061. 0 P/LP variants affected. Subtractive change (cannot create false P/LP).
Log strings now use CLASSIFIER_VERSION constant instead of hardcoded version.
Reference: HELIX-CR-2026-008, Abou Tayoun 2018 PMID:30192042.
PVS1: NMD_transcript_variant no longer blocks PVS1. VEP NMD_transcript_variant annotation means the transcript UNDERGOES nonsense-mediated decay, which is additional evidence FOR loss-of-function, not against it. The exclusion now targets NMD_escaping_variant (transcripts that escape NMD -- truncated protein expressed, may retain partial function).
Previous behavior: frameshift_variant,NMD_transcript_variant -> PVS1 blocked. New behavior: frameshift_variant,NMD_transcript_variant -> PVS1 applies. frameshift_variant,NMD_escaping_variant -> PVS1 blocked (correct).
Clinical trigger: HG2022_057 review (KCNN2 frameshift).
Reference: VEP documentation (Ensembl), Abou Tayoun et al. 2018 PMID:30192042.
De novo projection (PS2): Prospective computation of whether adding PS2 (confirmed de novo, Strong) would upgrade a VUS to LP or P. Identifies variants where trio confirmation has the highest clinical impact.
Excludes: non-VUS, non-heterozygous, BA1 active, compound het candidates (biologically exclusive with de novo), and conflicting evidence variants.
Minimum evidence guard (v3.11.2): PM2 alone is insufficient for de novo candidacy -- requires at least one additional pathogenic criterion beyond PM2.
De novo projection is informational only -- it does not change the current ACMG classification. It flags variants where parental testing would resolve VUS status.
disease_genes_clinvar CTE: Added review_stars >= 2 filter. The shared CTE that provides gene-level ClinVar disease association evidence (Source 1 of 5 in the disease association check) previously included all ClinVar P/LP genes without a star minimum. This created a circular reference: a 1-star ClinVar P/LP submission simultaneously triggered the ClinVar override AND provided the disease association evidence that allowed the v3.10.0 gate to pass. SHROOM2 (1-star Likely Pathogenic, zero disease association from Orphanet/ClinGen/AR_LOF_GENES/VCEP) self-validated its own gate.
The review_stars >= 2 threshold is consistent with PS1 (ps1_min_stars = 2): a single unreviewed ClinVar submission does not constitute gene-level disease association evidence. Two or more independent submitters with concordant interpretation (2+ stars) provide meaningful gene-disease validation.
Impact on PVS1 and PP3_Strong disease gates: minimal. Both gates reference the same CTE, but genes where PVS1 or PP3_Strong activate have redundant disease evidence from Orphanet, ClinGen haploinsufficiency, curated AR LoF gene list (~150 genes), or VCEP coverage. A gene relying solely on 1-star ClinVar for disease evidence is exactly the type of gene that should not receive PVS1 (Very Strong) or PP3_Strong.
Clinical trigger: SHROOM2 p.Gly211Ser remained LP after v3.10.0 deployment. Now correctly classified as VUS (ClinVar_gated).
ClinVar override: P/LP override with review_stars = 1 (single submitter, no peer review) now requires the gene to have an established disease association via at least one of five sources: ClinVar P/LP (gene-level), Orphanet entry, ClinGen haploinsufficiency score, curated AR LoF gene list (~150 genes), or VCEP coverage. Without this gate, a single ClinVar submission in a gene without any known Mendelian disease mechanism could promote a variant to Likely Pathogenic.
ClinVar override: P/LP assertions with 2+ review stars (multiple concordant submitters) bypass the disease association gate. Benign and Likely Benign ClinVar overrides are not affected by this gate.
ClinVar override: Gated variants (1-star P/LP in genes without disease association) fall through to standard ACMG scoring, which typically produces VUS when no ACMG criteria are activated.
gene_has_disease_association: New boolean column in the criteria evaluation stage. Reuses the same 5-source disease check as the PVS1 gate (v3.8.0) and PP3_Strong gate (v3.8.1). Zero new data sources, zero performance overhead.
clinvar_disease_gated: New audit trail boolean. Criteria string shows "ClinVar_gated" prefix when a ClinVar P/LP override was blocked by the disease association gate, enabling geneticists to identify gated variants.
Clinical trigger: SHROOM2 p.Gly211Ser in PD2025_060 -- classified as LP from a single ClinVar submission (1 star) in a gene with zero disease association across all five reference sources (Orphanet, ClinGen, HPO, VCEP, AR_LOF_GENES), zero constraint data (pLI/LOEUF/mis_z all NULL), and population frequency of 0.23%.
Reference: HELIX-CR-2026-007, Ghosh et al. 2018 PMID:30311383 (limitations of single-submitter ClinVar entries).
PVS1: Last-exon truncating variants downgraded from Very Strong (+8 points) to Strong (+4 points). ClinGen PVS1 Decision Tree (Abou Tayoun et al. 2018) Figure 1, Node 3: truncating variants in the last exon escape nonsense-mediated mRNA decay (NMD) because there is no downstream exon-exon junction complex.
PVS1: is_last_exon detection uses VEP exon_number column (format X/Y, last exon when X == Y). NULL or missing format defaults to Very Strong (conservative fallback). Single-exon genes (1/1) correctly receive PVS1_Strong.
PVS1: Criteria string shows PVS1 for Very Strong (non-last-exon) and PVS1_Strong for last-exon downgraded variants. Consistent with PP3_Strong, BP4_Moderate naming convention.
PVS1: PP3_splice double-counting guard unchanged -- uses original PVS1 condition covering both Very Strong and Strong paths.
PVS1: AR_LOF_GENES last-exon downgrade applies regardless of inheritance mode. NMD escape is a biophysical mechanism independent of AD/AR inheritance.
PVS1: Classification impact -- PVS1_Strong + PM2 = 6 points = Likely Pathogenic (most common minimal combination unaffected). PVS1_Strong + PS1 drops from Pathogenic to Likely Pathogenic.
Clinical trigger: PD2025_058 review (MYO6 p.Lys972GlufsTer5, exon 27/32 -- NOT last exon, unaffected by change).
Reference: Abou Tayoun et al. 2018 PMID:30192042.
BP1: ClinVar pathogenic missense guard -- BP1 no longer applies when ClinVar has a Pathogenic missense variant in the same gene. If ClinVar confirms pathogenic missense, the BP1 premise is falsified. Uses session-local ClinVar data.
BP1: Clinical trigger -- MCCC2 p.Val339Met in PD2025_099 received BP1 despite being ClinVar Pathogenic missense. AR enzyme genes have low mis_z but pathogenic homozygous missense.
PP3_Supporting: Criteria string label changed from PP3 to PP3_Supporting for consistency with PP3_Strong and PP3_Moderate. ClinGen SVI (Walker et al. 2023) recommends explicit evidence strength labeling.
PP3_Strong: Disease association gate -- PP3_Strong (BayesDel >= 0.518, Strong evidence level) now requires the gene to have an established disease association via at least one of five sources: ClinVar P/LP (gene-level), Orphanet gene-disease entry, ClinGen haploinsufficiency score, curated AR LoF gene list, or VCEP coverage.
PP3_Strong: Without this gate, PP3_Strong + PM2 could reach Likely Pathogenic in genes without any known Mendelian disease mechanism. Clinical trigger: PD2025_056 identified POU2F2 p.Val324Met and RYR3 p.Arg285Gln as false LP classifications.
PP3_Strong: Scientific basis -- Pejaver et al. 2022 calibrated BayesDel on ClinVar variants in disease genes; Walker et al. 2023 (ClinGen SVI) Section 3.2 restricts PP3/BP4 calibrated thresholds to genes with established disease mechanism.
PP3_Moderate and PP3_Supporting are NOT gated by disease association (cannot reach LP alone).
PP3_splice path unchanged (splice prediction independent of missense disease mechanism).
PVS1: Disease association gate -- PVS1 now requires the gene to have an established disease association via at least one of five sources: ClinVar P/LP (gene-level), Orphanet gene-disease entry, ClinGen haploinsufficiency score, curated AR LoF gene list (~150 genes), or VCEP coverage. Population constraint (pLI/LOEUF) alone is no longer sufficient.
PVS1: Genes with high pLI but no known Mendelian disease mechanism (e.g., TTC28, ABL1, HTT) are correctly excluded from PVS1. Gene audit across 5 validation cases: 250 genes blocked, 92 genes pass, ACMG SF v3.2 safety check passed (0 of 81 genes affected).
PVS1: Reference: Abou Tayoun et al. 2018 (ClinGen PVS1 Decision Tree) -- "PVS1 is applicable when loss of function is a known mechanism of disease." HELIX-CR-2026-003.
BS1: Constraint-implied AD fallback -- when no inheritance mode data exists from any source (Orphanet, ClinGen, HPO) but the gene is LoF-constrained (pLI > 0.9 or LOEUF < 0.35), BS1 uses the AD threshold (0.1%) instead of the AR default (5%). Resolves the architectural inconsistency where PVS1 treated a gene as AD-haploinsufficient but BS1 treated it as AR.
BS1: Explicit Orphanet AR and curated AR_LOF_GENES added as cascade priorities 5-6, ensuring explicit AR evidence overrides constraint-implied AD inference.
BS1: Cascade expanded from 5 to 8 levels: VCEP > ClinGen HI > Orphanet AD > Orphanet XLD > HPO AD > Orphanet AR > AR_LOF_GENES > Constraint-implied AD > Default AR.
BS1: Criteria string now includes BS1[Orphanet-AR] and BS1[constraint-AD] source annotations.
Reference databases: Orphanet/Orphadata and VCEP gene-specific specifications now documented as separate reference databases (9 total, previously 7).
Clinical validation: TTC28 p.Arg21Ter (PD2025_100) changes from LP to VUS. Discordance analysis: frequency_conflict CRITICAL findings reduced from 1 to 0.
BS1: Orphanet inheritance cascade for inheritance-aware frequency thresholds. 5-level priority: VCEP gene-specific -> ClinGen HI score=3 -> Orphanet AD -> Orphanet XLD -> HPO AD fallback -> AR default.
BS1: Orphanet AD coverage: 1,694 autosomal dominant genes (4x improvement over ~400 from ClinGen HI alone).
BS1: XLD/XLR distinction: X-linked dominant genes (90) receive AD threshold (0.1%); X-linked recessive genes (663) receive AR default (0.05%). XL-unspecified mapped to XLR as conservative default.
BS1: HPO AD fallback (HP:0000006) as Priority 4 catches AD genes not in Orphanet (e.g. KCNQ1, KCNH2 -- also covered by ClinGen HI).
BS1: Criteria string now includes inheritance source annotation: BS1[ClinGen-HI], BS1[Orphanet-AD], BS1[Orphanet-XLD], BS1[HPO-AD].
Reference DB: orphanet_gene_inheritance table added (3,197 genes, LEFT JOIN at classification time, <2ms overhead).
Data source: Orphadata CC-BY-4.0 (INSERM, France). Covers ~6,100 rare diseases with gene-disease-inheritance annotations.
Validation: 9/10 known AD genes confirmed (KCNJ11, GCK, HNF1A, FGFR3, SCN1A, etc.), 4/4 AR-only genes unchanged, 5/5 dual-inheritance genes correct.
BP1: Added mis_z < 2.0 guard to prevent BP1 in missense-constrained genes.
BP1: Genes with pLI < 0.1 AND mis_z >= 2.0 (e.g. KCNJ11, ABCC8, KCNQ1, SCN1A) have missense as primary disease mechanism -- BP1 ("primarily truncating variants cause disease") is factually incorrect for these genes.
Clinical validation: KCNJ11 p.Leu270Val correctly loses BP1 after this change.
PP3: Missense relevance guard added. PP3 (BayesDel) now requires evidence that missense variants cause disease in the gene before applying computational prediction.
PP3 guard has four gates: (1) mis_z > 2.0 for direct missense constraint, (2) mis_z > 1.5 AND pLI < 0.5 for AR genes like BRCA2/MLH1/ATM, (3) mis_z > 0.5 AND pLI > 0.9 for strong AD LoF-intolerant genes, (4) NULL fallback for genes without mis_z but with pLI > 0.5.
PP3_splice path unchanged (splice prediction independent of missense mechanism).
BP4 unchanged (benign prediction consistent with low missense constraint).
Scientific basis: Pejaver et al. 2022 calibrated BayesDel on missense-mechanism genes; applying to LoF-mechanism genes is methodologically incorrect.
PM1: Curated CRITICAL_PFAM_DOMAINS list (~60 domains) replaces generic Pfam presence check.
PM1: Only well-characterized functional domains with documented pathogenic variant clustering trigger PM1.
PM1: Excluded domains: generic structural (Caveolin, coiled-coil), DUF, transmembrane helices, repetitive regions.
PM1: Included domains: kinase catalytic (PF00069, PF07714), DNA-binding (PF00870, PF00505, PF00096, PF00533), ion channel pores (PF00520, PF00029, PF01365), GTPase (PF00071), enzyme active sites (PF00089), and ~40 additional VCEP-documented domains. Domain matching uses Pfam accession numbers since v3.14.0.
Scientific basis: ACMG 2015 requires PM1 for "critical and well-established functional domain without benign variation" -- not any Pfam annotation.
PP2: Now requires mis_z > 2.0 in addition to pLI > 0.5.
PP2: ACMG 2015 requires "missense variants are a common mechanism of disease" -- pLI alone measures LoF constraint, not missense constraint.
PP2: mis_z > 2.0 ensures gene is missense-constrained (Samocha et al. 2014, Karczewski et al. 2020).
Genes like CAV1 (pLI=0.846, mis_z=0.86) with LoF disease mechanism no longer incorrectly trigger PP2.
PVS1: Comprehensive AR LoF gene list expansion from ~35 to ~150 genes
AR LoF gene list covers all major AR disease categories: neurodegeneration, metabolic (amino acid, fatty acid oxidation, glycogen storage, lysosomal, peroxisomal), ciliopathies, sensory (hearing, vision), immune/hematologic, cardiac, neuromuscular, connective tissue, kidney, endocrine, liver, and respiratory
Every gene in the curated list has ClinGen Definitive/Strong validity or equivalent published evidence for biallelic LoF disease mechanism
VCEP AR genes (CDH23, GJB2, MYO7A, PAH, SLC26A4, USH2A) handled via VCEP pvs1_applicable gate, not duplicated in curated list
Homozygous reference genotype guard: variants with genotype hom_ref (0/0) excluded from ACMG classification
hom_ref variants arise from multi-allelic VCF sites where the sample has 0 ALT reads for a specific allele but was included due to another ALT allele at the same position
Without this guard, hom_ref frameshift/stop variants in LoF-intolerant genes received PVS1+PM2 = Likely Pathogenic despite not being present in the patient
20 false Pathogenic/Likely Pathogenic classifications eliminated in validation (Case 19)
PM2: NULL gnomAD allele frequency now correctly satisfies PM2 (absent from controls = never observed in ~800K individuals)
PM2: Previous behavior required frequency data to be present (non-NULL), incorrectly excluding truly absent variants
PVS1: Curated AR LoF gene list (~35 genes) replaces broad HPO-based bypass (v3.6 used HP:0000007 with 3,127 AR genes -- too broad, caused 43 false Likely Pathogenic)
PVS1: Curated list includes only genes with Definitive/Strong evidence for biallelic LoF disease mechanism (ClinGen Gene-Disease Validity, literature)
PVS1: Autosomal recessive genes bypass pLI/LOEUF constraint gate using HPO inheritance annotation (HP:0000007)
PVS1: Aligned with ClinGen PVS1 Decision Tree (Abou Tayoun et al. 2018) which does not require population constraint for genes with established LoF disease mechanism in recessive inheritance
HPO database now used for PVS1 AR gene identification in addition to PP4 phenotype matching
Note: HPO-based bypass replaced by curated gene list in v3.6.1 due to excessive breadth
ClinGen VCEP gene-specific specification overlay (optional, enabled by default)
BA1, BS1, PM2: gene-specific frequency thresholds from published VCEP specifications (~50-60 genes)
PVS1: gene-specific applicability gate (disabled for gain-of-function genes)
VCEP audit trail: criteria string includes [VCEP:Panel vX.Y] marker when gene-specific thresholds applied
VCEP toggle: can be enabled/disabled per case in case settings
Source: ClinGen Criteria Specification Registry (CSpec)
Bayesian point-based classification framework (Tavtigian et al. 2018, 2020) replaces sequential 18-rule evaluation
All 18 original ACMG combining rules produce identical results under the point system (backward compatible)
Point system fills gaps: evidence combinations not covered by original 18 rules now classified properly
PP3/BP4: BayesDel_noAF single-tool replaces weighted 6-predictor consensus (ClinGen SVI calibration, Pejaver et al. 2022)
PP3 evidence strength modulation: Strong (>= 0.518), Moderate (0.290-0.517), Supporting (0.130-0.289)
BP4 evidence strength modulation: Moderate (<= -0.361), Supporting (-0.360 to -0.181)
PM1 + PP3 double-counting guard: combined evidence capped at Strong equivalent (4 points)
Continuous confidence scores derived from point distance to classification threshold boundary
Conflicting evidence handled by point summation (more nuanced than binary VUS default)
High-confidence conflict safety check: Strong/Very Strong pathogenic vs. Strong benign flagged for manual review
Legacy predictors (SIFT, AlphaMissense, MetaSVM, DANN, PhyloP, GERP) retained as display data only
SpliceAI PP3 threshold aligned to ClinGen SVI 2023 recommendation (lowered from 0.5 to 0.2)
PP3_splice excluded when PVS1 applies (ClinGen SVI double-counting guard)
BP7 upgraded: synonymous + not splice_region + SpliceAI <= 0.1 (Walker et al. Figure 4)
BP7 conservation filter intentionally omitted per Walker et al. Table S13
Evidence strength modulation (PP3_Moderate for high SpliceAI) deferred pending VCEP specifications
SpliceAI integration: PP3_splice path for splice evidence
BP4 SpliceAI guard: requires max_score < 0.1 for benign consensus
Criteria string distinguishes PP3 (missense) from PP3_splice (splice)
Maximum sensitivity approach: removed all frequency and impact pre-filtering
All quality-passing variants proceed through classification
Clinicians decide clinical relevance using classification + annotations
Conflicting evidence priority: pathogenic + benign evidence produces VUS
BA1 stand-alone override: allele frequency > 5% always classified Benign
ClinVar override restricted to non-conflicting evidence only
SQL-based classification engine (100x performance improvement)
PM2: required non-NULL frequency data (corrected in v3.6.1)
References
Tavtigian SV, Greenblatt MS, Harrison SM, et al. Modeling the ACMG/AMP variant classification guidelines as a Bayesian classification framework.
Human Mutation. 2018;39(11):1485-1492.
PMID: 30311386Tavtigian SV, Harrison SM, Boucher KM, Biesecker LG. Fitting a naturally scaled point system to the ACMG/AMP variant classification guidelines.
Human Genetics. 2020;139(8):1057-1067.
PMID: 32666219Pejaver V, Byrne AB, Feng BJ, et al. Calibration of computational tools for missense variant pathogenicity classification and ClinGen recommendations for PP3/BP4 criteria.
American Journal of Human Genetics. 2022;109(12):2163-2177.
PMID: 36413997Richards S, Aziz N, Bale S, et al. 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.
Genetics in Medicine. 2015;17(5):405-424.
PMID: 25741868Walker LC, Hoya M, Wiggins GAR, et al. Using the ACMG/AMP framework to capture evidence related to predicted and observed impact on splicing: Recommendations from the ClinGen SVI Splicing Subgroup.
American Journal of Human Genetics. 2023;110(7):1046-1067.
PMID: 37352859Jaganathan K, Kyriazopoulou Panagiotopoulou S, McRae JF, et al. Predicting Splicing from Primary Sequence with Deep Learning.
Cell. 2019;176(3):535-548.e24.
PMID: 30661751Karczewski KJ, Francioli LC, Tiao G, et al. The mutational constraint spectrum quantified from variation in 141,456 humans.
Nature. 2020;581(7809):434-443.
PMID: 32461654Cheng J, Novati G, Pan J, et al. Accurate proteome-wide missense variant effect prediction with AlphaMissense.
Science. 2023;381(6664):eadg7492.
PMID: 37733863McLaren W, Gil L, Hunt SE, et al. The Ensembl Variant Effect Predictor.
Genome Biology. 2016;17(1):122.
PMID: 27268795Landrum MJ, Lee JM, Benson M, et al. ClinVar: improving access to variant interpretations and supporting evidence.
Nucleic Acids Research. 2018;46(D1):D1062-D1067.
PMID: 29165669Abou Tayoun AN, Pesaran T, DiStefano MT, et al. Recommendations for interpreting the loss of function PVS1 ACMG/AMP variant criterion.
Human Mutation. 2018;39(11):1517-1524.
PMID: 30192042Josephs KS, Roberts AM, Grozeva D, et al. Beyond gene-disease validity: capturing structured data on inheritance, allelic requirement, disease-relevant variant classes, and mechanism for inherited cardiac conditions.
Genome Medicine. 2023;15:86.
PMID: 37884986Cummings BB, Karczewski KJ, Kosmicki JA, et al. Transcript expression-aware annotation improves rare variant interpretation.
Nature. 2020;581(7809):452-458.
PMID: 32461655Questions About Our Methodology?
We welcome technical questions from clinical geneticists and laboratory directors. Transparency is foundational to clinical trust.