ERBB2 c.2584A>G, p.(Thr862Ala)

Reasoning

ERBB2 c.2584A>G (p.Thr862Ala / T862A) has been directly and functionally characterized in well-established assays in a single, rigorous study (PMID 22046346; Kavuri et al., PLoS One 2011). In HEK293 cells, T862A retained autokinase activity and, compared with WT ERBB2, displayed enhanced activation of JNK/SAPK and, to a lesser extent, ERK1/2; ligand-induced heterodimerization with EGFR/ERBB3 was not impaired. In NMuMg (non-transformed mouse mammary epithelial) cells, ERBB2-T862A formed more colonies than WT ERBB2 in soft agar, indicating enhanced transforming potential. In Ba/F3 cells, T862A showed a lapatinib IC50 of ~125 nM versus ~30 nM for WT (Table 1), classifying it as lapatinib-sensitive-but-relative-resistant (IC50 between 30 nM and 1 uM), and T862A was recovered as a secondary lapatinib-resistance mutation in an unbiased ENU mutagenesis + 2 uM lapatinib selection screen of WT-ERBB2 Ba/F3 cells. All assays included appropriate controls (WT ERBB2, untransfected, and EGFR/ERBB3 co-stimulation conditions); mutations were sequence-confirmed. The direction of effect (gain of function/activating) is concordant with OncoKB's variant-specific curation (biological effect: gain-of-function; oncogenicity: Oncogenic) and with the variant's somatic recurrence (COSMIC COSV54062276, n=27; statistically significant hotspot at T862). Under generic ACMG/AMP 2015 (PMID 25741868), well-established functional studies demonstrating a damaging effect support PS3; strength calibrated at strong per SVI PS3 guidance (Brnich et al. 2019, PMID 31112341) given multiple orthogonal, validated assay systems (signaling, transformation, drug sensitivity) with concordant results and controlled conditions. Caveat: functional data derive from somatic cancer model systems and a single publication; no ERBB2 VCEP specification exists to alter this calibration.

Rule

PS3: Well-established in vitro or in vivo functional studies supportive of a damaging effect on the gene or gene product (Richards et al. 2015, PMID 25741868); strength calibration per SVI PS3 recommendations (Brnich et al. 2019, PMID 31112341).

Evidence considered

• PMID:22046346 - Kavuri et al. 2011: ERBB2-T862A tested in HEK293 (autophosphorylation; JNK/SAPK and ERK1/2 activation vs WT; EGFR/ERBB3 heterodimerization controls), NMuMg soft-agar colony formation (enhanced vs WT), and Ba/F3 MTS drug-sensitivity assays (lapatinib IC50 ~125 nM vs WT ~30 nM; AEE788 IC50 ~191 nM vs WT ~257 nM).
• PMID:22046346 - T862A recovered as a secondary lapatinib-resistance mutation in an ENU mutagenesis + 2 uM lapatinib selection screen of WT-ERBB2 Ba/F3 cells (Figure S4).
• oncokb - Variant-specific curated annotation for ERBB2 T862A: biological effect gain-of-function; oncogenicity Oncogenic; cites PMID 22046346 among supporting literature.

Still missing

• Functional evidence rests on a single publication (PMID 22046346); independent replication of the T862A activating effect would strengthen confidence.
• Assays were performed in somatic cancer model systems (HEK293, NMuMg, Ba/F3); direct relevance to a germline disease mechanism is inferred, not demonstrated.
• No ERBB2 VCEP/gene-specific framework exists to define assay thresholds or PS3 strength; calibration follows generic ACMG/AMP 2015 with SVI PS3 strength guidance (PMID 31112341).

Sources

Reasoning

PM1 is met. Residue 862 lies within the ERBB2 tyrosine kinase domain in the activation-loop/ATP-site-base region (exon 21; PMID:22046346, Table 1) — a critical and well-established functional domain (protein tyrosine kinase catalytic domain). The residue is a statistically significant cancer mutational hotspot (cancerhotspots.org record 'ERBB2 T862'; hotspot-detection method per PMID:26619011), with 27 somatic occurrences of c.2584A>G in COSMIC (COSV54062276), and a variant-specific OncoKB entry curating T862A as Oncogenic/gain-of-function. The 'without benign variation' condition is satisfied: the variant is absent from gnomAD v2.1, v4.1 and gnomAD-Canada v1.0 (0 alleles). Caveat: the mutational-hotspot evidence is somatic in origin (cancer cohorts), so the primary support for PM1 is the functional-domain prong (kinase activation loop), which is germline-agnostic; the somatic hotspot data are used as corroboration. This is documented in warnings for human review.

Rule

PM1: Located in a mutational hotspot and/or critical and well-established functional domain (e.g., active site of an enzyme) without benign variation (Richards et al. 2015, PMID 25741868).

Evidence considered

• ERBB2 residue 862 lies in the protein tyrosine kinase domain, activation-loop/ATP-site-base region, exon 21 (PMID:22046346, Table 1).
• cancerhotspots.org records ERBB2 T862 as a statistically significant mutational hotspot (case evidence bundle, evidence.json 'hotspots' result; record_id 'ERBB2 T862'; hotspot-detection methodology per PMID:26619011).
• COSMIC: 27 somatic occurrences of ERBB2 c.2584A>G (COSV54062276) (case evidence bundle, evidence.json 'cosmic' result).
• No benign population variation: variant absent from gnomAD v2.1, gnomAD v4.1 and gnomAD-Canada v1.0 (source_registry keys: gnomad_v2, gnomad_v4, gnomad_canada).
• OncoKB variant-specific entry: ERBB2 T862A curated Oncogenic, gain-of-function, is_variant_specific true (source_registry key: oncokb), corroborating that residue 862 is functionally critical.
• Gene-level context: HER2 somatic mutations are recurrent, domain-clustered (extracellular, transmembrane, kinase) and mostly at known hotspot positions (PMID:29420467).

Still missing

• PM1 hotspot evidence derives from somatic cancer resources (cancerhotspots.org statistical hotspot; COSMIC n=27); germline applicability of somatic hotspot data is a judgment call, though the functional-domain prong (kinase activation loop) does not depend on it.
• No curated germline functional-domain residue list (e.g., ClinGen/VCEP) was available for ERBB2; domain annotation rests on the literature (PMID:22046346) and standard protein-domain knowledge (PTKc domain).

Sources

Reasoning

PM2 is met: the variant is absent (AF = 0) from gnomAD v2.1 exomes, gnomAD v4.1 exomes, and gnomAD-Canada v1.0 (HostSeq genomes), satisfying 'absent in controls' per Richards et al. 2015 (PMID 25741868). Strength is set to supporting (PM2_Supporting), consistent with current ClinGen SVI calibration practice. HUMAN REVIEW REQUIRED for applicability: c.2584A>G (p.Thr862Ala) is a recurrent somatic oncogenic ERBB2 mutation (OncoKB variant-specific curation: Oncogenic, gain-of-function; COSMIC COSV54062276 with 27 somatic observations; statistically significant cancer hotspot at residue T862). ERBB2 is a proto-oncogene without an established germline disease phenotype, and absence from population databases is expected for somatic hotspot variants, so PM2 may be non-informative for germline pathogenicity in this context. The population evidence itself (absence across three large datasets) is solid; whether PM2 should be counted in this germline classification given the somatic mechanism requires human adjudication.

Rule

PM2 (Richards et al. 2015, PMID 25741868): absent in controls (ExAC/gnomAD) or at extremely low frequency if recessive; applied at supporting strength (PM2_Supporting) per ClinGen SVI calibration practice.

Evidence considered

• gnomad_v2: variant absent from gnomAD v2.1 exomes (AF = 0)
• gnomad_v4: variant absent from gnomAD v4.1 exomes (AF = 0)
• gnomad_canada: variant absent from gnomAD-Canada v1.0 (HostSeq genomes; AF = 0)
• oncokb: ERBB2 T862A curated as Oncogenic with gain-of-function biological effect (variant-specific entry), establishing a somatic cancer context for the absence

Still missing

• Applicability of PM2 is uncertain for a recurrent somatic oncogenic hotspot in a gene (ERBB2) without an established germline disease phenotype; human review required.
• gnomAD v4.1 genomes were not queried; absence is confirmed in gnomAD v2.1 exomes, gnomAD v4.1 exomes, and gnomAD-Canada v1.0 genomes only.

Sources

Reasoning

Missense variant NM_004448.3:c.2584A>G (p.Thr862Ala) in ERBB2 exon 21 (c.2569-2754; variant at c.2584, deep exonic, not near splice consensus). No CSPEC/VCEP specification exists for ERBB2 germline interpretation, so generic ACMG/AMP 2015 (Richards et al., PMID 25741868) applies. REVEL v1.3 score 0.674 falls within the ClinGen SVI-calibrated PP3 supporting band (0.644-0.773; Pejaver et al., Am J Hum Genet 2022, PMID 36413997), providing one calibrated line of computational evidence for a deleterious effect; PP3 is therefore met at supporting strength. The splice-impact sub-path (SpliceAI) returned no score for this variant and could not contribute; the BayesDel noAF score (0.102536) was retrieved but is treated as insufficiently calibrated (no verified published threshold available to this pipeline) and was not used. REVEL is counted once and only for PP3 (no double-counting with BP4/BP7).

Rule

Generic ACMG/AMP PP3 (Richards et al. 2015, PMID 25741868) applied via ClinGen SVI calibrated REVEL thresholds (Pejaver et al. 2022, PMID 36413997): REVEL 0.644-0.773 = PP3 supporting.

Evidence considered

• REVEL v1.3 score 0.674 for NM_004448.3:c.2584A>G (p.Thr862Ala) falls in the PP3 supporting band (0.644-0.773) per ClinGen SVI REVEL calibration (Pejaver et al., Am J Hum Genet. 2022;109(12):2163-2177; PMID 36413997).
• SpliceAI Lookup returned no score (evidence.json results.spliceai, summary: 'No SpliceAI score returned'); no splice-impact prediction available to add to PP3.
• BayesDel noAF score 0.102536 (prefetch.json steps.bayesdel) excluded from PP3: treated as insufficiently calibrated for PP3/BP4 because no verified published threshold is available to this pipeline.

Still missing

• SpliceAI Lookup returned no score for NM_004448.3:c.2584A>G; the splicing-impact sub-path of PP3 could not be evaluated.
• Only one calibrated missense predictor (REVEL) was available: BayesDel is treated as insufficiently calibrated (no verified published threshold) and aGVGD data are not present in this case.

Sources

Reasoning

PS1 requires the same amino acid change (p.Thr862Ala) to have been established as pathogenic via a DIFFERENT nucleotide change. Two independent reasons this cannot be met: (1) no comparator variant exists — ClinVar has no record for ERBB2 c.2584A>G under any queried HGVS or genomic-coordinate form, and none of the reviewed publications (PMID:22046346, 26619011, 28572459, 29420467) documents p.T862A arising from any other nucleotide change; (2) structurally, the reference codon at residue 862 is ACA, and the only single-nucleotide substitution that produces p.T862A is c.2584A>G itself (A>G at the first codon position); every other single-nucleotide change at codon 862 yields a different amino acid or is synonymous. The residue's oncogenic evidence (OncoKB 'Oncogenic'; COSMIC somatic count; PMID:22046346 functional assays) concerns this exact nucleotide change and is somatic, not a germline pathogenic classification of a different allele, so it cannot satisfy PS1.

Rule

PS1: Same amino acid change as a previously established pathogenic variant regardless of nucleotide change (Richards et al. 2015, PMID 25741868).

Evidence considered

• ClinVar: no record for ERBB2 c.2584A>G under any queried form (ERBB2[GENE] AND c.2584A>G[VARNAME], NM_004448.3:c.2584A>G, NC_000017.10:g.37881392A>G) — no established pathogenic variant at this position (source_registry key: clinvar).
• Reference codon at residue 862 is ACA; the only single-nucleotide substitution producing p.T862A is c.2584A>G itself (A>G at the first codon position), so no 'different nucleotide change, same amino acid' PS1 scenario exists for this residue.
• Reviewed literature (PMID:22046346) reports only T862A itself at residue 862; no same-amino-acid change from an alternate nucleotide change is reported.

Still missing

• No established-pathogenic same-amino-acid comparator (p.T862A via a different nucleotide change) is documented in ClinVar or the reviewed literature.
• Given reference codon ACA at residue 862, c.2584A>G is the only single-nucleotide route to p.T862A, so a PS1 scenario is structurally impossible at this residue.

Sources

Reasoning

PM5 requires a DIFFERENT missense change at residue 862 that has been established as pathogenic. No such comparator is documented: ClinVar has no ERBB2 record at this position; the automated PM5 candidate harvest (pm5_candidates.json) returned 0 same-residue candidates with pm5_mode 'unknown' and classic same-residue harvesting skipped (per its hermes_instruction, PM5 was not applied from that artifact, and manual review of the bundle found no same-residue pathogenic comparator); and the reviewed literature (PMID:22046346) reports only T862A itself at residue 862, not an alternate missense (e.g., p.T862S/I/N) with established pathogenicity.

Rule

PM5: Novel missense change at an amino acid residue where a different missense change determined to be pathogenic has been seen before (Richards et al. 2015, PMID 25741868).

Evidence considered

• ClinVar: no record for ERBB2 c.2584A>G and no documented pathogenic alternate missense at residue 862 (source_registry key: clinvar).
• Automated PM5 candidate collection: total_candidates_seen 0, same_residue_candidates 0, pm5_mode 'unknown', classic PM5 search not eligible (source_registry key: pm5_candidates).
• Reviewed literature (PMID:22046346) reports no missense change at residue 862 other than T862A itself.

Still missing

• No alternate missense change at residue 862 (e.g., p.T862S/I/N) with established pathogenicity is documented in the case bundle.
• Automated PM5 candidate harvesting was skipped (pm5_candidates: pm5_mode 'unknown', 0 same-residue candidates), so the negative finding rests on manual review of ClinVar absence and the available literature.

Sources

Reasoning

PP2 requires BOTH a low rate of benign missense variation in the gene AND missense as a common mechanism of disease. The mechanism condition is not satisfied in the germline framework: ERBB2 has no established germline disease in which missense variants are a common mechanism — its documented disease relevance is somatic gain-of-function oncogenesis (activating missense, exon-20 insertions, amplification; OncoKB gain-of-function context; PMID:29420467 somatic HER2-mutant cancer context). The low-benign-missense-rate condition could not be quantified because gene-level missense constraint metrics (e.g., gnomAD missense Z-score / observed-expected ratio) are not present in the case bundle. Because a necessary condition of PP2 fails and the other is unverifiable, PP2 is not met.

Rule

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 (Richards et al. 2015, PMID 25741868).

Evidence considered

• ERBB2 germline disease with missense as a common mechanism is not established; the gene's documented disease relevance is somatic gain-of-function oncogenesis (source_registry keys: oncokb — gain-of-function biological effect; PMID:29420467 — somatic HER2-mutant cancer context).
• Gene-level missense constraint metrics (gnomAD missense Z-score / o-e) are not present in the case bundle; only variant-level absence in gnomAD v2.1/v4.1 is available (source_registry keys: gnomad_v2, gnomad_v4).

Still missing

• Gene-level missense constraint metrics (e.g., gnomAD missense Z-score / o-e ratio) are not present in the case bundle, so the 'low rate of benign missense variation' condition could not be quantified.
• No established germline disease for ERBB2 in which missense variants are a common mechanism is documented; the gene's disease mechanism is somatic gain-of-function oncogenesis.

Sources

Reasoning

PP5 requires a reputable source recently reporting the variant as pathogenic, applied here strictly for an exact-variant ClinVar expert-panel classification (Pathogenic/Likely pathogenic). The variant is absent from ClinVar entirely (evidence.json clinvar search_status: no_result, variation_ids: [], 'This variant is absent from ClinVar.'), so no exact-variant ClinVar expert-panel classification exists. OncoKB's curated 'Oncogenic' label was considered as a potential reputable source but is not a ClinVar expert-panel classification and must not trigger PP5 per the governing instruction. PP5 is not met.

Rule

PP5: Reputable source recently reports variant as pathogenic, but the evidence is not available to the lab to perform an independent evaluation (Richards et al. 2015, PMID 25741868). Applied only for an exact-variant ClinVar expert-panel Pathogenic/Likely pathogenic classification.

Evidence considered

• clinvar
• oncokb

Still missing

• No ClinVar record exists for c.2584A>G; if an expert-panel (e.g., ClinGen VCEP) classification is later deposited, PP5 should be re-evaluated.
• No ERBB2 ClinGen CSPEC/VCEP specification exists (generic ACMG fallback in use), so no VCEP-derived PP5 pathway is available.

Sources

Reasoning

BA1 requires an allele frequency greater than 5% in a large reference population (Richards et al. 2015, PMID 25741868; stand-alone benign). NM_004448.3:c.2584A>G (p.Thr862Ala) is absent (AF = 0) from gnomAD v2.1 exomes (GRCh37, 17-37881392-A-G), gnomAD v4.1 exomes (GRCh38, chr17-39725139-A-G), and gnomAD-Canada v1.0 (HostSeq genomes). With zero observed alleles in every queried population dataset, the allele frequency is far below any benign stand-alone threshold, so BA1 is not met. No ancestry-stratified frequencies exist because the variant is absent across the board; no threshold artifact is present (the absence is not a coordinate mismatch: GRCh37 and GRCh38 queries both matched the normalized variant coordinates).

Rule

BA1 (Richards et al. 2015, PMID 25741868): allele frequency >5% in reference population, stand-alone benign. Not met because the variant is absent (AF = 0) from all queried population databases.

Evidence considered

• gnomad_v2: variant absent from gnomAD v2.1 exomes (AF = 0; search_status absent)
• gnomad_v4: variant absent from gnomAD v4.1 exomes (AF = 0; search_status absent)
• gnomad_canada: variant absent from gnomAD-Canada v1.0 (HostSeq genomes; AF = 0)

Still missing

• Only exome datasets (gnomAD v2.1 exomes, gnomAD v4.1 exomes) plus gnomAD-Canada v1.0 genomes were queried; gnomAD v4.1 genomes were not queried. Outcome is threshold-independent here because AF = 0.

Sources

Reasoning

BS1 requires an allele frequency greater than expected for the disorder given disease prevalence and penetrance. The variant is absent (AF = 0) from gnomAD v2.1 exomes, gnomAD v4.1 exomes, and gnomAD-Canada v1.0, so it cannot exceed any disorder-expected frequency threshold regardless of disease model. Note that ERBB2 has no established germline Mendelian phenotype with defined prevalence/penetrance (this variant is a known somatic cancer mutation), which would make an expected-frequency anchor difficult to set, but the outcome is unchanged because AF = 0.

Rule

BS1 (Richards et al. 2015, PMID 25741868): allele frequency greater than expected for the disorder. Not met: AF = 0 in all queried population datasets.

Evidence considered

• gnomad_v2: variant absent from gnomAD v2.1 exomes (AF = 0)
• gnomad_v4: variant absent from gnomAD v4.1 exomes (AF = 0)
• gnomad_canada: variant absent from gnomAD-Canada v1.0 (HostSeq genomes; AF = 0)

Still missing

• Germline disease prevalence and penetrance for ERBB2 are not established, so a formal BS1 expected-frequency threshold cannot be anchored; outcome is threshold-independent here because AF = 0.

Sources

Reasoning

BS2 requires observation of the variant in healthy adult individuals (homozygous in recessive disorders; heterozygous/hemizygous in dominant or X-linked disorders) at a frequency consistent with benignity, with full penetrance expected at an early age. No such observations exist: the variant is absent from gnomAD v2.1 exomes, gnomAD v4.1 exomes, and gnomAD-Canada v1.0 (no heterozygotes and no homozygotes in any dataset), and it is absent from ClinVar (no germline clinical observations). BS2 is therefore not met.

Rule

BS2 (Richards et al. 2015, PMID 25741868): observed in a healthy adult individual; full penetrance expected at an early age. Not met: zero observations in population or ClinVar data.

Evidence considered

• gnomad_v2: no heterozygote or homozygote observations (variant absent from gnomAD v2.1 exomes)
• gnomad_v4: no heterozygote or homozygote observations (variant absent from gnomAD v4.1 exomes)
• gnomad_canada: no observations (variant absent from gnomAD-Canada v1.0 genomes)
• clinvar: no germline submissions or clinical observations for this variant

Still missing

• No curated healthy-control or case-control germline cohorts beyond gnomAD and ClinVar were reviewed for this group.

Sources

Reasoning

BS3 requires well-established functional studies demonstrating NO damaging effect on protein function or splicing. The available well-established functional data for ERBB2-T862A (PMID 22046346) demonstrate the opposite: enhanced JNK/SAPK and ERK1/2 activation, enhanced anchorage-independent growth (soft agar colony formation), and altered (relative-resistant) lapatinib sensitivity versus WT ERBB2, with T862A also recovered as a secondary lapatinib-resistance mutation. No study reports neutral or loss-of-function behavior for this variant, and no splicing effect is expected or demonstrated (missense, exon 21; SpliceAI returned no score). BS3 is therefore not met.

Rule

BS3: Well-established in vitro or in vivo functional studies show no damaging effect on protein function or splicing (Richards et al. 2015, PMID 25741868).

Evidence considered

• PMID:22046346 - functional data for ERBB2-T862A show a damaging (activating) effect, contradicting the 'no damaging effect' requirement of BS3.

Still missing

• No functional study demonstrating absence of a damaging effect for T862A exists in the reviewed evidence set; BS3 cannot be applied.
• SpliceAI returned no score for c.2584A>G, so a splicing-based BS3 rationale could not be assessed.

Sources

Reasoning

BP1 applies only to genes for which primarily truncating variants are known to cause disease. ERBB2 is not such a gene: no germline disease driven primarily by truncating ERBB2 variants is established, and the well-documented disease mechanism of ERBB2 is somatic gain-of-function alteration (activating missense, exon-20 insertions, amplification) in cancer. The variant itself is a kinase-domain missense with documented activating/transforming properties (PMID:22046346) and variant-specific OncoKB curation as Oncogenic/gain-of-function. The premise of BP1 therefore fails and the criterion is not met.

Rule

BP1: Missense variant in a gene for which primarily truncating variants are known to cause disease (Richards et al. 2015, PMID 25741868).

Evidence considered

• ERBB2 is not a gene for which primarily truncating variants are known to cause disease; the variant is a documented activating/oncogenic kinase-domain missense (source_registry keys: oncokb — Oncogenic/gain-of-function; PMID:22046346 — transforming T862A functional assays; PMID:29420467 — somatic HER2-mutant disease context).
• No germline disease phenotype driven primarily by ERBB2 truncating variants is documented anywhere in the case bundle (ClinVar has no ERBB2 c.2584A>G records; no CSPEC/VCEP germline specification exists for ERBB2).

Still missing

• No germline disease with primarily truncating ERBB2 variants is established; if such a mechanism were documented in the future, BP1 would require re-evaluation.

Sources

Reasoning

BP4 requires multiple lines of computational evidence suggesting no impact on the gene/gene product (conservation, evolutionary, splicing impact). No such evidence exists for this variant: REVEL v1.3 score 0.674 is far above the SVI-calibrated BP4 supporting upper threshold of 0.183 (Pejaver et al., PMID 36413997), i.e. the only calibrated missense predictor available points toward pathogenicity, not neutrality; SpliceAI returned no score, so no splice evidence of absent impact could be obtained; the BayesDel noAF score (0.102536) is treated as insufficiently calibrated (no verified published threshold available to this pipeline) and cannot support BP4. No calibrated computational evidence suggests a benign impact, therefore BP4 is not met.

Rule

Generic ACMG/AMP BP4 (Richards et al. 2015, PMID 25741868): requires computational evidence suggesting no impact; none of the consulted calibrated evidence (REVEL via PMID 36413997 thresholds; SpliceAI; BayesDel) supports a benign impact.

Evidence considered

• REVEL v1.3 score 0.674 is far above the SVI-calibrated BP4 supporting upper threshold of 0.183 (Pejaver et al., Am J Hum Genet. 2022;109(12):2163-2177; PMID 36413997); REVEL provides no benign-impact evidence for BP4.
• SpliceAI Lookup returned no score (evidence.json results.spliceai, summary: 'No SpliceAI score returned'); no splice evidence of absent impact available for BP4.
• BayesDel noAF score 0.102536 (prefetch.json steps.bayesdel) excluded from BP4: treated as insufficiently calibrated for PP3/BP4 because no verified published threshold is available to this pipeline.

Still missing

• SpliceAI returned no score for this variant; absence of a splice-impact effect could not be confirmed for BP4.
• BayesDel noAF score (0.102536) retrieved but treated as insufficiently calibrated for PP3/BP4 (no verified published threshold available to this pipeline); aGVGD data not present in this case.

Sources

Reasoning

BP6 requires a reputable source recently reporting the variant as benign, applied here strictly for an exact-variant ClinVar expert-panel classification (Benign/Likely benign). The variant is absent from ClinVar (evidence.json clinvar search_status: no_result, variation_ids: []), so no exact-variant ClinVar expert-panel classification exists. No source in the bundle reports this variant as benign. BP6 is not met.

Rule

BP6: Reputable source recently reports variant as benign, but the evidence is not available to the lab to perform an independent evaluation (Richards et al. 2015, PMID 25741868). Applied only for an exact-variant ClinVar expert-panel Benign/Likely benign classification.

Evidence considered

• clinvar

Still missing

• No ClinVar record exists for c.2584A>G; if an expert-panel Benign/Likely benign classification is later deposited, BP6 should be re-evaluated.

Sources

Reasoning

No de novo occurrence of NM_004448.3:c.2584A>G is reported anywhere in this case. ClinVar has no record for the variant (search_status: no_result), so no proband observation with confirmed paternity/maternity exists from that source, and no case report or cohort study in the retrieved literature documents a de novo event. De novo status can neither be established nor excluded from available evidence. Per generic ACMG/AMP 2015 (PMID 25741868), PS2 requires a de novo variant with confirmed parental relationships; insufficient evidence, so the criterion is not assessed.

Rule

ACMG/AMP 2015 (PMID 25741868): PS2 - de novo (paternity and maternity confirmed), strong

Evidence considered

• ClinVar search returned no record for NM_004448.3:c.2584A>G (search_status: no_result), so no de novo or family observations are available
• gnomAD v2.1/v4.1/Canada absence is population-frequency data only and does not inform de novo status
• Full texts of the four OncoKB-sourced publications were searched for de novo/segregation/germline/parental content with zero hits; none document a proband or parental genotypes

Still missing

• No proband/parental trio testing data available
• No confirmed de novo occurrence (paternity and maternity confirmed) reported in ClinVar or any retrieved literature

Sources

Reasoning

PS4 (case-control enrichment) cannot be evaluated: no case-control study comparing the prevalence of ERBB2 c.2584A>G (p.Thr862Ala) in affected individuals versus controls is present in the evidence bundle. The available signals are somatic-recurrence based, not case-control: COSMIC records 27 somatic occurrences (COSV54062276) and OncoKB curates the variant as Oncogenic/gain-of-function with a statistically significant hotspot annotation at residue T862. PMID:22046346 reports T862A in a primary gastric cancer cohort (Table 1) and functionally characterizes it, but provides no unaffected control arm. Population databases (gnomAD v2.1, v4.1, gnomAD-Canada) show the variant absent, which speaks to rarity rather than case-control enrichment. Per generic ACMG/AMP 2015 (PMID:25741868), PS4 requires a statistically significant increase in affected vs controls; somatic recurrence in tumor cohorts is not a PS4-grade case-control design. Evidence insufficient: status not_assessed.

Rule

PS4: The prevalence of the variant in affected individuals is significantly increased compared with controls (Richards et al. 2015, PMID 25741868).

Evidence considered

• clinvar
• oncokb
• PMID:22046346

Still missing

• No case-control study (affected vs control allele counts with statistics) for c.2584A>G exists in the evidence bundle.
• COSMIC somatic recurrence (n=27) and cancerhotspots.org T862 hotspot annotation are not case-control data; a cohort study with a defined control arm is required to apply PS4.
• Hotspot evidence has a parsing warning ('Regex extraction partial; verify screenshot') and is not a source_registry key.

Sources

Reasoning

PM3 (for recessive disorders, detected in trans with a pathogenic variant) cannot be adjudicated because the evidence bundle contains no proband-level observations, no allele-phase (trans/cis) information, and no pathogenic variant reported in trans with c.2584A>G. No CSPEC/VCEP specification exists for ERBB2 (cspec.found=false; vcep_materials empty), so the generic ACMG/AMP 2015 framework (PMID 25741868) governs. In this case the variant is a recurrent somatic cancer hotspot (OncoKB: Oncogenic, gain-of-function, variant-specific; COSMIC COSV54062276, n=27 somatic) and is absent from ClinVar and gnomAD v2.1/v4.1, so no germline biallelic disease context can be anchored. The pvs1_gene_context supporting sources do not establish an ERBB2 germline recessive phenotype (they are largely other-gene literature, e.g., ERBB3 biallelic syndrome, PMID:31752936). With no recessive disease context and no trans-phase observation of a pathogenic variant, PM3 is not assessable; the deciding data are entirely absent.

Rule

PM3 (ACMG/AMP 2015, PMID 25741868): for recessive disorders, detected in trans with a pathogenic variant. Not assessable here: no recessive disease context and no trans-phase pathogenic variant observation for c.2584A>G.

Evidence considered

• generic_acmg_combination_rules: governing framework is ACMG/AMP 2015 (PMID 25741868); no CSPEC/VCEP for ERBB2, so no gene-specific PM3 specification applies
• clinvar: variant absent from ClinVar; no germline submissions, no phase or family data
• gnomad_v2 / gnomad_v4: variant absent from population databases; no carrier observations to support any biallelic model
• oncokb: ERBB2 T862A curated as Oncogenic, gain-of-function (variant-specific) in a somatic cancer context; not germline phenotype evidence
• pvs1_gene_context: germline disease context for ERBB2 not established; supporting sources are mostly other-gene literature and cannot anchor a recessive inheritance model

Still missing

• No proband-level observation of c.2584A>G in any germline disease context
• No allele-phase (trans/cis) data relative to any second ERBB2 variant
• No pathogenic ERBB2 variant identified in trans with c.2584A>G
• No established recessive germline phenotype for ERBB2 anchored in the evidence bundle (CSPEC/VCEP absent; generic ACMG fallback)

Sources

Reasoning

No de novo occurrence of NM_004448.3:c.2584A>G (with or without parental confirmation) is reported in any case source. The variant is absent from ClinVar, and none of the retrieved publications contain proband or parental genotype data. PM6 (de novo without confirmed paternity/maternity) cannot be evaluated; insufficient evidence, so the criterion is not assessed.

Rule

ACMG/AMP 2015 (PMID 25741868): PM6 - de novo (paternity and maternity not confirmed), moderate

Evidence considered

• ClinVar search returned no record for NM_004448.3:c.2584A>G (search_status: no_result)
• gnomAD absence does not provide de novo evidence
• Keyword search of the four available full texts for de novo/segregation/germline/proband terms returned no matches

Still missing

• No proband observation with unconfirmed parental testing reported
• No literature or ClinVar case documenting the variant as de novo (parentage unconfirmed)

Sources

Reasoning

No segregation data exists for NM_004448.3:c.2584A>G. There are no affected family members tested, no meioses counted, and no familial case reports in ClinVar (variant absent) or in the retrieved literature, which is entirely somatic/functional in theme. PP1 (co-segregation with disease in affected family members) cannot be evaluated; insufficient evidence, so the criterion is not assessed.

Rule

ACMG/AMP 2015 (PMID 25741868): PP1 - co-segregation with disease in affected family members, supporting

Evidence considered

• ClinVar has no entry for the variant, so no familial segregation observations are available
• All four retrieved publications are functional/somatic cancer studies; full-text keyword search for segregation/familial/pedigree/germline content returned zero hits
• COSMIC and OncoKB entries describe somatic cancer observations, which do not constitute germline segregation evidence

Still missing

• No affected family members tested for the variant
• No segregation analysis or meiosis count reported in any source

Sources

Reasoning

PP4 (phenotype or family history highly specific for a disease with a single genetic etiology) cannot be evaluated: the case bundle contains no proband phenotype description, family history, or clinical indication for testing. No source in the bundle (ClinVar record absent; no clinical case report of this exact variant in a germline context) provides a phenotype that could be assessed for specificity. The variant is a well-known somatic cancer hotspot (OncoKB Oncogenic; COSMIC n=27), and germline disease association for this exact variant is not established by any source in the bundle. Evidence insufficient: status not_assessed.

Rule

PP4: Phenotype or family history highly specific for a disease with a single genetic etiology (Richards et al. 2015, PMID 25741868).

Still missing

• No proband phenotype or family history data are present in review_manifest.json / case_summary.json.
• No clinical case report describing the exact variant (c.2584A>G / p.Thr862Ala) in a germline diagnostic context with phenotypic detail.
• Germline disease phenotype for ERBB2 (single genetic etiology) is not defined for this variant in the available evidence.

Reasoning

No data on variant absence in affected family members (non-segregation) exists for NM_004448.3:c.2584A>G. BS4 requires documented lack of segregation, i.e., affected family members who do not carry the variant; no such observations are reported in ClinVar (variant absent) or any retrieved publication. Non-segregation can neither be demonstrated nor refuted; insufficient evidence, so the criterion is not assessed.

Rule

ACMG/AMP 2015 (PMID 25741868): BS4 - lack of segregation in affected members of a family, supporting (benign)

Evidence considered

• ClinVar has no entry for the variant (search_status: no_result), providing no familial genotype data
• Full texts of the four retrieved publications contain no segregation or family-genotype content (keyword search negative)
• gnomAD population absence is not evidence of lack of segregation within affected families

Still missing

• No affected family members genotyped, so lack of segregation cannot be evaluated
• No non-segregation observations reported in any source

Sources

Reasoning

BP2 (ClinGen SVI trans/cis formulation: observed in trans with a pathogenic variant for a fully dominant disease/mechanism, or observed in cis with a pathogenic variant in any inheritance pattern) cannot be adjudicated because the evidence bundle contains no allele-phase observations at all: no proband, no segregation data, no pathogenic variant in trans or in cis with c.2584A>G. The variant is absent from ClinVar and gnomAD v2.1/v4.1 and is a recurrent somatic cancer hotspot (OncoKB: Oncogenic, gain-of-function; COSMIC COSV54062276, n=27), and no fully dominant germline disease/mechanism for ERBB2 is established in the bundle (no CSPEC/VCEP; generic ACMG/AMP 2015, PMID 25741868, governs). Without any observed allele configuration relative to a pathogenic variant, BP2 is not assessable; the deciding data are entirely absent.

Rule

BP2 (ClinGen SVI trans/cis formulation; base framework ACMG/AMP 2015, PMID 25741868): observed in trans with a pathogenic variant for a fully dominant disease/mechanism, or observed in cis with a pathogenic variant in any inheritance pattern. Not assessable here: no allele-phase observations relative to any pathogenic variant.

Evidence considered

• generic_acmg_combination_rules: governing framework is ACMG/AMP 2015 (PMID 25741868); no CSPEC/VCEP for ERBB2, so no gene-specific BP2 specification applies
• clinvar: variant absent from ClinVar; no germline submissions, no phase or family data
• gnomad_v2 / gnomad_v4: variant absent from population databases; no observations enabling any cis/trans configuration analysis
• oncokb: ERBB2 T862A curated as Oncogenic, gain-of-function (variant-specific) in a somatic cancer context; not germline allele-phase evidence
• pvs1_gene_context: germline disease context for ERBB2 not established, so the fully-dominant trans-with-pathogenic arm of BP2 cannot be anchored

Still missing

• No observation of c.2584A>G in trans with a pathogenic variant in a fully dominant disease/mechanism context
• No observation of c.2584A>G in cis with any pathogenic variant
• No proband or family (segregation) data for this variant
• No established fully dominant germline disease/mechanism for ERBB2 anchored in the evidence bundle

Sources

Reasoning

BP5 (variant found in a case with an alternate molecular basis for disease) cannot be evaluated: no case-level data in the bundle describe a proband carrying this variant who also has a confirmed alternative molecular diagnosis. There is no proband clinical context at all for c.2584A>G. Evidence insufficient: status not_assessed.

Rule

BP5: Variant found in a case with an alternate molecular basis for disease (Richards et al. 2015, PMID 25741868).

Still missing

• No clinical case data indicating the variant was observed in an individual with a confirmed alternate molecular basis for their phenotype.
• No proband-level testing context (panel/WES findings, other causative variants) is present in the evidence bundle.