APC c.6907G>A, p.(Gly2303Arg)

Reasoning

Governing framework: InSiGHT/ClinGen APC VCEP specification v2.1 (gene-specific, takes precedence over generic ACMG), which sets BA1 (Benign Stand Alone) at a gnomAD Popmax Filtering Allele Frequency (AF) >= 0.1% (0.001) and explicitly instructs that this be assessed on the non-cancer dataset from gnomAD (v2.1.1). Using that designated source (gnomAD v2.1 non-cancer exomes; v2.1.1 is exome-based), the Popmax Filtering AF is 0.002158 (0.216%), which is ~2.2-fold above the stand-alone threshold, and the subset total AF is 0.000350957 (83/236,496 alleles, 1 homozygote). The signal is ancestry-concentrated rather than a scattered artefact: 80/30,518 South Asian alleles (AF 0.002621, 1 homozygote) versus 3/5,610 'remaining individuals' (AF 0.000535) and 0 alleles in the African/African American, Admixed American, East Asian, Finnish, non-Finnish European and Ashkenazi Jewish subsets of that non-cancer exome set. The same pattern is reproducible in the all-comers totals (gnomAD v2.1 Popmax FAF 0.002152, 80/30,608 South Asian; gnomAD v4.1 grpmax FAF 0.002444, 248/91,074 South Asian with 3 homozygotes, plus 1/1,179,878 non-Finnish European and 1/6,062 Middle Eastern), so the elevated frequency is not a single-dataset or single-call artefact, and the variant is confirmed in both v2.1 and v4.1 rather than being a liftover/calling error. The magnitude is incompatible with a variant that causes a dominant, highly penetrant adenomatous polyposis allele: the threshold is in the range calibrated against FAP/AFAP disease prevalence with reduced penetrance (the VCEP derivation assumes penetrance 0.8), whereas this allele is present in roughly 1 in 190 South Asian chromosomes in the non-cancer dataset and 1 in 370 overall. Threshold handling: the VCEP Popmax Filtering AF is the intended metric (per-population maximum, so the South Asian concentration is correctly treated as the population of concern rather than averaged away); both the non-cancer exome subset explicitly named by the VCEP and the all-comers v2.1/v4.1 totals exceed 0.1%, so no source-selection choice changes the verdict. BA1 is stand-alone: this criterion alone drives a Benign classification and requires no combination with other criteria. No material gap affects this call; gnomAD-Canada v1.0 returned HTTP 500 and could not be consulted, but a missing additional population dataset could only add carriers and cannot lower the observed frequency below the threshold.

Rule

APC VCEP v2.1 BA1 (Benign Stand Alone): GnomAD Popmax Filtering Allele Frequency (AF) >= 0.1% (0.001), assessed on the non-cancer dataset from gnomAD (v2.1.1).

Evidence considered

• APC VCEP specification v2.1 (InSiGHT Hereditary Colorectal Cancer/Polyposis Expert Panel) BA1 rule, listed strength Stand Alone: 'GnomAD Popmax Filtering Allele Frequency (AF) >= 0.1% (0.001).' The specification's instructionsToUse field for BA1 states: 'General recommendation: Use the non-cancer dataset from gnomAD (v2.1.1)'.
• The APC VCEP supplementary material (v2) independently states for BA1: 'Allele frequency >= 0.1% (preferably based on the P[opmax] ...)' and records a penetrance assumption of 0.8 in the same derivation table, confirming 0.1% Popmax is the governing stand-alone threshold rather than a generic 5% cut-off.
• gnomAD v2.1 non-cancer exomes (VCEP-designated source, GRCh37 position 5-112178198-G-A): total AF 0.000350957, 83/236,496 alleles, 1 homozygote; Popmax Filtering AF 0.00215798 (0.216%); South Asian 80/30,518 (AF 0.002621403, 1 homozygote); 'remaining individuals' 3/5,610 (AF 0.000535); zero alleles in all other ancestry groups of this subset.
• gnomAD v2.1 all-comers exomes (same URL, unfiltered totals): total AF 0.000330661, 83/251,012 alleles, 1 homozygote; Popmax FAF 0.00215168; South Asian 80/30,608 (AF 0.0026137, 1 homozygote); 'remaining individuals' 3/6,130.
• gnomAD v4.1 (GRCh38 chr5-112842501-G-A): total AF 0.000157382, 254/1,613,908 alleles, 3 homozygotes; grpmax FAF 0.00244403; South Asian 248/91,074 (AF 0.00272306, 3 homozygotes); non-Finnish European 1/1,179,878; Middle Eastern 1/6,062; remaining individuals 4/62,498.
• gnomAD v3.1 non-cancer genomes (the only v3 cohort type; v3 has no exome subset): total AF 0.0000541, 8/147,862 alleles, 0 homozygotes; South Asian 8/4,798 (AF 0.001667); grpmax FAF 0.00082419 driven by a 1KG:GIH subset of only 202 alleles. This genome-only non-cancer source does not by itself reach the 0.1% stand-alone threshold, but the VCEP specifies v2.1.1 (exome-based) for BA1, and the exome-based non-cancer source does reach it; the discrepancy is a sample-size/cohort-composition effect, not evidence against the exome-based call.
• Ancestry and penetrance interpretation: for a dominant, high-penetrance adenomatous polyposis allele the acceptable population frequency is far below 0.1%; an allele carried by 1 in ~190 South Asian chromosomes in a non-cancer adult dataset, with 3 homozygotes in gnomAD v4.1, is quantitatively incompatible with near-complete dominant penetrance at an FAP/AFAP prevalence well under 0.1%.
• The ACG hereditary gastrointestinal cancer syndromes guideline (PMID:25645574) supports the penetrance premise used in this reasoning: FAP arises from dominant mutations in APC with near-100% penetrance and biallelic inactivation of the tumor suppressor.

Still missing

• gnomAD-Canada v1.0 (HostSeq genomes) was unavailable (HTTP 500), so no Canadian genome-cohort frequency could be added; this cannot affect the verdict because the observed non-cancer Popmax FAF already exceeds the stand-alone threshold.

Sources

Reasoning

Governing framework: APC VCEP specification v2.1 sets BS1 (Benign Strong) at a gnomAD Popmax Filtering Allele Frequency (AF) >= 0.001% (0.00001), and instructs that this be assessed on the non-cancer dataset from gnomAD (v2.1.1). In the VCEP-designated source, gnomAD v2.1 non-cancer exomes, the Popmax Filtering AF is 0.002158 (0.216%) - more than two orders of magnitude above the 0.001% BS1 threshold; the same source's total AF is 0.000350957 (83/236,496 alleles) and its South Asian Popmax-equivalent frequency is 0.002621 (80/30,518 alleles). The criterion is therefore met at strong strength. Important combination caveat: BA1 is independently met on exactly the same allele-frequency evidence, and BA1 is stand-alone, so BS1 is redundant here and must not be counted as a second, independent benign criterion alongside BA1 (the VCEP combining rules place BA1 alone in Benign.Stand Alone and reach Benign before the BS1/BS2 Benign.Strong partition is consulted). The BS1 verdict is reported as met because the observed frequency genuinely exceeds its threshold, but the classification weight comes from BA1. Cross-dataset check: all-comers gnomAD v2.1 (Popmax FAF 0.002152) and v4.1 (grpmax FAF 0.002444) also exceed 0.001%, and even the lower-frequency non-cancer genome-only source (gnomAD v3.1 non-cancer, total AF 0.0000541 = 0.00541%, grpmax FAF 0.000824) exceeds the 0.001% BS1 threshold, so the call is robust to which non-cancer dataset is chosen. Threshold-artefact note: the v3.1 grpmax FAF (0.000824) rests on a single allele in a 202-allele 1KG:GIH subset, and that value still clears the BS1 bar, so the small-sample inflation in that subset is not load-bearing for the verdict; the v2.1 non-cancer source named by the VCEP treats the small-sample issue appropriately via the Popmax Filtering AF. No material gap affects this call.

Rule

APC VCEP v2.1 BS1 (Benign Strong): GnomAD Popmax Filtering Allele Frequency (AF) >= 0.001% (0.00001), assessed on the non-cancer dataset from gnomAD (v2.1.1).

Evidence considered

• APC VCEP specification v2.1 BS1 rule, listed strength Strong: 'GnomAD Popmax Filtering Allele Frequency (AF) >= 0.001% (0.00001).' The specification's instructionsToUse field for BS1 states: 'General recommendation: Use the non-cancer dataset from gnomAD (v2.1.1)'.
• The APC VCEP supplementary material (v2) states for BS1: 'Allele frequency >= 0.001% (preferably based on the [Popmax] ...)', with penetrance 0.8 shown in the adjacent derivation table - i.e. the very low 0.001% bar reflects the dominant, highly penetrant nature of APC-associated disease.
• gnomAD v2.1 non-cancer exomes: Popmax Filtering AF 0.00215798 (0.216%), total AF 0.000350957 (83/236,496 alleles, 1 homozygote), South Asian 80/30,518 (AF 0.002621403).
• gnomAD v2.1 all-comers exomes: Popmax FAF 0.00215168 (83/251,012 alleles). gnomAD v4.1: grpmax FAF 0.00244403 (254/1,613,908 alleles, 3 homozygotes). gnomAD v3.1 non-cancer genomes: total AF 0.0000541 (8/147,862 alleles, 0 homozygotes), grpmax FAF 0.00082419 from a 1KG:GIH subset of 202 alleles.
• APC VCEP v2 combining rules place BA1 alone in the Benign.Stand Alone partition (inference Benign) and BS1/BS2/BS3/BS4 in the Benign.Strong partition (inference Likely Benign), so BS1 and BA1 rest on the same frequency observation and must not be summed as independent benign evidence.

Still missing

• None material. The BS1 evidence is the same allele-frequency observation that satisfies BA1, so if BA1 is retained at stand-alone strength this criterion contributes no additional classification weight.

Sources

Reasoning

The APC VCEP v2.1 BP1 rule states that 'BP1 is applicable to APC with the exception of missense variants located in the first 15-amino acid repeat of the beta-catenin binding domain (codon 1021-1035).' The variant under assessment, NM_001127510.3:c.6907G>A p.(Gly2303Arg), is a missense variant at codon 2303, outside the 1021-1035 exception and also outside the VCEP-defined beta-catenin binding domain (codons 959-2129), so the criterion is applicable and no exception applies. APC is a gene in which truncating variants are the predominant mechanism of disease: over 90% of APC point mutations are truncating (nonsense, del/ins, splice-site), reported mutation hotspots are codons 1309, 1061, 213 and 1068, and the mutation cluster region lies in the 5' end of exon 15, with mutations 3' to codon 1700 accounting for only ~1% of cases. A rare missense substitution at codon 2303, a residue with no critical domain and no hotspot, therefore falls in the class the criterion is designed to de-weight, and BP1 is met at supporting strength.

Rule

BP1 (APC VCEP v2.1): missense variant in a gene for which primarily truncating variants cause disease; applicable to APC except missense variants in codons 1021-1035

Evidence considered

• APC VCEP v2.1 BP1 rule text: 'BP1 is applicable to APC with the exception of missense variants located in the first 15-amino acid repeat of the beta-catenin binding domain (codon 1021-1035).'
• The variant under assessment is a missense change, NP_001120982.1:p.(Gly2303Arg), at codon 2303 - outside the codon 1021-1035 exception and outside the beta-catenin binding domain (codons 959-2129) as defined in the same VCEP specification - so BP1's stated exception does not exclude its use here.
• PMID:21368914: APC point mutations are >90% truncating (nonsense, del/ins, splice sites), the vast majority lie in the 5' half of the gene, mutations 3' to codon 1700 are rare (~1%), and reported hotspots are codons 1309, 1061, 213 and 1068 - establishing that truncating variants, not missense variants such as p.(Gly2303Arg), are the predominant cause of APC-associated disease.
• PMID:25645574 (ACG hereditary GI cancer syndromes guideline): 'Although mutations have been found scattered throughout the APC gene, most are located in the 5' end of exon 15, often called the mutation cluster region' - codon 2303 lies well outside that cluster.
• PMID:25741868 (ACMG/AMP 2015 standards) is the source of the BP1 criterion itself: a missense variant in a gene for which primarily truncating variants are known to cause disease; it provides no variant-specific data on APC.
• The criterion was applied only at supporting strength, matching the APC VCEP's default strength for BP1, and the benign direction it contributes is consistent with the absence of a hotspot or critical domain at this codon.

Sources

Reasoning

Not met. NM_001127510.3:c.6907G>A (NC_000005.9:g.112178198G>A) is a single-nucleotide substitution that encodes NP_001120982.1:p.(Gly2303Arg), an amino-acid substitution (missense change). The APC InSiGHT VCEP specification (v2.1) defines PVS1 as a 'Null variant in a gene where LOF is a known mechanism of disease ... As per modified decision tree (Figure 1)', at Very Strong, Strong, Moderate or Supporting. A missense substitution is not a null variant: it is not a nonsense, frameshift, canonical +/-1,2 splice-site, initiation-codon or multi-exon deletion allele, which are the only triggers for PVS1. The APC VCEP PVS1 decision tree (Figure 1) sets out Lists A-E of eligible splice-region nucleotides at each strength tier (e.g. c.136-1G>A,C,T; c.136-2A>C,G,T; c.220+1G>A,C,T and c.220+2T>A,C,G in List A; c.220G>A,C,T in List B); c.6907G>A appears in none of these lists and is not a splice-consensus position. Transcript context confirms this: in NM_001127510.3 the variant lies well inside the large terminal exon (coding sequence of that exon runs from c.1959 to c.8532, with the transcript extending to c.*2113), roughly 5 kb from the exon's 5' boundary, with no downstream intron - so no canonical splice site is affected and no NMD-competent downstream exon exists. Indeed, because this terminal exon is the last exon, even a hypothetical truncating allele at codon 2303 of 2,843 would escape nonsense-mediated decay, which under the SVI PVS1 decision tree precludes full-strength PVS1; that NMD consideration is recorded here for completeness only, as no truncating allele is created by this variant. Loss of function is nevertheless an established disease mechanism for APC (germline APC LOF, autosomal dominant FAP), so PVS1 is applicable and was assessed on its merits rather than dismissed; it simply has no null-variant trigger. Because the variant is not a null, no downgrade rule is invoked and no strength tier is reported.

Rule

Null variant in a gene where LOF is a known mechanism of disease. As per modified decision tree (Figure 1) (APC VCEP v2.1); no null-variant trigger present for p.(Gly2303Arg).

Evidence considered

• Governing framework (cspec, ClinGen InSiGHT Hereditary Colorectal Cancer/Polyposis Expert Panel Specification for APC v2.1, doc 1529292791; searched the PVS1 entry): PVS1 is listed as Applicable and its strength rules are 'Null variant in a gene where LOF is a known mechanism of disease. As per modified decision tree (Figure 1) [Reference 1]' at Very Strong, Strong, Moderate and Supporting; the underlying framework is the ClinGen SVI PVS1 recommendations (Abou Tayoun et al., 2018).
• APC VCEP PVS1 decision tree (Fig.-1_APC_PVS1_decision-tree_2023-10-20.pdf): the figure's Lists A-E enumerate only specific splice-region nucleotide changes at each strength tier - List A (PVS1) includes c.136-1G>A,C,T and c.136-2A>C,G,T; List B (PVS1_Strong) includes c.220G>A,C,T; List C (PVS1_Moderate) includes c.1548G>A,C,T; List D (PVS1_Supporting) includes c.1549-1G>A,C,T and c.1549-2A>C,G,T; List E (N/A) includes c.-18-1G>A,C,T and c.934-1G>A,C,T. c.6907G>A does not appear in any list, so no PVS1 tier is assigned to it.
• Variant consequence data: NM_001127510.3:c.6907G>A is a G>A substitution at coding position 6907, predicted to produce NP_001120982.1:p.(Gly2303Arg) (LRG_130p2:p.(Gly2303Arg)); the affected codon is reported as 2303 within a 2,843-residue protein, with identical wild-type and mutant protein length, and the variant is classed as missense with canonical_splice_consensus = false and no protein-length change of any kind.
• Transcript/exon context: the NM_001127510.3 exon map places c.6907 inside the terminal large exon, whose coding portion spans c.1959-c.8532 (exon extending to c.*2113 in the c. coordinate list); VariantValidator annotates the variant's exonic position as exon 17 of NC_000005.9/NC_000005.10. The variant therefore sits ~4.9 kb from that exon's 5' boundary, within the last exon of the transcript, where there is no downstream intron and therefore no splice donor whose loss could trigger NMD.
• Gene-level eligibility (pvs1_gene_context): APC loss of function is the established germline disease mechanism (pvs1_gene_gate = eligible, supporting source cspec), so PVS1 was assessed as an applicable criterion.
• Splice prediction data (spliceai): the SpliceAI lookup for this variant returned no scores (query timed out; max_delta_score null), flagged in the case data as missing data rather than a score of 0.00; no SpliceAI-based assessment of a possible splice effect was therefore possible.
• PubMed literature screened for this criterion: PMID:21368914 (FAP/AFAP clinical utility gene card - >90% of APC point mutations are truncating and mutations 3' to codon 1700 are rare, ~1%) and PMID:25645574 (ACG hereditary GI cancer guideline - mutations cluster in the 5' end of exon 15, the mutation cluster region, with location-phenotype correlations); both place codon 2303 in the sparsely mutated 3' region of APC and neither reports this variant. PMID:25741868 (ACMG/AMP 2015 standards) supplied the generic PVS1 definition and its caveats on the extreme 3' end of a gene and on NMD escape. No publication in this case reports c.6907G>A or p.(Gly2303Arg).

Still missing

• SpliceAI returned no score for NM_001127510.3:c.6907G>A (the lookup timed out; max delta score is null, explicitly logged as missing data and not as 0.00), so a cryptic splice effect on this presumed missense variant cannot be formally excluded. The APC VCEP specification instructs that presumed missense variants be screened with splicing predictors. This does not alter the PVS1 verdict, which rests on the variant class (missense, not null, not at a canonical splice position), but leaves one in silico check unclosed.

Sources

Reasoning

Governing framework is the ClinGen InSiGHT Hereditary Colorectal Cancer/Polyposis Variant Curation Expert Panel specification for APC v2.1. Its PS1 rule requires that the variant under assessment result in the same amino-acid change as a previously established Pathogenic (PS1_Strong) or Likely Pathogenic (PS1_Moderate) APC variant. The specification states that there are currently only two Likely Pathogenic missense variants in APC, c.3077A>G p.(Asn1026Ser) and c.3084T>A p.(Ser1028Arg), and that no missense variant has been classified Pathogenic on current evidence. The variant under assessment is NM_001127510.3:c.6907G>A, NP_001120982.1:p.(Gly2303Arg): residue 2303, not 1026 or 1028, and Gly2303Arg is not the amino-acid change produced by either established comparator. No APC variant, missense or splice, has been established as Pathogenic or Likely Pathogenic for this particular amino-acid change: the exact-match ClinVar record (VCV000133533, 15 submissions, 0 expert-panel submissions, 1 star) carries only conflicting assertions dominated by Benign/Likely Benign, and the same-residue comparator harvest returned zero same-residue candidates. With no established APC variant to anchor to, PS1 is not met.

Rule

PS1 (APC VCEP v2.1): same amino-acid change as a previously established Pathogenic (Strong) or Likely Pathogenic (Moderate) APC variant

Evidence considered

• APC VCEP (InSiGHT) v2.1 PS1 rule text: for missense variants, PS1 applies 'when the variant under assessment results in the same amino acid change as previously established Pathogenic variant(s)'; the same rule notes 'There are currently only two Likely Pathogenic missense variants: c.3077A>G p.(Asn1026Ser) and c.3084T>A p.(Ser1028Arg)' and that 'No missense variant has been classified as Pathogenic based on current evidence.'
• The variant under assessment is NP_001120982.1:p.(Gly2303Arg) - residue 2303 with a Gly>Arg substitution. Neither established comparator (p.(Asn1026Ser), p.(Ser1028Arg)) shares this residue or this amino-acid change, so no same-amino-acid-change anchor exists in APC.
• ClinVar exact-match record VCV000133533 (NM_000038.6:c.6907G>A p.(Gly2303Arg); 15 submissions, 15 submitters, 0 expert-panel submissions, review status 'criteria provided, conflicting classifications', 1 star) contains no Pathogenic or Likely Pathogenic assertion for this amino-acid change; the submission mix is 5 Likely benign, 3 Uncertain significance, 3 Benign, 1 benign and 1 Likely Benign, with the largest submitters split between benign and uncertain assertions.
• The automated same-residue comparator harvest (ClinVar queries on APC G2303 / Gly2303) reviewed 15 records and returned zero same-residue candidates, so no alternate nucleotide change producing Gly2303Arg was found.

Sources

Reasoning

Governing framework is the ClinGen InSiGHT Hereditary Colorectal Cancer/Polyposis VCEP specification for APC v2.1 (CSPEC doc 1529292791), which overrides generic ACMG. The variant is a missense change, NM_001127510.3:c.6907G>A, p.(Gly2303Arg), codon 2303 (exon 17). Under the APC VCEP, PS3_VeryStrong/Strong/Moderate all require RNA assays demonstrating a premature stop codon or in-frame skipping of exon 13 or 14 (with absence of, or <10% of, full-length transcript); the PS3_Supporting branch additionally allows RNA assays showing in-frame skipping of other exons or over-expression of alternative transcripts (exons 10, 11 or 15), or protein assays showing increased beta-catenin-regulated transcription activity and/or decreased beta-catenin binding by surface plasmon resonance - but that protein-assay route is explicitly restricted to variants within the beta-catenin binding domain, codons 959-2129. Codon 2303 falls outside codons 959-2129, so the protein-assay route is not available for this variant, leaving RNA assays as the only possible route to PS3. No RNA assay (and no protein assay) for this variant exists in any consulted source: OncoKB reports an Unknown Oncogenic Effect with no variant-specific reviewed functional evidence; COSMIC recurrence (COSV57373649, n=4 somatic reports) is tumour recurrence data, not a validated functional assay; and none of the ten triaged publications mentions c.6907G>A or p.(Gly2303R) or reports any functional characterisation. In-silico predictor scores (REVEL 0.611, BayesDel 0.199863) are computational predictions, not functional-assay evidence, and are assessed under PP3/BP4. PS3 is therefore not met; no functional evidence supports a damaging effect on the gene product.

Rule

InSiGHT APC VCEP v2.1 (CSPEC doc 1529292791) PS3: RNA assays showing a premature stop codon or in-frame skipping of exon 13 or 14, with absence of full-length transcript (Very Strong) or <10% full-length transcript (Strong); RNA assay plus reports of exon deletion/skipping/loss or intronic insertion (Moderate); RNA assay showing in-frame skipping of other exons or alternative-transcript over-expression, OR a protein assay showing increased beta-catenin regulated transcription activity and/or decreased beta-catenin binding by SPR restricted to codons 959-2129 (Supporting).

Evidence considered

• APC VCEP specification v2.1 defines all PS3 strength tiers (Very Strong, Strong, Moderate) through RNA assays showing a premature stop codon or in-frame skipping of exon 13 or 14, with absence of full-length transcript (Very Strong), <10% full-length transcript (Strong), or reports of exon deletion/skipping/loss or intronic nucleotide insertion (Moderate).
• The PS3_Supporting tier permits RNA assays showing in-frame skipping of exons other than 13/14 or over-expression of alternative transcripts (exons 10, 11 or 15), or protein assays showing increased beta-catenin regulated transcription activity and/or decreased beta-catenin binding by surface plasmon resonance, with the protein-assay route limited to variants within the beta-catenin binding domain, codons 959-2129 of APC.
• Normalisation places the variant at codon 2303 (NM_001127510.3:c.6907G>A, NP_001120982.1:p.(Gly2303Arg), exon 17), i.e. outside the codon 959-2129 beta-catenin binding domain, so the framework's protein-assay evidence route cannot be applied.
• No RNA assay data for this variant appears in the case evidence: source retrieval produced no functional study, and OncoKB reports 'Unknown Oncogenic Effect' with no variant-specific reviewed functional evidence.
• COSMIC reports this variant as a somatic recurrence (COSV57373649, n=4); somatic recurrence is not a validated in vitro or in vivo functional assay and carries no PS3 weight.
• All ten publications triaged for this case were screened for functional content; none reports a functional assay of c.6907G>A / p.(Gly2303R), and none even mentions the variant (variant_mentioned=false throughout).
• In-silico scores available for this variant (REVEL 0.611, BayesDel 0.199863) are computational predictions applicable to PP3/BP4, not functional-assay evidence for PS3.

Still missing

• No RNA assay of NM_001127510.3:c.6907G>A exists in any consulted source; an RNA study demonstrating a premature stop codon or exon skipping is the only route by which PS3 could be met for this variant under the APC VCEP framework.
• The APC VCEP protein-assay route (increased beta-catenin-regulated transcription activity and/or decreased beta-catenin binding by SPR) is restricted to codons 959-2129 and cannot be applied at codon 2303, so a protein assay, even if performed, would not be framework-interpretable for this variant.
• Two ClinVar-cited references (PMID:20301519, PMID:26389258) have no full text or pre-computed extraction in this case; both are narrative resources (a GeneReviews chapter and a PDQ summary) and neither is a variant-level functional study.
• Absence of functional data is not evidence that the variant is functionally normal; it leaves the functional axis of the classification uninformative.

Sources

Reasoning

Governing framework: APC VCEP specification v2.1 keeps PM2 only at Supporting strength and defines it as 'Rare in controls, defined by an allele frequency <= 0.0003% (0.000003) if the allele count is > 1 OR by an allele frequency < 0.001% (0.00001) if the allele count is <= 1', with the instruction to use the total population of the non-cancer gnomAD dataset (v2.1.1). In the VCEP-designated source, gnomAD v2.1 non-cancer exomes, the allele count is 83 (> 1), so the applicable threshold is AF <= 0.000003; the observed AF is 0.000350957 (0.0351%, 83/236,496 alleles), roughly 117-fold above that threshold. The alternative AC <= 1 branch threshold (AF < 0.00001) is also exceeded by ~35-fold, so no reading of the rule makes this variant rare enough. The variant also exceeds the generic ClinGen SVI PM2 supporting cut-off often used as a fallback (AF <= 0.0001; PMID:25741868), and it fails PM2 by an even wider margin against the all-comers totals (gnomAD v2.1 AF 0.000330661; gnomAD v4.1 AF 0.000157382, 254/1,613,908 alleles). The second non-cancer source available, gnomAD v3.1 non-cancer genomes, gives AF 0.0000541 (8/147,862 alleles) - still above the 0.000003 threshold applicable at AC > 1 - so PM2 is not met under either non-cancer dataset. Ancestry does not create a rare-subpopulation argument in the opposite direction: the variant is most common in South Asians (0.262% in the non-cancer exome subset; 0.272% in v4.1), which is the population where a rare-allele claim would have to hold, and in that population it is common. Because PM2 fails on frequency by more than two orders of magnitude in the relevant source, this criterion cannot be rescued by any remaining gap; the unavailable gnomAD-Canada dataset could only add carriers. No other criterion in this group bears on PM2.

Rule

APC VCEP v2.1 PM2 (Supporting only): rare in controls, defined by allele frequency <= 0.0003% (0.000003) if the allele count is > 1, OR by allele frequency < 0.001% (0.00001) if the allele count is <= 1; assessed on the total population of the non-cancer gnomAD dataset (v2.1.1).

Evidence considered

• APC VCEP specification v2.1 PM2 rule, listed strength Supporting (no higher strength is applicable): 'Rare in controls, defined by an allele frequency <= 0.0003% (0.000003) if the allele count is > 1 OR by an allele frequency < 0.001% (0.00001) if the allele count is <= 1.' The specification's instructionsToUse field for PM2 states: 'General recommendation: Use the total population from the non-cancer dataset from gnomAD (v2.1.1)'.
• gnomAD v2.1 non-cancer exomes (VCEP-designated source, GRCh37 5-112178198-G-A): AF 0.000350957, 83/236,496 alleles (> 1 allele, so the 0.000003 threshold applies), 1 homozygote; South Asian 80/30,518 (AF 0.002621403); 'remaining individuals' 3/5,610 (AF 0.000535); 0 alleles in all other ancestry groups of this subset.
• gnomAD v2.1 all-comers exomes: AF 0.000330661, 83/251,012 alleles. gnomAD v4.1: AF 0.000157382, 254/1,613,908 alleles, 3 homozygotes (South Asian 248/91,074, AF 0.00272306).
• gnomAD v3.1 non-cancer genomes: AF 0.0000541, 8/147,862 alleles, 0 homozygotes (South Asian 8/4,798, AF 0.001667) - above the 0.000003 threshold that applies at AC > 1, so PM2 also fails on this non-cancer source.
• Generic fallback check (only if the VCEP were absent): the ClinGen SVI guidance to use PM2 at Supporting strength with an AF cut-off of <= 0.0001 (PMID:25741868) is likewise exceeded by ~3.5-fold on the non-cancer source and is not met; the VCEP's own stricter rule governs in any case.

Still missing

• gnomAD-Canada v1.0 (HostSeq genomes) was unavailable (HTTP 500); a missing population dataset cannot lower the observed frequency and therefore cannot change this result.

Sources

Reasoning

The APC VCEP v2.1 PM5 rule (Moderate) requires a missense change at an amino acid residue where a different missense change has been established as Pathogenic under the APC-specific modifications (Supporting when the established comparator is Likely Pathogenic), and it adds a Grantham-distance precondition - the variant under assessment must have a Grantham distance equal to or higher than the established comparator. The specification is explicit that only two missense variants are currently Likely Pathogenic in APC, c.3077A>G p.(Asn1026Ser) and c.3084T>A p.(Ser1028Arg), and that other different missense variants at those positions meet PM5_Supporting; no missense variant has been classified Pathogenic. The variant under assessment is p.(Gly2303Arg) at residue 2303, so neither qualifying residue (1026, 1028) is involved, and no Pathogenic or Likely Pathogenic missense variant at residue 2303 exists - the exact-match ClinVar record for this variant carries no pathogenic assertion, and the same-residue comparator harvest returned zero candidates for APC G2303. Because no qualifying same-residue comparator exists, the Grantham-distance precondition is never reached and PM5 is not met.

Rule

PM5 (APC VCEP v2.1): missense change at a residue where a different missense change was established Pathogenic (Moderate) or Likely Pathogenic (Supporting), with Grantham distance >= that of the comparator

Evidence considered

• APC VCEP v2.1 PM5 rule (Moderate): the reported missense variant must have been determined Pathogenic according to the APC-specific modifications; Supporting when it was determined Likely Pathogenic. The rule states 'There are currently only two Likely Pathogenic missense variants: c.3077A>G p.(Asn1026Ser) and c.3084T>A p.(Ser1028Arg). Other different missense variants at these positions meet PM5_supporting. No missense variant has been classified as Pathogenic based on current evidence.'
• The same APC VCEP PM5 rule imposes an additional precondition: 'Grantham's distance of the variant under assessment must have an equal or higher score than the reported variant' - a comparison that cannot be performed without a qualifying same-residue comparator.
• The variant under assessment is p.(Gly2303Arg) at residue 2303; the only residues with an established Likely Pathogenic APC missense comparator are 1026 and 1028, and a different amino acid change at residue 2303 has not been established as Pathogenic or Likely Pathogenic.
• ClinVar exact-match record VCV000133533 for c.6907G>A p.(Gly2303Arg) contains 15 submissions and no Pathogenic or Likely Pathogenic assertion, so it supplies no same-residue comparator.
• The PM5 comparator-harvest artifact for this case reports same_residue_candidates = 0 (15 records seen, 0 candidate comparators) with a soft failure from HTTP 429 rate limiting, i.e. no same-residue missense comparator was identified.

Still missing

• The automated same-residue comparator harvest returned zero candidates and errored with HTTP 429 (rate limiting), so it did not complete a full sweep; the verdict rests instead on the APC VCEP's closed list of established missense variants (p.(Asn1026Ser), p.(Ser1028Arg)) and on the exact-match ClinVar record showing no pathogenic same-residue comparator.

Sources

Reasoning

The governing specification is the InSiGHT/ClinGen Hereditary Colorectal Cancer and Polyposis VCEP for APC v2.1, whose PP3 supporting rule is the only applicable PP3 provision. The variant is a coding missense change (NM_001127510.3:c.6907G>A, NP_001120982.1:p.(Gly2303Arg)), so the VCEP missense arm governs: it directs that computational models of conservation/evolution must NOT be used for presumed missense variants and that in silico splicing predictors instead be used to reveal possible splicing effects. On the missense (protein-impact) path, REVEL 0.611 is below the ClinGen SVI PP3-supporting cutoff of 0.644 (Pejaver et al. 2022, PMID:36413997), i.e. an indeterminate gray-zone result that does not meet PP3 and does not license reaching across to a splice score. On the VCEP splice-effect path, the only VCEP-recommended predictor queried, SpliceAI, returned no result (lookup read timeout; max_delta_score is null rather than 0.00), so no SpliceAI-based splice effect is demonstrable; the Pangolin values returned alongside (splice gain 0.001, splice loss -0.003) are not among the VCEP-recommended programs (SpliceAI, MaxEntScan, VarSeak) and have no published ACMG-strength calibration, so they cannot satisfy the VCEP PP3 requirement. BayesDel 0.199863 has no established ACMG-strength calibration and is likewise unusable. PP3 is therefore not met: no applicable computational line supports a deleterious effect on the gene product.

Rule

VCEP APC v2.1 PP3_Supporting (Applicable): missense variants - in silico splicing predictors used to reveal possible splicing effects, conservation/evolution models not permitted; non-canonical splicing variants - multiple in silico splicing predictors support a deleterious effect.

Evidence considered

• APC/InSiGHT VCEP v2.1 PP3 supporting rule (cspec): 'Missense variants: Do not use computational prediction models for conservation, evolution, etc. In silico splicing predictors should be used for presumed missense variants to reveal possible splicing effects. Non-canonical splicing variants: Multiple in silico splicing predictors support a deleterious effect.'
• Variant consequence used for path selection: NM_001127510.3:c.6907G>A encodes NP_001120982.1:p.(Gly2303Arg), a coding missense change, so the VCEP missense arm (and, on the generic path, REVEL) governs PP3.
• REVEL score 0.611 (revel), below the ClinGen SVI PP3-supporting cutoff of 0.644 (Pejaver et al. 2022, Am J Hum Genet, PMID:36413997). Gray-zone result, so PP3 is not met on the generic missense path.
• BayesDel score 0.199863 (bayesdel): no generic ACMG-strength calibration is established for BayesDel, no VCEP cutoff is specified, and the VCEP excludes conservation/evolution models for missense PP3; BayesDel is therefore not available as PP3 support.
• SpliceAI lookup (spliceai) failed with a read timeout; max_delta_score is null rather than 0.00. Missing data, not evidence of absent splice impact, and it was not read as a negative splice prediction.
• Pangolin values recorded in the same SpliceAI query (splice gain 0.001, splice loss -0.003): Pangolin is not a VCEP-recommended program and has no published ACMG-strength calibration, so these values cannot substitute for the missing SpliceAI result.
• VCEP splice-prediction thresholds, quoted in the specification's PS1 instructions (cspec): for SpliceAI, loss of a native splice site is considered for scores between 0.8 and 1; gain of a cryptic splice site is strong for scores between 0.8 and 1 and moderate for a score between 0.2 and 0.8. For MaxEntScan a score >3 is required for credibility of a native site prediction and a threshold of -15% is considered for native splice site loss (Houdayer et al. 2012, PMID:22505045).

Still missing

• SpliceAI lookup failed with a read timeout, so max_delta_score is null; the VCEP splice-effect route for a presumed missense variant cannot be formally closed. Re-query SpliceAI (or run the other VCEP-recommended programs, MaxEntScan and VarSeak) before finalising PP3.
• No annotation of the variant's distance from the nearest exon/intron junction was present in the evidence reviewed, so the a priori plausibility of a splice effect at c.6907 could not be established from position alone.

Sources

Reasoning

PP5 requires a reputable source reporting the variant as pathogenic, and under this pipeline's rule it is applied only for an exact-variant ClinVar expert-panel classification: an expert-panel Pathogenic or Likely pathogenic assertion supports PP5, while ordinary laboratory submissions, aggregate ClinVar labels and non-expert assertions must never trigger it. The exact-variant ClinVar record (VariationID 133533, NM_000038.6(APC):c.6907G>A p.(Gly2303Arg)) contains zero expert-panel submissions out of 15 total, and the record-level review status is "criteria provided, single submitter" (one star, not the three-star expert-panel tier). Its content is an aggregate of ordinary clinical-laboratory assertions: 5 Likely benign, 3 Benign, 1 benign, 1 Likely Benign and 3 Uncertain significance clinical laboratories, with no Pathogenic or Likely pathogenic expert-panel assertion of any kind. Explicit comparison: qualifying exact-variant expert-panel pathogenic classifications = 0; 0 >= 1? No - the trigger is absent. The APC InSiGHT VCEP specification additionally lists PP5 as "Not Applicable" ("This criterion is not for use as recommended by the ClinGen SVI VCEP Review Committee"), so no VCEP-specific PP5 route exists either. PP5 is therefore not met.

Rule

PP5 (Supporting) applies only when an exact-variant ClinVar expert-panel classification is Pathogenic or Likely pathogenic; ordinary laboratory submissions, aggregate ClinVar labels and non-expert assertions do not trigger it. The APC InSiGHT VCEP additionally marks PP5 Not Applicable.

Evidence considered

• The exact-variant ClinVar submission audit reports expert_panel_submissions = 0 out of 15 submissions for NM_000038.6(APC):c.6907G>A (p.Gly2303Arg); every submitter is an ordinary clinical laboratory or curation group (e.g. Ambry Genetics, GeneDx, Labcorp/Invitae, Myriad, Color Health, Quest Diagnostics), and none is flagged is_expert_panel.
• The record-level ClinVar review status for this variant is "criteria provided, single submitter", i.e. one star, not the three-star expert-panel tier required for PP5.
• The exact-variant ClinVar assertions are mixed and non-pathogenic in aggregate: 5 clinical laboratories Likely benign, 3 Benign, 1 benign, 1 Likely Benign, 3 Uncertain significance, plus submissions with no classification provided; no Pathogenic or Likely pathogenic assertion exists.
• The APC InSiGHT VCEP specification lists PP5 as "Not Applicable" with the note "This criterion is not for use as recommended by the ClinGen SVI VCEP Review Committee"; it is therefore evaluated here only through the exact-variant expert-panel route, which is empty.
• Explicit comparison: 0 qualifying exact-variant expert-panel Pathogenic/Likely pathogenic classifications >= 1 required? No.
• The original ACMG/AMP 2015 framework defines PP5 as a reputable source reporting the variant as pathogenic without independent evaluation; no such pathogenic report exists for this variant in any source reviewed.

Sources

Reasoning

Governing framework: APC VCEP specification v2.1 sets BS2 (Benign Strong) at '>= 10 points for healthy individuals OR >= 2 times in homozygous state', with a supporting tier at '>= 3 points for healthy individuals'. The same specification's instructionsToUse for BS2 states that the non-cancer gnomAD dataset (v2.1.1) 'cannot be used for heterozygous healthy individuals, because of the limited phenotype information and since it is usually already used for BA1/BS1. However, the non-cancer dataset from gnomAD (v2.1.1) can be used to search for homozygous individuals.' Homozygote branch (VCEP-designated source): gnomAD v2.1 non-cancer exomes reports exactly 1 homozygote (South Asian; 83/236,496 alleles), and the second non-cancer source available for this variant, gnomAD v3.1 non-cancer genomes, reports 0 homozygotes (8/147,862 alleles). One homozygote is below the >= 2 required, so the homozygote branch is not satisfied on the VCEP-specified data. Healthy-individual points branch: 0 points can be assembled. The gnomAD non-cancer heterozygous carriers (83 alleles; 80 of them South Asian) cannot be scored as 'healthy individuals', because the VCEP instruction explicitly excludes the non-cancer dataset for heterozygous healthy individuals - doing so would both double-count the frequency evidence already used for BA1/BS1 and overstate phenotype information that gnomAD does not provide; and the 0.5-point keyword category ('control, non-cancer, normal, unaffected population') requires curated individual-level healthy status, which gnomAD carriers do not carry. Among the retrieved literature, the one genuinely healthy-adult cohort (PMID:24728327, whole-genome sequencing of 681 generally healthy adults across 158 cancer-susceptibility genes) contains no variant-level APC data and does not report c.6907G>A or p.(G2303R); no other retrieved full text reports this variant at all. With 0 points, neither the >= 10-point strong tier nor the >= 3-point supporting tier is met. Flag for human review - dataset scope, not the underlying biology, is the deciding factor: gnomAD v4.1 all-comers documents 3 homozygotes for this variant, all South Asian (248/91,074 South Asian alleles). If the v4.1 all-comers dataset is deemed admissible for the homozygote count, the '>= 2 times in homozygous state' condition would be satisfied and BS2 would be met at Strong. That dataset was not used here because the VCEP specifies the non-cancer gnomAD dataset (v2.1.1) for these criteria, no v4.1 non-cancer subset is available for this case, and v4.1 postdates the specification; the choice does not change the final classification because BA1 is already met stand-alone. Counts across overlapping datasets were not summed (the v2.1 non-cancer and v2.1 all-comers homozygote is the same individual, so the largest non-overlapping homozygote count is the single v2.1 non-cancer observation).

Rule

APC VCEP v2.1 BS2 (Benign Strong): >= 10 points for healthy individuals OR >= 2 times in homozygous state (supporting: >= 3 points for healthy individuals); non-cancer gnomAD (v2.1.1) usable only to search for homozygous individuals, not for heterozygous healthy individuals.

Evidence considered

• APC VCEP specification v2.1 BS2 rules: Strong - '>= 10 points for healthy individuals OR >= 2 times in homozygous state'; Supporting - '>= 3 points for healthy individuals'. A healthy individual is worth 1 point if aged >= 50 years with fewer than 5 adenomatous polyps at colonoscopy and no Table 1 features, or aged >= 50 years with colorectal cancer/polyposis not the testing indication; 0.5 points are given for individuals identified by keywords including control, non-cancer, normal, unaffected population.
• APC VCEP specification v2.1 BS2 instructionsToUse: 'The non-cancer dataset from gnomAD (v2.1.1) cannot be used for "heterozygous healthy individuals", because of the limited phenotype information and since it is usually already used for BA1/BS1. However, the non-cancer dataset from gnomAD (v2.1.1) can be used to search for homozygous individuals.'
• gnomAD v2.1 non-cancer exomes (GRCh37 5-112178198-G-A): 1 homozygote in 83/236,496 alleles (AF 0.000350957); the homozygote is in the South Asian subset (80/30,518 alleles, AF 0.002621403).
• gnomAD v3.1 non-cancer genomes: 0 homozygotes in 8/147,862 alleles (AF 0.0000541). gnomAD v2.1 all-comers exomes: 1 homozygote in 83/251,012 alleles. gnomAD v4.1 all-comers: 3 homozygotes in 254/1,613,908 alleles, all within the South Asian subset (248/91,074 alleles, AF 0.00272306).
• PMID:24728327 (whole-genome sequencing of 681 generally healthy adults, 158 cancer-susceptibility genes) mentions APC only at the gene level and reports no APC variant-level data; searching it for c.6907G>A, p.(G2303R), p.Gly2303Arg and G2303R returned no variant-level observations, so it contributes no healthy-individual points.
• No other retrieved full text reports this variant: the APC clinical utility gene card (PMID:21368914), the ACMG secondary-findings policy statement (PMID:25356965), the ACMG/AMP 2015 standards (PMID:25741868), the ACG hereditary GI cancer guideline (PMID:25645574) and the variant-scoring methodology paper (PMID:26467025) were each searched for the variant and contain no APC variant-level or healthy-individual data for it.

Still missing

• Dataset-scope question requiring human adjudication: gnomAD v4.1 (all-comers, no non-cancer subset available) reports 3 homozygotes, all South Asian, which would satisfy BS2 strong's '>= 2 times in homozygous state' if that dataset were admitted; the defining source per the APC VCEP v2.1 instruction is the non-cancer gnomAD v2.1.1 dataset, which shows 1 homozygote.
• No curated individual-level 'healthy adult' evidence exists for this variant: the only retrieved healthy-adult cohort study (PMID:24728327) does not report c.6907G>A / p.(G2303R), and gnomAD non-cancer heterozygous carriers are excluded from BS2 point scoring by the VCEP instructionsToUse.

Sources

Reasoning

Under the governing ClinGen InSiGHT APC VCEP specification v2.1 (CSPEC doc 1529292791), BS3_Strong requires an RNA assay of a synonymous or intronic variant in a constitutional patient sample demonstrating no mRNA aberration, together with demonstration of biallelic expression and/or use of nonsense-mediated decay inhibition. BS3_Supporting requires the same RNA assay without those controls, or a protein assay showing retention of beta-catenin regulated transcription activity comparable to wild-type - the latter explicitly limited to variants within the beta-catenin binding domain, codons 959-2129 of APC. The variant NM_001127510.3:c.6907G>A, p.(Gly2303Arg) is a missense change at codon 2303: it is neither synonymous nor intronic, so the RNA-assay branch of BS3 does not describe it, and codon 2303 is outside codons 959-2129, so the protein-assay branch is not framework-applicable. No RNA assay and no protein assay for this variant exists in any consulted source; OncoKB holds no variant-specific reviewed functional evidence, and none of the ten triaged publications characterises this variant functionally or even mentions it. BS3 is therefore not met. This is an absence of benign functional evidence, not a positive demonstration that the variant is functional; the alanine-replacement protein-assay route is unavailable at this codon.

Rule

InSiGHT APC VCEP v2.1 (CSPEC doc 1529292791) BS3: RNA assay of a synonymous or intronic variant in a constitutional patient sample showing no mRNA aberration AND biallelic expression and/or NMD inhibition (Strong); the same RNA assay without those controls, OR a protein assay showing retention of beta-catenin regulated transcription activity comparable to wild-type restricted to codons 959-2129 (Supporting).

Evidence considered

• APC VCEP specification v2.1 BS3_Strong: RNA assay of a synonymous or intronic variant in a constitutional patient sample demonstrating no mRNA aberration AND biallelic expression shown and/or nonsense-mediated decay inhibition used.
• APC VCEP specification v2.1 BS3_Supporting: the same RNA assay without demonstration of biallelic expression or NMD inhibition, OR a protein assay showing retention of beta-catenin regulated transcription activity comparable to wild-type, with that protein-assay option limited to variants within the beta-catenin binding domain, codons 959-2129 of APC.
• The variant is a missense substitution at codon 2303 (NM_001127510.3:c.6907G>A, p.(Gly2303Arg)), so it is neither synonymous nor intronic and the RNA-assay branches of BS3 do not describe it.
• Codon 2303 lies outside codons 959-2129, so the BS3_Supporting protein-assay branch is not framework-applicable to this variant irrespective of any assay result.
• No functional assay (RNA or protein) for this variant was found in any source; OncoKB reports 'Unknown Oncogenic Effect' with no variant-specific reviewed functional evidence.
• None of the ten triaged publications reports functional characterisation of c.6907G>A / p.(Gly2303R); the variant is not mentioned in any of them.
• COSMIC somatic recurrence (COSV57373649, n=4) provides no functional-assay evidence of retained or lost protein function and cannot support BS3.
• The absence of an assay is not affirmative evidence of a benign functional effect, so no BS3 weight (strong or supporting) is available.

Still missing

• No RNA assay of this variant exists; BS3_Strong and the RNA branch of BS3_Supporting are defined for synonymous or intronic variants, which this missense change is not.
• The BS3_Supporting protein-assay branch (retention of beta-catenin regulated transcription activity comparable to wild-type) is restricted to codons 959-2129 and cannot be applied to a variant at codon 2303.
• Absence of functional data must not be read as benign functional evidence: BS3 remains unestablished for this variant rather than refuted.
• Two ClinVar-cited references (PMID:20301519, PMID:26389258) have no full text or pre-computed extraction in this case and are narrative resources rather than variant-level assay studies.

Sources

Reasoning

BP2 is applicable for APC only at Benign Supporting strength, with the VCEP-modified rule: 'Observed in trans with a (Likely) Pathogenic APC variant OR >=3 times in an unknown phase with different (Likely) Pathogenic APC variants.' Neither limb is satisfied by the evidence available for c.6907G>A (p.Gly2303Arg). The case holds no proband, pedigree, or phase data: there is no affected-proband observation, no second APC variant, and no parental or offspring genotyping that would place this variant in trans (or, under the original ACMG clause, in cis) with a pathogenic APC variant. The ClinVar record (variation 133533; 15 submissions, none from an expert panel) contains no phase or co-occurrence assertion - no submitter reports BP2 or any other phase-based criterion, and the per-criterion evidence signals extracted from every audited submission are empty. The single submission with a validated PMID trail (SCV007112721, Quest Diagnostics, benign) cites Okawa et al., J Hepatol 2023 (PMID 36243179), whose abstract reports a gene-level enrichment of pathogenic APC variants among 1,292 biliary tract cancer cases versus 37,583 controls but says nothing about the phase of c.6907G>A; that PMID is not among this case's registered sources and is recorded here in prose only, never as a source key. gnomAD v4.1 lists 3 homozygotes for this variant, but population homozygosity is not an observation in trans with a pathogenic APC variant and does not trigger the VCEP rule. The two full-text papers consulted for this criterion (PMID:25741868, PMID:24728327) contain no variant-level co-occurrence or phase information. BP2 is therefore not met; evidence that could meet it would be case-level genotyping showing c.6907G>A alongside a (likely) pathogenic APC variant in the same individual.

Rule

APC VCEP v2.1, BP2 (Benign Supporting): 'Observed in trans with a (Likely) Pathogenic APC variant OR >=3 times in an unknown phase with different (Likely) Pathogenic APC variants.'

Evidence considered

• Governing specification (ClinGen InSiGHT Hereditary Colorectal Cancer/Polyposis VCEP, APC, version 2.1) applies BP2 at Benign Supporting with the rule: 'Observed in trans with a (Likely) Pathogenic APC variant OR >=3 times in an unknown phase with different (Likely) Pathogenic APC variants.' No stand-alone, very strong, strong, or moderate strength is available for BP2.
• The VCEP supplementary material shows the original ACMG/AMP BP2 row ('Observed in trans with a pathogenic variant for a fully penetrant dominant gene/disorder; or observed in cis with a pathogenic variant in any inheritance pattern') modified per Mester et al. 2018 (PMID:30311380) and Luo et al. 2019 (PMID:31648317); the cis limb is likewise unfulfilled here because no phase data exist.
• ClinVar record 133533 (15 audited submissions, no expert-panel submission) reports no phase-based or co-occurrence criterion: no submitter lists BP2, and the per-submission evidence signals harvested from the submission audit are empty for every accession.
• The only ClinVar submission carrying a validated PMID trail (SCV007112721, Quest Diagnostics, benign) cites Okawa et al., J Hepatol 2023 (PMID 36243179). Its abstract describes germline panel sequencing in 1,292 biliary tract cancer cases and 37,583 controls with gene-level enrichment of pathogenic APC variants; it contains no statement about the phase of c.6907G>A. This PMID is not a registered source for this case, so it is cited in prose only.
• gnomAD v4.1 reports 3 homozygotes and gnomAD v2.1 reports 1 homozygote for this variant; homozygosity in a population database does not demonstrate that the variant occurs in trans with a pathogenic APC variant in a patient, so neither limb of the VCEP BP2 rule is met by this observation.
• Full-text sources consulted for this criterion (PMID:25741868, PMID:24728327) provide no variant-level co-occurrence, cis/trans, or phase information for c.6907G>A; the healthy-adult WGS study (PMID:24728327) does not name the variant and restricts per-variant characterisation to BRCA1, BRCA2, TP53, KRAS and PTEN.
• ClinVar's consensus classification for this variant was not used as evidence for or against BP2; the criterion was evaluated only from primary phase/co-occurrence observations.

Still missing

• No proband, pedigree, parental, or offspring genotyping data exist in this case, so in-trans versus in-cis phase of c.6907G>A relative to any other APC variant could not be established directly.
• The second limb of the VCEP BP2 rule (>=3 occurrences in unknown phase with different (likely) pathogenic APC variants) requires case-level co-occurrence records; none are present in the registered sources for this variant.
• PMID 36243179, the only ClinVar-submission paper flagged as a criterion lead, is not contained in this case's source registry and has no fetched full text; it was screened at abstract level only, so any per-patient co-occurrence table it may hold could not be inspected.

Sources

Reasoning

BP6 requires a reputable source reporting the variant as benign, and under this pipeline's rule it is applied only for an exact-variant ClinVar expert-panel classification: an expert-panel Benign or Likely benign assertion supports BP6, while ordinary laboratory submissions, aggregate ClinVar labels and non-expert assertions must never trigger it. The exact-variant ClinVar record (VariationID 133533, NM_000038.6(APC):c.6907G>A p.(Gly2303Arg)) contains zero expert-panel submissions out of 15 total, and the record-level review status is "criteria provided, single submitter" (one star, not the three-star expert-panel tier). The benign-leaning labels in the record (3 Benign, 1 benign, 5 Likely benign, 1 Likely Benign clinical laboratories) are aggregate, non-expert, single-submitter clinical-laboratory classifications and are explicitly ineligible to trigger BP6; the remaining assertions are Uncertain significance or unclassified. Explicit comparison: qualifying exact-variant expert-panel benign classifications = 0; 0 >= 1? No - the trigger is absent. The APC InSiGHT VCEP specification also lists BP6 as "Not Applicable" ("This criterion is not for use as recommended by the ClinGen SVI VCEP Review Committee"). BP6 is therefore not met.

Rule

BP6 (Supporting) applies only when an exact-variant ClinVar expert-panel classification is Benign or Likely benign; ordinary laboratory submissions, aggregate ClinVar labels and non-expert assertions do not trigger it. The APC InSiGHT VCEP additionally marks BP6 Not Applicable.

Evidence considered

• The exact-variant ClinVar submission audit reports expert_panel_submissions = 0 out of 15 submissions for NM_000038.6(APC):c.6907G>A (p.Gly2303Arg); no submitter is flagged is_expert_panel, and the submitters are ordinary clinical laboratories and curation groups.
• The record-level ClinVar review status for this variant is "criteria provided, single submitter", i.e. one star, not the three-star expert-panel tier required for BP6.
• The exact-variant ClinVar benign labels (3 Benign, 1 benign, 5 Likely benign, 1 Likely Benign clinical laboratories) are ordinary aggregate laboratory submissions and must not trigger BP6; the remaining submissions are Uncertain significance or have no classification provided.
• The APC InSiGHT VCEP specification lists BP6 as "Not Applicable" with the note "This criterion is not for use as recommended by the ClinGen SVI VCEP Review Committee"; it is therefore evaluated here only through the exact-variant expert-panel route, which is empty.
• Explicit comparison: 0 qualifying exact-variant expert-panel Benign/Likely benign classifications >= 1 required? No.
• The original ACMG/AMP 2015 framework defines BP6 as a reputable source reporting the variant as benign without independent evaluation; the non-expert single-submitter ClinVar assertions do not meet that standard.

Sources

Reasoning

Governing specification: ClinGen InSiGHT Hereditary Colorectal Cancer/Polyposis Expert Panel (APC) v2.1, which scores PS2 by de novo score - >=4 de novo scores = Very Strong, 2-3.5 = Strong, 1-1.5 = Moderate - with the per-proband de novo score taken from Tables 1 and 2: a proband with >=1 phenotype point scores 2 de novo points when both parental relationships are confirmed and 1 when they are unconfirmed, and a proband with 0.5 phenotype points scores 1 or 0.5 respectively. This case contains no proband record, no phenotype description (no polyposis, desmoid, medulloblastoma, hepatoblastoma, CHRPE, gastric/duodenal adenoma or osteoma findings), no maternity/paternity testing and no de novo observation for APC c.6907G>A (p.Gly2303Arg) in any source. The six full-text papers extracted for this case (PMID:21368914, PMID:24728327, PMID:25356965, PMID:25645574, PMID:25741868, PMID:26467025) all record variant_mentioned = false, and none of the 15 ClinVar submissions reports de novo occurrence or parental testing for this variant. A de novo score therefore cannot be computed (neither the required phenotype point category nor the parental-confirmation status exists), so PS2 is not assessed rather than not met; note that a single confirmed de novo proband with >=1 phenotype point would already yield a de novo score of 2 and meet PS2 at Strong.

Rule

APC VCEP v2.1 PS2: >=4 de novo scores = Very Strong; 2-3.5 de novo scores = Strong; 1-1.5 de novo scores = Moderate. De novo score per proband from Tables 1&2: phenotype points >=1 -> 2 (confirmed parental relationships) or 1 (unconfirmed); phenotype points 0.5 -> 1 (confirmed) or 0.5 (unconfirmed).

Evidence considered

• APC VCEP v2.1 PS2 bands: >=4 de novo scores = Very Strong, 2-3.5 = Strong, 1-1.5 = Moderate ('2-3.5 _de novo_ scores. For curation of _de novo_ score see Tables 1 and 2.').
• APC VCEP Table 2 mapping used to convert proband phenotype points into de novo scores: phenotype points >=1 gives a de novo score of 2 with confirmed parental relationships or 1 with unconfirmed parental relationships; 0.5 phenotype points gives 1 or 0.5 respectively.
• APC VCEP Table 1 phenotype-point categories: 1 point for a phenotype highly specific for APC (e.g. typical colorectal phenotype 20-99 adenomas, desmoid without somatic CTNNB1 variant, WNT-subtype medulloblastoma without somatic CTNNB1 variant, hepatoblastoma without somatic CTNNB1 variant) and 0.5 point for a phenotype consistent with APC (e.g. =20 adenomas, unknown CTNNB1 status, multifocal/bilateral CHRPE, >=2 gastric or duodenal adenomas, osteomas, typical dominant FAP family history).
• APC VCEP supplementary material states the PS2 modification is based on the SVI recommendation for de novo criteria, i.e. de novo scoring replaces the plain 'confirmed parental relationships' claim.
• Case files reviewed for this criterion (prefetch.json, case_summary.json, evidence.json, clinvar_submission_audit.json, paper_extracts.json) contain no de novo occurrence, trio testing, or proband phenotype data for this variant; the case was submitted as a variant-level query NM_001127510.3:c.6907G>A with no clinical record attached.
• ClinVar variation 133533 holds 15 submissions (5 Likely benign, 3 Benign, 3 Uncertain significance, 1 Likely Benign, 1 benign from 15 submitters) but the retrievable submission records carry no de novo or parental-testing statements and are single-submitter assertions, so they cannot be used to score PS2 for a specific variant.
• Papers checked for de novo/trio content: PMID:25356965 (ACMG secondary-findings policy statement - no case data), PMID:26467025 (variant-scoring methodology over 11,894 variants - no variant-specific data), PMID:21368914 (FAP/AFAP clinical utility gene card), PMID:25645574 (ACG hereditary GI cancer guideline), PMID:25741868 (ACMG/AMP 2015 framework), PMID:24728327 (681 healthy adults WGS cohort). None names c.6907G>A or p.(G2303R) and none reports de novo occurrence.

Still missing

• No proband-level de novo observation for APC c.6907G>A is reported anywhere in the case evidence.
• No parental testing (confirmed maternity and/or paternity) is documented, so Table 1&2 parental-relationship status cannot be assigned.
• No proband phenotype is supplied, so the phenotype-point category (1 vs 0.5 vs 0) needed to convert phenotype points into a de novo score cannot be assigned.

Sources

Reasoning

The APC InSiGHT VCEP replaces the generic case-control PS4 with a proband-based phenotype-point score curated from Table 1, and its strength tiers are written as upward comparisons: Very Strong ">= 16 phenotype points", Strong "4-15.5 phenotype points", Moderate "2-3.5 phenotype points", Supporting "1-1.5 phenotype point" (1 point for a phenotype highly specific for APC, 0.5 point for a phenotype consistent with APC). Comparing the evidence available for this exact variant (c.6907G>A, p.(Gly2303Arg)): 0 >= 1 (the Supporting minimum)? No; 0 >= 16 (Very Strong)? No - no tier is reached. The reason is missing input rather than a near miss: no source in this case reports an individual carrying this variant with any phenotype at all. None of the ten screened publications mentions the variant (all variant_mentioned: false), the six full texts describe only gene-level APC/colorectal polyposis background, and the 15 exact-variant ClinVar submissions carry no phenotype detail (labels/indications only). No case-control odds ratio or relative risk with a confidence interval for this exact variant was found either, so the generic ACMG enrichment route is also unavailable. With no proband-level phenotype points assignable, PS4 is not assessed at any strength.

Rule

APC InSiGHT VCEP PS4 (phenotype points, Table 1): ">= 16 phenotype points" = Very Strong; "4-15.5" = Strong; "2-3.5" = Moderate; "1-1.5" = Supporting. This criterion strengthens as the point total goes UP (operator >=); no tier is reached when the curated total is 0.

Evidence considered

• The APC InSiGHT VCEP specification defines PS4 with a phenotype-point system: ">= 16 phenotype points" = Very Strong, "4-15.5 phenotype points" = Strong, "2-3.5 phenotype points" = Moderate, "1-1.5 phenotype point" = Supporting, with points curated per proband using Table 1.
• Table 1 of the APC VCEP materials assigns 1 phenotype point per proband for a phenotype highly specific for APC and 0.5 point for a phenotype consistent with APC, based on colorectal phenotype, desmoid(s), medulloblastoma, hepatoblastoma, CHRPE, multiple gastric/duodenal adenomas, osteoma(s) and typical FAP family history.
• The APC VCEP supplementary material states the PS4 modification follows Fortuno et al. 2020; no numeric case-control odds-ratio or relative-risk threshold is defined by the VCEP for APC.
• No source in this case reports an individual carrying c.6907G>A with a colorectal polyposis or other APC-associated phenotype: none of the ten screened publications mentions the variant, and the 15 exact-variant ClinVar submissions contain classification labels and testing indications only, with no patient-level phenotype.
• Explicit comparison performed against the governing operators: available phenotype points = 0; 0 >= 1 (Supporting tier, 1-1.5 points)? No; 0 >= 16 (Very Strong tier)? No. No PS4 strength tier is reached.
• PMID:36243179 (Okawa et al. 2023, J Hepatol), a paper referenced by one exact-variant ClinVar submission, is a biliary-tract-cancer case-control study (1,292 BTC cases vs 37,583 controls) reporting gene-level APC enrichment; its abstract provides no variant-specific APC polyposis phenotype and no phenotype points for this variant.
• PMID:21368914 (FAP/AFAP clinical utility gene card) and PMID:25645574 (ACG hereditary GI cancer guideline) were reviewed for proband-level case evidence and describe only gene-level mutation spectrum and disease mechanism; neither names c.6907G>A and neither supplies individual-level phenotype data.

Still missing

• Individual-level phenotype data (colonoscopy adenoma counts, desmoid/medulloblastoma/hepatoblastoma/CHRPE/gastric or duodenal adenoma/osteoma features, family history) for any proband carrying c.6907G>A are needed to curate APC VCEP Table 1 phenotype points.
• No case-control enrichment estimate (odds ratio or relative risk with confidence interval) for this exact variant is available.

Sources

Reasoning

Governing specification: APC VCEP v2.1, under which PM6 is scored from the same Tables 1 and 2 de novo scoring as PS2, with 2-3.5 de novo scores = Strong, 1-1.5 = Moderate and 0.5 = Supporting; the 0.5 de novo score corresponds to a proband with 0.5 phenotype points and unconfirmed parental relationships, i.e. the classical 'assumed de novo without confirmation of parentage' situation. No assumed-de novo observation of APC c.6907G>A (p.Gly2303Arg) is reported in any source available for this case: the six extracted full-text papers all record variant_mentioned = false, no ClinVar submission among the 15 states a de novo or assumed-de novo origin, and no proband or parental history exists in the case record. The minimum requirement for even PM6_Supporting (a de novo score of 0.5, i.e. at least one proband reported as de novo with a phenotype consistent with APC and unconfirmed parentage) is therefore not demonstrable, and PM6 is not assessed rather than not met.

Rule

APC VCEP v2.1 PM6: 2-3.5 de novo scores = Strong; 1-1.5 de novo scores = Moderate; 0.5 de novo scores = Supporting. De novo score per proband from Tables 1&2: phenotype points >=1 -> 2 (confirmed parental relationships) or 1 (unconfirmed); phenotype points 0.5 -> 1 (confirmed) or 0.5 (unconfirmed).

Evidence considered

• APC VCEP v2.1 PM6 bands: 2-3.5 de novo scores = Strong, 1-1.5 = Moderate, 0.5 = Supporting ('0.5 _de novo_ scores. For curation of _de novo_ score see Tables 1 and 2.').
• APC VCEP supplementary material states the PM6 modification is based on the SVI recommendation for de novo criteria, so PM6 in APC is driven by the same Table 1&2 de novo score as PS2, including the halved score when parental relationships are unconfirmed.
• APC VCEP Table 1&2: a proband with 0.5 phenotype points and unconfirmed parental relationships scores 0.5 de novo points, which is the exact configuration that would satisfy PM6 at Supporting; a proband with >=1 phenotype point and unconfirmed parentage scores 1 (PM6_Moderate).
• No source in this case reports an assumed-de novo occurrence of c.6907G>A: the extracted papers PMID:21368914, PMID:24728327, PMID:25356965, PMID:25645574, PMID:25741868 and PMID:26467025 all record variant_mentioned = false; GeneReviews PMID:20301519 and the PDQ record PMID:26389258 were not retrieved as full text (screenshot only) and are gene-level reviews that would not supply variant-specific de novo data in any case.
• ClinVar variation 133533: 15 submissions, none with retrievable criterion-level text describing a de novo or assumed-de novo observation; submissions are single-submitter and include conflicting classifications, so no de novo evidence can be drawn from them.

Still missing

• No report of a proband carrying c.6907G>A with assumed (unconfirmed) de novo origin in any case file, submitted record, or retrieved publication.
• No proband phenotype is supplied, so even the 0.5 phenotype-point category needed for a 0.5 de novo score cannot be assigned.

Sources

Reasoning

Governing specification: APC VCEP v2.1 PP1, which is scored by meioses rather than by families alone - Supporting for 3-4 meioses in >=1 family, Moderate for 5-6 meioses in >=1 family, Strong for >=7 meioses in >=2 families. Adenomatous polyposis is dominant with reduced penetrance, so this criterion requires an explicit pedigree in which informative affected and unaffected meioses can be counted. This case supplies no pedigree, no family members, no meiosis count and no co-segregation table for APC c.6907G>A (p.Gly2303Arg); the six extracted full-text papers all record variant_mentioned = false and none contains a family study of this variant, and no ClinVar submission reports co-segregation. Because zero meioses can be counted against a minimum requirement of 3-4 meioses, the criterion is not assessed rather than not met - absence of a pedigree is absence of evidence, not evidence of non-segregation.

Rule

APC VCEP v2.1 PP1: segregates in >=7 meioses in >=2 families = Strong; 5-6 meioses in >=1 family = Moderate; 3-4 meioses in >=1 family = Supporting.

Evidence considered

• APC VCEP v2.1 PP1 bands: 'Variant segregates in >= 7 meioses in >= 2 families' = Strong; 'Variant segregates in 5-6 meioses in >= 1 family' = Moderate; 'Variant segregates in 3-4 meioses in >= 1 family' = Supporting.
• APC VCEP supplementary material notes the PP1 modification is based on Luo et al., 2019 (PMID: 31042330), i.e. a quantified meiosis-count framework rather than the generic 'multiple affected family members' wording.
• No pedigree, family member genotypes, or meiosis counts appear in prefetch.json, case_summary.json, evidence.json, clinvar_submission_audit.json or the published-extract set for this case; the case was submitted as a variant-level query with no family record.
• None of the retrievable sources names the variant in a family context: PMID:21368914 (FAP/AFAP clinical utility gene card), PMID:24728327 (681 healthy adults), PMID:25356965 (ACMG secondary findings policy), PMID:25645574 (ACG guideline), PMID:25741868 (ACMG/AMP framework) and PMID:26467025 (scoring-methodology study) all record variant_mentioned = false.
• ClinVar variation 133533: 15 submissions (5 Likely benign, 3 Benign, 3 Uncertain significance, 1 Likely Benign, 1 benign); no submission record contains co-segregation data for this variant.

Still missing

• No pedigree or family structure is provided, so no meiosis count can be derived.
• No genotype data for any relative of any proband carrying c.6907G>A.
• No co-segregation study reporting this variant was retrieved (no variant-specific segregation data in any fetched paper).

Sources

Reasoning

Governing specification: APC VCEP v2.1 BS4, which is met at Strong when an affected family member without the variant scores at least 1 phenotype point (or when at least two affected members without the variant each score at least 0.5 phenotype points), and at Supporting when a single affected member without the variant scores at least 0.5 phenotype points, with phenotype points assigned from Table 1. This case contains no family members, no genotype results for any relative, and no phenotype-point assessment for anyone, so there is no affected non-carrier against which non-segregation could be demonstrated. The presence of 15 ClinVar submissions, of which 11 are classified Benign/Likely benign, does not satisfy BS4: BS4 requires family-level transmission evidence (an affected individual lacking the variant), and aggregate classification labels from other laboratories cannot be converted into a phenotype-point-bearing affected non-carrier, nor may a criterion be inferred from another source's final classification. BS4 is therefore not assessed rather than not met; no non-segregation statement (which would be required to meet it) and no segregating family (which would exclude it) exists in the record.

Rule

APC VCEP v2.1 BS4: affected member without the variant scoring >=1 phenotype point, or >=2 affected members without the variant each scoring >=0.5 phenotype points = Strong; a single affected member without the variant scoring >=0.5 phenotype points = Supporting.

Evidence considered

• APC VCEP v2.1 BS4 bands: 'Affected member without the variant must score at least 1 phenotype point or at least two affected members without the variant must each score at least 0.5 phenotype points' = Strong; 'Affected member without the variant must score at least 0.5 phenotype points' = Supporting.
• APC VCEP supplementary material records 'No comment' for BS4, i.e. the generic ACMG wording is retained but is operationalised through Table 1 phenotype points in the APC specification.
• APC VCEP Table 1 defines the phenotype-point values that a variant-negative affected relative would have to reach (1 point for a phenotype highly specific for APC, 0.5 point for a phenotype consistent with APC); no such individual or phenotype data exists in this case.
• ClinVar variation 133533: 15 submissions, 5 Likely benign, 3 Benign, 1 Likely Benign, 1 benign (11 benign-side), 3 Uncertain significance, 1 with no classification; submission records are single-submitter with no retrievable family-level non-segregation statements, and no expert-panel submission exists, so they provide neither an affected non-carrier nor usable BS4 evidence.
• No retrieved paper reports non-segregation of c.6907G>A: PMID:21368914, PMID:24728327, PMID:25356965, PMID:25645574, PMID:25741868 and PMID:26467025 all record variant_mentioned = false, and none is a family study.

Still missing

• No family members are reported, so no affected relative without the variant can be identified or scored.
• No phenotype-point assignment exists for any individual, so the >=1, two-times >=0.5, or single >=0.5 phenotype-point thresholds cannot be tested.
• No genotype data (presence/absence of c.6907G>A) for relatives of any carrier.

Sources

Reasoning

The APC InSiGHT VCEP defines BP5 at Supporting strength as applicable only when an alternate genetic basis explains a colorectal polyposis phenotype, specified as a (Likely) Pathogenic variant in another adenomatous polyposis gene (heterozygous POLD1 or POLE; biallelic MUTYH, NTHL1 or MSH3; and, for patients with onset in childhood/adolescence, biallelic MLH1, MSH2, MSH6 or PMS2), and it states this rule "is only applicable when a colorectal polyposis phenotype is present". The governing text for this criterion contains no numeric or likelihood-ratio value - the comparison direction warning does not apply here because there is no threshold to compare; the rule is categorical. For this case neither element can be established: no multi-gene testing result is reported for the tested individual (no alternate-gene variant of any class is documented), and no colorectal polyposis phenotype is documented anywhere in the case sources, which contain no patient-level clinical information at all. Because the required inputs are absent rather than affirmatively excluded, BP5 is not assessed rather than not met.

Rule

APC InSiGHT VCEP BP5 (Supporting): applicable only when a (Likely) Pathogenic variant in another specified adenomatous polyposis gene provides an alternate genetic basis for a colorectal polyposis phenotype; no numeric or likelihood-ratio threshold is defined.

Evidence considered

• The APC InSiGHT VCEP BP5 instruction states the criterion is applicable only for an alternate genetic basis of the colorectal polyposis phenotype, naming heterozygous POLD1 or POLE variants, biallelic MUTYH, NTHL1 or MSH3 variants, and (for childhood/adolescence onset) biallelic MLH1, MSH2, MSH6 or PMS2 variants, and that "This rule is only applicable when a colorectal polyposis phenotype is present."
• The BP5 strength rule is a single categorical statement - "Only applicable for an alternate genetic basis of the colorectal polyposis phenotype." - with no numeric value or likelihood-ratio threshold anywhere in the governing text, so no explicit inequality can be written for this criterion.
• No source consulted for this case documents multi-gene testing results for the tested individual, and no colorectal polyposis phenotype is recorded; the exact-variant ClinVar submissions carry classification labels only.
• The original ACMG/AMP 2015 definition of BP5 (variant found in a case with an alternate molecular basis for disease) is not satisfiable on the available evidence because no alternate molecular basis is documented.

Still missing

• Results of multi-gene testing for the tested individual (MUTYH, NTHL1, MSH3, POLD1, POLE and, for early-onset disease, MLH1/MSH2/MSH6/PMS2) are not available.
• A documented colorectal polyposis phenotype for the tested individual is required before BP5 can be applied at all.

Sources