POLE c.2792T>C, p.(Phe931Ser)

Reasoning

No ClinGen CSPEC/VCEP exists for POLE; the governing local Leon-Castillo et al. 2020 custom POLE framework customizes only PM1, PS4, PP3 and BP4 and states no population source or cutoff, so the generic SVI-supported PM2 threshold of observed allele frequency <=0.0001, applied at supporting strength only, governs, using the default all-comers gnomAD entries. NM_006231.4:c.2792T>C is present in gnomAD v2.1 at 3.88871e-05 (11/282,870 alleles) and in gnomAD v4.1 at 1.67279e-05 (27/1,614,072 alleles), both below 0.0001 by roughly 2.6-fold and 6-fold respectively, with 0 homozygotes in either dataset, and the reported gnomAD v4.1 grpmax FAF of 7.9e-07 is likewise far below threshold. The non-cancer subsets, although not required by the governing framework, give the same conclusion (gnomAD v2.1 non-cancer exomes 4.22037e-05, 10/236,946, 0 homozygotes; gnomAD v3.1 non-cancer genomes 2.02796e-05, 3/147,932, 0 homozygotes), so the result is not an artefact of the cancer-cohort composition of the all-comers totals. One ancestry caveat keeps this from being a routine call: the allele is confined to the Ashkenazi Jewish group, where the frequency is 9.6432e-04 in gnomAD v2.1 (10/10,370 alleles), 7.76817e-04 in gnomAD v4.1 (23/29,608 alleles) and 1.013993e-03 in gnomAD v2.1 non-cancer exomes (10/9,862 alleles) - roughly 8- to 10-fold above the 0.0001 threshold - while every other ancestry group carries 0 alleles in both datasets. Because the governing framework requires no popmax/population-specific reading, PM2 is met at supporting strength on the default all-comers frequency, but the criterion is flagged for human review so that a reviewer can decide whether the ancestry-concentrated frequency should defeat it.

Rule

Generic ACMG/AMP PM2 (supporting only): observed allele frequency <=0.0001, applied against the default all-comers gnomAD v2.1/v4.1 entries because the governing local POLE framework specifies no population source and no gene-specific PM2 cutoff.

Evidence considered

• No ClinGen CSPEC/VCEP for POLE was available (cspec not found); the governing specification is the local Leon-Castillo et al. 2020 custom POLE framework (custom_internal, official_clingen false, precedence rank 2), which customizes only PM1, PS4, PP3 and BP4. It neither specifies a non-cancer or exome-only population source nor a frequency cutoff for PM2, so the default all-comers gnomAD v2.1 and gnomAD v4.1 entries govern, with generic ACMG/AMP thresholds.
• Generic PM2 threshold used: observed allele frequency <=0.0001, applied at supporting strength only (ClinGen SVI recommendation to downgrade PM2 to supporting, based on Richards et al. 2015, PMID:25741868).
• gnomAD v2.1 all-comers: total AF 3.8887121292466504e-05 (11/282,870 alleles; exome 10/251,484, genome 1/31,386), 0 homozygotes. This is below the 0.0001 threshold.
• gnomAD v4.1 all-comers: total AF 1.6727878310261253e-05 (27/1,614,072 alleles; exome 24/1,461,882, genome 3/152,190), 0 homozygotes, reported grpmax FAF 7.900000000000001e-07. This is below the 0.0001 threshold.
• Sensitivity check on the available non-cancer subsets, which are not required by the governing framework but were examined because POLE is a hereditary cancer-predisposition gene and the all-comers totals include cancer cohorts: gnomAD v2.1 non-cancer exomes AF 4.220370886193479e-05 (10/236,946 alleles, 0 homozygotes) and gnomAD v3.1 non-cancer genomes AF 2.027958791877349e-05 (3/147,932 alleles, 0 homozygotes, grpmax FAF 0). Both are also below 0.0001, so PM2 is met whether the all-comers or the non-cancer source is used.
• Ancestry caveat: the allele is observed only in Ashkenazi Jewish individuals. Ashkenazi Jewish AF is 9.643201542912247e-04 (10/10,370 alleles, 0 homozygotes) in gnomAD v2.1, 7.768170764658201e-04 (23/29,608 alleles, 0 homozygotes) in gnomAD v4.1, and 1.013993104846887e-03 (10/9,862 alleles, 0 homozygotes) in gnomAD v2.1 non-cancer exomes; every other ancestry group reports 0 alleles in both datasets. These ancestry-specific values are roughly 8- to 10-fold above the 0.0001 generic threshold, which is why the criterion is flagged for human review rather than recorded silently.
• gnomAD-Canada v1.0 reported no result for this variant and contributed no frequency data.
• POLE germline predisposition is an adult-onset cancer-predisposition context with expected heterozygous carriers in population databases, so the 27 v4.1 and 11 v2.1 heterozygous observations do not by themselves argue against rarity; no homozygotes are reported in any dataset.

Still missing

• The variant is ancestry-concentrated: Ashkenazi Jewish AF is 7.76817e-04 (gnomAD v4.1, 23/29,608 alleles), 9.643201542912247e-04 (v2.1, 10/10,370) and 1.013993104846887e-03 (v2.1 non-cancer exomes, 10/9,862), i.e. about 8- to 10-fold above the generic 0.0001 PM2 threshold, while all other ancestry groups carry 0 alleles. Under a popmax/population-specific reading PM2 would not be met; the governing framework specifies no popmax requirement, so the all-comers frequency governed and PM2 was recorded as met at supporting strength. Human review should confirm which reading is intended.
• No POLE-specific PM2 frequency threshold derived from disease prevalence and penetrance is available; the generic SVI threshold was applied because the governing framework defines no PM2 rule.
• No individual-level or age/phenotype data accompany the gnomAD observations, so carrier frequency could not be interpreted against the penetrance of POLE-associated germline cancer predisposition.

Sources

Reasoning

Scope: NM_006231.4:c.2792T>C is a coding missense change (NP_006222.2:p.(Phe931Ser)), so PP3 is applicable (not a truncating/protein-length-changing variant). The governing POLE framework is the custom PP3 rule derived from Leon-Castillo et al. 2020 (vcep_db/POLE), which requires that the exact missense variant appear in Supplementary Table S2 or S3 with a REVEL class of 'likely disease causing' and <=1 benign in-silico results. The variant is ABSENT from all three supplementary tables: grep of PATH-250-323-s002.xlsx.txt (Table S1), PATH-250-323-s003.xlsx.txt (Table S2, TCGA in-silico results) and PATH-250-323-s004.xlsx.txt (Table S3, non-TCGA in-silico results) for 'c.2792T>C', 'p.Phe931Ser', 'p.F931S', 'F931' and '2792' returned no match, and the main PDF (PATH-250-323.pdf.txt) likewise contains no occurrence of the variant. The custom rule therefore explicitly directs: 'If the variant is absent from the supplementary tables, do not use this custom PP3 rule and fall back to the project's generic in-silico workflow.' Under that generic workflow the missense predictor REVEL = 0.749 (local REVEL v1.3 lookup, chr12:132661599 A>G). Against the published ClinGen SVI REVEL calibration (Pejaver et al. 2022, PMID:36413997) this falls in the PP3-supporting band (>=0.644) but below the moderate band (>=0.773) and below the strong band (>=0.932). PP3 is therefore met at supporting strength only. BayesDel = 0.145561 was available but is not used: no ACMG-strength calibration for BayesDel is established, and the governing VCEP specifies no BayesDel cutoff, so the score is treated as not_available for PP3. SpliceAI max delta = 0.008 does not contribute positively to PP3 (well below the 0.2 supporting / 0.5 moderate cutoffs) and is not used to argue benign here because the variant lies outside any splice region (nearest acceptor 81 bp, nearest donor 53 bp from SpliceAI DP values). Contextual caveat carried from the governing paper: Leon-Castillo et al. state that in-silico tools were validated for exonuclease-domain mutations, that >95% of POLE mutations outside the exonuclease domain are classified non-pathogenic by POLE-score, and p.Phe931Ser lies outside the exonuclease domain (which spans roughly codons 268-471); the generic REVEL-derived supporting call is therefore low-weight, which is why supporting (not moderate) strength is assigned and no upgrade is considered.

Rule

POLE custom PP3 rule not triggered (variant absent from PATH-250-323 Tables S2/S3); generic fallback: REVEL >= 0.644 => PP3_supporting (Pejaver et al. 2022, PMID:36413997).

Evidence considered

• PP3 scope check: c.2792T>C is a missense substitution p.(Phe931Ser), so the criterion is in scope; PP3 was not evaluated as a splice-impact call (SpliceAI max delta 0.008, nearest acceptor 81 bp and nearest donor 53 bp away, i.e. not a splice-region variant).
• Local REVEL v1.3 lookup returned revel_score = 0.749 at chr12:132661599 A>G (hg38; chr12:133238185 A>G hg37), found = true.
• Applied the published ClinGen SVI REVEL calibration verbatim: PP3 supporting >= 0.644, PP3 moderate >= 0.773, PP3 strong >= 0.932 (Pejaver et al. 2022, Am J Hum Genet, PMID:36413997). 0.749 >= 0.644 but < 0.773, giving PP3 at supporting strength.
• Local BayesDel lookup returned bayesdel_score = 0.145561 (BayesDel_170824_noAF), but no ACMG-strength calibration exists for BayesDel and the POLE framework specifies no BayesDel cutoff, so this score was treated as not_available for PP3.
• Governing VCEP custom PP3 rule (Leon-Castillo et al. 2020, PATH-250-323): PP3_Supporting requires the exact missense variant to appear in Supplementary Table S2 or S3 with REVEL class 'likely disease causing' and <=1 benign in-silico results. The rule itself directs the generic fallback when the variant is absent from those tables.
• Searched PATH-250-323-s002.xlsx.txt (Table S1), PATH-250-323-s003.xlsx.txt (Table S2), PATH-250-323-s004.xlsx.txt (Table S3) and PATH-250-323.pdf.txt for 'c.2792T>C', 'p.Phe931Ser', 'p.F931S', 'F931' and '2792': no occurrence in any file. Table S1 lists 100+ recurrent POLE EC variants (e.g. p.P286R, p.V411L, p.A456P, p.S297F) and Tables S2/S3 list the in-silico results for TCGA and non-TCGA variants; p.F931S is in none of them.
• Contextual weight limit from the governing paper: Leon-Castillo et al. used six in-silico tools for exonuclease-domain mutations and report that mutations outside the exonuclease domain are uncommonly associated with the ultramutated genomic phenotype; in-silico prediction was validated only for exonuclease-domain mutants. p.Phe931Ser (codon 931) lies outside the exonuclease domain.
• Generic ACMG/AMP 2015 definitions of PP3 (multiple lines of computational evidence support a deleterious effect) per Richards et al., PMID:25741868.

Still missing

• The variant is absent from all three VCEP supplementary tables (PATH-250-323-s002.xlsx Table S1, PATH-250-323-s003.xlsx Table S2, PATH-250-323-s004.xlsx Table S3) and from the PATH-250-323.pdf text; no VCEP-assigned PP3 code exists for it, so the framework's generic in-silico fallback was used instead. Searched each file for c.2792T>C, p.Phe931Ser, p.F931S, F931 and 2792 with zero matches.

Sources

Reasoning

PS1 requires that the same amino-acid substitution - p.Phe931Ser (p.F931S), whether reached by c.2792T>C or any other nucleotide change - already be established as pathogenic. No source in this case package establishes p.Phe931Ser as pathogenic or likely pathogenic: the only ClinVar record at POLE codon 931 is variation 240446 for this very variant, classified Uncertain significance (3 clinical laboratories) with one discordant Likely benign submission and no expert-panel review; the governing VCEP package's Supplementary Table S1 (recurrent POLE variants in COSMIC/TCGA endometrial carcinoma) contains no residue-931 entry, and Supplementary Tables S2/S3 (in silico results) likewise list no 931 variant; the paper text never names residue 931. Because no previously established pathogenic allele with the identical protein change exists, PS1 is not met at any strength.

Evidence considered

• ClinVar variation 240446 (the exact match for POLE c.2792T>C) is Uncertain significance with 5 submissions: 3 clinical laboratories reporting VUS and 1 reporting Likely benign; review status 'criteria provided, single submitter' with 0 expert-panel submissions.
• PATH-250-323-s002.xlsx Supplementary Table S1 was searched for 'c.2792T>C', 'p.Phe931Ser', 'p.F931S' and '931': the table lists 95 recurrent POLE variants from COSMIC/TCGA endometrial carcinoma (topped by p.P286R c.857C>G 181, p.V411L c.1231G>T 114, p.A456P c.1366G>C 20, p.S297F c.890C>T 18) and contains no residue-931 variant.
• PATH-250-323-s003.xlsx (Supplementary Table S2, 42 TCGA POLE variants with SIFT/PANTHER/SNAP2/PolyPhen2/PROVEAN/REVEL classes) and PATH-250-323-s004.xlsx (Supplementary Table S3, 39 non-TCGA EC variants) were searched and contain no F931 entry.
• PATH-250-323.pdf full text was searched for '931' and 'Phe931' with no hit; the paper's variant tables cover only the five established hotspot substitutions and the recurrent non-hotspot exonuclease-domain set.

Still missing

• No entry for c.2792T>C / p.Phe931Ser / p.F931S exists in any searched VCEP file (PATH-250-323.pdf, PATH-250-323-s002.xlsx, PATH-250-323-s003.xlsx, PATH-250-323-s004.xlsx) - this is 'no entry found', which is weaker than positive evidence that no such record exists anywhere.

Sources

Reasoning

The governing POLE framework is the local literature-derived León-Castillo et al. 2020 custom framework (customized criteria PM1, PS4, PP3, BP4), whose PS4 specification is a somatic-recurrence adaptation capped at Supporting. Its stated operator, quoted literally from the framework rule, is: the exact variant must be 'recurrent in both the COSMIC and TCGA endometrial carcinoma cohorts in Supplementary Table S1', have a 'combined EC count >=10', and belong to the established/pathogenic set (p.P286R, p.V411L, p.A456P, p.S297F). Explicit substitution: combined EC count of p.F931S >= 10? p.F931S has no row anywhere in Supplementary Table S1, so its combined EC count is 0; 0 >= 10? NO -> the custom PS4_Supporting rule does not fire. I searched the full 100-row Supplementary Table S1 (PATH-250-323-s002.xlsx.txt) and the main paper text (PATH-250-323.pdf.txt) for c.2792T>C, p.Phe931Ser and p.F931S and found no entry for this variant; the paper's pathogenic set is limited to P286R, V411L, S297F, A456P, S459F, and p.F931S appears in none of the supplementary in-silico or recurrence tables either. Generic ACMG PS4 (case-control enrichment / increased prevalence in affected individuals versus controls) is likewise unsupported: no germline case-control data exist for this variant, and the two case-observation papers triaged to PS4 (PMID:25394175, PMID:28873162) never name POLE or this variant. No numeric enrichment value (odds ratio, prevalence ratio) is available for this variant, so no >= threshold comparison can be made. PS4 is therefore not met at any strength.

Rule

POLE custom PS4_Supporting (somatic-recurrence adaptation, Supporting cap): exact variant recurrent in both COSMIC and TCGA EC cohorts in Supplementary Table S1 AND combined EC count >=10 AND in the established/pathogenic set {p.P286R, p.V411L, p.A456P, p.S297F}.

Evidence considered

• Searched the full POLE Supplementary Table S1 (PATH-250-323-s002.xlsx.txt, 100 variant rows) for 'c.2792T>C', 'p.F931S', 'p.Phe931Ser' and '2792': no row for this variant. Its combined COSMIC+TCGA endometrial-carcinoma count is therefore 0.
• Searched the main León-Castillo et al. 2020 paper text (PATH-250-323.pdf.txt) for 'c.2792T>C', 'p.F931S', 'p.Phe931Ser', 'Phe931' and '931': no mention; the paper designates only P286R, V411L, S297F, A456P and S459F as pathogenic hotspot mutations.
• Framework PS4 rule (verbatim operator): recurrent in both COSMIC and TCGA in Supplementary Table S1, combined EC count >=10, and within the established/pathogenic set. Explicit inequality: 0 >= 10? no.
• PMID:28873162 (Mandelker et al., MSK-IMPACT tumor-normal cohort, n=1040) was triaged as a PS4 candidate but the extraction records that neither POLE nor c.2792T>C is named anywhere in the paper.
• PMID:25394175 (ACMG/NSGC cancer-predisposition referral guideline) was triaged as a PS4 candidate but the extraction records that it never addresses POLE or this variant.
• No germline case-control enrichment statistic for this variant was found in any consulted source, so generic PS4 has no numeric value to compare against a threshold.

Still missing

• No germline case-control enrichment data (odds ratio / affected-vs-control prevalence) exist for POLE c.2792T>C p.F931S; neither PMID:25394175 nor PMID:28873162 names POLE or this variant.

Sources

Reasoning

The governing POLE framework is the local custom framework derived from Leon-Castillo et al. 2020 (framework_mode custom_internal, customized criteria PM1/PS4/PP3/BP4; no ClinGen CSPEC was available). Its PM1 specification is an exact-variant enumeration at three strengths: PM1_Strong for the five exonuclease-domain hotspots p.P286R, p.V411L, p.S297F, p.A456P, p.S459F; PM1_Moderate for the recurrent non-hotspot exonuclease-domain variants p.F367S, p.L424I, p.M295R, p.P436R, p.M444K, p.D368Y; PM1_Supporting for the uncertain tier p.A465V, p.L424V, p.T278M, p.A428T. p.F931S appears in none of those lists. Independently, residue 931 lies far outside the POLE exonuclease (proofreading) domain, which spans amino acids 268-471; all 14 variants the framework enumerates fall inside that domain. The framework's own source paper states that only 1/23 (4%) of tumours with non-exonuclease-domain mutations showed the POLE-ultramutated genomic signature, 'indicating that a large majority of mutations outside the exonuclease domain are not pathogenic', and that mutations outside the exonuclease domain 'appear to be passengers'. CancerHotspots returns no statistically significant hotspot row for POLE F931, COSMIC returns no record, and Supplementary Table S1 shows no somatic recurrence for this variant. PM1 is therefore not met at any strength (hotspot, critical domain, or benign-variation-poor region).

Evidence considered

• Governing framework: final_classification_framework.json, source local_custom_framework, version 'leon-castillo-2020-custom-framework-v1', official_clingen=false, customized_criteria [PM1, PS4, PP3, BP4]; PM1 rules are exact-variant lists: Strong = p.P286R, p.V411L, p.S297F, p.A456P, p.S459F; Moderate = p.F367S, p.L424I, p.M295R, p.P436R, p.M444K, p.D368Y; Supporting = p.A465V, p.L424V, p.T278M, p.A428T.
• p.F931S is absent from all three PM1 tiers, so neither the hotspot rule, the recurrent non-hotspot rule, nor the uncertain-tier rule triggers.
• The framework is explicitly not an official ClinGen CSPEC/VCEP rule set (notes: 'This is not an official ClinGen CSPEC/VCEP rule set'); case_summary records cspec.found=false, so generic ACMG/AMP applies outside the framework's explicit PM1 scope.
• Domain position: the POLE exonuclease (proofreading) domain spans amino acids 268-471 as recorded in the POLE gene-level context artifact (pvs1_gene_context), whose supporting text states 'Both missense variants occurred in the exonuclease domain (amino acids 268-471)'. Residue 931 is outside this domain.
• PMID 41673532/León-Castillo gene-level text: 'Only 1/23 (4%) tumours with non-EDM showed these genomic alterations, indicating that a large majority of mutations outside the exonuclease domain are not pathogenic' (Abstract of PATH-250-323.pdf), and 'those with POLE mutations outside the exonuclease domain: these appear to be passengers secondary to the hypermutator phenotype' (Discussion, PATH-250-323.pdf).
• CancerHotspots (cancerhotspots.org, queried POLE F931) returned no hotspot row: 'This variant does not lie in a statistically significant hotspot.' COSMIC (query POLE c.2792T>C) returned no record. Supplementary Table S1 shows no COSMIC/TCGA endometrial-carcinoma recurrence for any residue-931 variant.
• ClinVar variation 240446: Uncertain significance (3 laboratories) plus Likely benign (1 laboratory), 0 expert-panel submissions - no PM1-relevant expert assertion.

Still missing

• vcep_materials.json's domain_tables key is an empty object ({}), so no authoritative VCEP-extracted domain table existed to check residue 931 against; the domain conclusion here rests on the framework's enumerations (all inside 268-471) plus the exonuclease-domain boundary recorded in the gene-level context artifact.
• PATH-250-323-s001.tif is declared in the index as governing PM1 and PS4 but could not be converted to text (flagged unconvertible in vcep_file_coverage.json); it was not visually inspected, so a figure-only PM1 statement cannot be excluded.
• No entry for c.2792T>C / p.Phe931Ser / p.F931S was found in any searched VCEP file (PATH-250-323.pdf, PATH-250-323-s002.xlsx, PATH-250-323-s003.xlsx, PATH-250-323-s004.xlsx); that is 'no entry found', not proof of absence.

Sources

Reasoning

PM5 requires a different missense change at the same residue (codon 931 of POLE) that is already established as pathogenic or likely pathogenic. Automated comparator harvesting found no candidate: the PM5 candidate collection records found=false with 0 candidates seen, 0 same-residue candidates, 0 expert-panel P/LP candidates and 0 non-expert majority P/LP candidates (reason: classic same-residue PM5 semantics could not be confirmed safely). Manual checks agree: the only ClinVar record at codon 931 is the queried variant itself (variation 240446, Uncertain significance with one Likely benign submitter), and Supplementary Tables S1/S2/S3 list no residue-931 POLE variant (F931 neighbours in the tables are F990C and T906I, different codons). No comparator exists at this residue, so PM5 is not met at any strength.

Evidence considered

• pm5_candidates.json: found=false, pm5_mode 'unknown', eligible_for_classic_pm5_search=false, candidate_summary total_candidates_seen=0, same_residue_candidates=0, expert_panel_plp_candidates=0, non_expert_majority_plp_candidates=0; reason 'Unable to confirm classic same-residue PM5 semantics safely; skipping automatic candidate harvesting.'
• ClinVar variation 240446 (the only codon-931 record retrieved) is Uncertain significance/Likely benign - not pathogenic - and is the queried variant itself, so it cannot serve as the PM5 comparator.
• PATH-250-323-s002.xlsx Supplementary Table S1 contains no p.F931 entry; the nearest listed residues are p.T906I, p.F990C and p.A957V, all different codons.
• PATH-250-323-s003.xlsx and PATH-250-323-s004.xlsx list in-silico results for TCGA and non-TCGA EC variants respectively (T906I, F990C present) with no 931 residue entry.

Still missing

• No entry for c.2792T>C / p.Phe931Ser / p.F931S was found in any searched VCEP file (PATH-250-323.pdf, PATH-250-323-s002.xlsx, PATH-250-323-s003.xlsx, PATH-250-323-s004.xlsx) and none of those files lists any other codon-931 variant; this is 'no entry found' rather than proof of absence.
• No PM5-specific rule exists in the governing local POLE framework (customized criteria are PM1, PS4, PP3, BP4), so PM5 was evaluated against generic ACMG/AMP 2015 semantics.

Sources

Reasoning

Per this group's binding rule, PP5 may only be applied for an exact-variant ClinVar expert-panel classification (Pathogenic or Likely pathogenic); ordinary laboratory submissions, aggregate ClinVar labels, and non-expert assertions must never trigger it. Queried ClinVar VCV000240446 for NM_006231.4:c.2792T>C (p.Phe931Ser): expert_panels is an empty object and the submission audit records expert_panel_submissions = 0, with is_expert_panel = false for all five submissions (GeneDx, Ambry Genetics, Sema4, Sinai Health System, Labcorp Genetics). The aggregate record is 'Conflicting classifications of pathogenicity' / 'criteria provided, conflicting classifications' at 1 star, comprising 3 clinical laboratories reporting Uncertain significance and 1 clinical laboratory reporting Likely benign. There is no ClinGen VCEP or other expert-panel assertion for this exact variant, and the conflicting single-submitter VUS calls are exactly the class of ordinary submissions the rule excludes. No reputable-source pathogenic assertion exists, so PP5 is not met.

Rule

PP5 applied only for an exact-variant ClinVar expert-panel classification of Pathogenic / Likely pathogenic.

Evidence considered

• ClinVar VCV000240446 (variation ID 240446) for NM_006231.4:c.2792T>C (p.Phe931Ser): expert_panels empty, expert_panel_submissions = 0, is_expert_panel = false for all 5 submissions.
• Aggregate ClinVar classification 'Conflicting classifications of pathogenicity', review status 'criteria provided, conflicting classifications', 1 star; 3 laboratories Uncertain significance and 1 laboratory Likely benign.
• No exact-variant expert-panel (ClinGen VCEP) classification exists, so the PP5 trigger condition is absent; ordinary single-submitter assertions are explicitly excluded by the governing rule.

Sources

Reasoning

No ClinGen CSPEC/VCEP specification exists for POLE in this case; the governing document is the local Leon-Castillo et al. 2020 custom POLE framework (leon-castillo-2020-custom-framework-v1, precedence rank 2, not an official ClinGen framework), which customizes only PM1, PS4, PP3 and BP4 and defines no population-frequency rule for any strength. The generic ACMG/AMP stand-alone benign threshold of allele frequency >=0.05 (Richards et al. 2015, PMID:25741868) therefore applies, evaluated against the highest frequency observed in any ancestry group. NM_006231.4:c.2792T>C (NP_006222.2:p.Phe931Ser) is present in gnomAD v2.1 at an all-comers AF of 3.88871e-05 (11/282,870 alleles, 0 homozygotes), with a highest subpopulation AF of 9.6432e-04 in Ashkenazi Jewish individuals (10/10,370 alleles, 0 homozygotes), and in gnomAD v4.1 at an all-comers AF of 1.67279e-05 (27/1,614,072 alleles, 0 homozygotes), with a highest subpopulation AF of 7.76817e-04 in Ashkenazi Jewish individuals (23/29,608 alleles, 0 homozygotes; reported grpmax FAF 7.9e-07). The descriptive non-cancer subsets do not raise the frequency materially (gnomAD v2.1 non-cancer exomes 4.22037e-05, 10/236,946; gnomAD v3.1 non-cancer genomes 2.02796e-05, 3/147,932). The single highest value in any dataset, 1.013993e-03 (gnomAD v2.1 non-cancer exomes, Ashkenazi Jewish, 10/9,862 alleles), is still roughly 49-fold below 0.05, so BA1 is not met at any strength.

Rule

Generic ACMG/AMP BA1: stand-alone benign when allele frequency is >=0.05 in a population; no POLE-specific population-frequency specification was available, and the local custom POLE framework does not customize BA1.

Evidence considered

• No ClinGen CSPEC/VCEP for POLE was available in this case; the governing specification is the local Leon-Castillo et al. 2020 custom POLE framework (version leon-castillo-2020-custom-framework-v1, custom_internal, official_clingen false, precedence rank 2), whose customized_criteria list is limited to PM1, PS4, PP3 and BP4 and which explicitly directs generic ACMG/AMP fallback outside those criteria. No POLE-specific BA1/BS1/BS2/PM2 frequency rule is therefore available, and the generic ACMG/AMP fallback governs.
• The generic ACMG/AMP stand-alone benign threshold of allele frequency >=0.05 (Richards et al. 2015, PMID:25741868) was applied against the highest frequency observed in any ancestry group, which is the correct comparator for a benign frequency criterion.
• gnomAD v2.1 all-comers: total AF 3.8887121292466504e-05 (11/282,870 alleles), 0 homozygotes; exome AF 3.9763961126751603e-05 (10/251,484), genome AF 3.1861339450710506e-05 (1/31,386). Highest subpopulation: Ashkenazi Jewish AF 9.643201542912247e-04 (10/10,370 alleles, 0 homozygotes); all other ancestry groups observed 0 alleles at this position.
• gnomAD v4.1 all-comers: total AF 1.6727878310261253e-05 (27/1,614,072 alleles), 0 homozygotes; exome AF 1.6417193726990276e-05 (24/1,461,882), genome AF 1.9712201852946975e-05 (3/152,190). Highest subpopulation: Ashkenazi Jewish AF 7.768170764658201e-04 (23/29,608 alleles, 0 homozygotes); reported grpmax FAF 7.9e-07.
• The point is not dataset-dependent: the gnomAD v2.1 non-cancer exome subset reports AF 4.220370886193479e-05 (10/236,946 alleles, 0 homozygotes; Ashkenazi Jewish 1.013993104846887e-03, 10/9,862) and the gnomAD v3.1 non-cancer genome subset reports AF 2.027958791877349e-05 (3/147,932 alleles, 0 homozygotes; Ashkenazi Jewish 6.056935190793458e-04, 2/3,302; grpmax FAF 0). These subsets are described for completeness only: because no applicable specification requires a non-cancer or exome-only source, the default all-comers gnomAD v2.1 and v4.1 entries govern.
• gnomAD-Canada v1.0 returned no result for this variant (reported as not available).
• Maximum value observed anywhere across all sources is 1.013993104846887e-03, approximately 49-fold below the 0.05 stand-alone threshold, so BA1 is not reached on any reading of the data.

Sources

Reasoning

No POLE-specific ClinGen CSPEC/VCEP exists; the governing local Leon-Castillo et al. 2020 custom POLE framework customizes only PM1, PS4, PP3 and BP4, so BS1 falls to the generic ACMG/AMP fallback. The calibration block supplies a single generic BS1 (strong) threshold of allele frequency >=0.01, described as the generic SVI default with per-gene derivation preferred (Whiffin et al. 2017, PMID:28518168); no POLE-specific BS1 threshold has been derived, and no separate generic BS1-supporting tier is published. The highest frequency observed for NM_006231.4:c.2792T>C in any ancestry group in any source is 1.013993e-03 (gnomAD v2.1 non-cancer exomes, Ashkenazi Jewish, 10/9,862 alleles); in the default all-comers sources the highest subpopulation frequency is 9.6432e-04 (gnomAD v2.1, Ashkenazi Jewish, 10/10,370) and 7.76817e-04 (gnomAD v4.1, Ashkenazi Jewish, 23/29,608), against all-comers total allele frequencies of 3.88871e-05 (v2.1) and 1.67279e-05 (v4.1). Every observed value is at least ~9.9-fold below 0.01, so BS1 is not met at any strength.

Rule

Generic ACMG/AMP BS1 (strong): allele frequency >=0.01; no POLE-specific BS1 specification was available, and no generic BS1-supporting tier is defined in the supplied calibration data.

Evidence considered

• No ClinGen CSPEC/VCEP for POLE was available; the governing local Leon-Castillo et al. 2020 custom POLE framework customizes only PM1, PS4, PP3 and BP4 and provides no BS1 (or any benign frequency) specification, so the generic ACMG/AMP threshold was applied.
• Generic BS1 threshold used: allele frequency >=0.01, the generic SVI default for which per-gene derivation is preferred (Whiffin et al. 2017, PMID:28518168). No POLE-specific BS1 threshold was available in the governing framework, and the supplied calibration block defines no BS1-supporting tier, so only the 0.01 strong threshold was tested.
• gnomAD v2.1 all-comers: AF 3.8887121292466504e-05 (11/282,870 alleles); highest subpopulation Ashkenazi Jewish AF 9.643201542912247e-04 (10/10,370 alleles).
• gnomAD v4.1 all-comers: AF 1.6727878310261253e-05 (27/1,614,072 alleles); highest subpopulation Ashkenazi Jewish AF 7.768170764658201e-04 (23/29,608 alleles); reported grpmax FAF 7.9e-07.
• Descriptive non-cancer subsets, which are not required by any applicable specification here, give the same result: gnomAD v2.1 non-cancer exomes AF 4.220370886193479e-05 with Ashkenazi Jewish AF 1.013993104846887e-03 (10/9,862 alleles) as the single highest value in any source; gnomAD v3.1 non-cancer genomes AF 2.027958791877349e-05 with Ashkenazi Jewish AF 6.056935190793458e-04 (2/3,302 alleles).
• Even the maximum value observed anywhere (1.013993104846887e-03) is about 9.9-fold below the 0.01 threshold, and no ancestry group other than Ashkenazi Jewish carries this allele in any dataset, so BS1 cannot be met by any aggregation of the available frequency data.

Still missing

• No POLE-specific BS1 threshold derived from disease prevalence and penetrance is available (per-gene derivation as in Whiffin et al. 2017, PMID:28518168, would be preferred over the generic 0.01 default); the generic default was applied because the governing framework specifies no BS1 rule.
• No POLE disease-prevalence or penetrance estimate was present in the case evidence, so the gene-specific maximum credible allele frequency could not be calculated.

Sources

Reasoning

BS2 requires observation of the variant in a healthy adult individual, either homozygous for a recessive disorder or heterozygous for a dominant disorder in which full penetrance is expected at an early age. Neither premise is satisfied here. Zero homozygotes are reported in every dataset examined: gnomAD v2.1 all-comers (11/282,870 alleles, 0 homozygotes), gnomAD v4.1 all-comers (27/1,614,072 alleles, 0 homozygotes), gnomAD v2.1 non-cancer exomes (10/236,946 alleles, 0 homozygotes), and gnomAD v3.1 non-cancer genomes (3/147,932 alleles, 0 homozygotes), and the Ashkenazi Jewish subpopulations that carry the allele likewise report 0 homozygotes (v2.1 10/10,370; v4.1 23/29,608). For the dominant/heterozygous branch, the only observations available are population-database carrier counts, which carry no individual-level health status, phenotype, or age information and therefore cannot establish the healthy-adult observation the criterion requires; gnomAD recruitment excludes severe paediatric disease but does not document freedom from adult-onset cancer predisposition, and this allele is far too rare (all-comers AF 1.67e-05 to 3.89e-05) for the carrier counts to imply tolerance of a fully penetrant early-onset disorder. BS2 is therefore not met.

Rule

Generic ACMG/AMP BS2: observed in a healthy adult individual, homozygous for a recessive disorder or heterozygous for a dominant disorder with full penetrance expected at an early age; no POLE-specific BS2 rule was available.

Evidence considered

• No ClinGen CSPEC/VCEP for POLE was available; the governing local Leon-Castillo et al. 2020 custom POLE framework customizes only PM1, PS4, PP3 and BP4, so generic ACMG/AMP BS2 was applied.
• Homozygote counts are 0 in every source: gnomAD v2.1 all-comers total_hom 0 (11/282,870 alleles) and gnomAD v4.1 all-comers total_hom 0 (27/1,614,072 alleles), including 0 homozygotes in the Ashkenazi Jewish subpopulations (v2.1 10/10,370; v4.1 23/29,608).
• The gnomAD v2.1 non-cancer exome subset (total_hom 0, 10/236,946 alleles; Ashkenazi Jewish 10/9,862) and the gnomAD v3.1 non-cancer genome subset (total_hom 0, 3/147,932 alleles; Ashkenazi Jewish 2/3,302) likewise report no homozygotes.
• gnomAD and gnomAD-Canada records supply allele and homozygote counts only; they do not document individual health status, phenotype, age, or follow-up, so the heterozygous carriers reported cannot be treated as verified unaffected healthy adults for a disorder in which the BS2 premise is full penetrance at an early age.
• gnomAD-Canada v1.0 reported no result for this variant, so no additional unaffected-individual data were available from that source.

Still missing

• No POLE-specific BS2 specification (definition of a qualifying healthy individual or a homozygote point scale) was available; the generic ACMG/AMP premise was applied.
• Individual-level health status, phenotype and age for the 27 gnomAD v4.1 and 11 gnomAD v2.1 heterozygous carriers are not available, so a dominant-model BS2 assessment cannot be completed from population data alone.

Sources

Reasoning

BP1 applies only where truncating variants - not missense - are the established mechanism of disease for the gene, i.e. where a missense change is therefore expected to be tolerated. That premise does not hold for POLE. The governing POLE framework is built entirely on missense substitutions: its PM1 tiers enumerate 14 specific missense variants, and its pathogenic set is five exonuclease-domain missense hotspots (p.P286R, p.V411L, p.S297F, p.A456P, p.S459F) associated with the POLE-ultramutated phenotype. The gene-level mechanistic evidence in this package likewise identifies missense changes in the exonuclease (proofreading) domain (amino acids 268-471) as the functional/disease-relevant class. No source in this case package establishes truncating POLE variants as the primary disease mechanism. Because the criterion's premise fails, a missense change cannot be argued benign on BP1 grounds, and BP1 is not met.

Evidence considered

• PMID 25741868 (ACMG/AMP 2015) defines BP1 as a missense variant in a gene for which primarily truncating variants are known to cause disease; its scope is gene-level mechanism, not variant-level data.
• The governing POLE framework's pathogenic logic is missense-only: PM1_Strong covers five exonuclease-domain missense hotspots (p.P286R, p.V411L, p.S297F, p.A456P, p.S459F) and PM1_Moderate/Supporting cover ten further exonuclease-domain missense substitutions; no truncating-variant rule appears in the framework.
• PATH-250-323.pdf frames POLE pathogenicity around 'Pathogenic somatic missense mutations within the DNA polymerase epsilon (POLE) exonuclease domain' and the ultramutated signature (C>A >20%, T>G >4%, C>G <0.6%, indels <5%, TMB >100 mut/Mb), i.e. missense-driven, not truncation-driven.
• pvs1_gene_context records that the POLE proofreading/exonuclease domain spans amino acids 268-471 and that missense changes there are the functional class of interest; the queried variant is a missense at residue 931.
• ClinVar variation 240446 for this missense variant carries no assertion that truncating POLE variants cause the associated disease, so no BP1-style gene-mechanism justification is available.

Still missing

• No BP1-specific rule exists in the governing local POLE framework (customized criteria are PM1, PS4, PP3, BP4) and no ClinGen CSPEC was retrieved, so BP1 was evaluated against the generic ACMG/AMP 2015 definition.
• No entry for c.2792T>C / p.Phe931Ser / p.F931S was found in any searched VCEP file; the BP1 conclusion rests on gene-level mechanism evidence, not on a variant-specific record.

Sources

Reasoning

Scope: c.2792T>C is a coding missense variant, so BP4 is applicable. The governing POLE custom BP4 rule (Leon-Castillo et al. 2020, PATH-250-323) requires the exact missense variant to appear in Supplementary Table S2 or S3 with REVEL class 'Likely benign'/'Likely-benign' and >=4 benign in-silico results, and explicitly directs the generic in-silico fallback when the variant is absent from those tables. p.Phe931Ser is absent from Table S2 (PATH-250-323-s003.xlsx.txt), Table S3 (PATH-250-323-s004.xlsx.txt) and Table S1 (PATH-250-323-s002.xlsx.txt), so no VCEP-assigned BP4 exists for it. Under the generic missense path, BP4 is not supported: REVEL = 0.749 is far above every published BP4 threshold (supporting <=0.29, moderate <=0.183, strong <=0.016; Pejaver et al. 2022, PMID:36413997) and in fact lands in the PP3-supporting band, so the calibrated missense computational evidence points toward, not away from, a functional effect. BayesDel = 0.145561 provides no usable benign argument because no ACMG-strength BayesDel calibration exists and the VCEP specifies none. The SpliceAI score (max delta 0.008) does not rescue BP4 here: although 0.008 is below the <=0.1 SpliceAI BP4-supporting cutoff, the SpliceAI path is calibrated for splice-region/intronic variants, whereas this variant is a coding missense change 53 bp from the nearest donor and 81 bp from the nearest acceptor, is not in a splice region, and is absent from the VCEP tables; a low SpliceAI delta in a non-splice-region missense variant is the expected null result rather than independent benign functional evidence, and using it alongside a pathogenic-leaning REVEL would double-test the same variant in opposite directions. BP4 is therefore not met.

Rule

POLE custom BP4 rule not triggered (variant absent from PATH-250-323 Tables S2/S3); generic fallback: REVEL 0.749 > 0.29 (BP4_supporting cutoff) => BP4 not met.

Evidence considered

• BP4 scope check: missense variant, so BP4 is applicable; the variant is not protein-length-changing.
• Local REVEL v1.3 score = 0.749 compared against the published ClinGen SVI REVEL BP4 calibration applied verbatim: strong <= 0.016, moderate <= 0.183, supporting <= 0.29 (Pejaver et al. 2022, PMID:36413997). 0.749 exceeds all three, so no BP4 is supported by the missense computational evidence.
• Local BayesDel score = 0.145561: no ACMG-strength BayesDel calibration is established and the governing VCEP specifies none, so the value is not_available for BP4.
• SpliceAI max delta = 0.008 (DS_AG 0.007, DS_AL 0.0, DS_DG 0.008, DS_DL 0.002; DP_AG 81, DP_AL -5, DP_DG -53, DP_DL -4). The value is below the <=0.1 SpliceAI BP4-supporting cutoff (Jaganathan et al. 2019, Cell, PMID:30661751) but the SpliceAI path applies to splice-region/intronic variants; this variant is not in a splice region, so the score is the expected null result for a coding missense change and was not counted as independent benign evidence against a pathogenic-leaning REVEL score.
• Governing VCEP custom BP4 rule (Leon-Castillo et al. 2020, PATH-250-323): BP4_Supporting requires the exact missense variant in Supplementary Table S2 or S3 with REVEL class 'Likely benign'/'Likely-benign' and >=4 benign in-silico results, with an explicit generic fallback when the variant is absent from those tables.
• Searched PATH-250-323-s003.xlsx.txt and PATH-250-323-s004.xlsx.txt for 'c.2792T>C', 'p.Phe931Ser', 'p.F931S', 'F931' and '2792': no occurrence. The surrounding entries (e.g. F367S, F990C, S461L, T906I in Table S2) sit at codons well away from 931, and codon 931 is represented by no row in either table.
• Generic ACMG/AMP 2015 definition of BP4 (multiple lines of computational evidence suggest no impact on gene or gene product) per Richards et al., PMID:25741868.

Still missing

• No VCEP-assigned BP4 exists: the variant is absent from PATH-250-323-s003.xlsx (Table S2) and PATH-250-323-s004.xlsx (Table S3), the two tables the governing custom BP4 rule draws on, so the generic in-silico fallback was used. Searched all VCEP files for c.2792T>C, p.Phe931Ser, p.F931S, F931 and 2792 with zero matches.
• No established ACMG-strength calibration exists for BayesDel, and the POLE framework supplies no gene-specific BayesDel cutoff, so the available BayesDel score (0.145561) could not be converted into BP4 evidence in either direction.

Sources

Reasoning

BP6 (reputable source reports the variant as benign) is, per this group's binding rule, applicable only for an exact-variant ClinVar expert-panel classification of Benign or Likely benign; ordinary laboratory submissions and aggregate labels must never trigger it. ClinVar VCV000240446 for NM_006231.4:c.2792T>C (p.Phe931Ser) has expert_panels empty and expert_panel_submissions = 0, with is_expert_panel = false for all five submissions. The two Likely benign calls (Sema4, curation, excluded as non-clinical/unclear method; Labcorp Genetics/Invitae) are single-submitter laboratory assertions, not expert-panel classifications, and the aggregate record is 'Conflicting classifications of pathogenicity' (1 star; 3 laboratories Uncertain significance, 1 laboratory Likely benign). No exact-variant expert-panel benign classification exists, so the BP6 trigger condition is absent and BP6 is not met.

Rule

BP6 applied only for an exact-variant ClinVar expert-panel classification of Benign / Likely benign.

Evidence considered

• ClinVar VCV000240446 (variation ID 240446): expert_panels empty, expert_panel_submissions = 0, is_expert_panel = false for all 5 submissions.
• Likely benign single-submitter calls (Sema4 curation; Labcorp Genetics/Invitae clinical testing) are ordinary laboratory assertions and are excluded from triggering BP6 by the governing rule.
• Aggregate ClinVar status 'Conflicting classifications of pathogenicity', 1 star, with 3 laboratories Uncertain significance and 1 laboratory Likely benign; no exact-variant expert-panel benign classification exists.

Sources

Reasoning

PS2 requires a confirmed de novo occurrence in a patient with the disease phenotype, with biological parentage confirmed by parental testing. This case is a variant-level curation only: there is no proband, no disease phenotype, no parental testing result (neither locus-specific nor identity/kinship testing), and no family of any kind reported in the case evidence bundle, in prefetch.json, or in the POLE VCEP materials. No ClinGen CSPEC exists for POLE (cspec found = false), and the local Leon-Castillo et al. 2020 POLE framework customizes only PM1, PS4, PP3 and BP4, explicitly directing generic ACMG/AMP fallback for all other criteria; under generic ACMG/AMP 2015 (Richards et al.) PS2 cannot be triggered without an observed de novo event. The criterion is therefore not assessable rather than not met: there is simply no proband/parental data to evaluate. The absent de novo observation and parental testing are the gap.

Evidence considered

• Governing framework check: no ClinGen CSPEC/VCEP specification exists for POLE (case_summary cspec found = false); the local POLE literature package (Leon-Castillo et al. 2020, PATH-250-323) states it is not an official rule set, customizes only PM1, PS4, PP3 and BP4, and instructs fallback to generic ACMG/AMP outside those rules. Generic ACMG/AMP 2015 (Richards et al., PMID:25741868) therefore governs PS2.
• Generic ACMG/AMP 2015 defines PS2 as a confirmed de novo occurrence (with parental testing confirming biological parentage) in a patient with the disease phenotype; no such observation exists anywhere in the material reviewed for this case.
• The case contains variant-level data only (normalization, population frequency, in-silico, ClinVar/COSMIC/OncoKB lookups). No proband, phenotype, parental genotypes, pedigree, or family history field is present in the case evidence bundle, prefetch.json, or vcep_materials.json.
• The three full texts screened for this group (PMID:25741868, PMID:25394175, PMID:28873162) contain no de novo observation, parental testing, or segregation data for POLE c.2792T>C p.(F931S); the variant and the POLE gene are not named in PMID:25394175 or PMID:28873162.
• ClinVar record 240446 holds only classification submissions (Uncertain significance x3 laboratories, Likely benign x1 laboratory) with no de novo or parental-testing assertion attached.

Still missing

• No proband is identified and no clinical phenotype is recorded for this case; PS2 requires an affected individual in whom the variant arose de novo.
• No parental testing (neither targeted locus testing in both biological parents nor identity/paternity-maternity confirmation) is reported, so a de novo event cannot be confirmed.
• No segregation or family data were available in the case bundle, prefetch.json, or the POLE VCEP materials to substitute for direct parental testing.

Sources

Reasoning

Not assessed. The governing specification for this case is the local custom POLE framework derived from Leon-Castillo et al. 2020 (J Pathol 250:323-335), which customizes only PM1, PS4, PP3 and BP4 and contains no PS3 or BS3 rule; PS3 therefore falls back to generic ACMG/AMP 2015 (Richards et al., PMID:25741868), which requires well-established functional studies demonstrating a damaging effect on protein function, with documented assay validation, reproducibility and biological relevance. No such study exists for POLE c.2792T>C (p.Phe931Ser). OncoKB lists POLE F931S as 'Unknown Oncogenic Effect' and explicitly states that no variant-specific reviewed functional evidence was identified. The variant has no COSMIC somatic record, is not a statistically significant Cancer Hotspots residue, and is absent from the Leon-Castillo POLE hotspot set (p.P286R, p.V411L, p.S297F, p.A456P, p.S459F), from its recurrent pathogenic and POLE-score tiers, and from Supplementary Tables S1-S3, which were read directly and do contain polymerase-region variants (e.g. p.A1010T, p.A1140T, p.S1644L, p.V2025M). The framework's closest functional-type readout - the POLE-ultramutated base-substitution signature and tumour mutational burden - covers only the five exonuclease-domain hotspots plus six additional exonuclease-domain variants; p.F931 lies well outside that exonuclease-domain span (framework variants range from p.T278M to p.A465V) in the polymerase region, and the paper itself reports that only 1/23 (4%) of non-exonuclease-domain-mutant tumours showed the ultramutated signature. That argues against extrapolating hotspot functional evidence to p.F931S, but it is not an assay of p.F931S itself. Absence of functional data is not evidence of a damaging effect, so PS3 is not assessed rather than not met.

Rule

Generic ACMG/AMP 2015 (Richards et al., PMID:25741868): PS3 = well-established functional studies show a damaging effect on protein function. The governing local POLE framework (Leon-Castillo et al. 2020) specifies no PS3 rule.

Evidence considered

• Governing framework: this case runs in custom_gene_framework mode on the local POLE framework derived from Leon-Castillo et al. 2020 (J Pathol 250:323-335); the framework payload lists customized_criteria = PM1, PS4, PP3, BP4 only, so PS3/BS3 have no gene-specific specification and are evaluated under generic ACMG/AMP 2015 (Richards et al., PMID:25741868).
• ACMG/AMP 2015 defines PS3 as well-established functional studies showing a damaging effect, and its 'PS3 BS3 Functional studies' section requires that assay validity be justified by how closely the assay reflects the biological environment, whether it measures full biological function or only one component, and by validation, reproducibility and robustness data; none of these conditions can be met for p.F931S because no assay exists.
• OncoKB record for POLE F931S: classification 'Unknown Oncogenic Effect', is_variant_specific = false, not_reviewed = true, citations_pmids = [] - i.e. OncoKB identified no variant-specific reviewed functional evidence and supplied only gene-level POLE ultramutation context.
• COSMIC query for POLE c.2792T>C returned no record ('This variant has not previously been reported in somatic cancers'), and Cancer Hotspots reported no statistically significant hotspot row for POLE F931; neither source supplies functional impact data, but both show the variant lacks the somatic-recurrence context that carries functional weight in this gene.
• Leon-Castillo et al. 2020 full text (PATH-250-323.pdf) was read directly: the strings 'F931' and 'F931S' occur 0 times, the confirmed-pathogenic set is limited to five exonuclease-domain hotspots, and the paper states that only 1/23 (4%) tumours with non-exonuclease-domain mutations showed the ultramutated signature.
• Supplementary tables PATH-250-323-s002/s003/s004.xlsx were opened directly: Supplementary Table S1 has 101 recurrent variants with COSMIC/TCGA EC counts, Table S2 has 44 TCGA in-silico rows and Table S3 has 38 non-hotspot in-silico rows, several from the polymerase region (e.g. p.A1010T, p.A1140T, p.S1644L at c.4931C>T, p.V2025M, p.A957V); p.F931S is absent from all three tables, and none of the tables is a functional assay.
• Literature screened for functional evidence (PMID:25741868, PMID:25394175, PMID:28873162, PMID:28492532) contains no POLE functional assay and no mention of c.2792T>C or p.F931S at any of the retrieveable full-text or classifier levels.
• No numeric threshold applies to this criterion: PS3 is a qualitative evidence-existence criterion in ACMG/AMP 2015 and no gene-specific PS3 calibration exists for POLE in the provided materials.

Still missing

• No validated functional assay of POLE p.(Phe931Ser) - proofreading activity, polymerase extension fidelity, yeast-based mutator assay, or equivalent - was retrieved or reported by any source in this case.
• No patient-derived tumour readout (POLE-ultramutated base-substitution signature, tumour mutational burden) is available for a p.F931S carrier in the retrieved evidence, and such a readout would require formal validation before it could support PS3.
• The local Leon-Castillo POLE framework specifies no PS3/BS3 rule and no official ClinGen POLE CSPEC/VCEP specification was available; a published gene-specific PS3 specification would need to be applied if it becomes available.

Sources

Reasoning

PM3 requires that this variant be detected in trans with a pathogenic variant, which is only meaningful in a recessive disease context. The governing POLE package in this case is a local custom gene framework (no ClinGen CSPEC exists for POLE; cspec.found = false), and that package specifies rules only for PM1, PS4, PP3 and BP4 - it contains no PM3 rule and no POLE-specific statement of inheritance mode. Generic ACMG/AMP 2015 therefore governs, and PM3 remains an option because POLE is associated with both dominant phenotypes (colorectal cancer, susceptibility to, 12 / hereditary cancer-predisposing syndrome) and recessive phenotypes (facial dysmorphism-immunodeficiency-livedo-short stature / FILS and IMAGE-I syndrome), all of which appear among the ClinVar condition labels for this record. However, no proband-level genotype data are available for this case: the bundle contains no second POLE variant, no zygosity call, no parental or phase study, and no family pedigree or phenotype, and none of the five ClinVar submissions (all germline origin, three Uncertain significance and one Likely benign) reports a second pathogenic allele in trans or provides phase information. gnomAD v2.1 and v4.1 both list zero homozygotes (11/282870 and 27/1614072 alleles respectively), which is consistent with a recessive-trans hypothesis being untested rather than with trans configuration. Because the requisite in-trans observation with a pathogenic variant could not be made and the relevant inheritance mode for this proband is undetermined, this criterion is not assessed rather than met or refuted.

Evidence considered

• ClinVar record 240446 / VCV000240446 for NM_006231.4(POLE):c.2792T>C (p.Phe931Ser) is graded Uncertain significance by 3 clinical laboratories and Likely benign by 1 (5 submissions total, all origin = germline, 0 expert-panel submissions). No submission reports the variant in trans with a second POLE pathogenic variant, and the audit's evidence_signals for the submitted criteria captured only in-silico (PP3) statements - no PM3-type allelic/phase statement was present in any submission.
• ClinVar condition labels attached to this record span both inheritance modes: 'Colorectal cancer, susceptibility to, 12' and 'Hereditary cancer-predisposing syndrome' (dominant) as well as 'Facial dysmorphism-immunodeficiency-livedo-short stature syndrome' (FILS) and 'Intrauterine growth retardation, metaphyseal dysplasia, adrenal hypoplasia congenita, genital anomalies, and immunodeficiency' (IMAGE-I), which are biallelic/recessive POLE phenotypes. This makes the recessive-disorder precondition of PM3 relevant in principle but not resolved for this proband.
• The governing POLE custom package (PATH-250-323, Leon-Castillo et al. 2020, and its Supplementary Tables S1-S4) supplies contextual rules for PM1, PS4, PP3 and BP4 only, and explicitly states it is not an official ClinGen CSPEC/VCEP ACMG rule specification and should not be used to derive a germline ACMG call on its own; it contains no PM3 or BP2 rule and no statement of POLE inheritance mode.
• No CSPEC/variant-curation expert-panel specification was retrieved for POLE (cspec.found = false), so generic ACMG/AMP 2015 criteria (Richards et al.) provide the governing definitions for PM3 and BP2 in this case.
• Reproductive/population genotype context was checked in gnomAD v2.1 (11/282870 alleles, AF 3.88871e-05, homozygotes = 0; Ashkenazi Jewish 10/10370, AF 0.00096432) and gnomAD v4.1 (27/1614072 alleles, AF 1.67279e-05, homozygotes = 0; Ashkenazi Jewish 23/29608, AF 0.000776817): no homozygous or biallelic observations exist in either dataset, which neither supports nor refutes trans configuration in this proband.
• No case-level clinical file (pedigree, phenotype, zygosity, or genotype) is present in this case directory; the available case evidence is variant-level only.

Still missing

• No second POLE variant has been identified or tested in this proband, so a trans-partner pathogenic variant could not be evaluated.
• No phase information (parental genotyping, pedigree, or long-read/phasing data) is available, so cis/trans configuration cannot be assigned.
• The proband's clinical phenotype is not recorded, and POLE is associated with both dominant (colorectal cancer predisposition 12, hereditary cancer-predisposing syndrome) and recessive (FILS, IMAGE-I) conditions listed in ClinVar, so the applicable inheritance mode is undetermined.

Sources

Reasoning

PM6 requires an assumed de novo occurrence without confirmation of biological parentage, used when parental testing is unavailable but a proband is affected and the variant is absent from the parents by history or by testing of only one parent. No proband, no phenotype and no parental information of any kind are reported for this variant, so neither a confirmed nor an assumed de novo event can be recorded. No ClinGen CSPEC exists for POLE and the local Leon-Castillo et al. 2020 POLE framework does not specify PM6, so generic ACMG/AMP 2015 governs; under that framework there is no proband observation on which to base even an assumed de novo event. The criterion is not assessable, and the missing proband/parental data is the gap.

Evidence considered

• Frameworks consulted in order: no ClinGen CSPEC for POLE (cspec found = false); local custom POLE framework (Leon-Castillo et al. 2020 package) customizes only PM1/PS4/PP3/BP4 and directs generic ACMG/AMP fallback; generic ACMG/AMP 2015 (PMID:25741868) therefore governs PM6.
• Generic ACMG/AMP 2015 defines PM6 as an assumed de novo occurrence (patient affected, parentage not confirmed); no affected proband or parental information is documented anywhere in the case material reviewed.
• Case material is variant-level only: no pedigree, no proband, no parental genotypes, and no family history field in the case evidence bundle, prefetch.json, or vcep_materials.json.
• Neither of the two case-observation papers retrieved via ClinVar (PMID:25394175, PMID:28873162) names POLE or the c.2792T>C p.(F931S) variant, so neither supplies a de novo case for this variant.

Still missing

• No proband with a defined phenotype is identified for this variant, which PM6 requires before a de novo event can even be assumed.
• No parental clinical history or partial parental testing is available to support an assumed de novo occurrence (parentage unconfirmed).
• No POLE gene-specific PM6 specification exists (no CSPEC; local Leon-Castillo 2020 framework does not address PM6), so no gene-specific down-weighting or modification can be applied.

Sources

Reasoning

PP1 requires co-segregation of the variant with disease across multiple affected family members, scored by the number of informative meioses (with strength scaled by the number of segregations). No pedigree, affected relatives, family history, or meioses are reported for this variant anywhere in the case evidence bundle, prefetch.json, or the POLE VCEP materials. No ClinGen CSPEC exists for POLE and the local Leon-Castillo et al. 2020 POLE framework specifies no PP1 rule, so generic ACMG/AMP 2015 governs; without any family observations there is no meiosis to count and no co-segregation to score. The criterion is not assessable rather than not met: the absence of any pedigree or family-genotype data is the gap.

Evidence considered

• Framework precedence: no ClinGen CSPEC/VCEP specification for POLE; the local Leon-Castillo et al. 2020 POLE framework specifies no PP1 rule and directs generic ACMG/AMP fallback, so generic ACMG/AMP 2015 (PMID:25741868) governs PP1.
• Generic ACMG/AMP 2015 defines PP1 as co-segregation with disease in multiple affected family members, with strength increased by the number of informative meioses; no family or meioses data exist for this variant.
• No family history, pedigree, relative genotypes, or meioses are recorded in the case evidence bundle, prefetch.json, or vcep_materials.json; the case is a variant-level curation.
• PMID:25394175 (ACMG/NSGC referral-indications guideline) discusses family-history-based referral but provides no variant-level or family-level data for POLE c.2792T>C and does not name POLE, so it cannot support PP1 for this variant.

Still missing

• No pedigree or family structure is reported, so the number of affected family members available for co-segregation analysis is unknown (zero meioses countable).
• No genotype data from affected or unaffected relatives are available.
• No family history documentation is present in the case material to assess whether the variant tracks with disease in the family.

Sources

Reasoning

PP2 requires evidence that missense variation is a common mechanism of disease in POLE and that the rate of benign missense variation in the gene is low. The governing POLE framework is a local custom framework that customizes only PM1, PS4, PP3 and BP4 and supplies no PP2 specification; no ClinGen CSPEC exists for POLE in this case (cspec.found=false). The generic ACMG/AMP 2015 standards define PP2 but provide no gene-level missense constraint metric for POLE, and this case bundle contains no gnomAD missense constraint value (missense Z or observed/expected ratio) and no benign-missense-rate analysis for POLE. The VCEP package is explicitly diagnostic-context material for somatic endometrial carcinoma ('not by itself a germline ACMG/AMP rule specification'), and its domain logic is one of domain-restricted missense pathogenicity rather than gene-wide missense enrichment, so it cannot substitute for the gene-level data PP2 requires. Without a POLE-specific PP2 rule or gene-level benign-missense-rate data, PP2 is not assessable and is returned as not_assessed rather than met or not met.

Evidence considered

• PMID 25741868 (ACMG/AMP 2015) defines PP2 as missense in a gene with a low rate of benign missense variation and in which missense variants are a common mechanism of disease; the extracted paper content confirms the paper names neither POLE nor this variant and supplies no gene-level missense constraint value.
• final_classification_framework.json (custom_internal, leon-castillo-2020-custom-framework-v1) lists customized_criteria as PM1, PS4, PP3, BP4 only, and states it is not an official ClinGen CSPEC/VCEP rule set; no PP2 rule is defined.
• case_summary.json records cspec.found=false and framework_mode custom_gene_framework, so no VCEP PP2 specification exists to override generic ACMG/AMP.
• PATH-250-323.pdf index guidance states the package is 'not by itself a germline ACMG/AMP rule specification', and the paper's mechanistic claim is domain-restricted (pathogenicity concentrated in exonuclease-domain substitutions), which is not equivalent to a gene-wide low benign-missense rate.
• No gnomAD missense-constraint figure for POLE appears anywhere in the case bundle (the gnomAD results contain allele frequencies only). Residue 931 is outside the exonuclease domain (268-471), where the paper states mutations 'appear to be passengers', so a domain-based PP2 argument would not favour this variant.

Still missing

• No POLE-specific PP2 specification exists in the governing framework (customized criteria are only PM1, PS4, PP3, BP4) and no ClinGen CSPEC was retrieved (cspec.found=false).
• No gene-level missense constraint metric for POLE (e.g., gnomAD missense Z-score or observed/expected ratio) and no benign-missense-rate data are present in the case bundle; such data would be needed to apply generic PP2.
• PATH-250-323-s001.tif could not be converted to text, so any figure-level PP2-relevant content is unavailable.

Sources

Reasoning

PP4 requires that the patient's phenotype or family history be highly specific for a disease with a single genetic etiology. No proband-level phenotype, HPO term list, family history, or clinical indication was captured anywhere in this case bundle (case_summary.json, evidence.json and prefetch.json contain variant-level data only). The only phenotype-like information is the set of ClinVar condition labels attached to VCV000240446, and those are broad, non-specific cancer-predisposition labels ('Colorectal cancer, susceptibility to, 12', 'Facial dysmorphism-immunodeficiency-livedo-short stature syndrome', 'Hereditary cancer-predisposing syndrome', 'Intrauterine growth retardation, metaphyseal dysplasia, adrenal hypoplasia congenita, genital anomalies, and immunodeficiency', 'not provided'); the record is in fact a conflicting-classification record and these labels reflect different submitters' differentials rather than a single highly specific proband phenotype. Without a documented phenotype or family history there is no specificity to weigh, so PP4 cannot be evaluated; this is a missing-data gap rather than a negative finding, and it is closable if the clinical phenotype/HPO terms are supplied.

Evidence considered

• Reviewed the case bundle and prefetch.json for proband phenotype / clinical indication / HPO terms: none present; the case carries variant-level annotations only.
• ClinVar VCV000240446 condition labels are broad and non-specific cancer-predisposition terms from conflicting single submitters, not a phenotype highly specific for a single genetic etiology.

Still missing

• No proband phenotype, HPO terms, family history, or clinical indication was captured in this case (case_summary.json / evidence.json / prefetch.json); PP4 phenotype specificity is therefore unassessable.
• PATH-250-323-s001.tif (supplementary figure) could not be converted to text and was not visually reviewed; it is a somatic-tumour figure and is not expected to carry proband phenotype data.

Sources

Reasoning

Not assessed. BS3 under generic ACMG/AMP 2015 (the governing local POLE framework from Leon-Castillo et al. 2020 specifies no PS3/BS3 rule) requires well-established functional studies showing no damaging effect on protein function or splicing. No functional assay of POLE p.(Phe931Ser) - wild-type or variant - was identified in any retrieved source. OncoKB classifies POLE F931S as 'Unknown Oncogenic Effect' with no variant-specific reviewed evidence, which is a record of absence rather than a benign functional result; the absence of a COSMIC somatic record and the absence of a significant Cancer Hotspots residue are similarly absence-of-evidence observations with no functional-assay weight. Computational predictor outputs are in-silico evidence handled by the in-silico criteria group, not functional studies, and cannot satisfy BS3. The Leon-Castillo POLE-ultramutated signature data cover only exonuclease-domain variants and never name p.F931S, so they provide no benign functional readout for this residue either. Because no functional data exist in either direction, BS3 is not assessed rather than not met.

Rule

Generic ACMG/AMP 2015 (Richards et al., PMID:25741868): BS3 = well-established functional studies show no damaging effect on protein function or splicing. The governing local POLE framework (Leon-Castillo et al. 2020) specifies no BS3 rule.

Evidence considered

• Governing framework: the local POLE framework (Leon-Castillo et al. 2020, J Pathol 250:323-335) customizes PM1, PS4, PP3 and BP4 only, so BS3 is evaluated under generic ACMG/AMP 2015 (Richards et al., PMID:25741868), which requires well-established functional studies showing no damaging effect on protein function or splicing.
• OncoKB reports POLE F931S as 'Unknown Oncogenic Effect' with is_variant_specific = false and no citations - this is an explicit statement that no functional evidence exists, not evidence of normal function, and therefore cannot support BS3.
• COSMIC returned no record for POLE c.2792T>C and Cancer Hotspots found no significant hotspot at POLE F931; both are absence-of-evidence results and neither constitutes a functional assay showing retained protein function.
• Leon-Castillo et al. 2020 (PATH-250-323.pdf) was read directly and never mentions p.F931S (0 occurrences of F931/F931S); its ultramutated-signature evidence is restricted to exonuclease-domain variants, so it provides no benign functional readout for the polymerase-region residue p.F931.
• Supplementary Tables S1-S3 (PATH-250-323-s002/s003/s004.xlsx) were opened directly and do not contain p.F931S; they are somatic recurrence and in-silico prediction tables, not functional assays, and so cannot support a benign functional claim.
• Literature screened for functional evidence (PMID:25741868, PMID:25394175, PMID:28873162, PMID:28492532) contains no functional assay and no mention of c.2792T>C or p.F931S.
• No numeric threshold applies: BS3 is a qualitative evidence-existence criterion in ACMG/AMP 2015 and no gene-specific BS3 calibration exists for POLE in the provided materials.

Still missing

• No functional assay of POLE p.(Phe931Ser) demonstrating retained or normal proofreading/polymerase activity exists in the retrieved evidence, so the benign functional criterion cannot be satisfied.
• Negative or absent findings (no COSMIC record, no significant cancer hotspot, OncoKB 'Unknown Oncogenic Effect') are absence-of-evidence only and must not be converted into BS3.

Sources

Reasoning

BS4 requires demonstrated lack of segregation of the variant with disease in affected family members (for example, an affected relative carrying the variant who is phenotypically discordant, or an affected relative without the variant). No pedigree, no affected relatives, and no relative genotypes are reported for this variant in the case evidence bundle, prefetch.json, or the POLE VCEP materials, so non-segregation cannot be demonstrated. No ClinGen CSPEC exists for POLE and the local Leon-Castillo et al. 2020 POLE framework specifies no BS4 rule, so generic ACMG/AMP 2015 governs. Absence of family data is not evidence of non-segregation; the criterion is not assessable and the missing family/pedigree data is the gap.

Evidence considered

• Framework precedence: no ClinGen CSPEC/VCEP specification for POLE (cspec found = false); the local Leon-Castillo et al. 2020 POLE framework customizes only PM1/PS4/PP3/BP4 and specifies no BS4 rule, directing generic ACMG/AMP fallback; generic ACMG/AMP 2015 (PMID:25741868) governs BS4.
• Generic ACMG/AMP 2015 defines BS4 as lack of segregation in affected members of a family; no family or relative genotype data exist for this variant, so non-segregation cannot be established.
• No family history, pedigree, or relative genotypes appear in the case evidence bundle, prefetch.json, or vcep_materials.json.
• The variant is present in gnomAD (11/282,870 alleles in v2.1; 27/1,614,072 in v4.1, homozygotes = 0) and has no COSMIC somatic record, but population frequency and case counts are not family segregation evidence and do not satisfy BS4.

Still missing

• No pedigree or affected relatives are reported, so non-segregation of the variant with disease cannot be demonstrated.
• No genotypes from family members (affected or unaffected) are available for a segregation analysis.
• The uninformative population/genotype context in this case is not a substitute for family-based evidence of non-segregation.

Sources

Reasoning

BP2 covers two observations: the variant in trans with a pathogenic variant for a fully penetrant dominant gene/disorder, or the variant in cis with a pathogenic variant under any inheritance pattern. Both require an established second (pathogenic) POLE allele with a determined phase relationship to the queried variant. The governing POLE custom package (PATH-250-323, Leon-Castillo et al. 2020) defines no BP2 rule and no inheritance-mode statement; because no ClinGen CSPEC exists for POLE, generic ACMG/AMP 2015 governs. In this case there is no second POLE variant, no phase or parental data, no pedigree, and no proband phenotype anywhere in the case evidence, and the ClinVar submissions contain no cis/trans statement for this record - the only criterion signals captured from submitters were in-silico (PP3). The dominant POLE phenotype (colorectal cancer susceptibility 12 / hereditary cancer-predisposing syndrome) is also not a fully penetrant dominant disorder in the sense BP2 requires, and the recessive FILS/IMAGE-I phenotypes would require biallelic data that do not exist here. With no cis or trans configuration available to evaluate, BP2 is not assessed rather than met or refuted.

Evidence considered

• No cis/trans configuration data exist for c.2792T>C in this case: the case directory contains no pedigree, genotype, zygosity, or phasing information, and no second POLE variant has been reported.
• ClinVar record 240446/VariationID 240446 (NM_006231.4(POLE):c.2792T>C, p.Phe931Ser) has 5 submissions (3 Uncertain significance, 1 Likely benign as reported in the summary; all origin = germline, no expert panel). The submission audit extracted only in-silico (PP3) evidence signals and no statement placing the variant in cis or in trans with a pathogenic POLE allele, so the ClinVar record provides no BP2-bearing observation.
• The governing POLE custom framework (PATH-250-323 / Leon-Castillo et al. 2020 and its supplementary tables) contains rules only for PM1, PS4, PP3 and BP4 and is explicitly not a germline ACMG/AMP rule specification; it defines no BP2 rule and no statement that POLE-related disease is a fully penetrant dominant disorder.
• No ClinGen CSPEC/VCEP specification for POLE was available (cspec.found = false), so the generic ACMG/AMP 2015 definition of BP2 governs; the criteria leads flagged for review (SCV006053617 citing PMID:28873162/25741868 and SCV000289310 citing PMID:28492532) were checked and neither paper reports variant-level cis/trans or POLE genotype data.
• The papers retrieved for this case (Mandelker et al. 2017PMID:28873162 cohort; ACMG/NSGC referral guideline PMID:25394175; Richards et al. 2015 PMID:25741868) contain no POLE-specific or c.2792T>C-specific allelic, phase, or family data; the variant and gene are not named in any of them.

Still missing

• No pathogenic POLE variant co-occurring with c.2792T>C has been identified in this proband, so neither the cis nor the trans clause of BP2 can be tested.
• No phase information (parental genotypes, pedigree, or molecular phasing) is available to assign cis versus trans configuration.
• No proband phenotype or family history is recorded, so it is unknown whether the relevant POLE condition is a fully penetrant dominant disorder (the stated precondition of BP2's trans clause) or one of the recessive POLE phenotypes.
• The germline penetrance of POLE cancer predisposition is not addressed by the sources available for this case, so the 'fully penetrant dominant' precondition itself remains unevidenced.

Sources

Reasoning

BP5 requires that the variant be found in a case with an alternate molecular basis for disease (i.e. an affected individual carrying this variant who also has a separate, sufficient genetic explanation). No case-level genotype or phenotype data exist in this bundle: there is no proband record, no co-occurring pathogenic variant in another gene, and no documented alternate molecular diagnosis. The ClinVar submissions for VCV000240446 are germline single-submitter records with no condition labels or co-occurring-variant information attached (condition_labels empty in the submission audit). The governing POLE custom framework does not customize BP5 (customized criteria are limited to PM1, PS4, PP3, BP4), so no gene-specific BP5 operator or threshold exists; the generic ACMG BP5 definition applies, and no evidence for it was collected. This is a missing-data gap rather than a negative finding and is closable if case-level genotype/co-occurring-variant data are supplied.

Evidence considered

• Reviewed case bundle, prefetch.json and clinvar_submission_audit.json for any case with an alternate molecular basis for disease: none present; ClinVar submissions carry no condition labels indicating a co-occurring alternate cause.
• The governing POLE custom framework customizes only PM1, PS4, PP3 and BP4; BP5 is not gene-specified and generic ACMG BP5 requires case-level alternate-basis evidence that is absent.

Still missing

• No case-level clinical/genotype data (proband genotype, co-occurring pathogenic variant in another gene, or alternate molecular diagnosis) were captured in this case, so BP5 is unassessable.
• No POLE-specific BP5 rule exists in the governing framework (customized criteria are PM1, PS4, PP3, BP4 only) and no VCEP text specifies a numeric BP5 likelihood-ratio operator, so no threshold applies.

Sources