POLE c.2173+23G>T, p.?

Reasoning

NM_006231.4:c.2173+23G>T is an intronic substitution (p.?), so BP4 is evaluated on the SpliceAI path only; no missense path (REVEL) exists for this variant. SpliceAI Lookup returned a genuine result for this variant (spliceai_available=true) with DS_AG 0.00, DS_AL 0.003, DS_DG 0.00, DS_DL 0.001, max delta 0.003, and the nearest predicted cryptic-site changes are located 382/94/302/-116 bp from the variant, i.e. no plausible effect on the flanking canonical splice sites of this exon. 0.003 is far below the <=0.1 cutoff at which SpliceAI predicts no meaningful splice impact (Jaganathan et al. 2019, PMID:30661751), so multiple/calibrated computational evidence supports no splice-altering effect and BP4 is met at supporting strength, the maximum strength the SpliceAI calibration assigns. Pangolin (SG 0.001, SL -0.001) is concordant. The local POLE framework was searched first and contains no pre-assigned code for this variant: Supplementary Tables S2 and S3 (PATH-250-323-s003.xlsx, PATH-250-323-s004.xlsx) are protein-level missense tables with no c.2173+23G>T entry, and the custom BP4 rule is limited to missense variants present in those tables, so the generic in-silico workflow governs. This is the only computational benign code from this prediction; the same SpliceAI result is not reused for PP3 or BP7.

Rule

BP4: computational evidence suggests no impact on gene/gene product. Applied via the splice path only (intronic variant): SpliceAI max delta <=0.1 (Jaganathan et al. 2019, PMID:30661751) supports BP4 at supporting strength; no POLE-specific splice calibration exists.

Evidence considered

• SpliceAI Lookup (hg37) for NM_006231.4:c.2173+23G>T returned a genuine, non-missing result (search_status result_found, spliceai_available=true, variant_id 12-133244919-C-A): DS_AG 0.00, DS_AL 0.003, DS_DG 0.00, DS_DL 0.001, max delta score 0.003, with predicted delta positions DP_AG 382, DP_AL 94, DP_DG 302, DP_DL -116.
• SpliceAI BP4 cutoff used verbatim as supplied: <=0.1 (supporting) indicates no meaningful splice impact (Jaganathan et al. 2019, Cell; PMID:30661751). 0.003 is well below it.
• Pangolin (SG 0.001, SL -0.001) from the same lookup is concordant with no splice-altering effect; used as corroboration only.
• Only the SpliceAI path was used: REVEL returned found=false (non-coding variant) and no splice score was used to support a missense judgment, avoiding cross-path substitution.
• Custom POLE BP4_Supporting rule in the case framework requires an exact missense variant present in Supplementary Table S2/S3 with 'Likely benign' REVEL class and >=4 benign in-silico results; this intronic variant is absent from both tables, so the rule does not apply.
• Supplementary Tables S2 (PATH-250-323-s003.xlsx) and S3 (PATH-250-323-s004.xlsx) read in full: no intronic or c.2173 entry exists; PATH-250-323.pdf searched for 'c.2173', '+' notations and splice/intron terms with no match.
• No PMIDs were captured for this case (literature_pass total_pmids = 0; paper_extracts.json empty), so no published computational evidence for this variant is available or cited.

Still missing

• Search of the governing POLE framework files found no entry for c.2173+23G>T / p.?: PATH-250-323-s003.xlsx (Supplementary Table S2) and PATH-250-323-s004.xlsx (Supplementary Table S3) contain only protein-level missense rows, and PATH-250-323.pdf contains no splice-region or non-coding rule; no BP4 code is pre-assigned, so the generic SpliceAI-based path was used.
• BP4 rests solely on calibrated in-silico prediction; no RNA-level functional data (minigene or patient RNA) is available for this deep-intronic position.

Sources

Reasoning

Not met. NM_006231.4:c.2173+23G>T is a single-nucleotide intronic substitution 23 bp downstream of the exon 19 donor site (VariantValidator: NC_000012.11:g.133244919C>A and NC_000012.12:g.132668333C>A; variant_exonic_positions = '19i' for both assemblies; MANE Select / RefSeq Select transcript NM_006231.4, CCDS9278.1; predicted protein NP_006222.2:p.?, i.e. no amino-acid change). It is therefore none of the variant types to which PVS1 is restricted under the ClinGen SVI PVS1 recommendations (PMC6185798) - nonsense, frameshift, canonical +/-1 or 2 splice-consensus, initiation codon, or single/multi-exon deletion - and it is not at a canonical splice consensus site (canonical_splice_consensus = false; variant_bucket 'other'). The only route by which a deep-intronic substitution such as this could produce a null allele is a splice effect, and that path is negative: SpliceAI (basic, distance 500, hg37) gives DS_AG 0.000, DS_AL 0.003, DS_DG 0.000, DS_DL 0.001, max delta 0.003, with Pangolin SG 0.001 / SL -0.001. A max delta of 0.003 is far below both the recommended supporting cutoff of 0.2 and the 0.5 high-precision cutoff (Jaganathan et al. 2019, Cell, PMID:30661751), so no gain, loss or strengthening of a splice site is predicted and there is no computational basis for an aberrant, NMD-triggering or protein-truncating transcript. Position +23 also lies outside the standard splice-region window (last 3 exonic / first 8 intronic bases) discussed in ACMG/AMP 2015 (PMID:25741868), and that guideline states that for splice variants a null effect is assumed rather than established and that confirmation of impact requires RNA or protein analysis; no RNA or protein data exist for this variant in this case (and REVEL is not found, and would be irrelevant for a non-missense variant anyway). Because no PVS1-eligible consequence exists, no NMD assessment, exon-importance assessment or strength downgrade (e.g. a stop codon escaping NMD, or a non-critical exon) arises, so the strength field is null and no strength tier (very strong / strong / moderate / supporting) is written; the 'very strong' tier would be required only for a met, full-strength, un-downgraded null call. Framework precedence: no ClinGen CSPEC/VCEP exists for POLE (prefetch cspec.found = false) and the governing local POLE framework (Leon-Castillo et al. 2020, source_registry key final_classification_framework) customizes only PM1, PS4, PP3 and BP4 - it contains no PVS1 rule or decision tree - so PVS1 falls back to generic ACMG/AMP 2015 plus the ClinGen SVI PVS1 recommendations (pvs1_generic_framework, PMC6185798). The gene-level LoF gate was not accepted as PVS1 support (see gaps and warnings).

Rule

ClinGen SVI PVS1 recommendations (source_registry key pvs1_generic_framework, PMC6185798) applied under generic ACMG/AMP 2015 fallback: PVS1 requires a predicted-null variant; this non-canonical intronic substitution (23 bp into intron 19, SpliceAI max delta 0.003) has no predicted splice or protein effect and no RNA/protein confirmation, so PVS1 is not met and no strength tier is assigned.

Evidence considered

• Variant normalization (prefetch/variant_validator): NM_006231.4:c.2173+23G>T, genomic NC_000012.11:g.133244919C>A and NC_000012.12:g.132668333C>A; variant_exonic_positions = '19i' for both assemblies (intron 19); MANE Select / RefSeq Select transcript NM_006231.4, CCDS9278.1; predicted protein consequence NP_006222.2:p.?, i.e. no amino-acid change; the primary_assembly_loci entries confirm a single C>A substitution.
• SpliceAI Lookup (source_registry key 'spliceai'; basic, distance 500, hg37, variant 12-133244919-C-A): DS_AG 0.000, DS_AL 0.003, DS_DG 0.000, DS_DL 0.001; DP_AG 382, DP_AL 94, DP_DG 302, DP_DL -116; max_delta_score 0.003; Pangolin SG 0.001, SL -0.001; variant_consequence 'sequence_variant'; spliceai_available true. The capture reports 'no significant splice impact'; its human-readable sentence displays the max delta rounded to 0.00 while the per-annotation deltas give a true maximum of 0.003 - both are far below the SpliceAI 0.2 supporting and 0.5 high-precision cutoffs (Jaganathan et al. 2019, Cell, PMID:30661751), so the rounding does not affect the call.
• Variant-level PVS1 scaffold in the case data (source_registry key 'pvs1_variant_assessment'): canonical_splice_consensus = false, consequence_class 'unknown', variant_bucket 'other', apply_generic_pvs1_framework = false, with the rationale that the variant does not fall into the generic PVS1 null-variant buckets of nonsense, frameshift or canonical +/-1,2 splice consensus - consistent with no predicted-null consequence.
• ClinGen SVI PVS1 recommendations (source_registry key 'pvs1_generic_framework', PMC6185798): PVS1 is restricted to variants predicted to be null (nonsense, frameshift, canonical +/-1/2 splice, initiation codon, single- or multi-exon deletion) in a gene where loss of function is a disease mechanism, and for splice-region variants an aberrant transcript must be demonstrated rather than assumed before PVS1 can be applied.
• ACMG/AMP 2015 full text (PMID:25741868, PVS1 section): 'For splice site variants, the variant may lead to exon skipping, shortening or inclusion of intronic material... Although splice site variants are predicted to lead to a null effect, confirmation of impact requires functional analysis by either RNA or protein analysis.' The same section cautions that truncating variants in the last exon or the last 50 bp of the penultimate exon escape NMD and that alternatively spliced exons can make an assumed null benign - NMD relevance is moot here because no aberrant transcript is predicted at all.
• Governing framework check: prefetch cspec.found = false (no ClinGen CSPEC for POLE) and no VCEP specification exists; the local framework (source_registry keys 'final_classification_framework' and 'vcep_path_250_323', Leon-Castillo et al. 2020 J Pathol 250:323-335 plus Supplementary Tables S1-S3) is a somatic endometrial-carcinoma resource whose customized criteria are PM1, PS4, PP3 and BP4 only - all exonuclease-domain missense hotspot/substitution rules - and which contains no PVS1 or splice decision tree. Generic ACMG/AMP 2015 (source_registry key 'generic_acmg_combination_rules', PMID:25741868) plus the ClinGen SVI PVS1 recommendations therefore govern PVS1.
• Gene-level LoF gate audit: output/pvs1_gene_context.json (source_registry key 'pvs1_gene_context') sets germline_disease_context_found = true, lof_mechanism_supported = true and pvs1_gene_gate = 'eligible' citing five PMIDs (41530594, 35395208, 37071997, 41673532, 42194125) whose titles/abstracts concern MAN2B1 alpha-mannosidosis, FBXW7 neurodevelopmental syndrome, SRSF1 haploinsufficiency, CDKN2A-negative cutaneous melanoma and ASXL3 Bainbridge-Ropers syndrome; none mentions POLE, and one of them is even recorded as the POLE 'disease name'. This gate was not accepted as POLE loss-of-function mechanism evidence and PVS1 was not granted on its basis.
• Transcript relevance: the variant lies in intron 19 of the 49-exon canonical MANE Select transcript NM_006231.4, which is well upstream of the terminal exon and thus not in a position where a truncating event would mechanistically escape NMD - but since no null (or any) transcript consequence is predicted, no exon-importance or NMD-escape downgrade pathway is reached and no strength tier is assigned.

Still missing

• No RNA or protein-based splice assay exists for this intronic variant; a demonstrated aberrant (e.g. out-of-frame) transcript would be required to revisit PVS1, and no such data are present in this case.
• No gene-specific PVS1 decision tree is available: there is no ClinGen CSPEC/VCEP for POLE, and the local Leon-Castillo et al. 2020 package customizes only PM1/PS4/PP3/BP4.
• The gene-level LoF-mechanism justification present in the case data is not usable evidence (its supporting PMIDs concern unrelated genes); a genuine POLE LoF-mechanism literature basis was not retrievable in this run.

Sources

Reasoning

No CSPEC exists for POLE, so the governing specification is the local gene-specific POLE framework (León-Castillo et al. 2020 derived; precedence_rank 2, custom_internal), which is the only framework in this case that customizes PS4. Its PS4_Supporting specification reads: "Apply PS4_Supporting only when the exact POLE missense variant is recurrent in both the COSMIC and TCGA endometrial carcinoma cohorts in Supplementary Table S1, has a combined EC count >=10, and belongs to the established/pathogenic set highlighted by León-Castillo et al. 2020." Every element of that specification fails for NM_006231.4:c.2173+23G>T. The variant is intronic (c.2173+23G>T, 23 bp into intron 19, NP_006222.2:p.?), not a missense variant, so the custom rule's variant-type precondition is not satisfied. It is not one of the established pathogenic set (p.P286R, p.V411L, p.A456P, p.S297F, p.S459F) and is not any of the non-hotspot or POLE-score-3 variants named in the framework. A full read of Supplementary Table S1 (all 103 listed EC/COSMIC variants, extracted in full) shows no entry at or near codon 2173 and no intronic entry at all — the table lists coding substitutions only — and the case evidence records the variant as not present in COSMIC. Applying the rule's own numeric threshold as a written inequality: combined COSMIC+TCGA endometrial carcinoma count 0 >= 10? No. The other numeric operand, the COSMIC EC count 0 (recurrence in BOTH cohorts required), is likewise no. The comparison direction here is >= (higher recurrence strengthens), so a count of 0 plainly fails it. Generic ACMG PS4 fallback is also unsupported: PS4 requires case-control enrichment of the variant in affected individuals, and no germline POLE case-control or cohort-enrichment dataset exists anywhere in this case (the literature pass returned zero PMIDs; the only POLE literature in the case is a somatic endometrial-carcinoma classification paper that reports no germline case-control statistics). PS4 is therefore not met.

Rule

Local POLE custom framework (León-Castillo 2020 derived), PS4_Supporting: exact missense variant must be recurrent in BOTH COSMIC and TCGA EC cohorts in Supplementary Table S1 with combined EC count >=10 (>=, strengthens as count rises) AND belong to the established pathogenic set (p.P286R, p.V411L, p.A456P, p.S297F). Capped at Supporting.

Evidence considered

• The only POLE-specific criterion specification in this case is the local custom framework (León-Castillo 2020 derived, framework_mode custom_internal, official_clingen false), which customizes PS4 and states: 'PS4 in this framework is a custom somatic-recurrence adaptation capped at Supporting.'
• Custom PS4_Supporting rule text: 'Apply PS4_Supporting only when the exact POLE missense variant is recurrent in both the COSMIC and TCGA endometrial carcinoma cohorts in Supplementary Table S1, has a combined EC count >=10, and belongs to the established/pathogenic set highlighted by León-Castillo et al. 2020. Under the current package this supports p.P286R, p.V411L, p.A456P, and p.S297F.'
• Written inequality against the rule's own >= threshold: combined COSMIC+TCGA EC recurrence count = 0; 0 >= 10? No.
• Supplementary Table S1 (PATH-250-323-s002.xlsx) was read in full: its top rows are p.P286R (160 COSMIC + 21 TCGA = 181), p.V411L (101+13=114), p.A456P (18+2=20), p.S297F (15+3=18); every entry is a coding missense/nonsense/frameshift substitution and none is c.2173+23G>T or any intronic change.
• Variant identity: NM_006231.4:c.2173+23G>T, NP_006222.2:p.?, intronic position 19i (variant_exonic_positions start_exon/end_exon 19i), genomic NC_000012.12:g.132668333C>A; it is not in the framework's established pathogenic set (p.P286R, p.V411L, p.S297F, p.A456P, p.S459F), its moderate set (p.F367S, p.L424I, p.M295R, p.P436R, p.M444K, p.D368Y), or its supporting set (p.A465V, p.L424V, p.T278M, p.A428T).
• Somatic recurrence sources for this specific variant: COSMIC returned 'This variant has not previously been reported in somatic cancers (COSMIC)' and the OncoKB gene page returned no variant-level classification; OncoKB is a somatic-evidence resource and its POLE entry covers exonuclease-domain proofreading mutations only.
• León-Castillo et al. 2020 (PATH-250-323.pdf) is a somatic endometrial-carcinoma molecular-classification paper: it confirms pathogenicity-in-the-ultramutation-sense for five recurrent exonuclease-domain substitutions (P286R, V411L, S297F, A456P, S459F) and presents a POLE-score based on tumour mutation signature (C>A >20%, T>G >4%, C>G <0.6%, indels <5%, TMB >100 mut/Mb). It contains no germline case-control enrichment analysis and the word 'enrichment' appears only in the phrase 'with enrichment for pathogenic POLE mutations in the' tumour-classification sense, which is not case-control PS4 evidence.
• No cohort or case-control data for this variant exists in the case: literature_pass.json reports total_pmids 0 with the note 'No PMIDs found across evidence sources', and paper_extracts.json and fulltext_map.json are both empty objects.
• The variant is present in population databases at non-trivial frequency (gnomAD v4.1 AF 0.00030086, 465/1545548 alleles, 2 homozygotes; African/African American popmax AF 0.0056087), which is inconsistent with a recurrent pathogenic germline enrichment signal, although this is properly the domain of the population-frequency group rather than PS4.

Still missing

• PATH-250-323-s002.xlsx (Supplementary Table S1) was searched in full for c.2173+23G>T / p.? and contains no entry for this variant; the table is keyed by protein change for coding variants only and lists no intronic variants, so its absence is a negative finding rather than a missing row.
• PATH-250-323.pdf was searched in full for c.2173+23G>T, 2173, and p.? with no match; no intronic variant or germline case-control enrichment analysis appears anywhere in the paper.
• PATH-250-323-s001.tif is declared for PS4 but could not be converted to text (unconvertible per the case vcep_file_coverage manifest), so its content could not be searched.
• No germline POLE case-control enrichment dataset (affected vs. control carrier counts or odds ratio) exists in this case, so the generic ACMG PS4 specification could not be evaluated on its own terms either.

Sources

Reasoning

PM1 requires the variant to lie in a mutational hotspot and/or in a critical, well-established functional domain without benign variation. NM_006231.4:c.2173+23G>T is a deep intronic change 23 nucleotides into intron 19 of POLE (exon 19i; c.2173 corresponds to roughly codon 725), which places it far outside the POLE exonuclease domain and outside every hotspot the governing framework recognizes. The Leon-Castillo-derived local framework restricts PM1 to exact exonuclease-domain missense substitutions: PM1_Strong for the five established exonuclease-domain hotspots (p.P286R, p.V411L, p.S297F, p.A456P, p.S459F); PM1_Moderate for specified recurrent non-hotspot EDMs (p.F367S, p.L424I, p.M295R, p.P436R, p.M444K, p.D368Y); PM1_Supporting for the POLE-score-3 tier (p.A465V, p.L424V, p.T278M, p.A428T). None of these residues falls in intron 19, and the paper explicitly reports that a large majority of non-exonuclease-domain mutations are not pathogenic. vcep_materials.json domain_tables is an empty object, so there is no gene-approved critical-domain list that could place this intronic position inside a domain. Supplementary Table S1 (PATH-250-323-s002) contains no entry for this variant, and it is present in gnomAD v2.1 at AF 0.000665 (up to 0.593% in the African/African American population), a frequency profile inconsistent with a hotspot or an absence of benign variation. PM1 is not met.

Rule

PM1 (governing local POLE framework): hotspot / exonuclease-domain missense substitutions only (PM1_Strong for P286R, V411L, S297F, A456P, S459F; PM1_Moderate for specified recurrent EDMs). Variant is intronic and outside the exonuclease domain, so no strength applies.

Evidence considered

• Leon-Castillo et al. 2020 abstract and Table 1 discussion: five established pathogenic exonuclease-domain hotspot mutations (P286R, V411L, S297F, A456P, S459F) and the statement that a large majority of mutations outside the exonuclease domain are not pathogenic.
• Local POLE framework PM1 rules: PM1_Strong limited to the five exonuclease-domain hotspots; PM1_Moderate to p.F367S, p.L424I, p.M295R, p.P436R, p.M444K, p.D368Y; PM1_Supporting to p.A465V, p.L424V, p.T278M, p.A428T. No rule covers an intronic position.
• VariantValidator: c.2173+23G>T lies 23 nt into intron 19 (exon 19i), i.e. no codon/residue is affected (protein consequence p.?).
• gnomAD v2.1: total AF 0.000665 (155/233160 alleles), grpmax FAF 0.00505, African/African American AF 0.00593 (144/24282), 0 homozygotes; gnomAD v4.1 record also retrieved for the GRCh38 locus. The variant is not rare enough for a hotspot-like profile.

Still missing

• vcep_materials.json domain_tables is an empty object, so no VCEP-extracted critical-domain list was available to check the residue against; the exonuclease-domain/hotspot conclusion was therefore derived from the framework's PM1 rules and the paper's hotspot enumeration rather than from a supplied domain table.
• No entry for NM_006231.4:c.2173+23G>T in PATH-250-323-s002.xlsx Supplementary Table S1 (searched for c.2173+23G>T, 2173 and p.?); the table is protein-keyed and cannot contain an intronic allele with no protein change.
• PATH-250-323-s001.tif could not be converted to searchable text (visual supplementary figure only) and could not be machine-searched for this variant.

Sources

Reasoning

PM2 at supporting strength requires the variant to be absent from, or at a frequency at or below 0.0001 in, a large control/general population cohort (supplied generic calibration; ACMG/AMP 2015, PMID:25741868 as refined by the ClinGen SVI recommendation to downgrade PM2 to supporting). No governing POLE specification overrides this: there is no ClinGen CSPEC (cspec found = false) and the local Leon-Castillo-derived POLE framework customizes only PM1, PS4, PP3 and BP4 and specifies neither a PM2 cutoff nor a non-cancer/exome-only population source, so the generic cutoff and the all-comers gnomAD v2.1/v4.1 totals are the default. The variant is not that rare and is definitively not absent. Default sources: gnomAD v2.1 total AF 0.000664780 (155/233,160 alleles) and gnomAD v4.1 total AF 0.000300864 (465/1,545,548 alleles), with exome-only AFs of 0.000520364 (v2.1) and 0.000166520 (v4.1); grpmax FAF 0.00504945 (v2.1) and 0.00515979 (v4.1); highest ancestry African/African American 0.005930319 (v2.1, 144/24,282) and 0.005608733 (v4.1, 409/72,922). Alternative sources are all above the cutoff as well: gnomAD v2.1 non-cancer exomes AF 0.000501638 (94/187,386), grpmax FAF 0.00482092, African/African American 0.005798744 (133/22,936); gnomAD v3.1 non-cancer genomes AF 0.001527481 (226/147,956), grpmax FAF 0.00473164; gnomAD-Canada v1.0 AF 0.000542888 (10/18,420), African/African American 0.009803922 (10/1,020). Every global and popmax metric measured exceeds 0.0001, by roughly 1.7-fold (gnomAD v4.1 exome AF 0.000167) to about 15-fold (gnomAD v3.1 non-cancer AF 0.001527); the lowest value anywhere is the gnomAD v4.1 European (non-Finnish) ancestry-specific AF (3/1,144,940 = 0.0000026), but PM2 is assessed on the variant's overall rarity and popmax frequency, and the variant is carried on 465 alleles in v4.1 alone and reaches 0.56-0.98% in African-ancestry samples. PM2 is therefore not met, with the same result whether the all-comers default or the non-cancer subsets are used. ClinVar's classification was not used as evidence.

Rule

PM2 (supporting pathogenic): absent from, or at an allele/filtering frequency <= 0.0001 in, a large control/general population cohort (ACMG/AMP 2015 with ClinGen SVI supporting-strength calibration, PMID:25741868). No POLE-specific PM2 rule or population-source override exists. Not met at total AF 0.000665 (gnomAD v2.1).

Evidence considered

• Threshold applied: PM2 supporting pathogenic when the variant is absent or at an allele/filtering frequency <= 0.0001 in a large control/general population cohort (ACMG/AMP 2015, PMID:25741868; ClinGen SVI recommendation to downgrade PM2 to supporting), taken verbatim from the supplied published calibration.
• gnomAD v2.1 all-comers (default source): total AF 0.000664780 (AC 155 / AN 233,160), homozygotes 0; exome AF 0.000520364 (105/201,782); genome AF 0.001593473 (50/31,378); grpmax FAF 0.00504945; highest ancestry African/African American AF 0.005930319 (144/24,282).
• gnomAD v4.1 all-comers (default source): total AF 0.000300864 (AC 465 / AN 1,545,548), homozygotes 2; exome AF 0.000166520 (232/1,393,226); genome AF 0.001529654 (233/152,322); grpmax FAF 0.00515979; highest ancestry African/African American AF 0.005608733 (409/72,922).
• gnomAD v2.1 non-cancer exomes (sensitivity analysis; GNOMAD_V2_1_NON_CANCER reported by its exome-only figures): exome AF 0.000501638 (94/187,386), homozygotes 0; grpmax FAF 0.00482092; African/African American AF 0.005798744 (133/22,936).
• gnomAD v3.1 non-cancer genomes (sensitivity analysis; genomes only, v3 has no exome cohort): AF 0.001527481 (226/147,956), homozygotes 0; grpmax FAF 0.00473164.
• gnomAD-Canada v1.0 (HostSeq genomes): AF 0.000542888 (AC 10 / AN 18,420), homozygotes 1; African/African American AF 0.009803922 (10/1,020), homozygotes 1.
• The variant is not absent from any source: it is carried on 155 alleles in gnomAD v2.1, 465 alleles in gnomAD v4.1, 144 alleles in gnomAD v2.1 non-cancer exomes, 226 alleles in gnomAD v3.1 non-cancer genomes and 10 alleles in gnomAD-Canada, so read-depth/absence checking is not applicable and no source places it at or below 0.0001 overall.
• The lowest frequency value recorded anywhere is the gnomAD v4.1 European (non-Finnish) ancestry-specific AF of 0.0000026 (3/1,144,940); this was not used to satisfy PM2 because the criterion is assessed on the variant's overall/popmax rarity rather than a single low-frequency ancestry, and the same variant reaches 0.56-0.98% in African-ancestry samples in the same and other datasets.
• Source selection: the governing POLE framework (local Leon-Castillo-derived custom framework, official_clingen = false) specifies no non-cancer or exome-only population source for PM2, so the all-comers GNOMAD_V2_1/GNOMAD_V4_1 entries are the default; the v2.1 non-cancer exome and v3.1 non-cancer genome subsets were nevertheless evaluated and give the same not-met result.
• The single-submitter ClinVar classification (Likely benign, SCV002550158) was not used as evidence for PM2.

Sources

Reasoning

NM_006231.4:c.2173+23G>T is an intronic substitution 23 bp into an intron (protein consequence p.?, no amino-acid substitution), so PP3/BP4 must be taken from the SpliceAI splice path only; there is no missense protein-impact path and no REVEL score for this variant (REVEL lookup returned found=false, as expected for a non-coding change). SpliceAI Lookup was run for NM_006231.4:c.2173+23G>T (hg37) and returned a result (spliceai_available=true) with DS_AG 0.00, DS_AL 0.003, DS_DG 0.00, DS_DL 0.001, i.e. a maximum delta score of 0.003; the nearest predicted cryptic site changes are ~382/94/302/-116 bp away, far outside any plausible impact on the surrounding splice sites, and Pangolin likewise returned SG 0.001 / SL -0.001. This is below every published PP3 SpliceAI cutoff, including the >=0.2 supporting-level threshold of Jaganathan et al. 2019 (PMID:30661751), so PP3 is not met at any strength. The local POLE framework in this case (Leon-Castillo et al. 2020, J Pathol 250:323-335, and its supplementary in-silico tables S2/S3) was searched for this variant and contains no entry: both tables list protein-level changes only (e.g. P286R, V411L, A456P, S297F, F367S, plus a few frameshifts), and the custom PP3 rule is restricted to an exact missense variant appearing in those tables, so no code is pre-assigned and the generic in-silico workflow governs.

Rule

PP3: computational evidence supports a deleterious effect on gene/gene product. Applied via the splice path only (intronic variant): SpliceAI max delta >=0.2 (supporting) or >=0.5 (moderate) per Jaganathan et al. 2019 (PMID:30661751); no POLE-specific PP3 specification exists for non-missense variants. Not met at max delta 0.003.

Evidence considered

• Variant consequence: NM_006231.4:c.2173+23G>T is intronic (23 nt into the intron); normalization gives protein NP_006222.2:p.? with no missense substitution, so the SpliceAI path is the applicable computational path for PP3/BP4.
• SpliceAI Lookup (hg37) for NM_006231.4:c.2173+23G>T returned a genuine result (search_status result_found, spliceai_available=true): DS_AG 0.00, DS_AL 0.003, DS_DG 0.00, DS_DL 0.001; max delta score 0.003. The pipeline's summary sentence rounds this to '0.00', but the underlying max delta is 0.003.
• SpliceAI PP3 cutoffs used verbatim as supplied: PP3 moderate >=0.5, PP3 supporting >=0.2 (Jaganathan et al. 2019, Cell; PMID:30661751). 0.003 does not reach either.
• Pangolin scores from the same lookup (SG 0.001, SL -0.001) are concordant with no predicted splice impact; used only as corroboration, not as a separate ACMG criterion.
• REVEL lookup for this variant returned found=false / score=null; consistent with a non-coding variant and irrelevant to the splice path (no cross-path substitution was performed).
• Custom POLE framework (case_summary cspec block, derived from the local Leon-Castillo et al. 2020 package) restricts PP3_Supporting to an exact missense variant present in Supplementary Table S2 or S3 with a 'likely disease causing' REVEL class and <=1 benign in-silico result; c.2173+23G>T is not a missense variant and is absent from both tables, so the rule does not apply and the generic workflow governs.
• Supplementary Table S2 (PATH-250-323-s003.xlsx) and Table S3 (PATH-250-323-s004.xlsx) were read in full: both are keyed by protein substitutions (SIFT/PANTHER/SNAP2/PolyPhen2/PROVEAN/REVEL_class), with no intronic, splice-region, or c.2173 entries.
• PATH-250-323.pdf was searched for 'c.2173', the variant's intronic notation, and for splice/intron/synonymous terms; the paper covers somatic POLE exonuclease-domain missense hotspots (P286R, V411L, S297F, A456P, S459F) and the ultramutated signature, with no splice-prediction rule.
• No PMIDs were captured for this case (literature_pass total_pmids = 0; paper_extracts.json is empty), so no published paper-level computational evidence is available or cited here.

Still missing

• Search of the governing POLE framework files found no entry for c.2173+23G>T / p.?: PATH-250-323-s003.xlsx (Supplementary Table S2, TCGA EC in-silico results) and PATH-250-323-s004.xlsx (Supplementary Table S3, non-hotspot EC in-silico results) tabulate protein-level missense variants only, and PATH-250-323.pdf (Leon-Castillo et al. 2020) contains no intronic or splice-region variant rule; no PP3 code is pre-assigned for this variant, so the generic SpliceAI-based path was used instead.
• No functional/RNA assay (minigene, RT-PCR, or patient RNA) is available to corroborate the computational prediction of no splice impact.

Sources

Reasoning

PP5 applies only when a reputable source reports the exact variant as pathogenic, and per the governing rule for this group only an exact-variant ClinVar expert-panel classification of Pathogenic or Likely pathogenic may trigger it. The variant maps to ClinVar variation ID 1697411, whose aggregate classification is Likely benign with review status 'criteria provided, single submitter' — not Pathogenic or Likely pathogenic, so the PP5 direction fails on the label itself. Independently, the submission audit for this record shows expert_panel_submissions = 0, submissions_total = 0, usable_for_classification = 0, and criterion_lead_submissions = 0, so there is no expert-panel assertion of any kind, let alone a pathogenic one. A single-submitter laboratory entry is an ordinary laboratory submission and must never trigger PP5, and an aggregate ClinVar label is likewise explicitly excluded from triggering it. No VCEP, ClinGen expert panel, or other reputable-source assertion of pathogenicity for this exact variant is present in the case. PP5 is therefore not met.

Rule

PP5 triggers only for an exact-variant ClinVar expert-panel classification of Pathogenic or Likely pathogenic; ordinary laboratory submissions, aggregate ClinVar labels, and non-expert assertions never trigger it.

Evidence considered

• ClinVar record for this variant: Variation ID 1697411, aggregate classification 'Likely benign', review status 'criteria provided, single submitter' (source_registry key clinvar).
• ClinVar submission audit for that record: submissions_total 0, usable_for_classification 0, expert_panel_submissions 0, validated_pmids 0, criterion_lead_submissions 0; evidence_sentence 'This variant is present in ClinVar (Variation ID: 1697411); submission details unavailable.'
• Applying the governing group rule: only an exact-variant ClinVar expert-panel classification counts; the recorded label is Likely benign (not Pathogenic/Likely pathogenic) and it is a single-submitter laboratory submission, so PP5's trigger condition is absent on both counts.
• No expert-panel or VCEP assertion for this variant exists elsewhere in the case; the only POLE framework present is a local, non-ClinGen custom framework (official_clingen false, framework_mode custom_internal).

Still missing

• The ClinVar record for variation ID 1697411 was not confirmed as an exact-variant match (prefetch reports exact_match false with the warning 'No exact ClinVar XML match validated; using best available candidate'), so even the aggregate Likely benign label is unverified for this exact variant; this does not change the PP5 verdict because no expert-panel assertion is present in any case.
• No ClinGen VCEP or expert-panel specification for POLE was found in the case (no CSPEC; the local framework declares official_clingen false), so no expert-panel assertion of pathogenicity for this exact variant could be located.

Sources

Reasoning

BA1 (stand-alone benign) requires an allele frequency of at least 0.05 in a control/general population dataset (ACMG/AMP 2015, PMID:25741868). No governing POLE specification exists: there is no ClinGen CSPEC for POLE (cspec found = false), and the local Leon-Castillo-derived POLE framework that governs this case customizes only PM1, PS4, PP3 and BP4, specifies no population threshold and no non-cancer/exome-only population source, and directs generic ACMG/AMP fallback outside its own rules. The generic 0.05 threshold and the all-comers gnomAD v2.1/v4.1 totals were therefore used as the default. The variant is nowhere near 5% in any source. gnomAD v2.1 all-comers: total AF 0.000664780 (155/233,160 alleles), homozygotes 0, grpmax FAF 0.00504945, highest ancestry African/African American (AFR) AF 0.005930319 (144/24,282 alleles), homozygotes 0. gnomAD v4.1 all-comers: total AF 0.000300864 (465/1,545,548 alleles), homozygotes 2, grpmax FAF 0.00515979, highest ancestry AFR AF 0.005608733 (409/72,922 alleles), homozygotes 2. The non-cancer subsets are equivalent or higher but still far below 0.05: gnomAD v2.1 non-cancer exomes AF 0.000501638 (94/187,386), grpmax FAF 0.00482092, AFR 0.005798744 (133/22,936); gnomAD v3.1 non-cancer genomes AF 0.001527481 (226/147,956), grpmax FAF 0.00473164. gnomAD-Canada v1.0 (HostSeq genomes) AF 0.000542888 (10/18,420), AFR 0.009803922 (10/1,020), which is the highest ancestry-specific value seen in any dataset and is still only 0.98% on a 1,020-allele subgroup. The value closest to BA1 anywhere is the gnomAD v3.1 non-cancer 'highest population' flag, 1KG:LWK AF 0.0454545 (6/132 alleles), which is a small-sample artefact rather than a population frequency signal and in any case still falls below 0.05. Every adequately powered metric is below the 0.05 threshold by at least about 5-fold (gnomAD-Canada African ancestry 0.98%, 0.05/0.0098 = 5.1) and usually far more (about 8.4-fold for gnomAD v2.1 African ancestry 0.59%, about 75-fold for gnomAD v2.1 total AF 0.000665 and about 166-fold for gnomAD v4.1 total AF 0.000301), so BA1 is not met, and the verdict is unchanged whether the all-comers default or the non-cancer subsets are used. The single-submitter ClinVar classification was not used as evidence for this criterion.

Rule

BA1 (stand-alone benign): allele frequency >= 0.05 in a control/general population dataset (ACMG/AMP 2015, PMID:25741868). No POLE-specific BA1 rule or population-source override exists, so the generic threshold and the all-comers gnomAD v2.1/v4.1 totals apply. Not met at a maximum ancestry frequency of 0.98%.

Evidence considered

• Threshold applied: BA1 stand-alone benign at allele frequency >= 0.05 in a control/general population dataset (ACMG/AMP 2015, PMID:25741868), taken verbatim from the supplied published calibration; no governing POLE specification overrides it.
• gnomAD v2.1 all-comers (variant 12-133244919-C-A, GRCh37; default source because the governing POLE framework specifies no population source): total AF 0.000664780 (AC 155 / AN 233,160), homozygotes 0; exome AF 0.000520364 (105/201,782), homozygotes 0; genome AF 0.001593473 (50/31,378), homozygotes 0; grpmax FAF 0.00504945; highest ancestry African/African American AF 0.005930319 (144/24,282), homozygotes 0.
• gnomAD v4.1 all-comers (variant chr12-132668333-C-A, GRCh38; default source): total AF 0.000300864 (AC 465 / AN 1,545,548), homozygotes 2; exome AF 0.000166520 (232/1,393,226), homozygotes 2; genome AF 0.001529654 (233/152,322), homozygotes 0; grpmax FAF 0.00515979; highest ancestry African/African American AF 0.005608733 (409/72,922), homozygotes 2.
• gnomAD v2.1 non-cancer exomes (sensitivity analysis; GNOMAD_V2_1_NON_CANCER reported by its exome-only figures): exome AF 0.000501638 (94/187,386), homozygotes 0; grpmax FAF 0.00482092; African/African American AF 0.005798744 (133/22,936), homozygotes 0.
• gnomAD v3.1 non-cancer genomes (sensitivity analysis; v3 has no exome cohort, so this source is genomes-only): AF 0.001527481 (226/147,956), homozygotes 0; grpmax FAF 0.00473164.
• gnomAD-Canada v1.0 (HostSeq genomes): AF 0.000542888 (AC 10 / AN 18,420), homozygotes 1; African/African American AF 0.009803922 (10/1,020), homozygotes 1; European (non-Finnish) 0/11,742. This is the highest ancestry-specific frequency observed in any dataset and it remains about 5-fold below the 0.05 BA1 threshold.
• Ancestry pattern: the variant is consistently most frequent in African/African American samples (0.56-0.98% across datasets) and near-absent elsewhere (gnomAD v4.1 European non-Finnish 3/1,144,940 = 0.0000026; East Asian, Finnish, Ashkenazi Jewish and Amish 0 in v4.1). This is ancestry differentiation of a low-frequency variant, not a single-population founder signal and not a common polymorphism.
• The 'highest population' annotation for gnomAD v3.1 non-cancer (1KG:LWK, AF 0.0454545 = 6/132 alleles) rests on 6 alleles in a 132-allele subsample; it was treated as small-sample noise and not as a frequency signal for BA1 (grpmax FAF 0.00473164 was used instead).
• The single-submitter ClinVar classification (Likely benign, SCV002550158) was not used as evidence for BA1: a criterion is never inferred from another source's final classification.

Sources

Reasoning

BS1 (strong benign) requires an allele frequency greater than that expected for the disorder. No POLE-specific population threshold is available: no ClinGen CSPEC exists for POLE (cspec found = false) and the governing local POLE framework (leon-castillo-2020-custom-framework-v1, official_clingen = false) customizes only PM1, PS4, PP3 and BP4 and expressly falls back to generic ACMG/AMP outside those rules. The applicable threshold is therefore the supplied generic default of >= 0.01 (per-gene Whiffin-style derivation preferred, PMID:28518168), and a per-gene derivation could not be performed because the case materials contain no POLE-associated disease prevalence, penetrance or allelic-heterogeneity inputs. Against 0.01 the variant falls short in every source, but this time only narrowly in the highest-ancestry comparisons: gnomAD v2.1 all-comers total AF 0.000664780 (155/233,160) with grpmax FAF 0.00504945 and highest ancestry African/African American 0.005930319 (144/24,282); gnomAD v4.1 all-comers total AF 0.000300864 (465/1,545,548) with grpmax FAF 0.00515979 and AFR 0.005608733 (409/72,922); gnomAD v2.1 non-cancer exomes total AF 0.000501638 (94/187,386), grpmax FAF 0.00482092, AFR 0.005798744 (133/22,936); gnomAD v3.1 non-cancer genomes total AF 0.001527481 (226/147,956), grpmax FAF 0.00473164. The highest ancestry-specific value anywhere is gnomAD-Canada v1.0 AFR 0.009803922 (10/1,020 alleles), which is 0.98% against the 1.0% generic threshold - a margin of about 1.02-fold on a 1,020-allele subgroup - while the corresponding default-source value (gnomAD v2.1 AFR 0.005930319) sits about 1.7-fold below 0.01. Because the miss is narrow and the threshold is a generic default rather than a POLE-derived one, BS1 is recorded as not met but flagged for human review; a per-gene Whiffin-style threshold for this adult-onset, moderately penetrant cancer-predisposition phenotype could plausibly fall below the observed African-ancestry frequency and would change the call. The verdict is the same whether all-comers totals or non-cancer subsets are used. ClinVar's classification was not used as evidence.

Rule

BS1 (strong benign): allele frequency greater than expected for the disorder; generic SVI default >= 0.01 applied because no POLE-specific threshold exists and per-gene (Whiffin-style) derivation was not possible from the available materials (PMID:28518168, PMID:25741868). Not met at a maximum ancestry frequency of 0.98%.

Evidence considered

• Threshold applied: BS1 strong benign at allele frequency >= 0.01, generic SVI default with per-gene derivation preferred (Whiffin et al. 2017, PMID:28518168), taken verbatim from the supplied published calibration; no governing POLE specification overrides it and no per-gene derivation was possible.
• gnomAD v2.1 all-comers: total AF 0.000664780 (155/233,160); grpmax FAF 0.00504945; highest ancestry African/African American 0.005930319 (144/24,282). gnomAD v4.1 all-comers: total AF 0.000300864 (465/1,545,548); grpmax FAF 0.00515979; African/African American 0.005608733 (409/72,922).
• gnomAD v2.1 non-cancer exomes: total AF 0.000501638 (94/187,386); grpmax FAF 0.00482092; African/African American 0.005798744 (133/22,936). gnomAD v3.1 non-cancer genomes: total AF 0.001527481 (226/147,956); grpmax FAF 0.00473164.
• gnomAD-Canada v1.0 (HostSeq genomes): overall AF 0.000542888 (10/18,420); African/African American AF 0.009803922 (10/1,020) - the highest ancestry-specific frequency in this case and only about 1.02-fold below the 0.01 generic threshold.
• Governing framework check: no ClinGen CSPEC exists for POLE (cspec found = false); the local POLE framework derived from Leon-Castillo et al. 2020 customizes only PM1, PS4, PP3 and BP4 and states that outside its explicit rules the interpreter should fall back to generic ACMG/AMP, so it provides neither a gene-specific BS1 threshold nor a population-source specification.
• Disease-model context used for the 'greater than expected for the disorder' comparison: POLE germline disease is an adult-onset, incompletely penetrant cancer-predisposition/polyposis phenotype (gene-level context: germline heterozygous loss-of-function exonuclease-domain POLE variants cause polyposis and predispose to colorectal cancer). No prevalence or penetrance figures were available in the case materials, so this comparison remained qualitative.
• Ancestry caveat: the frequency signal is confined to African/African American samples (0.56-0.98%) and is near-absent in European non-Finnish, East Asian, Finnish, Ashkenazi Jewish and Amish samples (gnomAD v4.1 NFE 3/1,144,940); a threshold comparison based on a single high-frequency ancestry is the main source of uncertainty in this criterion.

Still missing

• No per-gene (Whiffin-style) BS1 threshold could be derived for POLE: the governing local framework defines no population cut-offs and the case materials contain no POLE-associated disease prevalence, penetrance or allelic-heterogeneity inputs, so the supplied generic default (>= 0.01) was applied. Unlike the other criteria in this group, the margin here is narrow (highest ancestry value 0.98% vs 1.0%; default-source maximum 0.59% vs 1.0%), so the not-met verdict is sensitive to a plausible gene-specific threshold recalibration and warrants human confirmation.
• gnomAD-Canada v1.0 is a comparatively small ancestry-resolved dataset (African/African American subgroup AN = 1,020), so its 0.98% African-ancestry frequency carries wide confidence limits and is the sole value in this case that approaches the 1% generic threshold.

Sources

Reasoning

BS2 requires the variant to be observed in a well-documented healthy adult individual for a recessive (homozygous), dominant (heterozygous) or X-linked (hemizygous) disorder that is fully penetrant at an early age (ACMG/AMP 2015, PMID:25741868). The disorder relevant to this case is the POLE germline cancer-predisposition/polyposis phenotype, which is autosomal dominant, adult-onset and incompletely penetrant (gene-level context: heterozygous germline loss-of-function exonuclease-domain POLE variants cause polyposis and predispose to colorectal cancer); the criterion's fully-penetrant-at-an-early-age precondition is therefore not satisfied, so the many heterozygous carrier alleles in gnomAD (155 alleles in v2.1, 465 in v4.1) are expected rather than benign evidence and do not trigger BS2. The recessive/homozygous reading was also examined because homozygous observations do exist: gnomAD v4.1 reports 2 homozygotes (total_hom 2, all in the African/African American group at 409/72,922 alleles) and gnomAD-Canada v1.0 reports 1 homozygote (African/African American, 10/1,020 alleles), while gnomAD v2.1 (0), gnomAD v2.1 non-cancer exomes (0) and gnomAD v3.1 non-cancer genomes (0) report none. Those homozygote counts are close to the Hardy-Weinberg expectation at the observed African-ancestry frequency (about 1-2 homozygotes among roughly 36,500 African-ancestry individuals in v4.1, and only about 0.05 expected in the 510-individual gnomAD-Canada African subgroup, where Poisson noise makes any Hardy-Weinberg comparison uninformative), gnomAD and HostSeq participants are not phenotyped for the POLE recessive phenotype (facial dysmorphism, immunodeficiency, livedo, short stature), and no dataset provides a documented healthy adult homozygous genotype for this disorder; the deep-intronic position (c.2173+23, SpliceAI max delta 0.003) also gives no independent support that homozygosity here is a penetrant genotype. BS2 is therefore not met under either reading, but the presence of three homozygous observations across two independent datasets is flagged for human review because the applicable inheritance model for this gene (dominant adult-onset versus biallelic early-onset) is not settled by the governing local framework. ClinVar's classification was not used as evidence.

Rule

BS2 (strong benign): observed in a well-documented healthy adult individual for a recessive (homozygous), dominant (heterozygous) or X-linked (hemizygous) disorder with full penetrance expected at an early age (ACMG/AMP 2015, PMID:25741868). Not met: the evaluated POLE phenotype is adult-onset and incompletely penetrant, and the population homozygote observations are unphenotyped.

Evidence considered

• BS2 criterion text applied per ACMG/AMP 2015 (PMID:25741868): the variant must be observed in a well-documented healthy adult individual for a recessive (homozygous), dominant (heterozygous) or X-linked (hemizygous) condition that is fully penetrant at an early age.
• Disease-model mismatch: the POLE germline phenotype under investigation in this case is an adult-onset, incompletely penetrant dominant cancer-predisposition/polyposis phenotype (gene-level context: germline heterozygous loss-of-function exonuclease-domain POLE variants cause polyposis and predispose to colorectal cancer), so the criterion's early-onset/full-penetrance precondition is not met and heterozygous carriers in population databases are expected.
• Homozygote observations: gnomAD v4.1 total_hom 2 (African/African American 409/72,922 alleles with 2 homozygotes; exome subset hom 2, genome subset hom 0); gnomAD-Canada v1.0 hom 1 (African/African American 10/1,020 alleles); gnomAD v2.1 all-comers hom 0; gnomAD v2.1 non-cancer exomes hom 0; gnomAD v3.1 non-cancer genomes hom 0.
• The gnomAD v4.1 homozygote count is consistent with Hardy-Weinberg expectation at the observed African-ancestry allele frequency (0.005609), i.e. about 1-2 homozygotes among the African-ancestry individuals sampled, rather than an excess indicating a tolerated hypomorphic genotype.
• Carrier alleles in gnomAD/HostSeq are not usable as 'well-documented healthy adults' for this phenotype: participants are unselected for and unphenotyped with respect to POLE cancer predisposition or the recessive POLE syndrome, and heterozygous carrier status in an adult is compatible with later-onset disease for this gene.
• Variant context relevant to the homozygous reading: c.2173+23G>T is deep intronic with no predicted splice effect (SpliceAI max delta 0.003), so homozygosity for this allele provides no functional basis for assuming a penetrant null genotype in the population homozygotes.
• Governing framework check: no ClinGen CSPEC exists for POLE and the local Leon-Castillo-derived POLE framework customizes only PM1, PS4, PP3 and BP4, defining no BS2 rule and no inheritance-model specification; generic ACMG/AMP therefore governs this criterion.
• The single-submitter ClinVar classification (Likely benign, SCV002550158) was not used as evidence for BS2.

Still missing

• The governing local POLE framework does not state which inheritance model BS2 should be applied to. POLE is documented for both a dominant adult-onset cancer-predisposition/polyposis phenotype and a rare biallelic early-onset syndrome (facial dysmorphism, immunodeficiency, livedo, short stature), and the case materials contain no prevalence, penetrance or age-of-onset data to adjudicate between them; under the dominant reading BS2 fails on its precondition, under the biallelic reading the three population homozygotes would need phenotyping to be usable.
• No phenotype information accompanies the homozygous observations in gnomAD v4.1 (2 homozygotes, African/African American exomes) or gnomAD-Canada v1.0 (1 homozygote, African/African American), so they cannot be treated as 'well-documented healthy adult' genotypes; a curator with access to individual-level data would be needed to use them.

Sources

Reasoning

BP6, as scoped for this group, applies only when a reputable source — operationalized here as an exact-variant ClinVar expert-panel classification — reports the variant as Benign or Likely benign. The variant's ClinVar record (variation ID 1697411) does carry an aggregate label of Likely benign, but the review status is 'criteria provided, single submitter' and the submission audit for that record shows expert_panel_submissions = 0 with submissions_total = 0 and usable_for_classification = 0. There is therefore no expert-panel assertion of benignity for this exact variant. An aggregate ClinVar label resting on an ordinary single-submitter laboratory submission must never trigger BP6, and the governing rule for this group explicitly excludes ordinary laboratory submissions and non-expert assertions. Because the qualifying expert-panel benign assertion does not exist, BP6 is not met, notwithstanding that the aggregate database label happens to read Likely benign. This negative BP6 result does not preclude benign-direction evidence being captured by the population-frequency group, where the variant's gnomAD frequency (AF 0.00030086 in v4.1, popmax 0.0056087, 2 homozygotes) is directly relevant.

Rule

BP6 triggers only for an exact-variant ClinVar expert-panel classification of Benign or Likely benign; ordinary laboratory submissions, aggregate ClinVar labels, and non-expert assertions never trigger it.

Evidence considered

• ClinVar record for this variant: Variation ID 1697411, aggregate classification 'Likely benign', review status 'criteria provided, single submitter' (source_registry key clinvar).
• ClinVar submission audit for that record: expert_panel_submissions 0, submissions_total 0, usable_for_classification 0, usable_for_criterion_leads 0, criterion_lead_submissions 0.
• Applying the governing group rule: BP6 requires an exact-variant ClinVar expert-panel Benign/Likely benign classification; a single-submitter aggregate label is an ordinary laboratory submission and is explicitly excluded from triggering BP6.
• No other reputable-source expert-panel benign assertion for this exact variant appears anywhere in the case materials.

Still missing

• The ClinVar record for variation ID 1697411 was not validated as an exact-variant match (prefetch exact_match false, warning 'No exact ClinVar XML match validated; using best available candidate'), so the aggregate Likely benign label is not confirmed to refer to this exact variant; the BP6 verdict does not depend on it because no expert-panel assertion exists either way.
• No ClinGen expert panel (VCEP/VUS Task Force) has published a benign classification for this variant that could be located in the case materials; no POLE CSPEC exists.

Sources

Reasoning

PS2 requires a confirmed de novo occurrence of the variant in a proband with the relevant phenotype, with parentage and parental genotype confirmation (generic ACMG/AMP 2015). The case contains no proband, no clinical/phenotype record, no parental testing result, and no trio genotype data for POLE NM_006231.4:c.2173+23G>T (NC_000012.11:g.133244919C>A). The available material is limited to population frequency (gnomAD v2.1/v4.1), a ClinVar record (Variation ID 1697411, single-submitter Likely benign), a SpliceAI lookup, and somatic POLE hotspot literature context; none of these address de novo occurrence or parentage. Parental samples and a confirmed de novo observation would be required before PS2 can be evaluated, so the criterion is not assessed rather than met or not met.

Rule

PS2 (Pathogenic Strong): confirmed de novo occurrence of the variant in a patient with the disease, with parental confirmation (ACMG/AMP 2015, PMID:25741868).

Evidence considered

• Generic ACMG/AMP 2015 criterion definition applied because the local POLE framework (Leon-Castillo et al. 2020 custom framework) customizes only PM1, PS4, PP3 and BP4 and specifies no PS2 rule.
• The consolidated case bundle, source_registry.json, evidence.json, case_summary.json and prefetch.json were searched for proband, parental testing, trio, de novo, family and phenotype content; no such records are present.
• Only population frequency (gnomAD v2.1 and v4.1), a ClinVar record (Variation ID 1697411, criteria provided single submitter, Likely benign), a SpliceAI lookup, and somatic POLE hotspot literature context are available; none carries de novo or parentage information.

Still missing

• No proband identified for this variant and no phenotype record linked to the variant.
• No parental genotype/testing results (trio or targeted parental testing) to confirm de novo status.
• No parentage confirmation (e.g. identity/relatedness testing) to support a claimed de novo event.

Reasoning

Not assessed. No ClinGen CSPEC/VCEP exists for POLE (prefetch steps/cspec found=false; case_summary cspec found=false), so the governing specification is the local custom framework derived from Leon-Castillo et al. 2020 (J Pathol 250:323-335). That framework customizes only PM1, PS4, PP3 and BP4, is explicitly not an official ClinGen rule set, and its own notes state it does not replace generic ACMG/AMP for 'functional assay criteria outside what is explicitly provided by the paper package'. PS3 therefore falls back to the generic ACMG/AMP 2015 definition (Richards et al., PMID:25741868): well-established functional studies showing a damaging effect on gene or protein function. NM_006231.4:c.2173+23G>T is an intronic substitution 23 bp into the intron downstream of the exon boundary at c.2173 (VariantValidator variant_exonic_positions = '19i' on both GRCh37 and GRCh38; genomic NC_000012.11:g.133244919C>A / NC_000012.12:g.132668333C>A; protein consequence NP_006222.2:p.?, no amino-acid change). The assay classes able to satisfy PS3 for this variant are therefore mRNA/splicing assays (direct RNA/cDNA analysis or an in vitro minigene splicing assay) or, for the protein, a POLE proofreading/polymerase-fidelity assay. No such assay of this variant exists in any source consulted. This case returned no literature at all (literature_pass total_pmids = 0; paper_extracts.json and fulltext_map.json are empty objects), and the functional case-evidence bundle contains zero occurrences of 'assay', 'in vitro' or 'minigene', so no published or in-house functional study of c.2173+23G>T was available to weigh. The governing local POLE package supplies no functional data either: the Leon-Castillo et al. 2020 full text contains zero occurrences of 'splice', 'intron', '2173' and 'assay', and is restricted to somatic endometrial-carcinoma exonuclease-domain missense hotspots (p.P286R, p.V411L, p.S297F, p.A456P, p.S459F) analysed by POLE-score and ultramutation signature; Supplementary Tables S1-S3 are somatic-recurrence and protein-level in-silico tables (Supplementary Table S2/S3 read in full) that contain no entry for this variant, and the framework's PM1/PS4/PP3/BP4 rules are all missense-only, so nothing in the package can be extended to this intronic allele. ClinVar VCV001697411 (single submitter, Rigshospitalet, 'Likely benign', 1 star) carries no assay data and its classification was not used to infer this criterion; COSMIC returned no record for this variant ('not previously reported in somatic cancers'); the OncoKB page failed to load, so no somatic functional annotation could be retrieved; gnomAD frequency data and SpliceAI are not functional evidence. The only functional-adjacent datum is computational: SpliceAI max delta 0.003 (DS_AG 0.000, DS_AL 0.003, DS_DG 0.000, DS_DL 0.001; Pangolin SG 0.001 / SL -0.001), a splice prediction rather than an assay result, which cannot establish a damaging effect. PS3 is therefore not assessable rather than negatively supported: the criterion's evidence base is absent, not contradicted.

Rule

Generic ACMG/AMP 2015 (Richards et al., PMID:25741868): PS3 = 'Well-established functional studies show a damaging effect on gene or protein function.' No POLE-specific PS3 specification exists: there is no ClinGen CSPEC/VCEP, and the governing local framework (Leon-Castillo et al. 2020) customizes only PM1, PS4, PP3, BP4 (missense-only rules). Not assessed for want of any functional assay of this intronic variant.

Evidence considered

• Governing framework: no ClinGen CSPEC/VCEP is available for POLE, so this case runs in custom_gene_framework mode on the local package derived from Leon-Castillo et al. 2020 (J Pathol 250:323-335). The framework payload lists customized_criteria = PM1, PS4, PP3, BP4 only, and its framework notes state it 'does not attempt to replace generic ACMG/AMP' for functional assay criteria; it is not an official ClinGen specification. PS3 has no gene-specific specification and falls back to generic ACMG/AMP 2015 (Richards et al., PMID:25741868).
• PS3 requirement (generic ACMG/AMP 2015): well-established functional studies showing a damaging effect on gene or protein function, with assay validity judged on biological relevance (full biological function versus one component), validation, reproducibility and robustness. No such study is available for this variant, so none of those conditions can be evaluated.
• Variant class driving the applicable assay type: NM_006231.4:c.2173+23G>T is intronic, 23 bp into the intron downstream of the exon boundary at c.2173 (VariantValidator variant_exonic_positions = 19i on GRCh37 and GRCh38), with predicted protein consequence NP_006222.2:p.? and no amino-acid substitution. It is not a missense change, so neither a REVEL score nor a missense protein-function assay is the relevant evidence class; the relevant classes are RNA/mRNA-processing assays or protein-level POLE function assays.
• Absence of any functional assay in the case evidence: the functional case-evidence bundle contains zero occurrences of 'assay', 'in vitro' and 'minigene', and the case generated no literature PMIDs (literature_pass total_pmids = 0; paper_extracts.json = {}; fulltext_map.json = {}), so no functional study of this variant is present to be weighed.
• Governing local POLE package searched for functional content: the Leon-Castillo et al. 2020 full text (PATH-250-323.pdf) contains zero occurrences of 'splice', 'intron', '2173' and 'assay'; it is a somatic endometrial-carcinoma study of exonuclease-domain missense hotspots (p.P286R, p.V411L, p.S297F, p.A456P, p.S459F) analysed via POLE-score and ultramutation signature, i.e. no germline functional assay of any kind, and no gene-specific PS3 rule exists in the framework.
• Supplementary tables searched: PATH-250-323-s002 (Table S1, recurrent POLE variants with COSMIC/TCGA EC counts), s003 (Table S2, TCGA in-silico results) and s004 (Table S3, non-TCGA in-silico results) were checked for this variant; all rows are protein-level missense (or frameshift) entries with zero occurrences of '2173', and none of the three tables is a functional assay, so none can evidence PS3.
• ClinVar VCV001697411 (SCV002550158, Center for Genomic Medicine, Rigshospitalet; criteria provided, single submitter; Likely benign; condition 'not specified'; last evaluated 2025-03-04) contains no functional assay data and no cited PMIDs; its classification was consulted only for completeness and was explicitly NOT used to infer this criterion.
• COSMIC returned no record for POLE c.2173+23G>T ('This variant has not previously been reported in somatic cancers'), and the OncoKB lookup failed to load (Page.wait_for_selector timeout), so neither source supplies a variant-level functional annotation in either direction.
• SpliceAI Lookup (hg37) for NM_006231.4:c.2173+23G>T returned a genuine result: DS_AG 0.000, DS_AL 0.003, DS_DG 0.000, DS_DL 0.001, max delta score 0.003 (Pangolin SG 0.001 / SL -0.001). This is a computational prediction of splice impact, not an assay result, and cannot satisfy PS3; it belongs to the in-silico evidence stream for PP3/BP4. No REVEL or BayesDel score exists for this non-missense variant, and neither could substitute for functional evidence.
• ACMG/AMP 2015 splice-variant requirement: although splice-region variants are predicted to lead to a null effect, confirmation of impact requires functional analysis by either RNA or protein analysis (Richards et al., PMID:25741868). No such RNA or protein analysis exists for c.2173+23G>T.

Still missing

• No variant-specific functional assay of POLE NM_006231.4:c.2173+23G>T was identified in any source consulted: no RNA/cDNA (RT-PCR) transcript analysis, no in vitro minigene splicing assay, and no POLE proofreading or polymerase-extension fidelity assay. Such an assay is the only route to PS3 for this intronic variant.
• Literature retrieval for this case returned no papers (literature_pass total_pmids = 0, paper_extracts.json empty), so a dedicated search for published functional or RNA studies of POLE c.2173+23G>T was not performed and the absence of a published assay is not conclusive.
• No assay controls, validation, reproducibility or robustness data of any kind exist for this variant, so PS3 strength calibration could not be considered.
• The OncoKB lookup failed with a page-load timeout, so any variant-specific somatic functional annotation it might have carried could not be checked.

Sources

Reasoning

Not assessed. PM3 requires a pathogenic variant detected in trans with this allele in a recessive gene/disorder, and no allelic (phase) information exists for NM_006231.4:c.2173+23G>T: the case contains no proband genotype, no parental or sibling genotypes, no pedigree, and no companion POLE variant in cis or trans. The ClinVar record retrieved (Variation ID 1697411) is an inexact candidate match, with no retrievable submissions, no criterion-level evidence and no PMIDs, so it cannot establish an allelic configuration; its reported classification was not used to infer this criterion. The applicable inheritance mode is also undefined: germline POLE disease includes both a dominant presentation (proofreading-associated polyposis and colorectal cancer predisposition) and a recessive one (facial dysmorphism, immunodeficiency, livedo and short stature), and no inheritance mode is recorded for this case. PM3 therefore cannot be evaluated and no strength can be assigned.

Rule

PM3 (ACMG/AMP 2015; Richards et al., PMID:25741868): For recessive disorders, detected in trans with a pathogenic variant. Not met, not excluded - no phase data.

Evidence considered

• No phase data are available for NM_006231.4:c.2173+23G>T: the case carries no proband genotype, no parental or sibling genotypes, no pedigree, and no second POLE variant that could be placed in cis or trans with this allele.
• The variant is intronic (c.2173+23G>T, 23 bp into intron 19 of NM_006231.4, chr12:132668333C>A on GRCh38), so no protein-level allelic inference is possible either.
• ClinVar Variation ID 1697411 is flagged as an inexact candidate match: no submissions or submitter details could be retrieved, no PMIDs are attached, and the condition field is 'not specified'; the record reports a classification of Likely benign (review status: criteria provided, single submitter), which was not used to infer any criterion in this group.
• The governing local POLE framework (derived from Leon-Castillo et al. 2020 and its supplementary tables) customizes only PM1, PS4, PP3 and BP4; PM3 is therefore evaluated against the generic ACMG/AMP 2015 criterion, under which a pathogenic variant must be detected in trans for a recessive disorder (Richards et al. 2015, PMID:25741868), and PM3 is not among the criteria the local POLE package specifies.
• No germline inheritance mode is recorded for POLE in the gene-level context assembled for this case (inheritance field empty), and the gene summary describes both a dominant presentation (polyposis and colorectal cancer predisposition) and a recessive presentation (facial dysmorphism, immunodeficiency, livedo and short stature), so the recessive-in-trans premise of PM3 cannot be fixed for this case.
• Population data give no allelic-configuration information: gnomAD v4.1 reports AF 0.000301 (465/1,545,548 alleles) with 2 homozygotes and gnomAD v2.1 reports AF 0.000665 (155/233,160 alleles) with 0 homozygotes; homozygosity in a population database says nothing about phase relative to a pathogenic POLE allele.

Still missing

• No phase information: no genotype data establishing whether this intronic allele lies in trans with a pathogenic POLE variant.
• No second POLE variant (pathogenic or otherwise) reported in the same individual, so no cis/trans configuration can be inferred.
• No affected-proband genotype or pedigree is available, and the inheritance mode relevant to this case (dominant proofreading-associated polyposis/CRC predisposition versus recessive facial dysmorphism-immunodeficiency-livedo-short stature) is not specified.
• The ClinVar record used (Variation ID 1697411) is a non-exact candidate match with no retrievable submissions, no review-status detail beyond 'criteria provided, single submitter', and no PMIDs.

Sources

Reasoning

PM6 requires an assumed de novo occurrence in a patient with the relevant phenotype in the absence of confirmed parental testing (generic ACMG/AMP 2015). This case provides no proband, no phenotype record, and no parental genotype or parental testing information for POLE NM_006231.4:c.2173+23G>T. The only case-specific data are population frequency (gnomAD v2.1/v4.1), a single-submitter ClinVar Likely benign record, a SpliceAI lookup, and somatic POLE hotspot literature context, none of which reports a de novo event. Without any proband observation, PM6 is not assessed.

Rule

PM6 (Pathogenic Moderate): assumed de novo occurrence in a patient with the disease, without confirmation of parentage/parental testing (ACMG/AMP 2015, PMID:25741868).

Evidence considered

• Generic ACMG/AMP 2015 criterion definition applied; the local POLE custom framework does not specify PM6.
• Searched the consolidated case bundle, source_registry.json, evidence.json, case_summary.json and prefetch.json for proband, parental testing, de novo and phenotype content; none is present.
• Available case material (gnomAD v2.1/v4.1 frequencies, ClinVar Variation ID 1697411 single-submitter Likely benign, SpliceAI lookup, POLE hotspot literature) contains no de novo or parent-of-origin information.

Still missing

• No proband or affected individual reported as carrying this variant.
• No statement of an assumed (unconfirmed) de novo occurrence and no parental testing history.
• No phenotype record to link the variant to a proband's disease presentation.

Reasoning

PP1 requires co-segregation of the variant with disease in multiple affected family members, supported by a pedigree and genotype data (generic ACMG/AMP 2015; ClinGen SVI segregation guidance). This case contains no pedigree, no family history, no genotype data from relatives, and no count of informative meioses for POLE NM_006231.4:c.2173+23G>T. The available evidence is population frequency (gnomAD v2.1/v4.1), a ClinVar record (Variation ID 1697411, single-submitter Likely benign), a SpliceAI lookup, and somatic POLE hotspot literature context. With no family structure or relative genotypes, segregation cannot be evaluated, so PP1 is not assessed.

Rule

PP1 (Pathogenic Supporting): co-segregation with disease in multiple affected family members (ACMG/AMP 2015, PMID:25741868); strength may be calibrated by segregation likelihood ratio under ClinGen SVI guidance.

Evidence considered

• Generic ACMG/AMP 2015 criterion definition applied; the local POLE custom framework customizes only PM1, PS4, PP3 and BP4 and specifies no PP1 rule.
• Searched the consolidated case bundle, source_registry.json, evidence.json, case_summary.json and prefetch.json for pedigree, family, segregation and meiosis content; no such data exist.
• The available sources (gnomAD v2.1/v4.1, ClinVar Variation ID 1697411 single-submitter Likely benign, SpliceAI lookup, somatic POLE hotspot literature) provide no family or segregation information.

Still missing

• No pedigree or family history accompanying this variant.
• No genotype data for affected or unaffected relatives.
• No count of informative meioses or segregation likelihood ratio that could support PP1 at any strength.

Reasoning

PP4 requires that the patient's phenotype or family history be highly specific for a disease with a single genetic etiology. Neither the local POLE custom framework nor any other material in this case specifies a PP4 rule, so the generic ACMG/AMP PP4 definition applies. Assessment is impossible because this case provides no proband-level clinical information at all: review_manifest.json and case_summary.json contain no HPO terms, no clinical indication, no tumour or polyposis history, no age at onset, no family history, and no affected/unaffected relative data. Nothing in evidence.json, the criteria evidence bundle, or the VCEP package records a patient phenotype that could be compared against the POLE germline disease spectrum. The gene-level context (POLE germline mutations cause polymerase proofreading-associated polyposis with colorectal adenomas/carcinoma, and a rare FACES-like syndrome of facial dysmorphism, immunodeficiency, livedo and short stature) is available, but PP4 is a patient-specific criterion and gene-level disease association alone cannot satisfy it. The only POLE clinical material in the case is León-Castillo et al. 2020, which concerns somatic endometrial-carcinoma classification rather than an individual's germline phenotype. Because the decisive input (the patient's phenotype) is entirely absent rather than merely uninformative, PP4 is recorded as not_assessed rather than not_met.

Evidence considered

• The local POLE custom framework customizes only PM1, PS4, PP3 and BP4 (customized_criteria list), so no gene-specific PP4 specification exists and the generic ACMG/AMP PP4 definition was used.
• Case files searched for proband phenotype content: review_manifest.json, case_summary.json (including compact_evidence), evidence.json, and the phenotype/prevalence evidence bundle. None contains HPO codes, a clinical indication, or family history fields.
• POLE germline disease context is documented at gene level only: germline loss-of-function in the exonuclease domain causes polyposis and colorectal cancer predisposition, and a rare syndrome of facial dysmorphism, immunodeficiency, livedo and short stature; this is gene-level association, not a patient-specific phenotype match.
• León-Castillo et al. 2020 (PATH-250-323.pdf) describes POLE-ultramutated endometrial carcinoma at the tumour level and explicitly frames the five recurrent exonuclease-domain substitutions as pathogenic in the somatic/ultramutation sense; it does not provide proband phenotype-specificity evidence.
• No PMIDs were retrieved for this case (literature_pass.json: total_pmids 0), so no case report or cohort describing this variant's phenotype could be consulted.

Still missing

• No proband clinical phenotype, HPO terms, clinical indication, age at onset, or family history is present in review_manifest.json, case_summary.json, or evidence.json.
• No case-level tumour/polyposis history is recorded, so the POLE germline phenotype spectrum (polymerase proofreading-associated polyposis / colorectal cancer, FACES-like syndrome) cannot be compared with the patient's presentation.
• PATH-250-323.pdf is a somatic endometrial-carcinoma study and provides no patient-specific germline phenotype data usable for PP4.

Reasoning

Not assessed. As with PS3, there is no ClinGen CSPEC/VCEP for POLE and the governing local framework (Leon-Castillo et al. 2020, J Pathol 250:323-335) customizes only PM1, PS4, PP3 and BP4 with missense-only rules, so BS3 falls back to the generic ACMG/AMP 2015 definition (Richards et al., PMID:25741868): well-established functional studies showing no damaging effect on protein function or splicing. NM_006231.4:c.2173+23G>T is an intronic substitution 23 bp into the intron downstream of the exon boundary at c.2173 (VariantValidator variant_exonic_positions = '19i'; protein consequence NP_006222.2:p.?, no amino-acid change), so a BS3-satisfying study would need to demonstrate preserved splicing or preserved POLE function directly - for example normal mRNA/cDNA products on RT-PCR, a minigene assay showing no aberrant transcript, or normal exonuclease/proofreading activity - together with assay validation and reproducibility data. No functional study of this variant exists in any source consulted: the case returned no literature at all (literature_pass total_pmids = 0; paper_extracts.json and fulltext_map.json empty), the functional bundle contains zero occurrences of 'assay', 'in vitro' or 'minigene', the Leon-Castillo et al. 2020 full text contains zero occurrences of 'splice', 'intron', '2173' and 'assay' (it is a somatic endometrial-carcinoma hotspot/ultramutation study with no functional assay), and Supplementary Tables S1-S3 contain no entry for this variant and are not assays. ClinVar VCV001697411 ('Likely benign', single submitter, Rigshospitalet) provides no assay data and its classification was not used to infer this criterion; COSMIC returned no record and the OncoKB page failed to load, so there is no somatic functional annotation either. The available SpliceAI result (max delta 0.003, DS_AL 0.003; DS_AG/DS_DG 0.000, DS_DL 0.001; Pangolin SG 0.001 / SL -0.001) is a computational prediction, not a functional demonstration, and the absence of a REVEL score simply reflects a non-missense variant; neither can satisfy BS3. BS3 is therefore not assessable: the absence of functional evidence must not be converted into a benign functional call.

Rule

Generic ACMG/AMP 2015 (Richards et al., PMID:25741868): BS3 = 'Well-established functional studies show no damaging effect on protein function or splicing.' No POLE-specific BS3 specification exists: there is no ClinGen CSPEC/VCEP, and the governing local framework (Leon-Castillo et al. 2020) customizes only PM1, PS4, PP3, BP4 (missense-only rules). Not assessed for want of any functional assay of this intronic variant.

Evidence considered

• Governing framework: no ClinGen CSPEC/VCEP for POLE; the local POLE framework derived from Leon-Castillo et al. 2020 customizes only PM1, PS4, PP3 and BP4 (missense-only rules, per its criteria.json and framework notes), contains no BS3 specification, and states it does not replace generic ACMG/AMP for functional assay criteria. BS3 therefore rests on generic ACMG/AMP 2015 (Richards et al., PMID:25741868).
• BS3 requirement (generic ACMG/AMP 2015): well-established functional studies showing no damaging effect on protein function or splicing, with the same assay-validity standards (biological relevance, full versus partial function, validation, reproducibility, robustness) applied as for PS3. No assay of this variant exists, so none of these conditions can be evaluated.
• Variant class: NM_006231.4:c.2173+23G>T lies 23 bp into the intron downstream of the exon boundary at c.2173 (VariantValidator variant_exonic_positions = 19i), with predicted protein consequence NP_006222.2:p.? and no amino-acid change. The BS3-relevant evidence class is a splicing/mRNA assay (RT-PCR/cDNA, minigene) or a protein-level POLE function assay showing a normal result; neither exists for this variant.
• Absence of any functional assay in the case evidence: the functional case-evidence bundle contains zero occurrences of 'assay', 'in vitro' and 'minigene', and the case generated no literature PMIDs (literature_pass total_pmids = 0; paper_extracts.json = {}; fulltext_map.json = {}), so no functional study of this variant - positive or negative - is available.
• Governing local POLE package searched for functional content: the Leon-Castillo et al. 2020 full text contains zero occurrences of 'splice', 'intron', '2173' and 'assay'; the paper is a somatic endometrial-carcinoma study of exonuclease-domain missense hotspots and POLE-ultramutated signatures, and it reports no assay demonstrating preserved POLE splicing or proofreading function for any intronic variant.
• Supplementary tables searched: PATH-250-323-s002 (Table S1) and PATH-250-323-s003/s004 (Tables S2/S3) contain no entry for this variant (zero occurrences of '2173') and are somatic-recurrence and protein-level in-silico tables; they are not functional assays and cannot support a benign functional claim.
• ClinVar VCV001697411 (single submitter, Rigshospitalet, 'Likely benign', 1 star, no cited PMIDs) provides no functional assay data; its classification was NOT used to infer BS3, per the rule that no criterion may be inferred from another source's final classification.
• COSMIC returned no record for this variant and the OncoKB page failed to load; both are absence-of-evidence (or unavailable) results and neither constitutes a functional assay showing retained protein function or normal splicing.
• SpliceAI for c.2173+23G>T: max delta 0.003 (DS_AG 0.000, DS_AL 0.003, DS_DG 0.000, DS_DL 0.001; Pangolin SG 0.001 / SL -0.001). This is an in-silico prediction; even at a benign-leaning value it cannot be counted as BS3 functional evidence, which requires an assay result showing no damaging effect. The absence of REVEL and BayesDel scores reflects the non-missense consequence type and likewise provides no functional evidence.
• ACMG/AMP 2015 mRNA-level guidance: direct RNA and/or cDNA analysis and in vitro minigene splicing assays 'can be highly informative' for variants in deeper intronic regions, but no such analysis has been performed for c.2173+23G>T, and confirmation of splice impact requires RNA or protein analysis (Richards et al., PMID:25741868).

Still missing

• No functional or RNA-level study of POLE NM_006231.4:c.2173+23G>T was identified that could demonstrate absence of a damaging effect: no RT-PCR/cDNA transcript analysis, no minigene splicing assay and no POLE proofreading or polymerase-activity assay, and therefore no assay controls, validation or reproducibility data.
• Literature retrieval for this case returned no papers (literature_pass total_pmids = 0; paper_extracts.json empty), so the absence of a published negative functional study cannot be treated as established and a targeted variant-specific literature search remains open.
• A benign-leaning computational splice prediction for this intronic variant informs BP4 (in-silico evidence) rather than BS3, which requires assay data; no such assay data are available.
• The OncoKB lookup failed with a page-load timeout, so a variant-specific somatic functional annotation that might have borne on retained function could not be checked.

Sources

Reasoning

BS4 requires evidence of lack of segregation, i.e. affected family members who do not carry the variant, documented in a pedigree with genotypes (generic ACMG/AMP 2015). This case contains no pedigree, no affected relatives, and no relative genotypes for POLE NM_006231.4:c.2173+23G>T, so neither segregation nor non-segregation can be established. The available evidence is population frequency (gnomAD v2.1/v4.1), a ClinVar record (Variation ID 1697411, single-submitter Likely benign), a SpliceAI lookup, and somatic POLE hotspot literature context. Non-segregation cannot be assessed without family data, so BS4 is not assessed.

Rule

BS4 (Benign Strong): lack of segregation in affected family members, i.e. a variant present in a proband but absent in an affected relative, based on pedigree and genotypes (ACMG/AMP 2015, PMID:25741868).

Evidence considered

• Generic ACMG/AMP 2015 criterion definition applied; the local POLE custom framework does not specify BS4.
• Searched the consolidated case bundle, source_registry.json, evidence.json, case_summary.json and prefetch.json for pedigree, family, non-segregation and affected-relative content; none is present.
• Available sources (gnomAD v2.1/v4.1, ClinVar Variation ID 1697411 single-submitter Likely benign, SpliceAI lookup, somatic POLE hotspot literature) contain no family segregation information.

Still missing

• No pedigree or family history of affected relatives.
• No genotype data showing an affected relative who lacks the variant.
• No informative meioses to evaluate a lack of segregation.

Reasoning

Not assessed. BP2 applies when a variant is observed in trans with a pathogenic variant for a fully penetrant dominant gene/disorder, or in cis with a pathogenic variant in any inheritance pattern. Neither configuration was tested or reported for NM_006231.4:c.2173+23G>T: there are no proband or parental genotypes, no pedigree, and no companion pathogenic POLE allele in this case, so the variant's allelic relationship to a pathogenic variant is unknown. The only database record retrieved (ClinVar Variation ID 1697411) is an inexact candidate match with no retrievable submissions or PMIDs and provides no allelic information; its classification was not used to infer this criterion. BP2 therefore cannot be evaluated, and this is a data gap rather than evidence that the variant is not in cis or trans with a pathogenic variant.

Rule

BP2 (ACMG/AMP 2015; Richards et al., PMID:25741868): Observed in trans with a pathogenic variant for a fully penetrant dominant gene/disorder, or observed in cis with a pathogenic variant in any inheritance pattern. Not met, not excluded - no allelic data.

Evidence considered

• BP2 requires observation of this variant in cis with a pathogenic variant (any inheritance pattern) or in trans with a pathogenic variant in a fully penetrant dominant gene/disorder; neither configuration is reported or testable for NM_006231.4:c.2173+23G>T.
• No companion POLE variant is present in this case: there is no proband genotype, no parental genotype, no pedigree and no co-occurring pathogenic allele, so no cis/trans relationship can be established either way.
• The variant is intronic (c.2173+23G>T, 23 bp into intron 19; chr12:132668333C>A, GRCh38), and no phase-resolved RNA or long-read data are available.
• ClinVar Variation ID 1697411 is an inexact candidate match with no retrievable submissions and no PMIDs (reported classification Likely benign, single submitter, condition 'not specified'); it documents no allelic configuration and was not used to infer this criterion.
• The governing local POLE framework (Leon-Castillo et al. 2020-derived) does not customize BP2 - it specifies only PM1, PS4, PP3 and BP4 - so BP2 was assessed under the generic ACMG/AMP 2015 definition (Richards et al. 2015, PMID:25741868); no CSPEC/VCEP specification for POLE was available.
• gnomAD v4.1 reports 2 homozygotes for this allele (AF 0.000301; African/African American AF 0.00561) and gnomAD v2.1 reports 0 homozygotes (AF 0.000665); biallelic occurrence in the general population provides no information about phase with a pathogenic POLE variant and is not BP2 evidence.

Still missing

• No phase data and no companion pathogenic POLE variant, so the variant cannot be placed in cis or trans with a pathogenic allele.
• No pedigree or parental genotypes, leaving the relevant inheritance pattern (fully penetrant dominant versus recessive) undefined for this case.
• The ClinVar candidate record (Variation ID 1697411) yielded no submissions or criterion-level evidence that could document a cis or trans configuration.

Sources

Reasoning

BP5 requires that the variant be found in a case with an alternate molecular basis for disease — i.e. an individual carrying this variant in whom the observed phenotype is explained by a different genetic cause (a co-occurring pathogenic variant in another disease gene, or another established molecular diagnosis). No POLE gene-specific BP5 specification exists in this case (the local framework customizes only PM1, PS4, PP3 and BP4; there is no CSPEC), so the generic ACMG/AMP BP5 definition applies. Evaluation is impossible because this case contains no case-level or genotype-level information whatsoever: there is no proband record, no list of other variants detected in the individual, no molecular diagnosis, and no affected-relative data in review_manifest.json, case_summary.json, evidence.json, or the criteria evidence bundle. Nothing in the POLE VCEP package or the retrieved literature supplies an individual carrying this variant with an alternate established diagnosis; León-Castillo et al. 2020 is a somatic tumour-classification study and reports no such case. Because the required case-level input is absent rather than negative, BP5 is recorded as not_assessed.

Evidence considered

• The local POLE custom framework customizes only PM1, PS4, PP3 and BP4 (customized_criteria list); no CSPEC exists for POLE, so the generic ACMG/AMP BP5 definition was used.
• Case files searched for case-level molecular data: review_manifest.json, case_summary.json (including compact_evidence), evidence.json, and the phenotype/prevalence evidence bundle. None records other detected variants, a co-occurring pathogenic variant, or an alternate molecular diagnosis for the individual.
• Somatic evidence sources are negative for this variant (COSMIC: 'This variant has not previously been reported in somatic cancers (COSMIC)'; OncoKB: no variant-level classification), so no case series or tumour record supplying an alternate-molecular-basis observation was available.
• No PMIDs were retrieved for this case (literature_pass.json: total_pmids 0), so no publication describing an individual carrying this variant could be consulted.

Still missing

• No proband-level genetic data (other variants detected, co-occurring pathogenic variants, or an established alternate molecular diagnosis) is present in any case file.
• No POLE-specific BP5 specification exists in this case; the local custom framework customizes only PM1, PS4, PP3 and BP4, and no CSPEC was found.
• PATH-250-323.pdf and its supplement cover somatic endometrial-carcinoma classification and contain no individual carrying this variant with an alternate molecular basis for disease.