ATM c.5868C>T, p.(Leu1956=)

Reasoning

Scope settled first: NM_000051.4:c.5868C>T is synonymous (NP_000042.3:p.(Leu1956=)), i.e. a non-missense variant for which splicing is the primary plausible damaging mechanism, so BP4 must be taken from the splicing sub-path (SpliceAI) and never from REVEL; REVEL is in fact absent for this variant (prefetch found=false). The governing ATM VCEP v1.6 BP4 supporting rule is 'Missense: REVEL score <=.249; Splicing: No predicted impact via splicing (SpliceAI <=0.1)', and the lab's curator note in the VCEP materials confirms that for non-missense variants where splicing is the primary plausible mechanism (synonymous, intronic outside the canonical +/-1,2 donor/acceptor positions) the splicing sub-path is used exactly as the VCEP specifies; the override therefore does not change the outcome for this variant relative to the VCEP's literal text. SpliceAI Lookup returns a maximum delta score of 0.053 (DS_AL 0.053, DS_AG 0.014, DS_DG 0.002, DS_DL 0.000), with no predicted gain (DS_AG/DS_DG essentially zero) or loss of a donor/acceptor site; 0.053 is at or below the <=0.1 no-predicted-impact cutoff, so BP4 supporting is met. The independent Pangolin scores in the same lookup (SG 0.008, SL -0.026) concur. The declared VCEP file Suppl_TableS1_PMID 40580951.xlsx contains the exact c.5868C>T row but assigns no BP4 code, and its 'Functional' Classification is a DeepATM model extrapolation (DeepATM_predicted=Yes, Confidence 'Computational prediction') that is at most weak, non-independent corroboration of the benign direction; it was not used to establish BP4 or to raise its strength. The ATM VCEP BP7 rule explicitly states that BP4 'may also be applied' alongside BP7 for synonymous variants, so this BP4 call is made deliberately and is not double-counted as a second independent prediction. Strength is capped at Supporting because a single predictor family (SpliceAI) supplies the whole benign computational signal and no RNA data confirm the negative call.

Rule

ATM VCEP v1.6 BP4 Supporting: Splicing - no predicted impact via SpliceAI <=0.1 (silent/intronic variants); for a synonymous variant the splicing sub-path is the applicable BP4 mechanism, and the ATM VCEP expressly permits BP4 to be applied together with BP7.

Evidence considered

• Variant consequence settled before scoring: NM_000051.4:c.5868C>T is synonymous (NP_000042.3:p.(Leu1956=)), a non-missense change for which splicing is the primary plausible mechanism, so BP4 was evaluated only through the SpliceAI path and no missense predictor was used; prefetch confirms REVEL found=false.
• ClinGen HBOP ATM VCEP v1.6 BP4 supporting rule: 'Missense: REVEL score <=.249; Splicing: No predicted impact via splicing (SpliceAI <=0.1).' The VCEP instructions state SpliceAI is its sole splice predictor and that PP3 applies at SpliceAI >=0.2 while BP4 applies at SpliceAI <=0.1 (Walker et al. 2023; Jaganathan et al. 2019, Cell, PMID:30661751).
• Lab curator note in the ATM VCEP materials (BP4 missense/splicing policy): for BP4 evaluate exactly one applicable mechanism per variant; missense variants use only the REVEL sub-path, while non-missense variants for which splicing is the primary plausible mechanism (synonymous, intronic outside the canonical +/-1,2 donor/acceptor positions) use the splicing sub-path as the VCEP specifies. This variant is synonymous, so the splicing sub-path is the correct choice and the override does not change the outcome.
• SpliceAI Lookup result (source_registry key 'spliceai'; hg37, distance 500, basic annotation, masked) returned DS_AG 0.014, DS_AL 0.053, DS_DG 0.002, DS_DL 0.000, max_delta_score 0.053, spliceai_available true; the pre-fetched sentence rounds the maximum to 0.05. Both values are at or below the <=0.1 BP4 cutoff.
• Pangolin scores from the same lookup (SG 0.008, SL -0.026) independently show no predicted splice gain or loss.
• ATM VCEP v1.6 BP7 instructions: 'May also apply BP4 to achieve Likely Benign' for synonymous and deep intronic variants - the combination of BP4 with BP7 applied here is explicitly sanctioned by the governing specification.
• Suppl_TableS1_PMID 40580951.xlsx exact row for c.5868C>T: Synonymous, exon_39, GnomAD_all_AF 1.301636218709e-05, ClinVar 'Likely benign', Combined_score 0.15021156126614, Classification 'Functional', Confidence 'Computational prediction', DeepATM_predicted 'Yes'. Used only as weak non-independent corroboration of the benign direction; no BP4 code is pre-assigned and the row is not a direct functional measurement.

Still missing

• No pre-assigned BP4 code exists for this variant: Suppl_TableS1_PMID 40580951.xlsx was searched for 'c.5868C>T', 'p.Leu1956=', 'p.L1956=' and 'L1956' and returns the exact synonymous row (line 18732) with NA predictor columns and no BP4 assignment.
• No RNA/cDNA splice assay or minigene result is available to confirm that the SpliceAI negative call reflects true absence of a splice effect; BP4 therefore rests on a single in-silico evidence line and is held at Supporting strength.

Sources

Reasoning

BP7 is in scope because NM_000051.4:c.5868C>T is a synonymous variant (NP_000042.3:p.(Leu1956=); one-letter p.L1956=), changing no amino acid. The ATM VCEP v1.6 BP7 supporting rule is: 'Use for synonymous and deep intronic variants defined as further than (but not including) +7 and further than (but not including) -21 at donor and acceptor sites, respectively', i.e. the variant must be synonymous and lie outside the immediate donor/acceptor splice region. Position check: hg38 chr11:108310265 falls in exon 39 of the governing ATM SNV table (Suppl_TableS1_PMID 40580951.xlsx assigns exon_39 to hg38 108310155-108310320, a 166-nt exon whose every position carries three SNV rows), placing the variant 110 nucleotides 3' of the intron 38/exon 39 acceptor boundary and 55 nucleotides 5' of the exon 39/intron 39 donor boundary - far outside the +7 donor / -21 acceptor window. Splice prediction agrees: SpliceAI Lookup gives a maximum delta score of 0.053 (DS_AL 0.053, DS_AG 0.014, DS_DG 0.002, DS_DL 0.000), below the VCEP's <=0.1 no-predicted-impact cutoff, so neither the native donor/acceptor nor a new cryptic site is predicted to be affected; there is no predicted impact to the splice consensus sequence and no predicted creation of a new splice site. The deeper-intronic alternative reading of this rule is also satisfied (the variant is exonic, well away from any splice junction), so BP7 supporting applies on either reading. No stronger BP7_Strong(RNA) or BP7_Moderate(RNA) level is assigned because no variant-specific RNA study demonstrating absence of an aberrant splice defect exists in this case. The declared VCEP table assigns no BP7 code for this variant; it confirms the synonymous annotation used here.

Rule

ATM VCEP v1.6 BP7 Supporting: use for synonymous and deep intronic variants defined as further than (but not including) +7 at the donor site and -21 at the acceptor site; BP7(RNA) at Strong/Moderate strength separately requires an observed lack of an aberrant RNA defect.

Evidence considered

• NM_000051.4:c.5868C>T normalises to NP_000042.3:p.(Leu1956=) (p.L1956=) - a synonymous, non-protein-changing change - which brings it inside BP7's scope.
• ClinGen HBOP ATM VCEP v1.6 BP7 supporting rule: 'Use for synonymous and deep intronic variants defined as further than (but not including) +7 and further than (but not including) -21 at donor and acceptor sites, respectively', with the VCEP's own instructions adding that BP7 'is not considered a conflicting piece of evidence against a body of evidence supporting a pathogenic splice defect' and that BP4 may be applied alongside it.
• Exon 39 in the governing ATM table (Suppl_TableS1_PMID 40580951.xlsx) spans hg38 chr11:108310155-108310320 (166 positions, 498 SNV rows, three per position, confirming the table is complete across the exon); the variant at hg38 chr11:108310265 is therefore exon position 111, i.e. 110 nt from the acceptor boundary and 55 nt from the donor boundary, well outside the +7/-21 splice region.
• SpliceAI Lookup result (source_registry key 'spliceai') returned DS_AG 0.014, DS_AL 0.053, DS_DG 0.002, DS_DL 0.000, max_delta_score 0.053, meeting the VCEP's <=0.1 no-predicted-impact expectation; no gain or loss of a donor/acceptor site is predicted. Pangolin SG 0.008 / SL -0.026 in the same lookup concur.
• Suppl_TableS1_PMID 40580951.xlsx row for c.5868C>T confirms the variant is Synonymous (protein L1956L, exon_39) but pre-assigns no BP7 (or PP3/BP4) code; its Classification 'Functional' with Confidence 'Computational prediction' and DeepATM_predicted 'Yes' is a model extrapolation only.
• No variant-specific RNA evidence (RT-PCR/capillary electrophoresis/minigene) demonstrating lack of an aberrant splice defect was found in the case sources, so no BP7_Strong(RNA)/BP7_Moderate(RNA) upgrade was possible.

Still missing

• Nucleotide-level conservation (e.g. phyloP/phastCons/GERP) was not supplied by any case source, so the conservation clause of the generic ACMG BP7 definition could not be checked directly; the governing ATM VCEP v1.6 BP7 rule is consequence- and position-based and does not require a separate conservation assessment, so BP7 was applied on the VCEP rule.
• No variant-specific RNA study is available, so BP7 is held at Supporting rather than the variable BP7(RNA) strengths.
• Suppl_TableS1_PMID 40580951.xlsx contains the exact c.5868C>T row but pre-assigns no BP7 code; exon 39 boundaries used for the position check were derived from the hg38 coordinates of that same governing VCEP table rather than from an independent transcript model.

Sources

Reasoning

PS1 is not met. The variant under assessment, NM_000051.4:c.5868C>T, is synonymous (NP_000042.3:p.(Leu1956=)) and therefore produces no amino-acid change, so the ATM VCEP standard PS1 sub-path - a different nucleotide change encoding the same amino-acid substitution as an established pathogenic/likely pathogenic variant - cannot be satisfied by definition. The alternative ATM PS1 splicing sub-path is also not satisfied. That table requires the variant under assessment to carry its own baseline computational/predictive splice code (PP3, or PVS1/PVS1_Strong/Moderate/Supporting at the donor/acceptor dinucleotide), with a predicted event that precisely matches the event of a known P/LP variant at the same position and at equal or greater prediction strength. SpliceAI for this variant gives a maximum delta of 0.053 (DS_AL 0.053, DS_AG 0.014, DS_DL 0.000, DS_DG 0.002) - below the VCEP's PP3 splice threshold of >=0.2 - so there is no predicted splice defect, no PP3 baseline code, and no entry point into the PS1 splicing table. Independently, no established P/LP variant exists at this residue: the VCEP's own SNV supplement (Sun et al. 2025, PMID 40580951, Table S1) records every alternative change at codon 1956 (L1956V, L1956I, L1956F, L1956R, L1956P, L1956H) as Uncertain significance or unclassified, none as pathogenic or likely pathogenic, and lists all three c.5868C> changes (L1956=) as Likely benign or unclassified. ACMG/AMP 2015 (PMID:25741868) additionally cautions that PS1's assumption of pathogenicity fails where a variant acts through the specific DNA change rather than through the amino-acid change - which is the only mechanism available to a synonymous change. No pre-assigned PS1 code governs this variant.

Rule

ATM PS1 (VCEP v1.5/v1.6): the missense sub-path requires a different nucleotide change giving the same amino-acid change as an established P/LP variant with splice defect ruled out for both; the splicing sub-path requires the variant under assessment's own PP3/PVS1 splice prediction code plus a predicted event that precisely matches a P/LP reference variant at the same position with equal or higher prediction strength.

Evidence considered

• ClinGen HBOP ATM VCEP v1.6 PS1: 'Missense: This rule may be applied only when a splice defect is ruled out for both alterations either by RNA analysis and/or in silico splice predictions'; 'Splicing: Use ATM PS1 Splicing table for splicing variants with similar predictions or observations of splice defect.'
• ATM PS1 Splicing table prerequisite (ATM_PS1_1.5.pdf, footnote *): the predicted event of the variant under assessment must precisely match the predicted event of the known (likely) pathogenic variant and the strength of the prediction for the variant under assessment must be of similar or higher strength; the reference variant must itself be classified under VCEP specifications.
• SpliceAI for NM_000051.4:c.5868C>T: max delta score 0.053 (DS_AL 0.053, DS_AG 0.014, DS_DG 0.002, DS_DL 0.0), i.e. no significant splice impact and no PP3-eligible splice prediction under the VCEP threshold of >=0.2.
• Normalized consequence is NP_000042.3:p.(Leu1956=) - a synonymous change with no amino-acid substitution, so there is no 'same amino acid change' to match against an established pathogenic variant.
• Sun et al. 2025 (PMID 40580951) Table S1 rows for residues 1956 and c.5868: all L1956 missense alternatives are 'Uncertain significance' or have no ClinVar classification, and c.5868C>A/G/T (all L1956=) are 'Likely benign' or unclassified - no established P/LP comparator at this codon.
• ACMG/AMP 2015 (PMID:25741868) PS1 caveat: the assumption of pathogenicity from a same-amino-acid-change variant fails if the variant acts through the specific DNA change rather than the amino acid change.

Still missing

• ATM_PS1_1.5.pdf (converted full text ATM_PS1_1.5.pdf.txt) was searched in full for c.5868C>T, p.Leu1956= and p.L1956= and contains no variant-specific entry; it is a generic PS1 decision table and no pre-assigned code governs this variant. The same was confirmed for ATM_PS1_1.6.pdf.txt.

Sources

Reasoning

The governing ATM VCEP PS3 specification assigns pathogenic functional evidence only when a variant fails to rescue an ATM-specific feature (e.g. phosphorylation of ATM substrates), with Supporting strength for failure of that ATM-specific feature alone and Moderate strength requiring failure of both an ATM-specific feature and radiosensitivity; Strong may not be used. The exact variant NM_000051.4:c.5868C>T (p.Leu1956=) is a synonymous change that does appear in Sun et al. Table S1 (PMID 40580951), but that row carries DeepATM_predicted=Yes and Confidence="Computational prediction", meaning it was not directly assayed and its "Functional" label is the trained model's own extrapolation rather than a functional measurement; the same row has CADD, boostDM, EVE, REVEL and AlphaMissense all NA, so it is not usable as functional evidence at any strength. The VCEP's own approved-assay calibration table (clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx) is methodology only and contains no per-variant entry for c.5868C>T / p.Leu1956= / p.L1956=, and none of the three papers it approves for ATM functional assays (Mitui 2009, PMID 18634022; Barone 2009, PMID 19431188; Scott 2002, PMID 11805335) was retrieved or reports this variant. With no direct, variant-specific assay showing failure to rescue an ATM-specific feature, PS3 is not met; flagged for human review because the only functional-labelled data point for this variant is an uncalibrated computational prediction. Splice-related RNA evidence is outside this criterion's scope for this group.

Rule

ATM VCEP v1.6 PS3: Supporting when a variant fails to rescue an ATM-specific feature (e.g. phosphorylation of ATM-specific targets) alone; Moderate when it fails to rescue both an ATM-specific feature and radiosensitivity; do not use Strong; radiosensitivity-only failure gets no weight.

Evidence considered

• The exact variant row in Sun et al. Table S1 is hg38 chr11:108310265 C>T, c.5868C>T, L1956L, exon_39, Variant_consequence Synonymous, with CADD.phred, boostDM_score, EVE_scores_ASM, REVEL and AlphaMissense all NA, GnomAD_all_AF 1.301636218709e-05, ClinVar_classification 'Likely benign', Combined_score 0.15021156126614, Classification 'Functional', Confidence 'Computational prediction', DeepATM_predicted 'Yes'.
• Per the governing instruction for that file, DeepATM_predicted=Yes with Confidence='Computational prediction' means the variant was NOT directly assayed; its Classification reflects the trained model's extrapolation and must not be presented as functional evidence, so it cannot support PS3 at any strength.
• The ATM VCEP PS3 rule text assigns Supporting when a variant fails to rescue an ATM-specific feature alone (e.g. phosphorylation of ATM-specific targets), Moderate when it fails to rescue both an ATM-specific feature and radiosensitivity, and states 'Do not use as strong'; radiosensitivity-only failure gets no weight because it is not specific to ATM deficiency.
• The VCEP approved-assay calibration file lists only Mitui 2009 (PMID:18634022), Barone 2009 (PMID:19431188) and Scott 2002 (PMID:11805335) for ATM kinase-activity assays (proposed strength PS3_Supporting/BS3_Supporting) and Mitui 2009 alone as an approved radiosensitivity assay (PS3_Moderate only when a kinase-activity PS3_Supporting is already met); searching its full converted text for c.5868C>T, p.Leu1956= and p.L1956= returned no variant-specific row.
• None of PMID:18634022, PMID:19431188 or PMID:11805335 is present in source_registry.json, fulltext_map.json or paper_extracts.json for this case, so no variant-specific result from an approved ATM functional assay is available.
• The five full texts retrieved for this case (PMID:25741868 ACMG/AMP 2015 standards; PMID:24366376, PMID:24366402, PMID:25394175, PMID:31429903 BRCA/breast-ovarian cancer risk-assessment guidelines) were searched for c.5868C>T, p.Leu1956= and p.L1956=; none mentions ATM or this variant and none reports a functional assay, so they contribute no PS3 evidence.
• The ClinVar submission audit for variation 453600 (7 submissions, all single-submitter, none expert-panel) lists no criterion leads, so no submitted functional-assay claim for this variant was available to route to PS3.

Still missing

• No entry for c.5868C>T, p.Leu1956=, or p.L1956= was found in clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx (searched in full); that file is an assay-methodology and strength-calibration table, not a per-variant results table.
• None of the three VCEP-approved assay papers (PMID:18634022 Mitui 2009, PMID:19431188 Barone 2009, PMID:11805335 Scott 2002) was retrieved for this case, so it is unknown whether this variant was ever tested in an approved kinase-activity or radiosensitivity assay; a variant-specific result in those papers is the missing evidence.
• The single variant-specific functional-labelled entry (Sun et al. 2025 Table S1) is a computational prediction (DeepATM_predicted=Yes), not a direct measurement, and is therefore not usable for PS3.

Sources

Reasoning

Governing specification: ClinGen Hereditary Breast, Ovarian and Pancreatic Cancer (HBOP) Expert Panel Specifications for ATM v1.6 (cspec), in which PS4 is Applicable at Pathogenic Strong with exactly one strength rule. Operator verification, quoted verbatim from the VCEP's own PS4 rule text (not inferred from another criterion in this group): 'Case-control studies; p-value ≤.05 **and** (Odds ratio, hazard ratio, or relative risk ≥2 **or** lower 95% CI ≥1.5).' This criterion strengthens as the ratio rises, so meeting it requires odds ratio / hazard ratio / relative risk ≥ 2 (or lower 95% CI ≥ 1.5) together with p ≤ 0.05. Explicit inequality check with the quantities that exist in this case: no odds ratio, hazard ratio, relative risk, or case-control p-value has been reported for ATM c.5868C>T, so there is no left-hand side to substitute and the required inequalities cannot be satisfied. The VCEP also explicitly disables the generic ACMG PS4 note-2 route: 'PS4_Moderate: Do not use. Proband counting for genes causing a common disorder need to be calibrated in a population-specific way before use.' No proband series for this variant was located either. All case-relevant literature retrieved for this variant (PMID:24366376, PMID:24366402, PMID:25394175, PMID:31429903) consists of BRCA1/BRCA2 risk-assessment and USPSTF screening recommendations or ACMG/NSGC referral criteria; the paper_extracts for each record variant_mentioned = false and none mentions ATM, so none supplies affected-versus-control data. ClinVar VCV000453600 holds seven germline submissions from clinical laboratories (five Likely benign, two Benign), all single-submitter, none a case-control or case-series report. The variant is rare in gnomAD v2.1 (9/251,084 alleles, AF 3.58e-05, grpmax FAF 1.53e-04), but rarity is not enrichment. PS4 is not met at any strength.

Rule

PS4 (Pathogenic Strong, cspec v1.6, Applicable): 'Case-control studies; p-value ≤.05 **and** (Odds ratio, hazard ratio, or relative risk ≥2 **or** lower 95% CI ≥1.5).' PS4_Moderate explicitly not to be used (proband counting not calibrated).

Evidence considered

• VCEP rule text for PS4 (cspec, ClinGen HBOP ATM v1.6, Pathogenic Strong, the only applicable strength): 'Case-control studies; p-value ≤.05 **and** (Odds ratio, hazard ratio, or relative risk ≥2 **or** lower 95% CI ≥1.5).' Operator as literally written for PS4 is '≥2' / '≥1.5' (strengthens as the value goes up); explicit check: no OR/HR/RR or case-control p-value exists for c.5868C>T, so the inequality has no substituend and cannot be met.
• VCEP instructionsToUse for PS4: 'PS4_Moderate: Do not use. Proband counting for genes causing a common disorder need to be calibrated in a population-specific way before use.' This removes the generic ACMG PS4 note-2 proband-counting path, leaving case-control statistics as the only route.
• ClinVar VCV000453600 (clinvar): 7 submissions, all from clinical testing laboratories and all with review status 'criteria provided, single submitter' (5 Likely benign, 2 Benign); conditions listed include ataxia-telangiectasia syndrome, familial cancer of breast, hereditary cancer-predisposing syndrome and 'not specified'. No case-control or case-series data; nothing reports affected-versus-control enrichment.
• paper_extracts for PMID:24366376, PMID:24366402, PMID:25394175 and PMID:31429903: each is a BRCA1/BRCA2 or USPSTF risk-assessment/screening recommendation or an ACMG/NSGC referral-criteria guideline; every record has variant_mentioned = false and none discusses ATM. No case-control enrichment estimate for c.5868C>T.
• gnomAD v2.1 (gnomad_v2): 9/251,084 alleles, AF 3.584e-05, grpmax FAF 1.53e-04, 0 homozygotes. The variant is rare, but no affected-versus-control comparison accompanies this frequency, so it provides no PS4 enrichment.

Still missing

• No case-control study and no case series reporting ATM c.5868C>T in affected versus control individuals was located in any retrieved source.
• PMID:26389258 (PDQ Cancer Genetics Risk Assessment and Counseling) was flagged as a PS4/PP5 candidate but only a screenshot was retrieved for this case, so its text could not be checked; it is a general PDQ summary and would not be expected to carry variant-level enrichment data.

Sources

Reasoning

Not met. The ClinGen HBOP ATM VCEP v1.6 PM2 rule requires a frequency of <=0.001% (1e-5) in the gnomAD subpopulation with the highest frequency, with a single narrow exception for a variant that exceeds 0.001% only as a single allele (n=1) in one subpopulation and is absent from all others. The variant's highest-subpopulation frequencies are 0.00029406 (South Asian, 9/30606) in gnomAD v2.1 all-comers exomes, 0.00029493 (South Asian, 9/30516) in v2.1 non-cancer exomes, and 0.000186694 (South Asian, 17/91058) in v4.1 - approximately 19- to 30-fold above the <=0.001% threshold. The n=1 exception plainly does not apply: the South Asian cluster comprises 9 alleles (v2.1) and 17 alleles (v4.1), and in v4.1 the variant is additionally present in the 'remaining' population (4 alleles), so it is not confined to a single subpopulation as a lone observation. Cohort-level frequencies are also above 1e-5 (v2.1 total AF 3.58446e-05; v4.1 total AF 1.30164e-05). gnomAD v3.1 non-cancer genomes carry 1/147740 alleles with a grpmax filtering AF of 0, and gnomAD-Canada v1.0 does not report the variant; neither offsets the high-frequency subpopulation finding. This is a genuinely uncommon but not ultra-rare allele at the VCEP's PM2 bar, so pathogenic rarity support is not awarded.

Rule

ClinGen HBOP ATM VCEP v1.6 PM2: Frequency <=.001% in gnomAD subpopulation with the highest frequency (pathogenic supporting); exception only when >0.001% but n=1 in a single subpopulation.

Evidence considered

• ClinGen HBOP ATM VCEP v1.6 PM2 rule as published in the cspec: 'Frequency <=.001% in gnomAD subpopulation with the highest frequency', with the stated exception that '>0.001% but n=1 in a single subpopulation (not present in any other subpopulation)' is sufficiently rare for PM2_supporting. This gene-specific threshold governs over the generic SVI PM2 supporting default of <=0.0001.
• gnomAD v2.1 all-comers exomes: highest subpopulation South Asian AF 0.00029406 (9/30606); total AF 3.58446e-05 (9/251084); 0 homozygotes; grpmax filtering AF 0.0001526.
• gnomAD v2.1 non-cancer exomes: highest subpopulation South Asian AF 0.00029493 (9/30516); total AF 3.80463e-05 (9/236554); 0 homozygotes.
• gnomAD v4.1: highest subpopulation South Asian AF 0.000186694 (17/91058); also present in the 'remaining' population (4/62478 alleles = 6.40225e-05), i.e. not confined to a single subpopulation; total AF 1.30164e-05 (21/1613354); 0 homozygotes.
• gnomAD v3.1 non-cancer genomes: 1/147740 alleles (AF 6.76865e-06), grpmax filtering AF 0; the single allele lies in the 204-allele 1KG:STU founder subpopulation (crude 0.49%), which is a small-sample artifact and not evidence of ultra-rarity or of subpopulation frequency at the level of the VCEP metric.
• gnomAD-Canada v1.0 (HostSeq genomes, GRCh38 chr11-108310265-C-T): variant absent (search_status 'absent'), so this additional population source provides no rarity support.
• Ancestry distribution: the variant is essentially South-Asian-restricted (0/9 non-SAS alleles in v2.1; 4 alleles in v4.1 'remaining', 0 in AFR/AMR/ASJ/EAS/NFE/FIN), consistent with a rare population-enriched synonymous allele rather than an ultra-rare pathogenic one.
• No retrieved publication for this case mentions ATM or c.5868C>T (p.Leu1956=), so no bibliographic source contributes to this criterion.

Sources

Reasoning

Governing specification: ClinGen HBOP ATM VCEP (cspec v1.6; declared file ATM_PM3_BP2_1.5.pdf, points table identical in ATM_PM3_BP2_1.6.pdf). PM3 scores points per unrelated Ataxia-Telangiectasia (A-T) proband carrying c.5868C>T: 4.0 with a confident phenotype plus a co-occurring P/LP ATM variant confirmed in trans, 2.0 phase unknown, 1.0 second variant unidentified or VUS, 2.0 homozygous (max 2 individuals), with half those values for a 'consistent' phenotype; PM3_Supporting requires >= 1.0 point. Two independent checks both fail. (1) Frequency gate: the VCEP general consideration states the variant 'may not exceed general population frequency >0.01%'. gnomAD v4.1 gives AC 21/1613354 (AF 1.30e-05) with grpmax filtering AF 0.011841% (South Asian popmax AF 0.018669%), and gnomAD v2.1 gives 9/251084 (AF 3.58e-05) with grpmax FAF 0.015260% (South Asian popmax 0.029406%); both grpmax values exceed the 0.01% ceiling. (2) Proband evidence: zero unrelated A-T probands carrying c.5868C>T were identified in any assembled source. No source reports this variant in an A-T-affected individual in trans, in cis, or homozygous. 0 points < 1.0 point threshold, so PM3 is not met. Note this is a negative by absence of proband data, not a positive benign observation, and the variant is a synonymous change (p.Leu1956=) for which the only VCEP-corpus hit is a functional-screen row (Suppl_TableS1_PMID 40580951.xlsx: exon 39, Synonymous, DeepATM_predicted=Yes, Confidence 'Computational prediction'), not a case observation.

Rule

See ATM PM3/BP2 table for approach to assign points per proband and final PM3 code assignment based on the sum of PM3-related points: PM3_Supporting = 1 point; PM3 = 2 points; PM3_Strong = 4 points; PM3_VeryStrong >= 8 points.

Evidence considered

• Governing specification (cspec v1.6 / ATM_PM3_BP2_1.5.pdf, 'Ataxia Telangiectasia PM3|BP2 table'): PM3 points per unrelated A-T proband are 4.0 (phenotype confident, confirmed in trans), 2.0 (phase unknown), 1.0 (second variant unidentified or VUS), 2.0 (homozygous, max 2 individuals), with half those values for a 'consistent' phenotype; conversion is PM3_Supporting = 1.0 point, PM3 (Moderate) = 2.0, PM3_Strong = 4.0, PM3_VeryStrong >= 8.0. General considerations state 'Variant may not exceed general population frequency >0.01%' and 'Multiple unrelated cases are additive'; footnote 1 requires that the VUA 'must be sufficiently rare (not meeting a benign population evidence code)', and footnote 2 requires that any co-occurring P or LP variant be classified under VCEP specifications rather than assumed.
• Frequency gate applied to this variant: gnomAD v4.1 reports AC 21/1613354 (AF 1.30164e-05; MAF 0.00130%) with grpmax filtering AF 0.00011841 (0.011841%) and highest population frequency in South Asians (17/91058, AF 0.018669%), homozygotes = 0; gnomAD v2.1 reports 9/251084 (AF 3.58446e-05; MAF 0.00358%) with grpmax FAF 0.0001526 (0.015260%) and South Asian AF 0.00029406 (0.029406%), homozygotes = 0; gnomAD-Canada v1.0 reports the variant absent. Both grpmax filtering AF values exceed the VCEP's 0.01% PM3 frequency ceiling, while neither approaches the ATM CSPEC benign frequency codes (BS1: grpmax filtering AF > 0.05%; BA1: grpmax filtering AF > 0.5%), so the variant is 'sufficiently rare' in the benign-code sense but above the PM3 ceiling.
• Search of the declared VCEP full-text corpus (/vcep_db/ATM/fulltext) for 'c.5868C>T', 'p.Leu1956=', 'p.L1956=', '5868' and '1956': the only hit for this exact variant is Suppl_TableS1_PMID 40580951.xlsx line 18732 — 108310265, C>T, c.5868C>T, L1956L, exon_39, Synonymous, GnomAD_AF 1.301636218709e-05, ClinVar_classification 'Likely benign', Classification 'Functional', Confidence 'Computational prediction', DeepATM_predicted 'Yes' — i.e. a model-extrapolated functional-screen row, not a proband with A-T. ATM_PM3_BP2_1.5.pdf, ATM_PM3_BP2_1.6.pdf and clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx contain no occurrence of this variant at all.
• Proband/phase search across the assembled case evidence: no source in the case bundle (prefetch, evidence, literature_pass, paper_extracts, ClinVar submission audit) records c.5868C>T in an A-T-affected proband, in trans or in cis with a P/LP ATM variant, or homozygous. publications.json reports _variant_mentioned_count = 0 across the retrieved corpus (terms searched included 'NP_000042.3:p.L1956=' and 'c.5868C>T'), and all 5 extracted full texts (PMID:25741868, PMID:24366376, PMID:24366402, PMID:25394175, PMID:31429903) were searched for 'c.5868C>T', 'p.Leu1956=', 'p.L1956=', 'L1956', '5868' and 'ATM' with no mention of the variant. With no qualifying proband the PM3 point total is 0, below the 1.0-point PM3_Supporting threshold.
• ClinVar (via the case's clinvar_submission_audit.json): 7 clinical-laboratory submissions for NM_000051.4(ATM):c.5868C>T (p.Leu1956=) — 5x Likely benign and 2x Benign, all 'criteria provided, single submitter', none expert-panel; no submission reports in-trans co-occurrence, phase, or proband phenotype information usable for PM3, and no submission's PubMed trail resolved.

Still missing

• ATM_PM3_BP2_1.5.pdf (searched for 'c.5868C>T', 'p.Leu1956=', 'p.L1956=', '5868', '1956'; and identically in ATM_PM3_BP2_1.6.pdf) contains only the generic ATM PM3/BP2 A-T points table — no pre-assigned code, row, or variant-specific entry for c.5868C>T (p.Leu1956=) exists in either file. There is therefore no governing pre-assigned PM3 code for this variant in the declared VCEP file.
• No unrelated A-T-affected proband carrying c.5868C>T (in trans, phase-unknown, or homozygous) was found in any source assembled for this case, so the PM3 point total is 0 by absence of proband reports rather than by a documented non-co-occurrence. The case literature pass retrieved 7 PMIDs (PubMed/ClinVar-sourced) and full text was obtained for 5, all BRCA-focused USPSTF/ACMG referral or guideline documents; no ATM A-T case-series paper was retrieved for this variant, so an exhaustive search of the A-T case literature was not performed within this case.
• Supplementary tables declared for the allelic group: clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx was searched for this variant and contains no variant-level rows (it calibrates approved functional assays only); Suppl_TableS1_PMID 40580951.xlsx contains this variant but only as a functional-screen row, not a PM3/BP2 proband record.

Sources

Reasoning

PM5 is not met. The ClinGen HBOP ATM VCEP v1.6 redefines PM5 and restricts its only applicable strength, PM5_Supporting, to a narrow variant class: 'DNA Variants: Apply to all NMD-prone, truncating variants receiving PVS1 (must be at VeryStrong) with premature termination codons upstream of p.Leu3048' and, for splice variants, only observed (not predicted) NMD-prone truncating variants receiving PVS1(RNA) at Very Strong with high-quality RNA data. The same specification instructs 'Do not use for missense changes'. NM_000051.4:c.5868C>T is a synonymous change, NP_000042.3:p.(Leu1956=): it introduces no premature termination codon, has no PVS1 evidence, and no truncating consequence, so none of the criterion's trigger conditions is satisfied. Independently, the classic same-residue comparator is also absent: the VCEP's SNV supplement (Sun et al. 2025, PMID 40580951, Table S1) lists all alternative changes at codon 1956 (L1956V, L1956I, L1956F, L1956R, L1956P, L1956H) as Uncertain significance or unclassified, none as pathogenic or likely pathogenic, and the case's PM5 comparator collection (pm5_candidates.json) recorded zero same-residue candidates with no eligible comparator search configured for this framework.

Rule

ATM PM5 is restricted to NMD-prone truncating variants receiving PVS1/VPS1(RNA) at Very Strong with a premature termination codon upstream of p.Leu3048, applied at Supporting strength; missense changes are excluded.

Evidence considered

• ClinGen HBOP ATM VCEP v1.6 PM5 rule (only applicable strength PM5_Supporting): 'DNA Variants: Apply to all NMD-prone, truncating variants receiving PVS1 (must be at VeryStrong) with premature termination codons upstream of p.Leu3048' and for splice variants only observed NMD-prone truncating variants receiving PVS1(RNA) at Very Strong with high-quality RNA data.
• The same specification's PM5 instructions state 'Do not use for missense changes: Multiple amino acid substitutions at the same residue can be pathogenic or benign and bioinformatic tools cannot yet confidently distinguish them'.
• The variant is synonymous, NP_000042.3:p.(Leu1956=), with no premature termination codon and no truncating consequence, so it cannot receive PVS1 at Very Strong.
• Sun et al. 2025 (PMID 40580951) Table S1 records L1956V/I/F/R/P/H as 'Uncertain significance' or unclassified and c.5868C>A/G/T (L1956=) as 'Likely benign' or unclassified; no P/LP variant at codon 1956 exists to serve as a comparator.
• The case PM5 comparator collection (pm5_candidates.json) reports found=false, eligible_for_classic_pm5_search=false, zero candidates seen, and recommendation 'not_applicable' with the note that ATM PM5 is a truncation-cutoff framework rule rather than classic same-residue missense PM5.

Still missing

• No variant-specific PM5 entry exists for c.5868C>T, p.Leu1956= or p.L1956= in the ATM VCEP directory; the governing PM5 rule is codon-class based (premature termination codon upstream of p.Leu3048), not variant-specific.
• No ClinVar P/LP variant at residue 1956 was identified among the collected comparators; the case's PM5 candidate collection returned zero same-residue candidates because this framework's PM5 operates on truncation position rather than same-residue missense comparators.

Sources

Reasoning

Scope settled first: NM_000051.4:c.5868C>T is a synonymous change (NP_000042.3:p.(Leu1956=)) in ATM exon 39, so the single applicable computational path for PP3 is the splice path (SpliceAI), not the missense path (REVEL). REVEL has no result for this variant (prefetch reports revel found=false) and would in any case be the wrong path for a synonymous variant, so it was not consulted. The governing ClinGen HBOP ATM VCEP v1.6 PP3 supporting rule is: 'Missense: REVEL >.7333; Splicing: Predicted impact via splicing (SpliceAI >=0.2) for silent, missense/in-frame and for intronic variants outside of donor and acceptor 1,2 sites', with SpliceAI used as the VCEP's sole splice predictor. The SpliceAI Lookup result for this variant gives DS_AG 0.014, DS_AL 0.053, DS_DG 0.002, DS_DL 0.000 and a maximum delta score of 0.053 (the pre-fetched summary sentence rounds this to 0.05). That maximum is below the ATM VCEP PP3 supporting threshold of >=0.2 (and below the generic SpliceAI moderate cutoff of >=0.5), so no computational evidence of a deleterious splice effect exists and PP3 is not met. The declared VCEP file Suppl_TableS1_PMID 40580951.xlsx does contain this exact variant (row: 108310265, C, T, c.5868C>T, L1956L, exon_39, Synonymous, GnomAD_all_AF 1.301636218709e-05, ClinVar 'Likely benign', Combined_score 0.15021156126614, Classification 'Functional', Confidence 'Computational prediction', DeepATM_predicted 'Yes'), but for synonymous rows the CADD, boostDM, EVE, REVEL and AlphaMissense columns are all NA and the file assigns no PP3 code; its 'Functional' call is a DeepATM model extrapolation (DeepATM_predicted=Yes, Confidence='Computational prediction'), i.e. the variant was not directly assayed, so it is at most weak secondary computational corroboration in the benign direction and cannot be inverted into PP3 support. PP3 is therefore not met.

Rule

ATM VCEP v1.6 PP3 Supporting: Splicing - predicted impact via SpliceAI >=0.2 for silent, missense/in-frame and intronic variants outside donor and acceptor 1,2 sites (SpliceAI is the VCEP's sole splice predictor; Jaganathan et al. 2019, Cell, PMID:30661751).

Evidence considered

• Variant consequence inspected before any predictor score: NM_000051.4:c.5868C>T normalises to NP_000042.3:p.(Leu1956=) (three-letter p.Leu1956=, one-letter p.L1956=) - a synonymous, non-protein-changing variant in exon 39 - so only the SpliceAI path was evaluated for PP3 and REVEL was deliberately not used.
• ClinGen HBOP ATM VCEP v1.6 (cspec, version 1.6) PP3 supporting rule: 'Missense: REVEL >.7333; Splicing: Predicted impact via splicing (SpliceAI >=0.2) for silent, missense/in-frame and for intronic variants outside of donor and acceptor 1,2 sites.' The VCEP's own instructions state it uses SpliceAI as a sole predictor and recommends the SVI thresholds PP3 for SpliceAI >=0.2 and BP4 for SpliceAI <=0.1 (Walker et al. 2023; SpliceAI source publication Jaganathan et al. 2019, Cell, PMID:30661751).
• SpliceAI Lookup result (source_registry key 'spliceai'; hg37, distance 500, basic annotation, masked; variant_id 11-108180992-C-T) returned DS_AG 0.014, DS_AL 0.053, DS_DG 0.002, DS_DL 0.000, with max_delta_score 0.053; the pre-fetched evidence sentence rounds this to 'max delta score = 0.05'. Both values sit well below the >=0.2 PP3 supporting cutoff.
• The same lookup reports Pangolin SG 0.008 and Pangolin SL -0.026, concordant with no gain or loss of splice signal; Pangolin was used only as a directional consistency check because no ACMG-strength Pangolin calibration is published.
• The declared governing VCEP file Suppl_TableS1_PMID 40580951.xlsx (source_registry key 'vcep_suppl_tables1_pmid_40580951') was searched for 'c.5868C>T', 'p.Leu1956=', 'p.L1956=' and 'L1956'; the exact row is present (line 18732) as a Synonymous exon_39 entry. Its CADD.phred, boostDM_score, EVE_scores_ASM, REVEL and AlphaMissense fields are all NA and it carries no PP3/BP4 code.
• That row is explicitly a model call, not a measurement: DeepATM_predicted = Yes and Confidence = 'Computational prediction' (Classification 'Functional', Combined_score 0.15021156126614). Per the file's declared use it cannot be presented as functional evidence, and as computational corroboration it is secondary only - it points in the benign direction and therefore does not support PP3.
• REVEL: prefetch reports found=false for this variant, and REVEL is not the applicable path for a synonymous variant in any case.

Still missing

• No pre-assigned PP3 code exists for this variant: the only per-variant governing table, Suppl_TableS1_PMID 40580951.xlsx, contains the exact c.5868C>T row but pre-assigns no PP3/BP4 code (its predictor columns are NA for this synonymous row), and the remaining declared VCEP files (ATM_PVS1_1.5/1.6, ATM_PS1_1.5/1.6, ATM_PM3_BP2_1.5/1.6, clingen_hbop_atm_supplementary_tables_1_and_2_v1) are PVS1/PS1/PM3-BP2 decision guides and an assay-calibration table, not per-variant computational lookup tables.
• No RNA/cDNA splice assay or minigene result is available to confirm the SpliceAI negative prediction; such data are not required for PP3 here but would raise confidence.

Sources

Reasoning

ClinGen HBOP ATM VCEP v1.6 (cspec) lists PP5 as 'Not Applicable for this VCEP' with the note 'This criterion is not for use as recommended by the ClinGen Sequence Variant Interpretation VCEP Review Committee.' Applying this group's governing rule, PP5 may be triggered only by an exact-variant ClinVar expert-panel classification of Pathogenic or Likely pathogenic; ordinary laboratory submissions, aggregate ClinVar labels and non-expert assertions never qualify. ClinVar VCV000453600 (NM_000051.4(ATM):c.5868C>T, p.Leu1956=) has seven germline submissions, all from clinical testing laboratories (Color Diagnostics, Ambry, Labcorp/Invitae, Fulgent, Myriad, Rigshospitalet, KCCC), each with review status 'criteria provided, single submitter'. The submission audit records expert_panel_submissions = 0, the ClinVar record carries no expert-panel entry, and its review status is 'criteria provided, multiple submitters, no conflicts' (2 stars, not 'reviewed by expert panel'). No exact-variant expert-panel assertion exists, and the only aggregate signal points the opposite way (Benign/Likely benign). The PP5 trigger condition is therefore absent; PP5 is not met. Note also that this is a synonymous change (p.Leu1956=) with no reported functional or case-level support that could substitute for an expert assertion.

Rule

PP5: Not Applicable for this VCEP (cspec v1.6, per ClinGen SVI VCEP Review Committee). Lab rule applied: PP5 only for an exact-variant ClinVar expert-panel classification of Pathogenic/Likely pathogenic.

Evidence considered

• cspec (ClinGen HBOP ATM VCEP v1.6), PP5 applicability field: 'Not Applicable for this VCEP', instructionsToUse: 'This criterion is not for use as recommended by the ClinGen Sequence Variant Interpretation VCEP Review Committee.' No PP5 strength rules are defined.
• clinvar (VCV000453600, NM_000051.4(ATM):c.5868C>T p.Leu1956=): 7 submissions, all clinical testing laboratories, all review status 'criteria provided, single submitter'; aggregate label 'Benign/Likely benign'; API review status 'criteria provided, multiple submitters, no conflicts' (2 stars); the record's expert_panels field is empty. Zero expert-panel submissions = no PP5 trigger.
• ClinVar submission audit for this case: expert_panel_submissions = 0, criterion_lead_submissions = 0, usable_for_criterion_leads = 0 out of 7 submissions; the two PMIDs cited by submitters (25741868, 28492532) were rejected as unresolved and neither is an expert-panel classification.
• No literature source in this case reports the variant at all: paper_extracts for PMID:24366376, PMID:24366402, PMID:25394175, PMID:31429903 and PMID:25741868 all carry variant_mentioned = false, so there is no reputable-source pathogenic assertion to weigh even at supporting strength.

Still missing

• PP5 requires an exact-variant ClinVar expert-panel Pathogenic/Likely pathogenic classification; none exists for VCV000453600, so this criterion is unobtainable for this variant regardless of other evidence.

Sources

Reasoning

Not met. The governing specification is the ClinGen Hereditary Breast, Ovarian and Pancreatic Cancer (HBOP) Expert Panel ATM VCEP v1.6 (cspec doc 639508985), whose BA1 rule is 'Grpmax Filtering AF >.5% in gnomAD' - a stand-alone benign threshold of 0.5%. The highest grpmax filtering AF observed for ATM c.5868C>T (NM_000051.4, p.(Leu1956=)) in any dataset queried is 0.000153 (0.0153%, gnomAD v2.1 non-cancer exomes, South Asian); the all-comers figures are equally low: gnomAD v2.1 exomes grpmax FAF 0.0001526 (total AF 3.58446e-05, 9/251084 alleles, homozygotes 0) and gnomAD v4.1 joint grpmax FAF 0.00011841 (total AF 1.30164e-05, 21/1613354 alleles; exome 20/1461246, genome 1/152108, homozygotes 0). gnomAD v3.1 non-cancer genomes report a grpmax filtering AF of 0. Every value is roughly 33-fold or more below the 0.5% stand-alone threshold, so BA1 is not met under either the all-comers or the non-cancer/exome-only population sources. The variant does cluster by ancestry (all 9 v2.1 alleles and 17 of 21 v4.1 alleles are South Asian; v2.1 SAS AF 0.00029406, 9/30606; v4.1 SAS AF 0.000186694, 17/91058), but even that highest-frequency subpopulation sits ~170-fold below 0.5%. No homozygotes are reported in any dataset.

Rule

ClinGen HBOP ATM VCEP v1.6 BA1: Grpmax Filtering AF >.5% in gnomAD (stand-alone benign).

Evidence considered

• ClinGen HBOP ATM VCEP v1.6 BA1 rule as published in the cspec: 'Grpmax Filtering AF \>.5% in gnomAD' - applied in preference to the generic ACMG BA1 stand-alone default (0.05), because an applicable gene-specific specification exists.
• gnomAD v2.1 all-comers exomes (GRCh37 11-108180992-C-T): total AF 3.58446e-05 (9/251084), grpmax filtering AF 0.0001526, highest subpopulation South Asian AF 0.00029406 (9/30606), homozygotes 0.
• gnomAD v2.1 non-cancer exomes (same GRCh37 variant): AF 3.80463e-05 (9/236554), grpmax filtering AF 0.000153, South Asian AF 0.00029493 (9/30516), homozygotes 0 - indistinguishable from the all-comers figure.
• gnomAD v4.1 (GRCh38 chr11-108310265-C-T): total AF 1.30164e-05 (21/1613354; exome 20/1461246, genome 1/152108), joint grpmax filtering AF 0.00011841, South Asian AF 0.000186694 (17/91058), homozygotes 0.
• gnomAD v3.1 non-cancer genomes: 1/147740 alleles (AF 6.76865e-06), grpmax filtering AF reported as 0.
• The ATM VCEP does not name a non-cancer or exome-only population source for BA1/BS1/PM2/BS2, so the all-comers gnomAD v2.1/v4.1 entries are the default; the non-cancer subsets were also evaluated and yield the same conclusion at every threshold.
• No retrieved publication for this case mentions ATM or c.5868C>T (p.Leu1956=), so no bibliographic source contributes to this criterion.

Sources

Reasoning

Not met. The ClinGen HBOP ATM VCEP v1.6 BS1 rule is 'Grpmax Filtering AF >.05% in gnomAD'. All grpmax filtering AFs for ATM c.5868C>T fall well short: 0.000153 (0.0153%) in gnomAD v2.1 non-cancer exomes, 0.0001526 (0.01526%) in v2.1 all-comers exomes, 0.00011841 (0.011841%) in v4.1, and 0 in v3.1 non-cancer genomes - i.e. about 3.3-fold or more below the 0.05% threshold. The highest single-subpopulation frequency (South Asian, 9/30606 = 0.0294% in v2.1; 17/91058 = 0.0187% in v4.1) is likewise below 0.05%. One apparent near-miss deserves explicit discounting: gnomAD v3.1 non-cancer records this variant once in the 1KG:STU subpopulation, whose 204 alleles give a crude 0.49% frequency that would sit just under BA1 and above BS1 if read naively. That is a single-allele small-sample artifact in a tiny founder sample; the dataset's own grpmax filtering AF - the metric the VCEP actually specifies - is 0 for that dataset, so no BS1 support arises. BS1 is therefore not met from any queried gnomAD source.

Rule

ClinGen HBOP ATM VCEP v1.6 BS1: Grpmax Filtering AF >.05% in gnomAD (benign strong).

Evidence considered

• ClinGen HBOP ATM VCEP v1.6 BS1 rule as published in the cspec: 'Grpmax Filtering AF \>.05% in gnomAD' (benign strong) - the gene-specific ATM threshold, used in place of the generic SVI BS1 default of >=0.01 (Whiffin et al. 2017, PMID:28518168).
• gnomAD v2.1 all-comers exomes: grpmax filtering AF 0.0001526 (South Asian 9/30606 = 0.00029406); total AF 3.58446e-05 (9/251084).
• gnomAD v2.1 non-cancer exomes: grpmax filtering AF 0.000153 (South Asian 9/30516 = 0.00029493); total AF 3.80463e-05 (9/236554).
• gnomAD v4.1: joint grpmax filtering AF 0.00011841 (exome-only grpmax FAF 0.00011551); South Asian 17/91058 = 0.000186694; total AF 1.30164e-05 (21/1613354).
• gnomAD v3.1 non-cancer genomes: grpmax filtering AF 0; single carrier in the 204-allele 1KG:STU subpopulation (crude 0.49%) treated as a small-sample/threshold artifact rather than as subpopulation evidence, since the VCEP criterion is defined on grpmax filtering AF.
• Homozygote counts are 0 in gnomAD v2.1, v2.1 non-cancer, v4.1 and v3.1 non-cancer; the variant is absent from gnomAD-Canada v1.0.
• No retrieved publication for this case mentions ATM or c.5868C>T (p.Leu1956=), so no bibliographic source contributes to this criterion.

Sources

Reasoning

The governing ATM VCEP BS3 specification assigns benign functional evidence when a variant rescues an ATM-specific feature (e.g. phosphorylation of ATM-specific targets) or cellular radiosensitivity, with Supporting strength for rescue of either and Moderate strength for rescue of both. The exact variant NM_000051.4:c.5868C>T (p.Leu1956=) is a synonymous change whose only variant-specific functional-labelled entry is in Sun et al. Table S1 (PMID 40580951), where it is recorded as Classification 'Functional' but with DeepATM_predicted=Yes and Confidence='Computational prediction' - a model extrapolation, not a rescue measurement - so it cannot be read as an assay demonstrating preserved ATM function and cannot support BS3. The VCEP's approved-assay calibration file is methodology-only and contains no per-variant entry for this variant, and none of the three approved assay papers (PMID:18634022, PMID:19431188, PMID:11805335) was retrieved or reports this variant. With no direct variant-specific functional assay showing rescue of ATM function or radiosensitivity, BS3 is not met; flagged for human review because the closest data point is an uncalibrated computational prediction rather than a measurement.

Rule

ATM VCEP v1.6 BS3: Supporting when a variant rescues either an ATM-specific feature or radiosensitivity; Moderate when it rescues both; radiosensitivity alone or conflicting kinase/radiosensitivity results get no weight.

Evidence considered

• The exact variant row in Sun et al. Table S1 is hg38 chr11:108310265 C>T, c.5868C>T, L1956L, exon_39, Variant_consequence Synonymous, with Combined_score 0.15021156126614, Classification 'Functional', Confidence 'Computational prediction', DeepATM_predicted 'Yes', and GnomAD_all_AF 1.301636218709e-05.
• Because DeepATM_predicted=Yes, the 'Functional' label is the DeepATM model's own extrapolation rather than a prime-editing measurement of this variant, so it must not be treated as functional evidence for BS3 (at most weak computational corroboration for the in-silico group).
• The ATM VCEP BS3 rule text assigns Moderate strength 'when a variant rescues both an ATM specific feature (e.g. phosphorylation of ATM-specific targets) AND radiosensitivity', and Supporting strength 'when a variant rescues EITHER an ATM specific feature OR rescues radiosensitivity'.
• The VCEP calibration file was searched in full for c.5868C>T, p.Leu1956= and p.L1956= and contains no variant-specific entry; it lists approved kinase assays (Mitui 2009/PMID:18634022, Barone 2009/PMID:19431188, Scott 2002/PMID:11805335, proposed BS3_Supporting) and an approved radiosensitivity assay (Mitui 2009, BS3_Supporting addable only with a kinase result; no weight for radiosensitivity alone or for conflicting kinase/radiosensitivity results).
• The five retrieved full texts (PMID:25741868, PMID:24366376, PMID:24366402, PMID:25394175, PMID:31429903) were searched for the variant in all notations and contain no ATM variant data or functional assay result.
• The ClinVar submission audit for variation 453600 records 7 submissions (5 Likely benign, 2 Benign) with no criterion leads and no validated functional evidence, so no assay-based benign claim for this variant was available to route to BS3.

Still missing

• No entry for c.5868C>T, p.Leu1956=, or p.L1956= was found in clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx (searched in full); that file calibrates approved assay strength but reports no variant-specific results.
• No variant-specific BS3-eligible result exists from a VCEP-approved assay (Mitui 2009 PMID:18634022, Barone 2009 PMID:19431188, Scott 2002 PMID:11805335); none of those papers was retrieved and none is known to report this variant.
• The Table S1 'Functional' label for this variant derives from a computational prediction (DeepATM_predicted=Yes) and therefore cannot be used as BS3 support; human review is needed before any strength could be considered.

Sources

Reasoning

Governing specification: ClinGen HBOP ATM VCEP (cspec v1.6; declared file ATM_PM3_BP2_1.5.pdf, points table identical in ATM_PM3_BP2_1.6.pdf). In this recessive A-T framework BP2 is scored per unaffected (non-A-T) adult proband older than 18 years who carries c.5868C>T together with a P/LP ATM variant: -4.0 when confirmed in trans, -2.0 when phase is unknown, and -2.0 (laboratory setting) or -1.0 (database setting) when homozygous, maximum -2.0 for the homozygous branch. Conversion: BP2 = -1 point (Supporting), BP2_Moderate = -2 points, BP2_Strong <= -4 points. Every branch requires an actual observation, and none exists for this variant: no unaffected adult carrier in trans with a P/LP ATM variant is reported in any assembled source (the case's literature corpus has _variant_mentioned_count = 0 and ClinVar's 7 submissions provide no phase or co-occurrence data), and the homozygous/database branch is likewise empty because gnomAD reports 0 homozygotes in both v2.1 (9/251084 alleles) and v4.1 (21/1613354 alleles) — which footnote 4 would in any case exclude from BP2 points. Total 0 points, so the <= -1 point BP2 Supporting threshold is not reached.

Rule

See ATM PM3/BP2 table for approach to assign points per proband and final BP2 code assignment based on the sum of BP2-related points: BP2 = -1 point (Supporting); BP2_Moderate = -2 points; BP2_Strong <= -4 points. When assessing homozygous or in trans variants for possible downgrade in an unaffected individual, the individual should be 18 years or older with no evidence of A-T.

Evidence considered

• Governing specification (cspec v1.6 / ATM_PM3_BP2_1.5.pdf): BP2 is scored as points per unaffected (non-A-T) adult proband older than 18 years — -4.0 confirmed in trans with a P/LP variant, -2.0 phase unknown, -2.0 homozygous in a laboratory setting or -1.0 homozygous in a database setting (homozygous maximum -2.0) — converted as BP2 = -1 point (Supporting), BP2_Moderate = -2 points, BP2_Strong <= -4 points. The specification adds that when assessing homozygous or in trans variants for possible downgrade in an unaffected individual, 'the individual should be 18 years or older with no evidence of A-T' (footnote 3), and that a P or LP co-occurring variant must itself be classified under VCEP specifications rather than assumed. Footnote 4 confines the homozygous branch to phenotyped individuals from clinical or research cohorts and states that >= 2 gnomAD homozygotes are already captured by the BA1/BS1 frequency codes, so database homozygote observations should not be double-counted as BP2.
• Homozygous/database branch of BP2: gnomAD v4.1 reports 0 homozygotes for c.5868C>T (AC 21/1613354, AF 1.30164e-05, grpmax filtering AF 0.011841%, homozygotes = 0) and gnomAD v2.1 also reports 0 homozygotes (AC 9/251084, AF 3.58446e-05, grpmax FAF 0.015260%, homozygotes = 0); gnomAD-Canada v1.0 reports the variant absent. There is therefore no homozygous adult observation to score, and no >= 2-homozygote gnomAD observation exists for this variant.
• Trans/in-cis branch of BP2: no source assembled for this case records c.5868C>T in an unaffected (non-A-T) adult together with a P/LP ATM variant, in trans or in cis. publications.json reports _variant_mentioned_count = 0 for the retrieved corpus, and each of the 5 extracted full texts (PMID:25741868, PMID:24366376, PMID:24366402, PMID:25394175, PMID:31429903) was searched for 'c.5868C>T', 'p.Leu1956=', 'p.L1956=', 'L1956' and 'ATM' without any mention of the variant or of biallelic ATM carriers.
• ClinVar (via the case's clinvar_submission_audit.json): 7 single-submitter clinical-laboratory submissions for NM_000051.4(ATM):c.5868C>T (p.Leu1956=) — 5x Likely benign, 2x Benign, none from an expert panel — with no reported criteria signals and no usable data on phase or co-occurrence with a P/LP ATM variant (all PubMed trails unresolved). These submissions document benign classifications by other laboratories but supply no BP2 point-bearing observation, and per the group instructions another source's classification is not used to infer this criterion.
• Search of the declared VCEP full-text corpus for the variant tokens ('c.5868C>T', 'p.Leu1956=', 'p.L1956=', '5868', '1956') returned only the functional-screen row in Suppl_TableS1_PMID 40580951.xlsx (exon 39, L1956L, Synonymous, DeepATM_predicted = Yes, Confidence 'Computational prediction'); ATM_PM3_BP2_1.5.pdf, ATM_PM3_BP2_1.6.pdf and clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx contain no occurrence of this variant, so no pre-assigned BP2 code or BP2 calibration entry applies.

Still missing

• ATM_PM3_BP2_1.5.pdf (searched for 'c.5868C>T', 'p.Leu1956=', 'p.L1956=', '5868', '1956'; identical result in ATM_PM3_BP2_1.6.pdf) is the generic ATM PM3/BP2 A-T points table and carries no pre-assigned BP2 code or variant-specific row for this variant.
• No unaffected adult (>18 years, no evidence of A-T) carrying c.5868C>T with a P/LP ATM variant in trans was identified in any assembled source, and no such biallelic individual is documented as homozygous; the case literature pass retrieved no ATM/A-T cohort paper for this variant (7 PMIDs retrieved, 5 full texts obtained, all BRCA-focused guidelines). The BP2 result is therefore 0 points by absence of qualifying observations rather than by a documented co-occurrence in cis or in a healthy adult.

Sources

Reasoning

ClinGen HBOP ATM VCEP v1.6 (cspec) lists BP6 as 'Not Applicable for this VCEP' ('This criterion is not for use as recommended by the ClinGen SVI VCEP Review Committee'). Applying this group's governing rule, BP6 may be triggered only by an exact-variant ClinVar expert-panel classification of Benign or Likely benign; ordinary laboratory submissions, aggregate ClinVar labels and non-expert assertions must never trigger it. ClinVar VCV000453600 (NM_000051.4(ATM):c.5868C>T, p.Leu1956=) has seven submissions from clinical testing laboratories only (5 Likely benign, 2 Benign), every one with review status 'criteria provided, single submitter'; expert_panel_submissions = 0 and the record contains no expert-panel entry. The variant's aggregate ClinVar label, 'Benign/Likely benign' at 2 stars ('criteria provided, multiple submitters, no conflicts'), is precisely an aggregate of non-expert laboratory submissions and therefore does not satisfy the BP6 trigger, even though its direction (benign) matches the criterion's polarity. No exact-variant expert-panel benign classification exists, so BP6 is not met; the 2-star aggregate label was deliberately not converted into BP6 evidence.

Rule

BP6: Not Applicable for this VCEP (cspec v1.6, per ClinGen SVI VCEP Review Committee). Lab rule applied: BP6 only for an exact-variant ClinVar expert-panel classification of Benign/Likely benign.

Evidence considered

• cspec (ClinGen HBOP ATM VCEP v1.6), BP6 applicability field: 'Not Applicable for this VCEP', instructionsToUse: 'This criterion is not for use as recommended by the ClinGen Sequence Variant Interpretation VCEP Review Committee.' No BP6 strength rules are defined.
• clinvar (VCV000453600): 7 submissions, all from clinical testing laboratories, all review status 'criteria provided, single submitter' (Color Diagnostics, Ambry Genetics, Labcorp/Invitae, Fulgent Genetics: Likely benign; Myriad Genetics, KCCC/NGS Laboratory: Benign); ClinVar API review status 'criteria provided, multiple submitters, no conflicts' = 2 stars; expert_panels field empty. Aggregate 'Benign/Likely benign' label is an aggregate of non-expert submissions and is excluded by the BP6 rule; expert-panel submission count is 0, so BP6 is not triggered.
• ClinVar submission audit for this case: expert_panel_submissions = 0, validated_pmids = 0 (PMID:25741868 and PMID:28492532 rejected as 'pubmed_metadata_unresolved'), criterion_lead_submissions = 0 - none of the seven submissions carries an expert-panel assertion usable for BP6.
• The direction of the aggregate label (benign) was noted but not treated as BP6 evidence, because the governing rule requires an exact-variant expert-panel classification rather than a multi-laboratory aggregate.

Still missing

• No exact-variant ClinVar expert-panel Benign/Likely benign classification exists for VCV000453600, so BP6 cannot be triggered for this variant even though seven non-expert submissions report a benign direction.

Sources

Reasoning

PP1 (co-segregation with disease in multiple affected family members) is applicable for ATM under the governing ClinGen HBOP ATM VCEP v1.6, but only through its autosomal-recessive (ataxia-telangiectasia) path, with the prerequisite that affected relatives must have both variants identified in the proband; the VCEP explicitly instructs "AD Condition - Do not use: Co-segregation analysis in lower-penetrance genes can lead to false positive results (PMID 32773770)." The VCEP strength ladder for the AR condition is: segregation in 1 affected relative = PP1 Supporting, 2 affected relatives = PP1 Moderate, >=3 affected relatives = PP1 Strong. This case is a single-variant laboratory assessment of ATM c.5868C>T (p.Leu1956=), a synonymous change (NP_000042.3:p.(Leu1956=)). The consolidated case evidence contains no proband, no pedigree, no parental or relative genotypes, no affected relative count, no meioses and no phenotype-consistency statement: an exhaustive search of review_manifest.json, case_summary.json, source_registry.json, evidence.json, literature_pass.json, fulltext_map.json and paper_extracts.json returns zero occurrences of segregation, co-segregation, family, affected relative, meioses or proband-level data for this variant. The ClinVar record (variation 453600) holds 7 single-submitter entries (5 Likely benign, 2 Benign) with no expert-panel submission, and the submission audit reports 0 usable criterion leads, 0 validated PMIDs and 2 rejected PMIDs, so no segregation observation can be recovered from that route. With zero informative affected relatives available, PP1 can neither be met nor refuted; it is therefore not_assessed rather than not_met. A pedigree with genotype data for affected and unaffected relatives, in which both ATM alleles are characterised in the proband and the second (in trans) variant is itself classified under VCEP specifications, would be required to score PP1 at Supporting strength or above.

Rule

ClinGen HBOP ATM VCEP v1.6, PP1 (AR condition): PP1_Strong = segregation in >=3 affected relatives; PP1_Moderate = 2 affected relatives; PP1 (Supporting) = 1 affected relative; prerequisite - affected relatives must have both variants identified in the proband; AD co-segregation path not to be used (PMID 32773770).

Evidence considered

• Governing framework: ClinGen HBOP ATM VCEP specification v1.6 (cspec doc 639508985), where PP1 is Applicable with default strength Pathogenic Supporting.
• PP1 strength ladder per v1.6: PP1_Strong = "AR Condition: Segregation in >=3 affected relatives"; PP1_Moderate = "AR Condition: Segregation in 2 affected relatives"; PP1 (Supporting) = "AR Condition: Segregation in 1 affected relative".
• PP1 prerequisite text from the VCEP: "AR Condition: Affected relatives must have both variants identified in proband." and "AD Condition - Do not use: Co-segregation analysis in lower-penetrance genes can lead to false positive results (PMID 32773770)."
• Variant context: ATM NM_000051.4:c.5868C>T, NP_000042.3:p.(Leu1956=) / p.(L1956=), GRCh38 chr11:108310265C>T; a synonymous change, so no protein-level or splice mechanism is asserted by this group.
• Searched the consolidated case evidence for total-segregation or co-segregation content; there are no hits for segregation, co-segregation, family, affected relatives, meioses or proband-level data associated with this variant in review_manifest.json, case_summary.json, source_registry.json, evidence.json, literature_pass.json, fulltext_map.json or paper_extracts.json.
• ClinVar variation 453600: 7 submissions (5 Likely benign, 2 Benign), all "criteria provided, single submitter", no expert-panel submission; the submission audit records 0 criterion leads usable for criteria, 0 validated PMIDs and 2 rejected PMIDs, so no segregation claim is attributable to those submissions.
• All five retrieved full texts (PMID:25741868, PMID:24366376, PMID:24366402, PMID:25394175, PMID:31429903) were checked: none mentions ATM c.5868C>T, p.(Leu1956=), the ATM gene, or any family segregation data.

Still missing

• No proband, pedigree, parental testing, affected-relative genotypes, or family history of ataxia-telangiectasia or ATM-related cancer is present anywhere in the case evidence, so no informative meioses or co-segregation events exist to count.
• The VCEP prerequisite that affected relatives must have both ATM variants identified (with the second variant classified under VCEP specifications) cannot be verified, because no second variant or phasing data for the proband are provided.
• PP1 can only be applied through the AR (ataxia-telangiectasia) path; the AD co-segregation path is prohibited by the VCEP (PMID 32773770), so any future family data must include A-T-affected relatives with biallelic ATM genotypes.
• This variant is a synonymous change (p.Leu1956=) that ClinVar submitters report as Benign/Likely benign; absent family data there is no segregation dataset that could offset that.

Sources