ATM c.858A>G, p.(Gln286=)

Reasoning

Scope: NM_000051.4:c.858A>G is a synonymous change (NP_000042.3:p.(Gln286=)/p.(Q286=), codon 286 in exon 7; variant_consequence 'Synonymous' in the governing VCEP table). PP3 is therefore evaluated through the splice path only, using SpliceAI; the REVEL missense path is not applicable to a synonymous variant and no REVEL score exists for it. The ATM VCEP v1.6 specification assigns PP3 (default Pathogenic Supporting) for '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' - silent (synonymous) variants are expressly in scope. SpliceAI Lookup for NM_000051.4:c.858A>G (distance 500, mask 0) reports a maximum delta score of 0.208, driven by a predicted acceptor gain (DS_AG 0.208 at DP_AG +1, i.e. a cryptic acceptor created one nucleotide downstream of the variant); the other event deltas are negligible (DS_AL 0.018, DS_DG 0.001, DS_DL 0.002). Because 0.208 >= 0.2, the VCEP PP3 supporting threshold is met. Two qualifications are noted: the margin over the threshold is only 0.008, and Pangolin shows no corroborating signal (SG 0.083, SL -0.049); the code is nevertheless assigned because the VCEP rule is an explicit numerical cutoff. The VCEP caveat that PP3 for splice predictions must not be applied together with PVS1/PVS1_Variable(RNA) does not obstruct this code, since PVS1 was not applied to this variant (consequence_class synonymous, not a canonical +/-1,2 splice consensus change). The exact c.858A>G / Q286Q row in Suppl_TableS1_PMID 40580951 (line 2740) records CADD, boostDM, EVE, REVEL and AlphaMissense all as NA, so that table supplies no pre-calibrated PP3 value for this variant; its Classification/Combined_score/Confidence/DeepATM_predicted fields are functional-screen outputs, not a computational PP3 code, and are not used here.

Rule

ATM VCEP v1.6 PP3 Supporting (splice path): SpliceAI max delta >= 0.2 for silent variants.

Evidence considered

• Variant scope: NM_000051.4:c.858A>G is synonymous (NP_000042.3:p.(Gln286=)/p.(Q286=)), codon 286, exon 7, so PP3 takes the splice path (SpliceAI) and not the missense REVEL path.
• ATM VCEP v1.6 PP3 (Pathogenic Supporting): '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.'
• SpliceAI Lookup (NM_000051.4:c.858A>G, hg37 11-108115710-A-G, distance 500, mask 0) returned max delta score 0.208: DS_AG 0.208 at DP_AG +1 (cryptic acceptor gain one nucleotide downstream), DS_AL 0.018 at DP_AL -195, DS_DG 0.001 at DP_DG -108, DS_DL 0.002 at DP_DL +43.
• 0.208 >= 0.2, so the ATM VCEP PP3 supporting threshold is met; the margin is only 0.008 and Pangolin gives no corroborating signal (SG 0.083, SL -0.049).
• The VCEP restriction that PP3 for splice predictions may not be applied in addition to PVS1 or PVS1_Variable(RNA) does not apply here: this variant was not assigned PVS1 (consequence_class synonymous, canonical_splice_consensus false).
• The exact c.858A>G row in Suppl_TableS1_PMID 40580951 (converted full text line 2740) shows Q286Q, exon_7, Synonymous, with CADD, boostDM, EVE, REVEL and AlphaMissense all recorded as NA; that table therefore assigns no pre-calibrated PP3 code for this variant.

Sources

Reasoning

Governing specification: ClinGen HBOP ATM VCEP v1.6 (cspec), BS3 rules - BS3_Moderate when a variant rescues both an ATM-specific feature (e.g. phosphorylation of ATM-specific targets) AND radiosensitivity, BS3_Supporting when it rescues either. The VCEP's approved-assay calibration file (clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx) names only Mitui 2009 (PMID 18634022), Barone 2009 (PMID 19431188) and Scott 2002 (PMID 11805335), and contains no per-variant entries: searching it for c.858A>G, p.Gln286= and p.Q286= returned nothing, so no approved-assay benign call exists for this variant. The one direct functional measurement available is Sun et al. 2025 (PMID 40580951) Table S1: c.858A>G (p.Q286Q, synonymous, exon 7) has DeepATM_predicted=No - meaning the call derives from the prime-editing saturation genome-editing measurement itself, not from the DeepATM model - with Confidence=Medium-high, Combined_score 0.0240282777808958 and Classification=Functional. In the same table, directly-measured rows labelled Non-functional have Combined_score <= -1.360 while Functional rows are >= -0.912, so this variant sits squarely in the function-retaining band, i.e. the assay shows no damaging effect on ATM function. That is the benign direction and is the only BS3-supporting evidence, so it is assigned at the VCEP's lowest tier, BS3_Supporting (rescue of one ATM-related readout). Two caveats keep confidence low and require human review: (1) this assay is not one of the VCEP's three approved assays and the ATM VCEP has not calibrated a strength for it, so the strength assignment is provisional and the variant must be flagged; and (2) the same screen mislabels the adjacent truncating variant c.856C>T (p.Gln286*) as Functional (Combined_score -0.421), an obvious false-negative in the immediate region, and 81 nonsense variants overall are labelled Functional, so a Functional call at this codon carries reduced discriminative weight. If a reviewer declines to treat an uncalibrated assay as BS3 evidence, this criterion should fall back to not_assessed.

Evidence considered

• Governing specification consulted: ClinGen HBOP ATM VCEP v1.6 (cspec), BS3 rules - BS3_Moderate 'Use when a variant rescues both an ATM specific feature (e.g. phosphorylation of ATM-specific targets) AND radiosensitivity'; BS3_Supporting 'Use when a variant rescues EITHER an ATM specific feature OR rescues radiosensitivity'.
• Searched clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx for 'c.858A>G', 'p.Gln286=' and 'p.Q286=': no match. The file supplies only the approved-assay methodology and proposed strengths (kinase assays: PS3_Supporting/BS3_Supporting, approved for Mitui 2009, Barone 2009, Scott 2002; radiosensitivity: adds no weight on its own and none if it conflicts with the kinase result), plus validation-controls that name other variants, not c.858A>G.
• Searched Suppl_TableS1_PMID 40580951.xlsx full text for 'c.858A>G', '858A', 'p.Gln286=' and 'p.Q286=': one row matched - hg38 108244983 / A / G / c.858A>G / Q286Q / exon_7 / Synonymous / residue 286 / Combined_score 0.0240282777808958 / Classification Functional / Confidence Medium-high / DeepATM_predicted No. DeepATM_predicted=No marks this as a direct result of the prime-editing saturation genome-editing screen rather than a DeepATM model extrapolation, which is the condition the VCEP file sets for routing a result to PS3/BS3.
• Band boundaries read directly from the supplied table (not an external publication): among DeepATM_predicted=No rows, Non-functional rows span Combined_score <= -1.360, Intermediate -1.360 to -0.912, Functional >= -0.912. The variant's 0.024 is 1.4 points above the Functional/Intermediate boundary, i.e. clearly in the function-retaining band - a benign-direction result suited to BS3, not PS3.
• Assay-reliability check performed on the same table: 81 directly-assayed nonsense variants are labelled Functional, including the immediately adjacent c.856C>T (p.Gln286*, Combined_score -0.421, Confidence Medium-high, DeepATM_predicted=No), which is a truncating variant and cannot plausibly retain ATM function. The Functional call for c.858A>G therefore has limited discriminative power at this codon and supports only the lowest BS3 tier, with human review.
• Cross-check of the table's own Confidence field: Medium-high, and on the same row the computational predictor columns (CADD.phred, boostDM_score, EVE_scores_ASM, REVEL, AlphaMissense) are all 'NA' for this synonymous change, so no predictor score contributes to the functional call.
• The other papers carried in this group's bundle (PMID 25741868, PMID 26467025, PMID 17508274, PMID 24418350) contain no variant-specific functional data: the paper_extracts entries all record variant_mentioned=false for c.858A>G / p.Q286=, so none contributes functional evidence for or against BS3.

Still missing

• clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx was searched for 'c.858A>G', 'p.Gln286=' and 'p.Q286=' and has no entry for this variant - it reports only assay methodology, controls and strength calibration, so it provides no variant-level benign result and cannot confirm c.858A>G was tested in an approved assay.
• The three VCEP-approved assay papers (PMID 18634022, PMID 19431188, PMID 11805335) were not retrieved in this case (absent from fulltext_map.json, paper_extracts.json and publications/), so they could not be searched for c.858A>G at the paper level.
• The full text of PMID 40580951 was not fetched; only Table S1 was available. The specific endpoint measured by the prime-editing screen (ATM substrate phosphorylation versus radiosensitivity/viability) is therefore unverified, so it cannot be decided whether the result corresponds to BS3_Supporting (rescue of either readout) or BS3_Moderate (rescue of both) under the VCEP rules.
• The ATM VCEP has not calibrated a strength for the PMID 40580951 assay, which is not among its three approved assays; the BS3_Supporting call is provisional pending human review of assay acceptability.

Sources

Reasoning

The governing ClinGen HBOP ATM VCEP PS1 guide has two sub-paths. (1) Missense: PS1 requires the variant to encode the SAME amino acid change as a previously established P/LP variant (splice defect ruled out for both). c.858A>G is a synonymous substitution (NM_000051.4:c.858A>G -> NP_000042.3:p.(Gln286=)), so no amino-acid change is produced and the missense path cannot apply. (2) Splicing-region path (ATM PS1 Splicing table): the VUA's own baseline predictive code and its position relative to the donor/acceptor dinucleotide are compared against a reference P or LP variant at the same position, and the prerequisite is that the VUA's predicted splice event precisely matches the reference variant's predicted event with equal-or-higher prediction strength. SpliceAI max delta for this variant is 0.208 (DS_AG 0.208), which meets the VCEP's SpliceAI >=0.2 cutoff and therefore gives the VUA a PP3 baseline code, so the splicing path is the theoretically applicable route. However, c.858 sits deep within exon 7 (exon 7 spans c.663-901), well outside both the exon 7 donor motif (last 3 exonic bases c.899-901 plus 6 intronic nt) and the acceptor motif (exon start c.663 plus 20 upstream nt), and no established P/LP ATM variant is documented at the same nucleotide or within that donor/acceptor motif. At the same nucleotide position the only other changes are c.858A>C and c.858A>T, both encoding p.Gln286His (missense) and both scored function-neutral in the Sun et al. saturation screen; the only ClinVar record for c.858A>G itself (variation 185146) is conflicting benign/VUS, not P/LP. With no reference P or LP variant to anchor the comparison, the splicing path's prerequisite is unmet and PS1 cannot be assigned at any strength.

Rule

ATM PS1 (ClinGen HBOP ATM VCEP v1.6): missense path requires the same amino-acid change as an established P/LP variant; splicing path requires an established P/LP reference variant at the same position with a matching predicted splice event.

Evidence considered

• CSpec ATM v1.6 PS1 rule: 'Use for missense changes as long as splicing is ruled-out for both alterations. Use ATM PS1 Splicing table for splicing variants with similar predictions or observations of splice defect.' Default strength Pathogenic Strong.
• ATM_PS1_1.6 / ATM_PS1_1.5 splicing table: for a VUA located outside the donor/acceptor +/-1,2 dinucleotide with baseline code PP3, a same-nucleotide P reference gives PS1 and an LP reference gives PS1_Moderate; a reference within the same donor/acceptor motif gives PS1_Moderate/PS1_Supporting. Prerequisite for all rows: the VUA's predicted splice event must precisely match the known (likely) pathogenic variant's predicted event and the VUA's prediction strength must be equal to or higher.
• SpliceAI lookup (NM_000051.4:c.858A>G): max delta score 0.208 (DS_AG 0.208, DS_AL 0.018, DS_DG 0.001, DS_DL 0.002; DP_AG +1) -> satisfies the VCEP PP3 cutoff of SpliceAI >= 0.2 for the variant's baseline predictive code.
• c.858 is in exon 7 (c.663-901 of NM_000051.4), remote from both the exon 7 donor motif (c.899-901 + 6 intronic nt) and the acceptor motif (c.663 + 20 upstream nt), i.e. located outside the donor/acceptor +/-1,2 dinucleotide positions.
• ClinVar variation 185146 (c.858A>G) is conflicting (5x Likely benign, 4x Uncertain significance, 1x Benign); no expert-panel submission and no P/LP classification.
• Sun et al. 2025 (PMID 40580951) Table S1: c.858A>C and c.858A>T both give p.Gln286His and are classified functional (no P/LP variant at c.858 to serve as a reference).

Still missing

• The ATM PS1 Splicing table (ATM_PS1_1.5.pdf.txt / ATM_PS1_1.6.pdf.txt) is a categorical rule table and contains no per-variant entry for c.858A>G, p.Gln286=, or p.Q286=; searched all three notations (c.858A>G, p.Gln286=, p.Q286=) with no match.
• No established pathogenic or likely pathogenic ATM variant (classified under the VCEP specification) was identified at the same nucleotide (c.858) or within exon 7's donor/acceptor motif, so the PS1 splicing-path prerequisite cannot be satisfied from the collected evidence.

Sources

Reasoning

Governing specification: ClinGen HBOP ATM VCEP v1.6 (cspec, doc 639508985), which restricts PS3 to functional assays and states 'Do not use as strong' for ATM; PS3_Moderate requires the variant to fail to rescue BOTH an ATM-specific feature (e.g. phosphorylation of ATM targets) AND radiosensitivity, and PS3_Supporting requires failure to rescue an ATM-specific feature alone (VCEP text explicitly excludes radiosensitivity-only failure, as it is not specific to ATM deficiency). The VCEP's own approved-assay calibration file (clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx) names three approved assays - Mitui 2009 (PMID 18634022), Barone 2009 (PMID 19431188) and Scott 2002 (PMID 11805335), all stable transfection of ATM cDNA constructs into ATM-null LCLs with Western-blot readout of ATM substrate phosphorylation +/- radiosensitivity. Searching that file for c.858A>G, p.Gln286= and p.Q286= returned no entry: the file contains only assay methodology and strength calibration ('Proposed strength PS3_Supporting/BS3_Supporting'), with no per-variant results, and the validation-control lists it reports (e.g. c.7636del9, c.8147T>C/p.V2716A, c.8546G>C/p.R2849P, c.8599G>C/p.G2867R, c.7987delGTT/p.V2662del; c.5908C>T/p.Q1970*) do not include this variant, so it cannot be confirmed from that file that c.858A>G was ever tested in an approved assay. The three approved papers themselves were not retrieved in this case (no entry in fulltext_map.json or paper_extracts.json and no file under publications/), so their full texts could not be searched directly. The only direct functional measurement available for this variant is Sun et al. 2025 (PMID 40580951) Table S1, row hg38 11:108244983 A>G / c.858A>G / p.Q286Q, DeepATM_predicted=No (i.e. measured in the prime-editing saturation genome-editing screen itself, not extrapolated), Confidence=Medium-high, Combined_score 0.0240282777808958, Classification=Functional. A Functional call is evidence of retained, not impaired, ATM function, so it argues against PS3 rather than for it. No publication, assay, or database entry reviewed reports reduced ATM function for this variant. PS3 is therefore not met: there is no functional evidence of a damaging effect, and none of the VCEP-approved assays reports this variant.

Evidence considered

• Governing specification consulted: ClinGen HBOP ATM VCEP v1.6 (cspec), PS3 rules - 'Do not use as strong'; PS3_Moderate when a variant fails to rescue both an ATM-specific feature (e.g. phosphorylation of ATM-specific targets) AND radiosensitivity; PS3_Supporting when a variant fails to rescue an ATM-specific feature only, with radiosensitivity-only failure explicitly given no weight.
• Searched clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx (both sheets, 'ATM kinase activity' and 'ATM radiosensitivity') for 'c.858A>G', 'p.Gln286=' and 'p.Q286=': no match. The file lists only assay descriptions, controls and proposed strengths (PS3_Supporting/BS3_Supporting for the three approved kinase assays; PS3_moderate only if a kinase-assay PS3_supporting is already met for Mitui 2009 radiosensitivity), and its validation-control rows name other variants (c.7636del9; c.8147T>C p.V2716A; c.8546G>C p.R2849P; c.8599G>C p.G2867R; c.7987delGTT p.V2662del; c.5908C>T p.Q1970*), none of which is c.858A>G.
• Searched Suppl_TableS1_PMID 40580951.xlsx full text for 'c.858A>G', '858A', 'p.Gln286=' and 'p.Q286=': exactly one row matched - 108244983 / A / G / c.858A>G / Q286Q / exon_7 / Synonymous / residue 286 / GnomAD_all_AF 6.20197346795e-06 / Combined_score 0.0240282777808958 / Classification Functional / Confidence Medium-high / DeepATM_predicted No. DeepATM_predicted=No means the call comes from the prime-editing saturation genome-editing measurement itself rather than the DeepATM model, so it is direct functional data, not a computational prediction.
• Band context read directly from the supplied table (not an external publication): across all DeepATM_predicted=No rows, rows labelled Non-functional have Combined_score <= -1.360, Intermediate -1.360 to -0.912, and Functional >= -0.912 (Functional n=17505, Intermediate n=1549, Non-functional n=5480). The variant's score of 0.024 sits well inside the Functional band, i.e. the assay reads this variant as retaining ATM function - the opposite direction from PS3.
• Within exon 7 alone, 5031 directly-assayed synonymous SNVs are labelled Functional versus 252 Non-functional and 185 Intermediate, and c.858A>G is reported at the same Medium-high confidence as its neighbours - consistent with a benign (function-retaining) call, and offering no sign of a damaging functional effect.
• No functional evidence of impaired ATM function for this variant was found in any other source: literature_pass.json themes 'functional' (PMID 25741868 - ACMG/AMP framework only; PMID 26467025 - generic scoring-system methodology) contain no variant-specific functional data, and the paper_extracts entries for PMID 17508274, PMID 24418350, PMID 25741868 and PMID 26467025 all record variant_mentioned=false for c.858A>G / p.Q286=.
• The prime-editing assay that produced the Functional call is not one of the three approved assays in the VCEP's PS3/BS3 calibration table, so it could not have generated PS3 evidence even had it been positive; the variant's actual result is in the benign direction.

Still missing

• clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx was searched for 'c.858A>G', 'p.Gln286=' and 'p.Q286=' and contains no entry for this variant - the file holds only assay methodology and strength calibration, no per-variant results, so it cannot establish whether c.858A>G was tested in one of the three VCEP-approved assays.
• The three VCEP-approved PS3/BS3 papers (PMID 18634022 / Mitui 2009, PMID 19431188 / Barone 2009, PMID 11805335 / Scott 2002) were not retrieved in this case and therefore could not be searched for this variant at the paper level.
• The full article text of PMID 40580951 (Sun et al. 2025) was not fetched; only its Table S1 was available, so the assay's readout (which ATM-dependent endpoint and under what treatment) could not be verified against the VCEP's requirement that PS3 evidence reflect an ATM-specific feature such as substrate phosphorylation.

Sources

Reasoning

Governing framework is the ClinGen HBOP ATM VCEP v1.6 (cspec), which specifies PS4 (default strength Pathogenic Strong) with the literal text: 'Case-control studies; p-value <=.05 and (Odds ratio, hazard ratio, or relative risk >=2 or lower 95% CI >=1.5).' Per the read-the-operator-first rule, the exact operators written in this criterion's own source text are: p <= .05, and Odds ratio / hazard ratio / relative risk >= 2, or lower 95% CI >= 1.5. These strengthen as the enrichment value goes UP (value >= threshold). No case-control, cohort-enrichment, odds-ratio, hazard-ratio or relative-risk estimate exists anywhere in this variant's evidence, so there is no value to substitute into the inequality. The variant under assessment is ATM NM_000051.4:c.858A>G, p.(Gln286=), a synonymous (p.(=)) change at genomic NC_000011.9:g.108115710A>G. The two case-observation sources retrieved as PS4/PP5 candidates (PMID:17508274, an NSGC hereditary breast/ovarian cancer counseling guideline; PMID:24418350, an EFNS/ENS chronic-ataxia consensus) are gene/disease-level clinical guidance that do not name the ATM gene or the c.858A>G (p.Q286=) variant and report no ATM case-control or enrichment data. Because no comparison value can be written, the inequality (value >= threshold) cannot be satisfied; PS4 is not met. No further gap remains to close: absence of a case-control study is a genuine negative, not a retrieval failure.

Rule

VCEP ATM v1.6 PS4 (Pathogenic Strong): Case-control studies; p-value <=.05 and (Odds ratio, hazard ratio, or relative risk >=2 or lower 95% CI >=1.5).

Evidence considered

• Governing specification: ClinGen HBOP ATM VCEP v1.6 (cspec), which provides a PS4 rule at Pathogenic Strong strength; exact operator text quoted: p-value <=.05 AND (Odds ratio, hazard ratio, or relative risk >=2 OR lower 95% CI >=1.5), i.e. met requires value >= threshold for OR/HR/RR.
• Variant context applied to the rule: NM_000051.4:c.858A>G is a synonymous substitution, NP_000042.3:p.(Gln286=), genomic NC_000011.9:g.108115710A>G (GRCh38 chr11:108244983 A>G).
• No case-control, cohort, or case-enrichment study reporting an odds ratio, hazard ratio, relative risk or 95% CI for ATM c.858A>G was present in the case evidence; therefore no numeric value could be substituted into the PS4 inequality and the criterion is not satisfied.
• PMID:17508274 (NSGC HBOC risk-assessment/counseling recommendations) was reviewed as a PS4 candidate source; it is a gene/disease-level guideline restricted to BRCA1/BRCA2 and does not mention ATM or this variant.
• PMID:24418350 (EFNS/ENS chronic-ataxia consensus) was reviewed as a PS4 candidate source; it discusses ataxia-telangiectasia as a clinical entity but does not name the ATM gene or the c.858A>G (p.Q286=) variant and reports no variant-level data.

Sources

Reasoning

The governing framework is the ClinGen HBOP ATM VCEP v1.6 (cspec), which applies PM2 at Supporting strength when the frequency is <=0.001% in the gnomAD subpopulation with the highest frequency, with an exception when the frequency exceeds 0.001% but the variant is observed exactly once (n=1) in a single subpopulation and is absent from all others. The VCEP does not designate a non-cancer or exome-only population source for PM2, so the all-comers gnomAD entries are the default. The variant's highest-frequency subpopulation is African/African American in both all-comers datasets: gnomAD v2.1 AF 8.07559e-05 (0.00808%; 2/24,766 alleles) and gnomAD v4.1 AF 5.33903e-05 (0.00534%; 4/74,920 alleles). Both exceed the 0.001% threshold, and neither satisfies the n=1 exception because the allele is counted twice in v2.1 and four times in v4.1 within that subpopulation. PM2 is therefore not met. The conclusion is flagged for human review because the exception clause is dataset-dependent: in the gnomAD v3.1 non-cancer genomes subset the variant is seen exactly once and only in the African/African American group (1/41,138; total 1/147,964), which would satisfy the n=1 exception and yield PM2_Supporting if a non-cancer subset were treated as the governing source, which this VCEP does not specify.

Rule

ATM HBOP VCEP v1.6 PM2 (Pathogenic Supporting): frequency <= 0.001% in the gnomAD subpopulation with the highest frequency; exception applies if >0.001% but n=1 in a single subpopulation absent from all others.

Evidence considered

• ATM HBOP VCEP v1.6 applies PM2 (Pathogenic Supporting) at frequency <=0.001% in the gnomAD subpopulation with the highest frequency; exception: if >0.001% but n=1 in a single subpopulation and absent from all others, PM2_Supporting still applies.
• The VCEP does not specify a non-cancer or exome-only source for PM2, so the all-comers gnomAD v2.1 and v4.1 entries are used.
• gnomAD v2.1: highest subpopulation African/African American 8.07559e-05 (0.00808%, 2/24,766 alleles) - above the 0.001% threshold and an allele count of 2, so the n=1 exception does not apply.
• gnomAD v4.1: highest subpopulation African/African American 5.33903e-05 (0.00534%, 4/74,920 alleles) - above the 0.001% threshold and an allele count of 4, so the n=1 exception does not apply; grpmax filtering AF 1.746e-05.
• gnomAD v2.1 non-cancer exomes: total AF 8.4982e-06 (2/235,344 alleles); African/African American 8.5397e-05 (2/23,420 alleles) - still above 0.001% with an allele count of 2.
• gnomAD v3.1 non-cancer genomes: total AF 6.7584e-06 (1/147,964 alleles, 0 homozygotes); the single allele falls in the African/African American group (2.43084e-05, 1/41,138) and is absent from every other subpopulation, which would meet the VCEP's n=1 exception if that subset were the governing source.
• The variant is a synonymous substitution (NP_000042.3:p.(Gln286=)); PM2 depends solely on population frequency, not consequence type.

Still missing

• The VCEP's n=1 exception would be satisfied if the gnomAD v3.1 non-cancer genome subset were the governing source (single African/African American allele, 1/147,964; absent elsewhere), so the met/not-met outcome for PM2 depends on which gnomAD dataset is treated as authoritative; the VCEP says only 'gnomAD'.
• gnomAD-Canada v1.0 returned no data (query timed out), so no additional ancestry-specific frequency could be obtained.
• gnomAD v2.1 does not report a grpmax filtering allele frequency (null); PM2 was assessed using the maximum ancestry allele frequency in each dataset.

Sources

Reasoning

The governing specification for ATM PM3 is the ClinGen HBOP ATM VCEP points-based PM3/BP2 table (ATM_PM3_BP2_1.5.pdf and ATM_PM3_BP2_1.6.pdf, both read in full), which applies regardless of variant class and assigns points per unrelated A-T proband: phenotype 'confident' = 4.0 in trans / 2.0 phase unknown / 1.0 second variant unidentified or VUS / 2.0 homozygous; phenotype 'consistent' = half those values; conversion PM3_Supporting = 1.0 point, PM3 = 2.0, PM3_Strong = 4.0, PM3_VeryStrong = 8.0. The table is a generic scoring matrix with no variant-specific row, so there is no pre-assigned code for c.858A>G (p.Gln286=) in either version. Applying it: the total accumulated points for this variant are 0. No unrelated A-T proband carrying c.858A>G was identified in any source - neither in trans with a VCEP-classified P/LP ATM variant (confirmed or phase unknown), nor with a second variant unidentified/VUS, nor as a homozygous affected individual, and there is therefore no inference available from a consanguineous union or from bilateral ATM-related cancer family history. ClinVar variation 185146 has 9 single-submitter submissions (5 Likely benign, 4 Uncertain significance, 1 Benign in the scraped summary; conflicting classifications at 1 star) but the submission audit yields 0 criterion leads, no phase or zygosity information, and no expert-panel assertion usable here. Rarity, the other precondition of the table, is not limiting: c.858A>G has MAF 0.00107% in gnomAD v2.1 (3/281212 alleles) and 0.00062% in gnomAD v4.1 (10/1612390 alleles), both below the >0.01% general-population ceiling, with zero homozygotes in gnomAD v2.1, v4.1 and gnomAD-Canada. With 0 points versus the 1.0-point PM3_Supporting floor, PM3 is not met. The variant is a synonymous change (p.Gln286=) and PM3 under this specification is not variant-class-restricted, but a synonymous change contributes no additional allelic weight of its own.

Rule

ATM VCEP v1.6 PM3: 0 points accumulated on the ATM PM3/BP2 table; 0 < 1.0 point (PM3_Supporting threshold), so PM3 is not met.

Evidence considered

• ClinGen HBOP ATM VCEP v1.6 PM3 specifies PM3_Supporting = 1 point, PM3 = 2 points, PM3_Strong = 4 points, PM3_VeryStrong >= 8 points, with points assigned per unrelated A-T proband via the ATM PM3/BP2 table.
• ATM PM3/BP2 table (v1.5 and v1.6 identical for this purpose): points per unrelated A-T proband = 4.0 confirmed in trans, 2.0 phase unknown, 1.0 second variant unidentified or VUS, 2.0 homozygous (max 2 individuals) for a confident phenotype, halved for a consistent phenotype.
• Table general consideration: the variant under assessment may not exceed a general population frequency of 0.01%; c.858A>G at MAF 0.00107% (gnomAD v2.1) and 0.00062% (gnomAD v4.1) satisfies this and is not excluded on rarity grounds.
• Search of the full PM3/BP2 text files for c.858A>G, p.Gln286= and p.Q286= returned no variant-specific row; the file is a generic points matrix with no pre-assigned variant code.
• ClinVar variation 185146: 9 submissions (5 Likely benign, 4 Uncertain significance, 1 Benign), review status 'criteria provided, conflicting classifications', 1 star, no expert-panel assertion, and 0 usable criterion leads with no phase or zygosity information.
• gnomAD v2.1: 3/281212 alleles, MAF 0.00107%, homozygotes 0, highest subpopulation African/African American AF 8.08e-05. gnomAD v4.1: 10/1612390 alleles, MAF 0.00062%, homozygotes 0, grpmax FAF 1.746e-05. gnomAD-Canada v1.0: not available.
• PMID:17508274 (NSGC HBOC risk-assessment guideline), PMID:24418350 (EFNS/ENS chronic ataxia consensus), PMID:25741868 (ACMG/AMP 2015 framework) and PMID:26467025 (rules-based scoring methodology) were each searched for c.858A>G / p.Q286=; none mentions the variant, the ATM gene, or any proband carrying it.

Still missing

• No proband-level A-T observation of c.858A>G exists in any source searched, so the PM3 points total is a true zero rather than a phase-unknown or second-variant-unidentified entry: ClinVar variation 185146 (9 submissions) provides no phase/zygosity data and 0 criterion leads; the four fetched full texts (PMID:17508274, PMID:24418350, PMID:25741868, PMID:26467025) do not name the variant; gnomAD v2.1, v4.1 and gnomAD-Canada report zero homozygotes.
• The governing files ATM_PM3_BP2_1.5.pdf and ATM_PM3_BP2_1.6.pdf were searched in full and contain NO variant-specific entry for c.858A>G (or p.Gln286= / p.Q286= / c.858A>G). They are generic points tables, not a per-variant lookup, so no pre-assigned code exists to outrank this assessment.
• Two literature_pass entries triaged to theme 'case_observation' as PM3-relevant candidates (PMID:24432435, PMID:42258614) have no full text retrieved and no paper_extract, so any A-T proband carrying c.858A>G reported in that literature remains unexamined - this is the only unresolved retrieval gap for PM3.
• PMID:20050888 and PMID:28492532 are guideline/methodology papers with no ATM or variant-level content; they neither support nor exclude PM3.

Sources

Reasoning

The governing ATM VCEP restricts PM5 (default Pathogenic Moderate) to two situations, both requiring PVS1: (a) NMD-prone truncating DNA variants receiving PVS1 at Very Strong with a premature termination codon upstream of p.Leu3048, and (b) observed (RNA-confirmed) NMD-prone truncating splice variants receiving PVS1(RNA) at Very Strong with a PTC upstream of p.Leu3048. The VCEP explicitly states this is not applicable for predicted splice impact without RNA data or PVS1(RNA) at reduced strength. c.858A>G is a synonymous change (p.Gln286=): it is not a truncating/null variant, contains no premature termination codon, and receives no PVS1, so it fails both branches of the VCEP PM5 rule. The classic same-residue missense PM5 interpretation is also unavailable: a synonymous substitution produces no amino-acid change, and no established P/LP missense variant exists at residue 286 (c.858A>C and c.858A>T both yield p.Gln286His and are scored function-neutral).

Rule

ATM PM5 (ClinGen HBOP ATM VCEP v1.6): truncating-only / PVS1-dependent rule (PTC upstream of p.Leu3048).

Evidence considered

• CSpec ATM v1.6 PM5 rule: apply to NMD-prone truncating DNA variants receiving PVS1 at Very Strong with PTC upstream of p.Leu3048, and to observed (not predicted) NMD-prone truncating splice variants receiving PVS1(RNA) at Very Strong; 'This is not applicable for predicted splice impact without RNA data or PVS1(RNA) at reduced strength.'
• Normalized consequence is synonymous p.(Gln286=); no truncating variant and no PVS1 assignment, so the VCEP PM5 branches are not triggered.
• pm5_candidates.json: found=false, zero same-residue candidates, residue context not parseable (non-missense).
• Sun et al. 2025 (PMID 40580951) Table S1: residue 286 alternate substitutions c.858A>C and c.858A>T (both p.Gln286His) are scored functional, i.e. not established P/LP comparators.

Still missing

• The PM5 comparator collection (pm5_candidates.json) returned zero same-residue candidates and marked classic PM5 search ineligible (found=false; reason: 'PM5 is a truncation-cutoff or non-missense framework rule rather than classic same-residue missense PM5').

Sources

Reasoning

The governing ClinGen HBOP ATM VCEP v1.6 specification lists PP5 as 'Not Applicable'. Under the standing global policy for this pipeline, PP5 (Supporting strength) is applied only when there is an exact-variant ClinVar expert-panel (ClinGen VCEP/EP) classification, i.e. a 3-star review status of Pathogenic or Likely pathogenic, and ordinary laboratory submissions, aggregate ClinVar labels, and non-expert assertions must never trigger PP5. For ATM c.858A>G (p.Gln286=), ClinVar VCV000185146 is rated 1 star with review status 'criteria provided, conflicting classifications' and has no expert-panel submissions (expert_panels field empty; expert_panel_submissions = 0). Its 9 clinical-laboratory submissions are classified Likely benign (5), Uncertain significance (4) and Benign (1) — none pathogenic, and all non-expert single-submitter assertions. No exact-variant expert-panel Pathogenic/Likely pathogenic assertion exists, so the PP5 trigger is absent. PP5 is not met.

Rule

Global PP5: Supporting only when an exact-variant ClinVar expert-panel (3-star) classification is Pathogenic/Likely pathogenic. VCEP ATM v1.6 lists PP5 as Not Applicable.

Evidence considered

• ClinGen HBOP ATM VCEP v1.6 (cspec) lists PP5 applicability as 'Not Applicable'.
• Global PP5 policy: applies at Supporting strength only for an exact-variant ClinVar expert-panel (3-star) Pathogenic/Likely pathogenic classification; laboratory submissions and aggregate labels do not qualify.
• ClinVar VCV000185146 (ClinVarID 185146) shows review status 'criteria provided, conflicting classifications', 1 star, 9 clinical-laboratory submissions (5 Likely benign, 4 Uncertain significance, 1 Benign) and 0 expert-panel submissions; no pathogenic expert assertion for this exact variant exists.

Sources

Reasoning

The governing framework is the ClinGen HBOP ATM VCEP v1.6 (cspec), which assigns BA1 (Benign Stand-Alone) when the Grpmax Filtering AF in gnomAD exceeds 0.5%. This VCEP does not designate a non-cancer or exome-only population source for BA1, so the all-comers gnomAD entries are used. The variant is extremely rare in every queried dataset: gnomAD v2.1 total AF 1.06681e-05 (3/281,212 alleles, homozygotes 0) with the highest ancestry frequency in African/African American at 8.07559e-05 (0.00808%; 2/24,766 alleles, homozygotes 0); gnomAD v4.1 total AF 6.20197e-06 (10/1,612,390 alleles, homozygotes 0), African/African American 5.33903e-05 (0.00534%; 4/74,920) and grpmax filtering AF 1.746e-05. The highest observed population frequency is ~0.008%, more than 60-fold below the 0.5% stand-alone threshold, and no homozygote was observed anywhere. BA1 is therefore not met.

Rule

ATM HBOP VCEP v1.6 BA1 (Benign Stand Alone): Grpmax Filtering AF > 0.5% in gnomAD.

Evidence considered

• ATM HBOP VCEP v1.6 sets BA1 (Benign Stand Alone) at a Grpmax Filtering AF >0.5% in gnomAD; the specification does not restrict BA1 to a non-cancer or exome-only dataset, so all-comers gnomAD v2.1 and v4.1 are used.
• gnomAD v2.1: total AF 1.06681e-05 (3/281,212 alleles, 0 homozygotes); highest ancestry African/African American 8.07559e-05 (0.00808%, 2/24,766 alleles, 0 homozygotes).
• gnomAD v4.1: total AF 6.20197e-06 (10/1,612,390 alleles, 0 homozygotes); highest ancestry African/African American 5.33903e-05 (0.00534%, 4/74,920 alleles, 0 homozygotes); grpmax filtering AF 1.746e-05.
• gnomAD v4.1 exome grpmax filtering AF is 2.376e-05 and the joint grpmax filtering AF is 1.746e-05; both are orders of magnitude below the 0.5% stand-alone threshold.
• The variant is a synonymous substitution (NP_000042.3:p.(Gln286=)); BA1 depends solely on population frequency, not consequence type.

Still missing

• gnomAD-Canada v1.0 returned no data (query timed out); it therefore contributed no frequency evidence.
• gnomAD v2.1 does not report a grpmax filtering allele frequency (null), so the maximum ancestry allele frequency was used as the comparator for the v2.1 dataset.

Sources

Reasoning

The governing framework is the ClinGen HBOP ATM VCEP v1.6 (cspec), which assigns BS1 (Benign Strong) when the Grpmax Filtering AF in gnomAD exceeds 0.05%. This VCEP does not designate a non-cancer or exome-only population source for BS1, so the all-comers gnomAD entries are used. The variant is vanishingly rare in every queried dataset: gnomAD v2.1 total AF 1.06681e-05 (3/281,212 alleles) with the maximum ancestry frequency in African/African American at 8.07559e-05 (0.00808%; 2/24,766 alleles); gnomAD v4.1 total AF 6.20197e-06 (10/1,612,390 alleles), African/African American 5.33903e-05 and grpmax filtering AF 1.746e-05. The highest observed population frequency (~0.008%) is roughly six-fold below the 0.05% strong threshold, and no homozygotes were observed. BS1 is therefore not met.

Rule

ATM HBOP VCEP v1.6 BS1 (Benign Strong): Grpmax Filtering AF > 0.05% in gnomAD.

Evidence considered

• ATM HBOP VCEP v1.6 sets BS1 (Benign Strong) at a Grpmax Filtering AF >0.05% in gnomAD; the specification does not restrict BS1 to a non-cancer or exome-only dataset, so all-comers gnomAD v2.1 and v4.1 are used.
• gnomAD v2.1: total AF 1.06681e-05 (3/281,212 alleles, 0 homozygotes); highest ancestry African/African American 8.07559e-05 (0.00808%, 2/24,766 alleles, 0 homozygotes).
• gnomAD v4.1: total AF 6.20197e-06 (10/1,612,390 alleles, 0 homozygotes); highest ancestry African/African American 5.33903e-05 (0.00534%, 4/74,920 alleles, 0 homozygotes); grpmax filtering AF 1.746e-05.
• No dataset reaches the 0.05% level, and no homozygote for this variant was observed in any queried dataset, so the benign-strong frequency argument is not supported.
• The variant is a synonymous substitution (NP_000042.3:p.(Gln286=)); BS1 depends solely on population frequency, not consequence type.

Still missing

• gnomAD-Canada v1.0 returned no data (query timed out); it therefore contributed no frequency evidence.
• gnomAD v2.1 does not report a grpmax filtering allele frequency (null), so the maximum ancestry allele frequency was used as the comparator for the v2.1 dataset.

Sources

Reasoning

BP2 under the ClinGen HBOP ATM VCEP is assigned from the same points-based PM3/BP2 table (ATM_PM3_BP2_1.5.pdf and ATM_PM3_BP2_1.6.pdf, both read in full): points per unaffected (non-A-T) adult proband aged >18 with no evidence of A-T are -4.0 confirmed in trans with a P/LP ATM variant, -2.0 phase unknown, and -2.0 (laboratory setting) or -1.0 (database setting) for a homozygous observation capped at -2.0; conversion BP2 = -1 point, BP2_Moderate = -2 points, BP2_Strong <= -4 points; observations in cis must not be used, and a co-occurring P/LP variant must itself be classified under VCEP specifications. Applying this to c.858A>G (p.Gln286=): the accumulated BP2 score is 0. No unaffected adult carrying c.858A>G with a VCEP-classified P/LP ATM variant in trans was identified, and no phase-unknown co-occurrence was reported; the ClinVar record (variation 185146, 9 single-submitter submissions, conflicting classifications, 0 criterion leads) provides no phase, zygosity or co-occurrence information, and no cis observation with a pathogenic variant is documented. Nor is the -1.0 'database setting' homozygous path available: gnomAD v2.1, gnomAD v4.1 and gnomAD-Canada each report zero homozygotes for this variant, so there is no homozygous unaffected-adult observation to score. Although the generic BP2 definition also covers cis with a pathogenic variant in any inheritance pattern, no such cis configuration is reported for this variant either. Since the strongest BP2 tier (<= -4 points) is not approached and 0 points falls short of even the -1-point BP2 threshold, BP2 is not met.

Rule

ATM VCEP v1.6 BP2: 0 points accumulated on the ATM PM3/BP2 table; 0 > -1 point (BP2 threshold) is not reached, so BP2 is not met.

Evidence considered

• ClinGen HBOP ATM VCEP v1.6 BP2 (Benign Supporting): -1 point = BP2, -2 points = BP2_Moderate, <= -4 points = BP2_Strong, assigned via the ATM PM3/BP2 table.
• ATM PM3/BP2 table: points per unaffected (non-A-T) adult (>18 years, no evidence of A-T) proband carrying the variant with a pathogenic or likely pathogenic ATM variant = -4.0 confirmed in trans, -2.0 phase unknown, -2.0 laboratory setting or -1.0 database setting for homozygous (max -2.0); 'Do not use observations in cis'.
• BP2 table footnotes: the variant under assessment should not be bioinformatically predicted or experimentally proven to have a clinically important effect on protein or mRNA splicing; the co-occurring P/LP variant must be classified under VCEP specifications; the database-setting homozygous path excludes gnomAD-type frequency data, which are already captured by BA1/BS1.
• Search of the two governing PM3/BP2 text files for c.858A>G, p.Gln286= and p.Q286= returned no variant-specific row or pre-assigned code.
• ClinVar variation 185146: 9 submissions (5 Likely benign, 4 Uncertain significance, 1 Benign), review status 'criteria provided, conflicting classifications', 1 star, no expert-panel assertion, 0 usable criterion leads; no phase, zygosity or co-occurring-variant information.
• gnomAD v2.1 (3/281212 alleles, MAF 0.00107%, homozygotes 0), gnomAD v4.1 (10/1612390 alleles, MAF 0.00062%, homozygotes 0) and gnomAD-Canada v1.0 (not available) all lack a homozygous observation.
• PMID:17508274, PMID:24418350, PMID:25741868 and PMID:26467025 were each searched for c.858A>G / p.Q286=; none mentions the variant or reports an unaffected carrier.

Still missing

• No unaffected (non-A-T) adult observation of c.858A>G with a co-occurring P/LP ATM variant was found in any source, so the BP2 score is a true zero rather than a phase-unknown entry: ClinVar variation 185146 (9 submissions) reports no phase/zygosity/co-occurrence data and 0 criterion leads, and the four fetched full texts do not name the variant.
• The -1.0-point 'database setting' homozygous BP2 path cannot be exercised because gnomAD v2.1, gnomAD v4.1 and gnomAD-Canada all report zero homozygotes for c.858A>G; no unaffected homozygous adult is documented.
• The governing files ATM_PM3_BP2_1.5.pdf and ATM_PM3_BP2_1.6.pdf were searched in full for c.858A>G, p.Gln286= and p.Q286= and contain no variant-specific entry - they are generic points tables, so no pre-assigned BP2 code exists for this variant.
• PMID:24432435 and PMID:42258614 (both triaged as case_observation candidates) have no retrieved full text, so any unaffected-carrier co-occurrence data in those papers remains unexamined.
• No cis-with-pathogenic-variant observation (the other generic BP2 trigger) was found in any searched source; cis observations are explicitly excluded from BP2 scoring under this VCEP anyway.

Sources

Reasoning

BP4 for this variant is evaluated only through the SpliceAI path, because c.858A>G is a synonymous change and the ATM VCEP specification gives BP4 as 'Missense: REVEL score <=.249' or 'Splicing: No predicted impact via splicing (SpliceAI <=0.1)'; for a synonymous variant only the splice arm can apply, and no REVEL score exists in any case. SpliceAI reports a maximum delta score of 0.208 for this variant, which is above the <=0.1 BP4 cutoff; indeed it is above the >=0.2 PP3 cutoff used by the same VCEP specification, so the prediction is in the direction of splice impact rather than no impact. BP4 is therefore not met. This is not a gap or an unresolved result: the applicable predictor returned a score, and that score falls outside the benign range, so the criterion is affirmatively not met rather than not assessed. No RNA data are available, so BP4 cannot be combined with BP7(RNA) in either direction, and the Sun et al. Table S1 row for c.858A>G lists all of its computational predictors as NA, providing no pre-assigned BP4 code.

Rule

ATM VCEP v1.6 BP4 Supporting (splice path): SpliceAI max delta <= 0.1.

Evidence considered

• ATM VCEP v1.6 BP4 (Benign Supporting): 'Missense: REVEL score <=.249' and 'Splicing: No predicted impact via splicing (SpliceAI <=0.1).'
• Because c.858A>G is synonymous, only the splicing arm of BP4 can apply; the missense REVEL arm is not available for this consequence type.
• SpliceAI max delta score for NM_000051.4:c.858A>G is 0.208 (DS_AG 0.208 at DP_AG +1), which is above the <=0.1 BP4 cutoff and also above the >=0.2 PP3 cutoff, so no 'no predicted impact' conclusion can be drawn.
• The exact c.858A>G row in Suppl_TableS1_PMID 40580951 records all computational predictor columns (CADD, boostDM, EVE, REVEL, AlphaMissense) as NA, so the governing table assigns no pre-calibrated BP4 code for this variant.

Sources

Reasoning

The governing ClinGen HBOP ATM VCEP v1.6 specification lists BP6 as 'Not Applicable'. Under the standing global policy, BP6 (Supporting strength) is applied only when there is an exact-variant ClinVar expert-panel (ClinGen VCEP/EP) classification, i.e. a 3-star review status of Benign or Likely benign; ordinary laboratory submissions, aggregate ClinVar labels, and non-expert assertions must never trigger BP6. For ATM c.858A>G (p.Gln286=), ClinVar VCV000185146 is rated only 1 star with review status 'criteria provided, conflicting classifications' and has zero expert-panel submissions. Its aggregate label is Likely benign and its individual submissions include Likely benign (5) and Benign (1) from clinical laboratories, but these are non-expert single-submitter assertions and by the stated rule cannot trigger BP6. No exact-variant expert-panel Benign/Likely benign assertion exists, so the BP6 trigger is absent. BP6 is not met.

Rule

Global BP6: Supporting only when an exact-variant ClinVar expert-panel (3-star) classification is Benign/Likely benign. VCEP ATM v1.6 lists BP6 as Not Applicable.

Evidence considered

• ClinGen HBOP ATM VCEP v1.6 (cspec) lists BP6 applicability as 'Not Applicable'.
• Global BP6 policy: applies at Supporting strength only for an exact-variant ClinVar expert-panel (3-star) Benign/Likely benign classification; laboratory submissions and aggregate labels do not qualify.
• ClinVar VCV000185146 (ClinVarID 185146) shows 1-star review status 'criteria provided, conflicting classifications' with 9 non-expert clinical-laboratory submissions (5 Likely benign, 4 Uncertain significance, 1 Benign) and 0 expert-panel submissions; the aggregate Likely-benign label is therefore not an expert assertion.

Sources

Reasoning

Scope: BP7 applies to synonymous and qualifying deep-intronic variants, and c.858A>G is synonymous (p.Gln286=, codon 286, exon 7), so BP7 is in scope and is adjudicated rather than set aside. The ATM VCEP v1.6 BP7 rule reads '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', and the underlying ACMG/AMP criterion requires that splicing prediction algorithms show no impact on the splice consensus sequence and no creation of a new splice site (PMID:25741868). Here the SpliceAI prediction contradicts that requirement: the maximum delta score is 0.208, produced by a predicted acceptor gain (DS_AG 0.208 at DP_AG +1), i.e. a predicted new splice site one nucleotide downstream of the variant, and that same 0.208 is what qualifies the variant for a supporting PP3 splice code above. A variant cannot simultaneously carry a PP3 predicted-splice-impact code and a BP7 no-splice-impact code from the same prediction, so BP7 is not met. This conclusion does not depend on any RNA data, and no RNA-based splice assay for this variant was found; the RNA arm of BP7 (BP7_Variable(RNA)/BP7_RNA) is therefore also unavailable, not satisfied. The deep-intronic positional wording of the rule is not reached, since the variant fails the no-predicted-impact requirement regardless of its distance from the donor/acceptor sites.

Rule

ATM VCEP v1.6 BP7 Supporting: synonymous/deep-intronic variants with no predicted splice impact (SpliceAI <= 0.1 and no new splice site).

Evidence considered

• Consequence scope: NM_000051.4:c.858A>G is a synonymous substitution (NP_000042.3:p.(Gln286=)/p.(Q286=)), so BP7 is applicable rather than out of scope.
• ATM VCEP v1.6 BP7 (Benign Supporting) rule as declared by the expert panel: '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'; the panel also notes BP7 is not considered conflicting evidence against a body of evidence supporting a pathogenic splice defect.
• The ACMG/AMP source of BP7 (PMID:25741868) requires that splicing prediction algorithms predict no impact to the splice consensus sequence and no creation of a new splice site.
• SpliceAI for c.858A>G predicts the creation of a new acceptor site (DS_AG 0.208 at DP_AG +1; max delta 0.208), so the no-splice-impact requirement of BP7 is not satisfied.
• No RNA splice assay data for this variant were located in the case evidence, so BP7_Variable(RNA)/BP7_RNA cannot be applied; the DeepATM prime-editing classification for c.858A>G in Suppl_TableS1_PMID 40580951 ('Functional', Medium-high, DeepATM_predicted=No) is a functional-screen output, not an RNA splice assay, and was not used here.

Still missing

• No RNA-based splice assay (BP7_Variable(RNA)) data were found for ATM c.858A>G, so the observed-splice-defect arm of the ATM VCEP BP7 rule could not be evaluated; this does not affect the supporting-strength BP7 conclusion, which fails on the predicted splice impact.

Sources

Reasoning

PP1 is applicable at ATM: the ClinGen HBOP ATM VCEP v1.6 (cspec) specifies it for this autosomal recessive condition as co-segregation with disease in affected relatives, graded 1 affected relative = Supporting, 2 affected relatives = Moderate, and >=3 affected relatives = Strong. Evaluating it requires a proband together with genotyped affected relatives in whom the variant co-segregates, and a countable number of informative meioses. None of these exist for this case: there is no proband, no pedigree or family history, no affected relative, and no meiosis/segregation data in any source; all available evidence is variant-level (population frequency, ClinVar, in-silico splice prediction, functional screens). Because the criterion can be neither met nor shown unmet without family data, it is returned as not_assessed, and the residual gap is the complete absence of pedigree and segregation information rather than a negative segregation result.

Rule

ATM VCEP v1.6 PP1 (AR condition): segregation in 1 affected relative = Supporting; in 2 affected relatives = Moderate; in >=3 affected relatives = Strong.

Evidence considered

• ClinGen HBOP ATM VCEP v1.6 specifies PP1 for this autosomal recessive condition as segregation in 1 affected relative (Supporting), 2 affected relatives (Moderate), or >=3 affected relatives (Strong).
• The case contains no proband, no pedigree, no family history and no meioses data; no affected relative carrying c.858A>G (p.Gln286=) is reported in any source.
• The variant is a synonymous change (NM_000051.4:c.858A>G, p.Gln286=) documented only at the variant level (population frequency, ClinVar submissions, splice prediction), with no accompanying clinical family.
• PMID:26467025 uses segregation as one input of a rules-based variant scoring system but reports no segregation data for ATM or for this variant.

Still missing

• No proband, pedigree, family history, or genotyped affected relatives are available, so segregation with disease cannot be counted; PP1 requires at least 1 affected relative co-carrying the variant for Supporting strength.
• No non-segregation data (unaffected carriers) are available either, so PP1 versus BS4 cannot be resolved from the case data.

Sources