Five-step scale from benign to pathogenic. This variant is uncertain. Select a step to read what it means.
This variantScenario
BenignPathogenic
Uncertain.The recorded classification is uncertain significance.
In plain language
Gene-level background from the recorded interpretation.
Words with a dotted underline have a short definition. Hover or tap them.
About ATM
ATM encodes a kinase that acts as a master controller of the cellular DNA damage response: it detects double-strand breaks and coordinates DNA repair, cell cycle arrest, or apoptosis through downstream targets including p53, BRCA1, and CHK2. Loss-of-function mutations in both copies of ATM cause ataxia telangiectasia, an autosomal recessive disorder characterized by neurological problems, immune deficiency, and cancer predisposition. ATM functions as a tumor suppressor: individuals with ataxia telangiectasia are prone to childhood lymphomas, leukemias, and breast cancer, while carriers of a single altered copy have increased risk of breast, pancreatic, prostate, and other cancers. Somatic ATM mutations also occur in lymphoid malignancies and solid tumors, and ATM-deficient cancers are often especially sensitive to DNA-damaging treatments.
LYFE Sciences is an AI system, and it can make mistakes.
Criteria may be applied incorrectly, sources may be misread, and a
confident-looking classification can still be wrong.
Double-check every criterion and its underlying evidence
before relying on any call.
01The reasoning
How it was classified
The recorded framework is ClinGen ATM VCEP v1.6. The criteria and strengths shown below come from the saved assessment; this page does not recalculate the classification.
Recorded applied criteria
The applied criteria and their strengths are shown as assessed under the stated framework.
Met at supporting: gnomAD v4.1 group-max filtering AF 0.000292% in European (non-Finnish) is below the VCEP 0.001% threshold, with zero homozygotes.
Benign criteria applied · 0
No benign criterion marked as applied in the recorded assessment.
Recorded classification: Uncertain significance
VUS: only PM2 (supporting) is met, and no criteria-combination rule in the ATM VCEP v1.6 framework is satisfied by that single supporting criterion.
02The sources
Evidence at a glance
Evidence summaries and captured source records, where available. Missing entries should not be interpreted as negative findings.
gnomAD v4.1
gnomAD v2.1
Population frequency
11 / 1,611,514 alleles
gnomAD v4.1: 11 / 1,611,514 alleles. Recorded AF 6.82588e-06. Recorded homozygote count 0.
gnomAD v2.1
This variant is present in gnomAD v2.1 (AF= 3.97924e-06; MAF= 0.00040%, 1/251304 alleles, homozygotes = 0) and has highest observed frequency in the European (non-Finnish) population (AF= 8.80034e-06; MAF= 0.00088%, 1/113632 alleles, homozygotes = 0).
OncoKB did not identify variant-specific reviewed functional evidence for this variant; gene-level curated context is available for reviewer follow-up. ATM, a kinase involved in the DNA damage response, is mutated in various solid and hematologic malignancies.
ACMG/AMP criteria under ClinGen ATM VCEP v1.6, with the data and sources behind each one.
28 criteria with recorded statuses: 1 met · 10 not met · 3 not assessed · 14 not applicable. Open any row for the full reasoning, the rule and the evidence behind it.
All three criteria in this group are inapplicable to NM_000051.4:c.6317A>T because the variant is a mid-exonic missense substitution (NP_000042.3:p.(Asn2106Ile), exon 43, SpliceAI max delta 0.005, no truncation and no protein-length change): PVS1 is restricted by the ATM VCEP to null variants, PM4 is restricted by the ATM VCEP to stop-loss variants, and BP3 is declared 'Not Applicable' by the same VCEP specification. This group therefore contributes no pathogenic or benign evidence to the final classification.
The VCEP PVS1 decision guide documents were searched in full for this variant's notations and contain no per-variant PVS1 code; the variant's only occurrence in the converted VCEP materials is the Sun et al. 2025 functional-screen supplement, which corroborates its missense consequence.
This group contributes no evidence, in either direction, to the classification of NM_000051.4:c.6317A>T (NP_000042.3:p.(Asn2106Ile), exon 43). Three of the five criteria (PM1, PP2, BP1) are declared 'Not applicable' by the governing ClinGen HBOP ATM VCEP v1.6 specification (cspec doc 639508985): PM1 because benign and pathogenic variants co-occur within ATM domains and no germline hotspots are defined, PP2 because ATM has no defined low rate of benign missense variation, and BP1 because missense pathogenic variants are known for ATM. A fourth (PM5) is applicable in principle but, as specified by this VCEP, is restricted to NMD-prone truncating variants (DNA, or RNA-observed splice) with premature termination codons upstream of p.Leu3048 and explicitly excludes missense changes - p.Asn2106Ile cannot receive PVS1 and therefore cannot enter either branch.
The one criterion in this group that can genuinely fail on this variant (PS1) does fail: the standard missense path requires the same amino acid change to have been previously established as pathogenic, but p.Asn2106Ile is uniquely encoded by c.6317A>T, and that change is itself only a 2-star Uncertain significance entry in ClinVar (variation 135768, five submitters, no conflicts). The reduced-strength splicing path is unreachable because SpliceAI max delta is 0.005 (threshold for PP3(splicing) eligibility 0.2) and no known P/LP reference variant exists at the same position; the ATM PS1 splicing table is a class-level lookup with no variant rows.
Both VCEP PS1 files and the two supplementary-table documents were searched in full for the variant's notations (c.6317A>T, p.Asn2106Ile, p.N2106I, and the bare residue number 2106): the only occurrence of this variant anywhere in the converted ATM VCEP materials is the Sun et al. 2025 functional-screen supplement (Suppl_TableS1_PMID 40580951.xlsx, row for hg38_pos 108317491), which confirms it is a Missense substitution at residue 2106 in exon 43, flags DeepATM_predicted = No with Confidence = High (a directly measured rather than model-extrapolated row), and records its ClinVar_classification as Uncertain significance. That supplement also contains the neighbouring codon-2106 substitutions (N2106Y, N2106H, N2106D, N2106T, N2106S, N2106K), none of which carries a ClinVar classification - so no established pathogenic same-residue comparator exists in the evidence collected. Functional weighting of the Sun et al. row belongs to the PS3/BS3 group, not here.
Net effect: zero pathogenic points and zero benign points from the codon/residue/domain group, so the final classification rests entirely on the other criteria groups.
PS3 not met. The ClinGen HBOP ATM VCEP v1.6 permits PS3 only when a variant fails to rescue ATM function (Moderate if both an ATM-specific feature and radiosensitivity fail, Supporting if the ATM-specific feature alone fails), and the VCEP's approved-assay table recognises three assays (Mitui 2009/PMID 18634022, Barone 2009/PMID 19431188, Scott 2002/PMID 11805335). No result from any of those assays exists for c.6317A>T, and the only functional measurement available (Sun et al. 2025, PMID 40580951, Suppl. Table S1: Combined_score -0.754, 'Functional', DeepATM_predicted No, Confidence High) reports retained function, which is the opposite of the damaging effect PS3 requires.
BS3 not met, flagged for human review. The same Sun et al. 2025 measurement is directionally BS3-consistent - the variant is directly assayed and classified Functional, comfortably inside that table's Functional band (-0.912 to +8.287) and well above its Intermediate (-1.360 to -0.912) and Non-functional bands - but the prime-editing saturation-genome-editing screen is not one of the VCEP's three approved PS3/BS3 assays, measures neither of the VCEP's specified readouts (ATM substrate phosphorylation, radiosensitivity), and postdates the VCEP v1.5 calibration. Assigning BS3_Supporting or BS3_Moderate from it would be an uncalibrated strength, so BS3 is recorded as not met and the single uncalibrated benign-directional result is left for human adjudication.
Net functional-group contribution: no pathogenic functional evidence (PS3 not met) and no VCEP-calibrated benign functional evidence (BS3 not met); the only functional data point is one direct but unapproved assay reporting retained ATM function, which a reviewer may wish to weigh as exploratory evidence.
Population-frequency evidence does not support a benign interpretation: gnomAD v4.1 group-max filtering AF is 2.92e-06 (0.000292%) with a total frequency of 6.83e-06 (11/1,611,514 alleles, 0 homozygotes), leaving BA1 (>0.5%) and BS1 (>0.05%) unmet by orders of magnitude under the ATM VCEP v1.6 thresholds.
BS2 is not applicable for ATM by specification because ATM-related disease is incompletely penetrant; no homozygous carrier is present in any gnomAD dataset.
PM2 is met at Supporting strength: the variant is at or below 0.001% in the highest-frequency gnomAD population on every applicable source (v4.1 group-max filtering AF 0.000292% in European non-Finnish; v2.1 highest population frequency 0.00088%), with zero homozygotes, though the raw 0.00481% frequency in gnomAD v4.1's heterogeneous Remaining individuals bin warrants curator confirmation.
No criteria match.
Met · 1
PM2SupportingMet at supporting: gnomAD v4.1 group-max filtering AF 0.000292% in European (non-Finnish) is below the VCEP 0.001% threshold, with zero homozygotes.
Details ▾Hide ▴
Reasoning
The ATM VCEP v1.6 assigns PM2 at Supporting strength only and requires a frequency <=0.001% (1e-5) in the gnomAD subpopulation with the highest frequency, with an exception for a variant that exceeds 0.001% solely through n=1 in a single subpopulation not present elsewhere. On gnomAD v4.1 the variant is present at a total frequency of 6.82588e-06 (11/1,611,514 alleles, 0 homozygotes, all alleles called in the exome dataset), with a group-max filtering allele frequency of 2.92e-06 (0.000292%; exome grpmax FAF 3.1e-06), and gnomAD designates European (non-Finnish) as the group carrying that maximum - 3.4-fold below the 0.001% threshold. On gnomAD v2.1 (exome data) the total AF is 3.97924e-06 (1/251,304) and the highest population frequency is 8.80034e-06 (0.00088%, European non-Finnish), also below 0.001%. The VCEP does not specify a non-cancer or exome-only population source, so the all-comers datasets are the default; the non-cancer subsets were nevertheless checked and do not change the result (gnomAD v2.1 non-cancer exomes 1/236,766 alleles, highest population frequency 9.74507e-06 = 0.000975%; gnomAD v3.1 non-cancer genomes absent; gnomAD-Canada v1.0 absent). Across all four sources the variant sits below 0.001% in the highest-frequency population and no homozygote has been observed in roughly 2.1 million gnomAD alleles, which is the relevant observation for a recessive condition. The single value above the threshold is the raw frequency of gnomAD v4.1's 'Remaining individuals' bin (3/62,362 = 4.81e-05 = 0.00481%). That bin is a heterogeneous catch-all rather than one of gnomAD's defined ancestry groups, and it is not the group gnomAD itself identifies as carrying the maximal filtering allele frequency for this variant (European non-Finnish, with 8 alleles in 1,178,890); the VCEP's own note that the filtering allele frequency 'accounts for population size and founder populations' supports using the filtering-AF estimate rather than a 3-allele raw frequency from a 62,000-allele bin. PM2 is therefore met at Supporting strength, but the call is flagged for human review because a strictly literal reading of 'subpopulation with the highest frequency' using raw allele frequency would instead require the exception test, which would fail here (n=3, and the variant is present in more than one subpopulation). The VCEP's additional condition that coverage at the locus be >30X was not independently verifiable from the supplied data; the variant's detection in the gnomAD v4.1 exome callset (AN 1,460,382) indicates the site is covered and callable, but a per-locus depth measurement should be confirmed before final reporting.
Rule
PM2 Supporting, ATM VCEP v1.6: frequency <=0.001% in the gnomAD subpopulation with the highest frequency; exception when >0.001% but n=1 in a single subpopulation not present in any other subpopulation.
Evidence considered
ATM VCEP v1.6 PM2 rule: frequency <=0.001% in the gnomAD subpopulation with the highest frequency, applied as PM2_Supporting (not moderate), with the exception that >0.001% attributable to n=1 in a single subpopulation not present elsewhere is still sufficiently rare for PM2_Supporting.
gnomAD v4.1: total AF 6.82588e-06 (11/1,611,514 alleles), 0 homozygotes; exome AC 11/AN 1,460,382; genome AC 0; group-max filtering AF 2.92e-06 (joint) and 3.1e-06 (exome), maximum carried by European (non-Finnish).
gnomAD v2.1 (exome data): total AF 3.97924e-06 (1/251,304 alleles), 0 homozygotes; highest population frequency 8.80034e-06 (European non-Finnish, 1/113,632 alleles).
gnomAD v4.1 subpopulation detail: European (non-Finnish) 8/1,178,890 (6.78604e-06); Remaining individuals 3/62,362 (4.81062e-05); all other ancestry groups (African/African American, Admixed American, Ashkenazi Jewish, Amish, East Asian, Finnish, Middle Eastern, South Asian) show 0 alleles.
Non-cancer subsets, checked although the VCEP does not specify them: gnomAD v2.1 non-cancer exomes 1/236,766 alleles (4.22358e-06) with highest population frequency 9.74507e-06 (European non-Finnish, 0.000975%) and 0 homozygotes; gnomAD v3.1 non-cancer genomes absent; gnomAD-Canada v1.0 absent.
Homozygote observations: zero homozygotes in every dataset that contains the variant, i.e. no homozygous carrier has been observed among roughly 2.1 million gnomAD alleles - the pertinent frequency observation for a recessive condition such as ataxia telangiectasia.
ACMG/AMP 2015 (PMID:25741868) frames PM2 as observation at or below the frequency expected for the disorder in a race-matched general population cohort, and requires that read depth be sufficient for an accurate absence call; the ATM VCEP modifies this to Supporting strength with a 0.001% subpopulation ceiling.
Still missing
A strictly literal reading of the VCEP's 'subpopulation with the highest frequency' using raw allele frequency would use gnomAD v4.1 Remaining individuals (3/62,362 = 0.00481%), which exceeds 0.001% and is not rescued by the VCEP's n=1 exception (n=3, and the variant is present in more than one subpopulation); curator confirmation of the filtering-allele-frequency interpretation used here is advised.
The VCEP's condition that coverage at the locus be >30X could not be verified from the captured population datasets and was not measured independently; only the fact that the variant is called in the gnomAD v4.1 exome callset (AN 1,460,382) indicates an adequately covered, callable site.
PS1Not met: no established pathogenic variant shares the p.Asn2106Ile change (ClinVar 135768 is a 2-star VUS) and SpliceAI max delta 0.005 bars the splicing table.
Details ▾Hide ▴
Reasoning
The governing specification is the ClinGen HBOP ATM VCEP PS1 guide (cspec v1.6 doc 639508985 declares PS1 at default strength Pathogenic Strong; the VCEP index declares ATM_PS1_1.5.pdf as the governing PS1 file, and ATM_PS1_1.6.pdf carries materially identical text). Two sub-paths exist. (1) Standard missense PS1: 'Same amino acid change as a previously established pathogenic variant regardless of nucleotide change ... 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.' NM_000051.4:c.6317A>T produces NP_000042.3:p.(Asn2106Ile). Codon 2106 is AAC; the only single-nucleotide route to isoleucine (ATC) is c.6317A>T itself - the alternatives c.6317A>C (ACC, Thr) and c.6317A>G (AGC, Ser) encode different residues, and c.6316A>cT and c.6318C>cA/G/T encode Tyr/Lys/Lys, none of which is Ile. There is therefore no second, independently established nucleotide change at this codon that could serve as the 'previously established pathogenic variant', and ClinVar treats c.6317A>T itself as Uncertain significance (VCV000135768, variation 135768, 2-star, 'criteria provided, multiple submitters, no conflicts', five clinical laboratories, last updated 2026-06-27). Searches of every converted ATM VCEP text for 'c.6317A>T', 'p.Asn2106Ile', 'p.N2106I' and '2106' returned no per-variant PS1 entry, and the ATM PS1 splicing table is a class-level decision table with no variant rows. (2) Splicing-region PS1: this path requires the variant under assessment to have its own baseline predictive code (PP3(splicing), i.e. SpliceAI >= 0.2 by ATM VCEP's own threshold, or a PVS1 code at the donor/acceptor +/-1,2 dinucleotide positions) and requires its predicted event to precisely match that of a known (likely) pathogenic variant at the same position, with prediction strength equal to or higher than the reference. For this variant SpliceAI max delta is 0.005 (DS_AG 0.005, DS_AL 0.002, DS_DG 0.0, DS_DL 0.0; Pangolin SG 0.001, SL -0.006), well below the 0.2 threshold, so no predicted splice event exists to match; and no known pathogenic or likely pathogenic reference variant at the same nucleotide is documented in any VCEP file or in ClinVar (the codon-2106 alternates N2106Y/H/D/T/S/K appear in the Sun et al. 2025 saturation-screen supplement with an empty ClinVar classification column, i.e. not reported in ClinVar). PS1 is not met at any strength (Strong, Moderate or Supporting all fail for the same reason: there is no established pathogenic reference variant). The reduced-strength tiers of the splicing table are not reachable, so the absence of a reference variant is decisive rather than merely a tier downgrade.
Rule
ATM VCEP PS1 (v1.5/v1.6, cspec doc 639508985): Pathogenic Strong for 'same amino acid change as a previously established pathogenic variant regardless of nucleotide change', applicable to missense only when a splice defect is ruled out for both alterations, with reduced strengths (PS1_Moderate/PS1_Supporting) available only via the ATM PS1 splicing table when the VUA's predicted splice event matches a reference P/LP variant's. No such same-amino-acid or same-position reference variant exists; PS1 not met at any strength.
Evidence considered
ATM VCEP PS1 rule (cspec v1.6 doc 639508985): default strength Pathogenic Strong; instructions 'Use as ascribed for missense changes as long as a splice defect is ruled out for both variants; Use ATM PS1 Splicing table for splicing variants with similar predictions or observations of splice defect. (PMID: 36865205)'.
ATM_PS1_1.5.pdf (declared by the VCEP index as the governing PS1 file) page 1: 'PS1 Same amino acid change as a previously established pathogenic variant regardless of nucleotide change. 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.' Page 2 prerequisite: 'The predicted event of the VUA must precisely match the predicted event of the known (likely) pathogenic variant ... AND the strength of the prediction for the VUA must be of similar or higher strength than the strength of the prediction for the known (likely) pathogenic variant. (Likely) pathogenic variant should be assigned classification using VCEP specifications.' The ATM_PS1_1.6.pdf decision table is identical in structure and content.
The ATM PS1 splicing table contains no variant-specific rows - it is a class-level lookup keyed on the VUA's baseline code (PP3 or PVS1 tier), the VUA's position relative to the donor/acceptor +/-1,2 dinucleotide, and whether the reference variant is P or LP. No row can be populated for this variant because there is no reference P/LP variant at the same position.
Variant identity and codon arithmetic: NM_000051.4:c.6317A>T -> NP_000042.3:p.(Asn2106Ile), exon 43, chr11:108317491 (GRCh38). Codon 2106 is AAC; c.6317A>T yields ATC (Ile), which is the only single-nucleotide substitution producing isoleucine at this codon.
ClinVar (variation 135768, VCV000135768): 'Uncertain significance', review status 'criteria provided, multiple submitters, no conflicts' (2 stars), 5 submitters (Color Diagnostics, Natera, Ambry, Labcorp, GeneDx), conditions including ataxia-telangiectasia syndrome and hereditary cancer-predisposing syndrome - the variant itself is not an established pathogenic or likely pathogenic variant, so it cannot serve as the PS1 reference for any other change.
SpliceAI lookup for NM_000051.4:c.6317A>T: DS_AG 0.005, DS_AL 0.002, DS_DG 0.0, DS_DL 0.0, max delta 0.005; Pangolin SG 0.001, SL -0.006 - no predicted splice impact, so splicing is effectively ruled out for this allele and the VUA cannot be assigned the baseline PP3(splicing) code needed to enter the PS1 splicing sub-path.
Sun et al. 2025 (PMID 40580951) Table S1: the c.6316A>T (N2106Y), c.6317A>C (N2106T), c.6317A>G (N2106S) and c.6318T>A/G (N2106K) rows all have an empty ClinVar_classification field, whereas the c.6317A>T row is annotated 'Uncertain significance' - no alternate change at residue 2106 is reported as pathogenic in ClinVar.
The pipeline's PM5 comparator collection (output/pm5_candidates.json) reports total_candidates_seen = 0 and same_residue_candidates = 0, consistent with there being no established pathogenicity at this residue.
Still missing
Searched all converted ATM VCEP text at /Users/gs_agent/LYFESCI_HERMES_v9/vcep_db/ATM/fulltext (ATM_PS1_1.5.pdf.txt, ATM_PS1_1.6.pdf.txt, ATM_PVS1_1.5/1.6.pdf.txt, ATM_PM3_BP2_1.5/1.6.pdf.txt, clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx.txt, index.txt.txt) for 'c.6317A>T', 'p.Asn2106Ile', 'p.N2106I' and '2106': none of these files contains a per-variant entry for this variant, so there is no pre-assigned PS1 code to read off. This is a statement about the VCEP documents, not about absence of evidence.
No ClinVar-listed variant at ATM codon 2106 other than this one could be identified from the collected evidence: the pipeline's PM5 comparator search (output/pm5_candidates.json) returned 0 same-residue candidates, and the codon-2106 alternate substitutions in the Sun et al. 2025 supplement carry an empty ClinVar_classification column. A definitive ClinVar-wide query for every coding change at codon 2106 was not independently re-run for this group.
PS3Not met: the only functional result is the Sun 2025 prime-editing screen (PMID 40580951), which scores this variant Functional, and no VCEP-approved kinase or radiosensitivity assay was reported.
Details ▾Hide ▴
Reasoning
PS3 requires a functional assay demonstrating a damaging (loss-of-function) effect on ATM. The governing specification is the ClinGen HBOP ATM VCEP v1.6 (cspec), whose PS3 rules permit Moderate strength only when a variant fails to rescue BOTH an ATM-specific feature (e.g. phosphorylation of ATM-specific targets) AND radiosensitivity, Supporting strength when only the ATM-specific feature fails, and explicitly state 'Do not use as strong'; a radiosensitivity-only result gets no weight. The VCEP's own approved-assay calibration table (clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx) names exactly three approved assays - Mitui 2009 (PMID 18634022), Barone 2009 (PMID 19431188) and Scott 2002 (PMID 11805335), all stable transfection of ATM cDNA constructs into an ATM-null LCL line with Western-blot readout of ATM substrate phosphorylation - and this table carries no per-variant rows, so it cannot itself supply a variant-level PS3 answer. None of those three publications is present in this case (the publications directory holds only 25741868, 24418350 and 25394175) and c.6317A>T appears in none of them, so no VCEP-approved kinase-activity or radiosensitivity assay result exists for this variant. The only functional measurement located anywhere for this variant is in Sun et al. 2025 (PMID 40580951) Supplementary Table S1, where hg38_pos 108317491 / c.6317A>T / N2106I is scored Combined_score -0.754 with Classification 'Functional', Confidence 'High' and DeepATM_predicted 'No'. DeepATM_predicted=No identifies this as a direct prime-editing saturation-genome-editing measurement rather than a model extrapolation; within that table the Functional band spans -0.912 to +8.287 while Intermediate spans -1.360 to -0.912, so -0.754 sits inside the function-retained range. A retained-function result is the opposite of the loss-of-function effect PS3 requires and therefore cannot support PS3 in any direction. ClinVar (ClinVarID 135768, Uncertain significance, 5 clinical-laboratory submissions, no expert-panel review) and OncoKB (no variant-specific reviewed functional evidence) are consistent with the absence of PS3-supporting data.
Rule
ClinGen HBOP ATM VCEP v1.6 PS3: PS3_Moderate = variant fails to rescue both an ATM-specific feature (e.g. phosphorylation of ATM-specific targets) AND radiosensitivity; PS3_Supporting = variant fails to rescue an ATM-specific feature only; PS3 at Strong is not permitted; radiosensitivity-only results receive no weight.
Evidence considered
Governing specification: ClinGen HBOP ATM VCEP v1.6 (cspec). PS3 rules: PS3_Moderate = the variant fails to rescue BOTH an ATM-specific feature (e.g. phosphorylation of ATM-specific targets) AND radiosensitivity; PS3_Supporting = the variant fails to rescue an ATM-specific feature only; PS3 at Strong is explicitly disallowed ('Do not use as strong'); radiosensitivity-only has no weight because it is not a feature specific to ATM deficiency.
Generic ACMG/AMP (PMID 25741868) defines PS3 as requiring a well-established functional study showing a damaging effect; that paper contains no ATM-specific or variant-specific data (its illustrative genes are MYH7, CFTR, NIPBL and APOE), so it is used only as the definitional source for PS3/BS3.
The VCEP's approved-assay calibration table (clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx) lists three approved kinase-activity assays (Mitui 2009/PMID 18634022, Barone 2009/PMID 19431188, Scott 2002/PMID 11805335; stable transfection of ATM cDNA constructs into an ATM-null LCL line with Western blotting of ATM substrates S1981/SMC1/NBS1/CHK2/p53 after irradiation) and radiosensitivity assays (Mitui 2009 approved; Scott 2002 x2), proposing PS3_Supporting for kinase-activity failure and PS3_Moderate for radiosensitivity only when a kinase-activity result is also positive, with no weight if the two conflict.
Searched that approved-assay table for every notation of this variant ('c.6317A>T', 'p.Asn2106Ile', 'p.N2106I'): it has no per-variant rows at all (47 lines of methodology/calibration), so it declares which assays the VCEP accepts but does not report whether c.6317A>T was ever tested in them.
None of the three approved-assay papers is available in this case (source_registry/publications contain only PMID:25741868, PMID:24418350 and PMID:25394175), and c.6317A>T appears in none of the fetched papers; therefore no result from a VCEP-approved assay can be read off for this variant.
DeepATM_predicted 'No' with Confidence 'High' means the Functional call comes from the underlying prime-editing saturation-genome-editing screen (direct measurement), not from the DeepATM model's extrapolation; across the table 24,534 rows are DeepATM_predicted=No and 4,421 are Yes/Computational prediction, so the distinction is material.
Calibration of that score from the table itself: Classification-wise, Functional rows span Combined_score -0.912 to +8.287 (n=20,738), Intermediate -1.360 to -0.912 (n=1,859) and Non-functional -20.804 to -1.360 (n=6,358). The variant's -0.754 therefore lies inside the Function-retained band, about 0.16 above the Functional/Intermediate boundary and far above the Non-functional boundary; a nonsense control in the same table, c.5908C>T (Q1970*), scores -2.81 with Classification 'Non-functional'.
Interpretation for PS3: the sole functional result is positive for retained function, so there is no dysfunction signal of any kind to convert into PS3; the assay is also not one the VCEP has approved or calibrated, and no approved-assay result exists. PS3 is therefore not met, and the direction of the only available functional evidence is against it.
ClinVar (key 'clinvar', ClinVarID 135768) reports Uncertain significance based on 5 clinical-laboratory submissions with no expert-panel submission, and OncoKB (key 'oncokb') explicitly reports that it 'did not identify variant-specific reviewed functional evidence for this variant'; neither source contributes PS3 evidence.
Still missing
No VCEP-approved functional assay has been performed for c.6317A>T: the variant does not appear in any of the three approved-assay publications (Mitui 2009 PMID 18634022, Barone 2009 PMID 19431188, Scott 2002 PMID 11805335), and none of those three papers is available in this case (publications/ contains only 25741868, 24418350, 25394175). The absence of a VCEP-approved assay is a gap in the evidence base, not proof that the variant has normal function.
Searched vcep_db/ATM/fulltext/clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx.txt for 'c.6317A>T', 'p.Asn2106Ile' and 'p.N2106I' (and the whole vcep_db/ATM tree): that file contains no variant-level rows whatsoever - it is assay methodology and strength calibration only - so no entry for this variant exists there.
Searched vcep_db/ATM/fulltext/Suppl_TableS1_PMID 40580951.xlsx.txt for 'c.6317A>T', 'p.Asn2106Ile' and 'p.N2106I': a single matching row exists (hg38_pos 108317491) and is the only functional measurement found for this variant.
No RNA-based or patient-cell-based functional data (ATM protein level, kinase activity in patient lymphoblasts, radiosensitivity of patient cells) was found in the case materials or in any fetched publication.
PS4Not met: no case-control odds ratio, hazard ratio or relative risk is reported for ATM c.6317A>T, so the VCEP's OR >=2 (or lower CI >=1.5) threshold has nothing to satisfy it.
Details ▾Hide ▴
Reasoning
Governing specification: ClinGen Hereditary Breast, Ovarian and Pancreatic Cancer (HBOP) Expert Panel Specifications to the ACMG/AMP Guidelines for ATM v1.6 (cspec doc 639508985), which marks PS4 'Applicable' with exactly one strength. The PS4 (Strong) rule, quoted verbatim from that criterion's own text: 'Case-control studies; p-value <=.05 **and** (Odds ratio, hazard ratio, or relative risk >=2 **or** lower 95% CI >=1.5).' Operator direction, taken from this criterion's own operator and not from any neighbouring criterion in this group: the enrichment statistic strengthens as the value goes UP, so the test is 'value >= 2' (or 'lower 95% CI >= 1.5'), with p <= 0.05 as a co-requirement. Explicit substitution: no odds ratio, hazard ratio, relative risk or confidence interval is reported anywhere for ATM c.6317A>T (p.Asn2106Ile), so the inequality has no left-hand side and cannot be satisfied ('OR >= 2? no value available'). The four PMIDs triaged as PS4 candidates (PMID:20301317, PMID:20301790, PMID:24418350, PMID:25394175) are narrative guidelines and disease overviews (chronic-ataxia work-up, cancer-referral indications), not case-control studies; the two retrieved in full text (PMID:24418350, PMID:25394175) never mention c.6317A>T or p.(N2106I) and report no enrichment data for any ATM variant. The generic ACMG fallback (ACMG PS4 Note 2, prior observation in multiple unrelated patients with absence in controls) is unavailable here because the VCEP states 'PS4_Moderate: Do not use. Proband counting for genes causing a common disorder need to be calibrated in a population-specific way before use.' The only variant-level quantitative data are population frequencies with no case/control contrast (gnomAD v2.1 total AF 3.98e-06, 1/251,304 alleles, 0 homozygotes; gnomAD v4.1 total AF 6.83e-06, 11/1,611,514 alleles, 0 homozygotes), and ClinVar VCV000135768 carries only Uncertain significance assertions from five ordinary laboratories. PS4 is therefore not met.
Rule
PS4 (Strong) = Case-control studies; p-value <=.05 and (Odds ratio, hazard ratio, or relative risk >=2 or lower 95% CI >=1.5). PS4_Moderate: Do not use.
Evidence considered
HBOP ATM VCEP v1.6 PS4 rule quoted verbatim: 'Case-control studies; p-value <=.05 **and** (Odds ratio, hazard ratio, or relative risk >=2 **or** lower 95% CI >=1.5).' Direction read from this criterion's own text: value must be >= 2, or lower 95% CI >= 1.5; the statistic strengthens upward.
No case-control, cohort or enrichment study reports ATM c.6317A>T (p.Asn2106Ile): the four PMIDs flagged as PS4 candidates are a chronic-ataxia consensus guideline, an ataxia diagnostic guideline and two cancer/ataxia guideline documents; the two retrieved in full text (PMID:24418350, PMID:25394175) contain no mention of the variant and no odds ratio, hazard ratio or relative risk.
VCEP instruction 'PS4_Moderate: Do not use. Proband counting for genes causing a common disorder need to be calibrated in a population-specific way before use' removes the generic ACMG multiple-unrelated-patient fallback for ATM.
Variant-level rarity only, with no case/control contrast: gnomAD v2.1 total AF 3.97924e-06 (1/251,304 alleles, 0 homozygotes) and gnomAD v4.1 total AF 6.82588e-06 (11/1,611,514 alleles, 0 homozygotes).
ClinVar VCV000135768 contains only Uncertain significance assertions (5 clinical laboratories), i.e. no observed enrichment in affected individuals from any source.
Still missing
No case-control, cohort or case-series enrichment study of ATM c.6317A>T was identified; the variant remains a single-allele gnomAD observation with no affected-versus-control data.
GeneReviews chapters PMID:20301317 (Hereditary Ataxia Overview) and PMID:20301790 (Ataxia-Telangiectasia) were triaged as PS4/PP5 candidates but had no retrievable full text in this case; both are narrative disease overviews that would not contain case-control enrichment data for a specific ATM missense variant.
No patient-level phenotype or family history accompanies this variant in the case record, so no genotype-phenotype enrichment can be assembled locally.
PP3Not met: missense REVEL 0.539 sits in the gray zone, below the ATM VCEP PP3 threshold of REVEL >0.7333.
Details ▾Hide ▴
Reasoning
NM_000051.4:c.6317A>T is a missense change (NP_000042.3:p.(Asn2106Ile)), so the applicable computational path for PP3 is the missense protein-impact path (REVEL), not splicing. The governing specification is ClinGen HBOP ATM VCEP v1.6 (framework_mode=vcep), whose PP3 rule for missense is REVEL > 0.7333. The REVEL score for this variant is 0.539, from the pipeline's local REVEL v1.3 lookup (source_registry key 'revel'), and the same value (0.539) is reported in the VCEP-declared supporting table Suppl_TableS1_PMID 40580951 (row c.6317A>T / N2106I, exon 43, Missense), so the two independent reads agree exactly. 0.539 is well below the VCEP threshold of 0.7333 and also below the generic ClinGen SVI supporting threshold of 0.644 (Pejaver et al. 2022, PMID 36413997), so PP3 is not met at any strength. The variant's SpliceAI result (max delta 0.005-0.01, no predicted impact) was not used here: it belongs to the splice path and says nothing about tolerance of the Asn2106Ile substitution; conversely it does not create BP4 support either. Secondary computational columns in the VCEP table are concordantly non-supportive for a pathogenic call (AlphaMissense 0.2578 - ambiguous/low range, EVE 0.695, BayesDel -0.028517, all treated as corroboration only, never as the primary source), and none of them changes the REVEL-based verdict.
Variant consequence settled first: NM_000051.4:c.6317A>T = NP_000042.3:p.(Asn2106Ile), a missense variant, so the REVEL path applies and the SpliceAI path does not.
Local REVEL v1.3 lookup returned 0.539 for this variant (source_registry key 'revel').
VCEP-declared supporter searched for all three notations with grep -rn over /Users/gs_agent/LYFESCI_HERMES_v9/vcep_db/ATM/fulltext: 'c.6317A>T', 'p.Asn2106Ile', 'p.N2106I'. Only Suppl_TableS1_PMID 40580951.xlsx.txt matched (line 20199; row: hg38 108317491 A>T, c.6317A>T, N2106I, exon_43, Missense, REVEL 0.539, AlphaMissense 0.2578, CADD 21.5, EVE 0.695, GnomAD_all_AF 7.53e-06, ClinVar Uncertain significance, Combined_score -0.754, Classification Functional, Confidence High, DeepATM_predicted No).
That table pre-assigns no PP3 or BP4 code for this variant; it carries only functional-classification and computational-predictor columns, so no code was available to apply directly. Its REVEL column corroborates the local lookup rather than replacing it.
Generic ClinGen SVI REVEL calibration consulted and also not met (supporting PP3 requires >=0.644; Pejaver et al. 2022, PMID 36413997).
SpliceAI result for this variant (max delta 0.005-0.01) was deliberately not used as evidence for this missense variant's PP3, per the one-path rule.
Still missing
The full text of PMID 40580951 (Sun et al. 2025) was not fetched for this case; only the VCEP-declared supplement table Suppl_TableS1_PMID 40580951.xlsx was available and searched. Its per-variant REVEL column (0.539) agrees with the independent local REVEL lookup, so the PP3/BP4 missense verdict does not depend on the unfetched manuscript.
PP5Not met: ClinVar VCV000135768 is Uncertain significance at two stars with five ordinary laboratory submissions and zero expert-panel submissions.
Details ▾Hide ▴
Reasoning
Under the applied rule PP5 is triggered only by an exact-variant ClinVar expert-panel classification of Pathogenic or Likely pathogenic; aggregate ClinVar labels, ordinary laboratory submissions and non-expert assertions must never trigger it. The exact-match ClinVar record for NM_000051.4:c.6317A>T (p.Asn2106Ile) is VCV000135768 (ClinVar ID 135768), classified Uncertain significance with review status 'criteria provided, multiple submitters, no conflicts' (2 stars). All five submissions come from ordinary clinical-testing laboratories: Color Diagnostics (SCV000682322), Natera (SCV002083429), Ambry Genetics (SCV000660683), Labcorp Genetics/Invitae (SCV000166130) and GeneDx (SCV001990228); each is flagged is_expert_panel = false, the expert-panel submission count is 0, and the expert-panels map is empty. No submission asserts Pathogenic or Likely pathogenic, and none of the discordant-labelled or no-criteria records qualifies as an expert assertion either. The governing HBOP ATM VCEP additionally lists PP5 as 'Not Applicable for this VCEP' ('This criterion is not for use as recommended by the ClinGen Sequence Variant Interpretation VCEP Review Committee'); the lab's expert-panel override was checked and its precondition (an exact-variant expert-panel P/LP classification) is absent, so the VCEP exclusion is not what decides this call — the absence of any expert-panelled assertion is. PP5 is therefore not met.
Rule
Not Applicable for this VCEP (SVI VCEP Review Committee). Lab override: apply PP5 at Supporting only when an exact-variant ClinVar expert-panel classification is Pathogenic or Likely pathogenic.
Evidence considered
cspec: PP5 applicability 'Not Applicable for this VCEP' - 'This criterion is not for use as recommended by the ClinGen Sequence Variant Interpretation VCEP Review Committee.'
Applied rule: PP5 requires an exact-variant ClinVar expert-panel classification of Pathogenic/Likely pathogenic; ordinary laboratory submissions, aggregate ClinVar labels and non-expert assertions never trigger it.
ClinVar VCV000135768 (variation ID 135768, exact transcript-HGVS match): classification Uncertain significance; review status 'criteria provided, multiple submitters, no conflicts' (2 stars); 5 submissions from 5 submitters.
All five submissions are ordinary clinical laboratories (Color Diagnostics, Natera, Ambry Genetics, Labcorp/Invitae, GeneDx); is_expert_panel = false for every submission; expert-panel submissions = 0; expert-panels map empty.
No Pathogenic or Likely pathogenic assertion for this exact variant exists from any expert panel, so the PP5 trigger condition is absent.
Still missing
If an expert panel later issues an exact-variant Pathogenic/Likely pathogenic classification for c.6317A>T, PP5 would then apply at Supporting strength and this call should be revisited.
BA1Not met: gnomAD v4.1 group-max filtering AF 0.000292% is roughly 1,700-fold below the VCEP stand-alone benign threshold of 0.5%.
Details ▾Hide ▴
Reasoning
The governing specification is the ClinGen HBOP ATM VCEP v1.6 (cspec doc 639508985), which defines BA1 as a stand-alone benign criterion requiring a Grpmax Filtering AF >0.5% in gnomAD. gnomAD v4.1 reports a group-max filtering allele frequency of 2.92e-06 (0.000292%), with gnomAD designating European (non-Finnish) as the group carrying that maximum; the total frequency is 6.83e-06 (11/1,611,514 alleles, 0 homozygotes, all in the exome callset). gnomAD v2.1 (exome data; no grpmax FAF reported) yields a total AF of 3.98e-06 (1/251,304) and a highest-population AF of 8.80e-06 (0.00088%, European non-Finnish). Even the highest raw subpopulation frequency available in any dataset - gnomAD v4.1 Remaining individuals, 3/62,362 = 0.00481% - remains more than 100-fold below the 0.5% threshold. ATM-related disease prevalence and penetrance do not alter this: the variant is far too rare in every population to constitute stand-alone benign frequency evidence. BA1 is not met.
Rule
BA1 stand-alone benign, ATM VCEP v1.6: Grpmax Filtering AF >0.5% in gnomAD.
Evidence considered
ATM VCEP v1.6 BA1 rule: 'Grpmax Filtering AF >.5% in gnomAD', stand-alone benign strength, with the specification noting that the filtering allele frequency is a statistical model that accounts for population size and founder populations.
gnomAD v4.1: total allele frequency 6.82588e-06 (11/1,611,514 alleles), 0 homozygotes, genomes 0 alleles (all 11 observed in exomes, exome AN 1,460,382); group-max filtering AF 2.92e-06 (joint; exome grpmax FAF 3.1e-06) with the maximum carried by European (non-Finnish).
gnomAD v2.1 (exome data only): total AF 3.97924e-06 (1/251,304 alleles), 0 homozygotes; highest population frequency 8.80034e-06 (European non-Finnish, 1/113,632 alleles).
Highest raw subpopulation frequency across all captured datasets: gnomAD v4.1 Remaining individuals 4.81062e-05 (3/62,362 alleles, 0 homozygotes), which is still >100-fold below the 0.5% BA1 threshold.
ACMG/AMP 2015 (PMID:25741868) defines stand-alone benign support as an allele frequency above 5% in a control population, consistent with the VCEP threshold being applied here.
No population-frequency evidence approaches the BA1 threshold; disease prevalence for ATM-related cancer predisposition is far below the frequency that would be required for a stand-alone benign call.
Still missing
gnomAD v2.1 does not report a grpmax filtering allele frequency, so the VCEP's grpmax rule was applied to gnomAD v4.1; v2.1's highest population frequency (0.00088%) is likewise orders of magnitude below the threshold, so the verdict is independent of dataset choice.
BS1Not met: the highest gnomAD frequency, group-max filtering AF 0.000292%, is about 170-fold below the VCEP strong benign threshold of 0.05%.
Details ▾Hide ▴
Reasoning
The ATM VCEP v1.6 defines BS1 (strong benign) as a Grpmax Filtering AF >0.05% in gnomAD. gnomAD v4.1 gives a group-max filtering AF of 2.92e-06 (0.000292%), roughly 170-fold below the threshold, with a total frequency of 6.83e-06 (11/1,611,514 alleles, 0 homozygotes). gnomAD v2.1, which reports no grpmax FAF, gives a total AF of 3.98e-06 (1/251,304) and a highest-population AF of 8.80e-06 (0.00088%) - also far below 0.05%. The maximum raw subpopulation value in any captured dataset (gnomAD v4.1 Remaining individuals, 0.00481%) is itself still about 10-fold below 0.05%. The variant is therefore too rare to support a strong benign frequency argument, and the threshold is not approached under any dataset or subpopulation selection.
Rule
BS1 strong benign, ATM VCEP v1.6: Grpmax Filtering AF >0.05% in gnomAD.
Evidence considered
ATM VCEP v1.6 BS1 rule: 'Grpmax Filtering AF >.05% in gnomAD', strong benign strength.
gnomAD v4.1: group-max filtering AF 2.92e-06 (0.000292%); total AF 6.82588e-06 (11/1,611,514 alleles); 0 homozygotes; all alleles in the exome callset.
gnomAD v2.1 (exomes): total AF 3.97924e-06 (1/251,304 alleles); highest population frequency 8.80034e-06 (European non-Finnish); 0 homozygotes.
Non-cancer subsets checked for completeness (the VCEP does not specify them): gnomAD v2.1 non-cancer exomes total AF 4.22358e-06 (1/236,766) with highest population frequency 9.74507e-06 (European non-Finnish, 0.000975%); gnomAD v3.1 non-cancer genomes absent. Neither approaches 0.05%.
ACMG/AMP 2015 (PMID:25741868) frames BS1 as an allele frequency greater than expected for the disorder; at ~0.0003% this variant is far below any frequency expected for ATM-related disease.
Still missing
gnomAD v2.1 reports no grpmax filtering allele frequency, so the VCEP's grpmax rule was applied to gnomAD v4.1; the v2.1 highest-population frequency (0.00088%) is also far below the threshold.
BS3Not met: the prime-editing screen directly measures this variant as Functional (Combined score -0.754, above the -0.912 Functional boundary), but that assay is not a VCEP-approved or calibrated ATM test.
Details ▾Hide ▴
Reasoning
BS3 requires a well-established functional assay showing no damaging effect - for ATM, rescue of ATM function - converted to the strength the VCEP has calibrated. ClinGen HBOP ATM VCEP v1.6 specifies BS3_Moderate when a variant rescues BOTH an ATM-specific feature (e.g. phosphorylation of ATM-specific targets) AND radiosensitivity, and BS3_Supporting when it rescues EITHER. The VCEP's approved-assay calibration table names only three assays (Mitui 2009/PMID 18634022, Barone 2009/PMID 19431188, Scott 2002/PMID 11805335; stable transfection into ATM-null LCLs with ATM-substrate phosphorylation readout, plus radiosensitivity for Mitui 2009 and Scott 2002), and none of those three publications is available in this case and none reports c.6317A>T, so no VCEP-approved rescue result exists for this variant. The only functional evidence located is the prime-editing saturation-genome-editing screen of Sun et al. 2025 (PMID 40580951, Supplementary Table S1), in which c.6317A>T / N2106I is directly measured (DeepATM_predicted = No, Confidence = High) and classified 'Functional' with Combined_score -0.754, inside the table's Functional band of -0.912 to +8.287 and clearly distinct from the Intermediate (-1.360 to -0.912) and Non-functional bands. Directionally that result is BS3-consistent, since it indicates function is retained rather than lost. However, that assay is not one of the VCEP's three approved PS3/BS3 assays and measures neither readout the VCEP requires - ATM substrate phosphorylation nor cellular radiosensitivity - and it postdates the VCEP v1.5 calibration on which the approved-assay table was built, so the VCEP has assigned it no strength. Applying BS3 at any strength would therefore amount to inventing an uncalibrated strength, including the VCEP's own prohibition on giving radiosensitivity-only data weight. No VCEP-approved kinase-activity or radiosensitivity data is available to corroborate or contradict the SGE result, so BS3 is not met at a VCEP-calibrated strength, and the call is flagged for human review because the single uncalibrated functional result points benign-ward.
Rule
ClinGen HBOP ATM VCEP v1.6 BS3: BS3_Moderate = variant rescues both an ATM-specific feature (e.g. phosphorylation of ATM-specific targets) AND radiosensitivity; BS3_Supporting = variant rescues either an ATM-specific feature OR radiosensitivity. Radiosensitivity alone carries no weight, and conflicting kinase/radiosensitivity results carry no weight.
Evidence considered
Governing specification: ClinGen HBOP ATM VCEP v1.6 (cspec). BS3 rules: BS3_Moderate = the variant rescues BOTH an ATM-specific feature (e.g. phosphorylation of ATM-specific targets) AND radiosensitivity; BS3_Supporting = the variant rescues EITHER. Only these two strengths are specified.
Approved-assay calibration table (clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx): the 'ATM kinase activity' sheet approves Mitui 2009 (PMID 18634022, qualitative S1981/SMC1 Western blot, normal readout per WT construct pMAT1), Barone 2009 (PMID 19431188, densitometric SMC1-S966/NBS1-S343/CHK2-T68/p53-S15/S1981, normal = WT level, abnormal = non-detectable or reduced) and Scott 2002 (PMID 11805335, p53-S15 in vivo/in vitro), each with positive (WT cDNA) and negative (ATM-null or empty vector) controls and no biological replicates formally met for Barone/Scott; all three are proposed at PS3_Supporting/BS3_Supporting. The 'ATM radiosensitivity' sheet approves only Mitui 2009 (survival fraction; normal SF >36%, radiosensitive SF <21%) and lists two Scott 2002 readouts (induced chromosome aberrations, ~1 ICA normal vs 2.98-3.20 ICAs abnormal; and viability), with the explicit notes 'No weight should be applied if only radiosensitivity is available' and 'No weight should be applied if radiosensitivity and kinase assay are conflicting'.
Searched that approved-assay table for every notation of this variant ('c.6317A>T', 'p.Asn2106Ile', 'p.N2106I'): no entry exists, because the file reports assay methodology and calibration rather than per-variant results. Stated explicitly so the absence of a hit is not mistaken for evidence of a benign functional result.
No VCEP-approved assay data for this variant: none of PMID 18634022, PMID 19431188 or PMID 11805335 is present in this case's publications directory (which holds only 25741868, 24418350 and 25394175), and the variant is not mentioned in any fetched publication, so both the rescue-of-ATM-specific-feature and rescue-of-radiosensitivity arms of the BS3 rules are unassessed for c.6317A>T.
Only functional evidence located: Sun et al. 2025 (PMID 40580951) Supplementary Table S1, row hg38_pos 108317491 -> c.6317A>T / N2106I, exon_43, Missense, Combined_score -0.754, Classification 'Functional', Confidence 'High', DeepATM_predicted 'No' (all other fields of that row: CADD.phred 21.5, boostDM 0.0725, EVE 0.695, REVEL 0.539, AlphaMissense 0.2578, GnomAD_all_AF 7.53e-06, ClinVar_classification 'Uncertain significance').
The Functional/Intermediate/Non-functional banding of that table supports the 'Functional' label for this variant: Functional rows span Combined_score -0.912 to +8.287 (n=20,738), Intermediate -1.360 to -0.912 (n=1,859), Non-functional -20.804 to -1.360 (n=6,358); -0.754 is inside the Functional band, ~0.16 above the Functional/Intermediate boundary, and the table's own nonsense example c.5908C>T (Q1970*) sits at -2.81/Non-functional. DeepATM_predicted 'No' (24,534 rows) versus 'Yes'/Computational prediction (4,421 rows) confirms this row was directly assayed rather than model-extrapolated.
Reason BS3 is recorded as not met despite a directionally benign result: the prime-editing saturation-genome-editing assay measures neither of the VCEP's two specified readouts (ATM substrate phosphorylation, cellular radiosensitivity), it is not among the VCEP's three approved assays, and the approved-assay table predates it (VCEP v1.5). The VCEP has therefore calibrated no strength for it, and applying BS3_Supporting or BS3_Moderate from it would be an uncalibrated assignment. The result is retained as functional evidence in evidence_used and flagged for human review rather than being converted into a strength.
ClinVar (key 'clinvar', ClinVarID 135768) is Uncertain significance (5 clinical-laboratory submissions, no expert-panel review) and OncoKB (key 'oncokb') reports no variant-specific reviewed functional evidence; both are consistent with the absence of an approved-assay BS3 result.
Generic ACMG/AMP (PMID 25741868) supplies the definition of BS3 ('well-established functional studies show no damaging effect'); it contains no ATM-specific or variant-specific content and is cited only as the definitional source.
Still missing
No VCEP-approved rescue assay exists for c.6317A>T: it does not appear in Mitui 2009 (PMID 18634022), Barone 2009 (PMID 19431188) or Scott 2002 (PMID 11805335), and none of those three papers is available in this case, so the BS3_Supporting (rescue of either readout) versus BS3_Moderate (rescue of both) determination cannot be made.
Searched vcep_db/ATM/fulltext/clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx.txt for 'c.6317A>T', 'p.Asn2106Ile' and 'p.N2106I': the file has no variant-level entries (it is methodology/calibration only), so no BS3 strength can be read off it for this variant.
Searched vcep_db/ATM/fulltext/Suppl_TableS1_PMID 40580951.xlsx.txt for 'c.6317A>T', 'p.Asn2106Ile' and 'p.N2106I': one matching row was found (hg38_pos 108317491, Classification 'Functional', Confidence 'High', DeepATM_predicted 'No') and it is the sole functional evidence for this variant.
Human review required: decide whether the Sun et al. 2025 prime-editing assay (PMID 40580951) can be accepted as an established functional assay for ATM BS3, and if so what strength it warrants. The ATM VCEP has not calibrated this assay, and it postdates the v1.5 approved-assay table, so no defensible strength can be assigned automatically.
No ATM protein-level, kinase-activity or patient-cell radiosensitivity data for this variant was found in the case materials or any fetched publication, so the SGE functional call cannot be corroborated by a second, independent assay.
BP4Not met: missense REVEL 0.539 exceeds the ATM VCEP BP4 cutoff of REVEL <=0.249, and the clean SpliceAI score cannot substitute.
Details ▾Hide ▴
Reasoning
Same scope determination as PP3: c.6317A>T is a missense variant (p.(Asn2106Ile)), so benign computational support must come from the missense REVEL path. ClinGen HBOP ATM VCEP v1.6 BP4 rule for missense is REVEL <= 0.249. The measured REVEL is 0.539 (local REVEL v1.3 lookup; identical 0.539 in the VCEP-declared Table S1), which is far above both the VCEP cutoff of 0.249 and the generic ClinGen SVI supporting cutoff of <= 0.29 (Pejaver et al. 2022, PMID 36413997). BP4 is therefore not met - this is a gray-zone REVEL result, which explicitly means BP4 is not met and does not license reaching across to the splice path. The variant's clean SpliceAI result (max delta 0.005-0.01, below the <= 0.1 BP4 splice cutoff) cannot be used as BP4 support because SpliceAI measures splice impact, not missense tolerance; splicing and protein impact are separate mechanisms and the VCEP instructs against applying the wrong type of computational evidence. No VCEP supporting material pre-assigns a BP4 code for this variant.
Local REVEL v1.3 lookup returned 0.539 (source_registry key 'revel'), independent of and identical to the REVEL column value 0.539 in the VCEP-declared Table S1 row for c.6317A>T / N2106I.
Generic ClinGen SVI REVEL calibration also not met (supporting BP4 requires <= 0.29; Pejaver et al. 2022, PMID 36413997).
SpliceAI max delta 0.005-0.01 is below the VCEP splice BP4 cutoff of <= 0.1, but was not applied: the variant is missense, so the splice path is out of scope for this criterion and a clean splice prediction is not BP4 support for a missense change.
No PP3/BP4 code for this variant was found in any VCEP-declared file after grep of 'c.6317A>T', 'p.Asn2106Ile' and 'p.N2106I' across /Users/gs_agent/LYFESCI_HERMES_v9/vcep_db/ATM/fulltext.
Still missing
No RP/BP pre-assigned code for this variant exists in the VCEP-declared Suppl_TableS1_PMID 40580951.xlsx, so BP4 rests entirely on the missense REVEL path, which is indeterminate at 0.539.
BP6Not met: no expert-panel submission or Benign classification exists for c.6317A>T; VCV000135768 is Uncertain significance at two stars.
Details ▾Hide ▴
Reasoning
Under the applied rule BP6 is triggered only by an exact-variant ClinVar expert-panel classification of Benign or Likely benign; ordinary laboratory submissions and aggregate ClinVar labels must never trigger it. The exact-match record VCV000135768 (ClinVar ID 135768) for NM_000051.4:c.6317A>T (p.Asn2106Ile) is classified Uncertain significance at 2 stars ('criteria provided, multiple submitters, no conflicts'). All five submissions are from ordinary clinical-testing laboratories (Color Diagnostics, Natera, Ambry Genetics, Labcorp/Invitae, GeneDx), every one flagged is_expert_panel = false, expert-panel submissions = 0, and the expert-panels map is empty; no submission reports Benign or Likely benign. Because a benign expert-panel classification is the sole trigger, and none exists for this exact variant, BP6 is not met. The governing HBOP ATM VCEP also lists BP6 as 'Not Applicable for this VCEP' (SVI VCEP Review Committee recommendation), so the criterion could not be applied at a VCEP-calibrated strength in any event.
Rule
Not Applicable for this VCEP (SVI VCEP Review Committee). Lab override: apply BP6 at Supporting only when an exact-variant ClinVar expert-panel classification is Benign or Likely benign.
Evidence considered
cspec: BP6 applicability 'Not Applicable for this VCEP' - 'This criterion is not for use as recommended by the ClinGen Sequence Variant Interpretation VCEP Review Committee.'
Applied rule: BP6 requires an exact-variant ClinVar expert-panel classification of Benign/Likely benign; ordinary laboratory submissions and aggregate ClinVar labels never trigger it.
ClinVar VCV000135768: classification Uncertain significance; review status 'criteria provided, multiple submitters, no conflicts' (2 stars); 5 submissions, 5 submitters.
All five submissions are ordinary clinical laboratories (Color Diagnostics, Natera, Ambry Genetics, Labcorp/Invitae, GeneDx); is_expert_panel = false for every submission; expert-panel submissions = 0; no Benign or Likely benign assertion appears in any submission.
No benign expert-panel classification for c.6317A>T therefore exists to support BP6.
Still missing
If an expert panel later classifies this exact variant as Benign or Likely benign, BP6 would apply at Supporting strength and this call should be revisited.
PM3Not assessed: no A-T proband carries c.6317A>T with a second ATM pathogenic variant in trans, so the VCEP table's minimum 1.0 PM3 point is unattainable.
Details ▾Hide ▴
Reasoning
The governing framework is the ClinGen HBOP ATM VCEP specification (CSPEC v1.6, mirrored in ATM_PM3_BP2_1.5.pdf and ATM_PM3_BP2_1.6.pdf), which replaces the generic PM3 rule with a points table keyed to Ataxia-Telangiectasia (A-T) probands and the phase/classification of the co-occurring ATM allele: confident phenotype = 4.0 points confirmed in trans, 2.0 phase unknown, 1.0 second variant unidentified or VUS, 2.0 homozygous (max 2 individuals); consistent phenotype = half those values; points convert to PM3_Supporting at 1.0, PM3 (Moderate) at 2.0, PM3_Strong at 4.0 and PM3_VeryStrong at 8.0. The variant is eligible on rarity: gnomAD v4.1 total AF 6.83e-06 (11/1,611,514 alleles, 0 homozygotes) with highest subpopulation AF 4.81e-05 (3/62,362, Remaining individuals) and grpmax FAF 2.92e-06, all below the VCEP's 0.01% general-population-frequency ceiling, and gnomAD v2.1 AF 3.98e-06 (1/251,304). However, the case contains no qualifying proband observation at all: ClinVar variation 135768 carries five clinical-laboratory submissions, all Uncertain significance, none reporting an A-T phenotype, a second ATM variant, or phase (no expert-panel submission), and the co-occurring variant would in any case have to be classified under the VCEP specifications rather than assumed. No affected proband homozygous for c.6317A>T, no confirmed/unknown-phase trans observation, and no affected individual from a consanguineous union or with bilaterally ATM-consistent cancer family history is documented, so the summed PM3 points are zero and no strength tier can be reached. The variant's own somatic occurrence in COSMIC (2 records) is oncological and does not bear on germline phase in A-T. PM3 therefore cannot be assessed rather than being affirmatively refuted by negative evidence. The files ATM_PM3_BP2_1.5.pdf and ATM_PM3_BP2_1.6.pdf were searched under c.6317A>T, p.Asn2106Ile and p.N2106I and contain no variant-specific entry (they are generic per-proband points tables, not variant-keyed assertions), and the supplied literature extracts likewise report no exact-variant observation.
Rule
ClinGen HBOP ATM VCEP PM3/BP2 points table (ATM_PM3_BP2_1.5.pdf / ATM_PM3_BP2_1.6.pdf; CSPEC v1.6): PM3_Supporting = 1.0 point, PM3 (Moderate) = 2.0 points, PM3_Strong = 4.0 points, PM3_VeryStrong >= 8.0 points, summed over unrelated A-T probands by phenotype certainty and phase/zygosity of the co-occurring ATM allele; the variant must not exceed 0.01% general population frequency.
Evidence considered
ClinGen HBOP ATM VCEP PM3/BP2 table: confident A-T phenotype = 4.0 points confirmed in trans, 2.0 phase unknown, 1.0 second variant unidentified or VUS, 2.0 homozygous (max 2 individuals); consistent phenotype = half those values; PM3_Supporting = 1.0 point, PM3 = 2.0, PM3_Strong = 4.0, PM3_VeryStrong = 8.0 points.
VCEP general consideration for PM3: the variant may not exceed a general population frequency of 0.01%, and multiple unrelated cases are additive; a co-occurring pathogenic or likely pathogenic variant must itself be classified under VCEP specifications rather than assumed.
gnomAD v4.1: total AF 6.83e-06 (11/1,611,514 alleles, 0 homozygotes), highest subpopulation AF 4.81e-05 (3/62,362; Remaining individuals), grpmax FAF 2.92e-06; gnomAD v2.1: AF 3.98e-06 (1/251,304 alleles, 0 homozygotes) - rarity eligibility for the table is satisfied but this supplies no phase information.
ClinVar variation 135768 (NM_000051.4(ATM):c.6317A>T p.Asn2106Ile): 5 clinical-laboratory submissions all classified Uncertain significance, no expert-panel submission; the submission texts address only computational prediction, absence from population databases and absence of functional studies - none reports a second ATM variant, phase, or an A-T phenotype.
Searches of ATM_PM3_BP2_1.5.pdf and ATM_PM3_BP2_1.6.pdf for c.6317A>T, p.Asn2106Ile and p.N2106I returned no variant-specific row; both files are gene-level points tables assigning points per unrelated A-T proband rather than per-variant pre-assigned codes.
COSMIC records this variant somatically (COSV99069667, n = 2), which is tumour-level and provides no germline cis/trans phase evidence for an A-T proband.
No full-text paper in this case reports the exact variant c.6317A>T / p.(N2106I): the extracted findings for PMID:25741868, PMID:24418350 and PMID:25394175 all record variant_mentioned = false and contain no case, phase or co-occurrence data.
Still missing
No affected Ataxia-Telangiectasia proband record for c.6317A>T (p.Asn2106Ile) with a second ATM pathogenic/likely pathogenic variant, phase determination, or qualifying homozygosity is present in the case evidence.
No A-T phenotype classification (confident vs consistent per the VCEP criteria) is documented for any carrier of this variant.
ClinVar variation 135768 has 5 submissions, all Uncertain significance, none expert-panel, none reporting a co-occurring ATM variant or phase.
Governing file ATM_PM3_BP2_1.5.pdf (and ATM_PM3_BP2_1.6.pdf) was searched for c.6317A>T, p.Asn2106Ile and p.N2106I and contains no variant-specific entry; it is a generic points table, so no pre-assigned PM3 code exists for this variant.
Papers PMID:20301317, PMID:20301790 and PMID:20050888 (ClinVar-cited gene/disease-level reviews) had no full text available and no extraction entry, so any variant-level observation they might contain could not be inspected.
PP1Not assessed: no affected-relative segregations are documented for this variant, versus the ATM VCEP threshold of one affected relative for PP1 supporting.
Details ▾Hide ▴
Reasoning
The ClinGen HBOP ATM VCEP specification version 1.6 permits PP1 for this autosomal-recessive condition based on segregation in affected relatives, graded Supporting for 1 affected relative, Moderate for 2 affected relatives, and Strong for at least 3 affected relatives, with the requirement that affected relatives carry both variants identified in the proband. The case materials contain no family pedigree, no affected-relative genotypes, no parental testing, no phase information, and no meiosis count for NM_000051.4:c.6317A>T (NP_000042.3:p.(Asn2106Ile)). The variant is not mentioned in any of the reviewed full-text papers, and the ClinVar record (VCV000135768) contains five submissions, all Uncertain significance, none from an expert panel, and none providing family or segregation data. With zero documented affected-relative segregations, no PP1 strength can be assigned and the criterion is not assessed.
Rule
AR Condition: segregation in 1 affected relative = Supporting; segregation in 2 affected relatives = Moderate; segregation in at least 3 affected relatives = Strong. Affected relatives must have both variants identified in the proband.
Evidence considered
The ClinGen HBOP ATM VCEP specification (version 1.6, retrieved as cspec) lists PP1 as applicable with AR-condition grading: segregation in 1 affected relative = Supporting, 2 affected relatives = Moderate, at least 3 affected relatives = Strong, and requires that affected relatives have both variants identified in the proband.
The ClinVar record VCV000135768 (ClinVarID 135768) reports five submissions, all Uncertain significance, with no expert-panel review and no family, phase, or segregation information.
PMID:25741868 (ACMG/AMP 2015 standards) defines the generic PP1 co-segregation criterion but is a classificatory framework and reports no ATM-specific or variant-level segregation data; c.6317A>T, p.(N2106I) and N2106I are not mentioned.
PMID:24418350 (EFNS/ENS chronic-ataxia consensus) addresses ATM only at the gene/disease level and reports no variant-level data, case observations or segregation; c.6317A>T and p.(N2106Ile) are not mentioned.
PMID:25394175 (ACMG/NSGC cancer-predisposition referral guideline) does not name ATM or the variant and reports no segregation, case-observation or family data.
No pedigree, parental testing, meioses, or affected-relative genotypes for c.6317A>T appear anywhere in the case materials.
Still missing
No family pedigree, affected-relative genotypes, parental testing results, phase information, or meiosis count were available for this case.
The variant c.6317A>T / p.(Asn2106Ile) was not mentioned in any reviewed full-text paper, so no variant-specific segregation evidence could be extracted from the literature.
No ClinVar submission (VCV000135768, five submitters, no expert panel) provides family, phase, or segregation information for this variant.
Full text was unavailable for the two ClinVar-cited GeneReviews chapters (PMID:20301790 'Ataxia-Telangiectasia' and PMID:20301317 'Hereditary Ataxia Overview'); only screenshots were present, so they could not be searched for variant-specific family data.
BP2Not assessed: no unaffected adult carries c.6317A>T in trans or cis with a pathogenic ATM variant, so no negative BP2 points from the VCEP table can be assigned.
Details ▾Hide ▴
Reasoning
BP2 in the governing ClinGen HBOP ATM VCEP specification is the condition-specific benign mirror of the ATM PM3/BP2 points table (CSPEC v1.6; ATM_PM3_BP2_1.5.pdf and ATM_PM3_BP2_1.6.pdf), which assigns negative points for an unaffected, non-A-T adult (>=18 years, no evidence of A-T) who carries this variant together with a pathogenic or likely pathogenic ATM variant: -4.0 for confirmed trans, -2.0 for phase unknown, and for a homozygous observation -2.0 in a laboratory setting or -1.0 in a database setting, converting to BP2_Supporting at -1.0, BP2_Moderate at -2.0 and BP2_Strong at <=-4.0. No such individual is documented anywhere in the case: ClinVar variation 135768 has five submissions, all Uncertain significance with no unaffected-adult carrier, co-occurring pathogenic ATM variant or phase information, and gnomAD shows zero homozygotes in both v2.1 (1/251,304 alleles) and v4.1 (11/1,611,514 alleles) - and in any event the VCEP explicitly restricts the homozygous path to phenotyped individuals from clinical or research cohorts and excludes observations used for frequency-based codes. There is likewise no in-cis configuration (e.g. an allele carrying both this variant and a pathogenic ATM variant on the same haplotype) reported. With no qualifying observation to which the table's negative points can be applied, the summed BP2 points are zero, so BP2 cannot be assessed rather than being refuted. ATM_PM3_BP2_1.5.pdf and ATM_PM3_BP2_1.6.pdf were searched under c.6317A>T, p.Asn2106Ile and p.N2106I with no variant-specific entry found.
Rule
ClinGen HBOP ATM VCEP BP2/PM3 points table (ATM_PM3_BP2_1.5.pdf / ATM_PM3_BP2_1.6.pdf; CSPEC v1.6): BP2_Supporting = -1.0 point, BP2_Moderate = -2.0 points, BP2_Strong <= -4.0 points, summed over qualifying unaffected non-A-T adults (>=18 years) carrying the variant with a pathogenic/likely pathogenic ATM variant in trans, or in cis anywhere; the co-occurring variant must be classified under VCEP specifications.
Evidence considered
ClinGen HBOP ATM VCEP BP2/PM3 points table: -4.0 points for a pathogenic or likely pathogenic ATM variant in confirmed trans, -2.0 for phase unknown, -2.0 (laboratory setting) or -1.0 (database setting) for a homozygous observation; BP2_Supporting = -1.0 point, BP2_Moderate = -2.0 points, BP2_Strong <= -4.0 points.
VCEP restriction: the homozygous BP2 path applies only to phenotyped individuals from clinical or research cohorts and must not be applied to data used to assign frequency-based codes, because >=2 homozygotes in gnomAD are already captured by BA1/BS1; the individual must be >=18 years old with no evidence of A-T.
gnomAD v4.1 shows 0 homozygotes (11/1,611,514 alleles, AF 6.83e-06) and gnomAD v2.1 shows 0 homozygotes (1/251,304 alleles, AF 3.98e-06), so no database-setting homozygous carrier exists to score.
ClinVar variation 135768: 5 submissions, all Uncertain significance, none from an expert panel; no submission reports an unaffected adult carrier, a co-occurring pathogenic ATM variant, or cis/trans phase.
Searches of ATM_PM3_BP2_1.5.pdf and ATM_PM3_BP2_1.6.pdf for c.6317A>T, p.Asn2106Ile and p.N2106I returned no variant-specific row assigning a pre-computed BP2 code.
The exact variant appears in no full-text paper supplied with this case (PMID:25741868, PMID:24418350, PMID:25394175 all record variant_mentioned = false), so no publication contributes a cis/trans or unaffected-carrier observation.
Still missing
No unaffected non-A-T adult aged >=18 carrying c.6317A>T with a pathogenic/likely pathogenic ATM variant in confirmed trans or phase-unknown configuration is documented in the case evidence.
No qualifying homozygous observation exists in a phenotyped clinical or research cohort; gnomAD reports zero homozygotes in v2.1 and v4.1, and frequency-derived observations are excluded from the BP2 homozygous path by the VCEP.
No in-cis configuration (this variant and a pathogenic ATM variant on the same haplotype) is reported.
Governing file ATM_PM3_BP2_1.5.pdf (and ATM_PM3_BP2_1.6.pdf) was searched for c.6317A>T, p.Asn2106Ile and p.N2106I and contains no variant-specific entry; it is a generic points table, so no pre-assigned BP2 code exists for this variant.
Not applicable · 14· recorded as not applicableShow reasons ▾Hide reasons ▴PVS1N/ANot applicable: c.6317A>T is a mid-exonic missense change (p.Asn2106Ile, exon 43, SpliceAI max delta 0.005), not a null variant covered by the ATM VCEP PVS1 decision tree.
Details ▾Hide ▴
Reasoning
NM_000051.4:c.6317A>T is a single-nucleotide substitution that changes residue 2106 from asparagine to isoleucine: NM_000051.4(NP_000042.3):p.(Asn2106Ile). VariantValidator places it within coding exon 43 of the default transcript NM_000051.4/NP_000042.3, at chr11:108317491 (GRCh38) = chr11:108188218 (GRCh37), and reports a missense protein consequence with no frameshift, no premature termination codon, no initiation-codon disruption, no canonical +/-1,2 splice-site involvement (SpliceAI max delta score 0.005 across DS_AG 0.005, DS_AL 0.002, DS_DG 0.0, DS_DL 0.0; Pangolin SG 0.001, SL -0.006), and no exon-level deletion. The ATM VCEP PVS1 specification (ATM_PVS1_1.5.pdf, identical text in ATM_PVS1_1.6.pdf) states PVS1 as 'Null variant (nonsense, frameshift, canonical +/-1 or 2 splice sites, initiation codon, single or multi-exon deletion) in a gene where loss of function (LOF) is a known mechanism of disease', and the VCEP PVS1/PVS1(RNA) decision tree is a null-variant decision tree (protein-truncating events and observed splice defects). A mid-exonic missense substitution producing a full-length protein with one amino-acid change is not a null variant and cannot enter that decision tree, so no NMD assessment, exon-importance assessment, or 3'-end/HEAT/FATKIN domain reasoning is reachable for it. PVS1 is therefore not applicable to this variant rather than met at any strength; no strength tier is assigned and no downgrade rule is invoked, because the criterion's precondition (a null/truncating consequence) is not satisfied. The VCEP PVS1 document itself contains no per-variant entry for c.6317A>T / p.Asn2106Ile / p.N2106I (it is a decision guide, not a variant table), and no other declared VCEP file carries a pre-assigned PVS1 code for this variant.
Rule
ATM VCEP v1.5/v1.6 PVS1 specification (ATM_PVS1_1.5.pdf / ATM_PVS1_1.6.pdf), adapted from PMID 30192042: PVS1 applies to null variants (nonsense, frameshift, canonical +/-1,2 splice sites, initiation codon, single/multi-exon deletion). Missense substitutions are outside the decision tree; criterion not applicable.
Evidence considered
ATM VCEP specification (cspec v1.6, doc 639508985) lists PVS1 with default strength Pathogenic Very Strong and instructs 'Use ATM PVS1 Decision Tree' for Very Strong/Strong/Moderate/Supporting.
ATM_PVS1_1.5.pdf (declared by the VCEP index as governing PVS1) defines PVS1 as a null variant - nonsense, frameshift, canonical +/-1 or 2 splice sites, initiation codon, single or multi-exon deletion - in a gene where LOF is a known mechanism; its notes address NMD escape, the contiguous FAT/PI3K/FATC (FATKIN) domain, HEAT-repeat exon loss, and the most 3' pathogenic residue p.R3047, all in the context of protein-truncating or splice-defect events.
The ATM_PVS1_1.6.pdf text is materially identical to 1.5 on this point (same PVS1 null-variant definition and same domain/NMD notes).
Variant consequence per VariantValidator/Mutalyzer in the case normalization data: NM_000051.4:c.6317A>T -> NP_000042.3:p.(Asn2106Ile), exonic position exon 43 (NC_000011.10 and NC_000011.9), a missense substitution; no truncating or splice-consequence annotation is reported.
SpliceAI lookup (NM_000051.4:c.6317A>T): DS_AG 0.005, DS_AL 0.002, DS_DG 0.0, DS_DL 0.0, max delta 0.005; Pangolin SG 0.001, SL -0.006 - no splice impact, so no canonical +/-1,2 or spliceogenic consequence that could route the variant into the PVS1/PVS1(RNA) decision tree.
Sun et al. 2025 (PMID 40580951) Table S1 row for hg38_pos 108317491: c.6317A>T, N2106I, exon_43, Variant_consequence = Missense, DeepATM_predicted = No, Confidence High - independently confirms the variant is a missense change, i.e. not a null allele.
Still missing
Searched /Users/gs_agent/LYFESCI_HERMES_v9/vcep_db/ATM/fulltext (all converted VCEP text, incl. ATM_PVS1_1.5.pdf.txt and ATM_PVS1_1.6.pdf.txt, ATM_PM3_BP2_1.5/1.6, ATM_PS1_1.5/1.6, clingen_hbop_atm_supplementary_tables_1_and_2_v1.xlsx.txt) for 'c.6317A>T', 'p.Asn2106Ile', 'p.N2106I' and '2106': no ATM VCEP PVS1 file contains a per-variant entry for this variant - the file is a null-variant decision guide only, and no pre-assigned PVS1 code exists to read off. This is a statement about the VCEP document, not about absence of evidence.
The only occurrence of this variant anywhere in the converted VCEP texts is in the Sun et al. 2025 functional screen supplement (Suppl_TableS1_PMID 40580951.xlsx.txt), which annotates it as Missense - this confirms variant consequence but carries no PVS1-relevant content.
PS2VCEPNot applicable: the ATM VCEP v1.6 prohibits PS2 for ATM disease because informative de novo occurrences have not been observed.
Details ▾Hide ▴
Reasoning
The ClinGen HBOP ATM VCEP specification version 1.6 marks PS2 not applicable and instructs that it not be used for ATM-related autosomal dominant or autosomal recessive disease, because informative de novo occurrences have not yet been observed and de novo autosomal-recessive disease is unlikely to be informed by phase. This governing gene-specific specification overrides the generic ACMG/AMP PS2 criterion, and no case-specific de novo observation (with or without confirmed maternity and paternity) is documented for NM_000051.4:c.6317A>T (NP_000042.3:p.(Asn2106Ile)) in any case material.
VCEP rule
Do not use for AD or AR disease: informative de novo occurrences have not yet been observed and de novo AR conditions are unlikely to be informed by phase.
Evidence considered
The ClinGen HBOP ATM VCEP specification (version 1.6, retrieved as cspec) lists PS2 with applicability 'Not applicable' and the instruction 'Do not use for AD or AR disease: Informative de novo occurrences have not yet been observed and de novo AR conditions are unlikely to be informed by phase'.
The generic ACMG/AMP 2015 PS2 definition requires a confirmed de novo occurrence (maternity and paternity confirmed) in a patient with the disease and no family history; the gene-specific VCEP prohibition takes precedence.
No de novo observation or parental testing result is documented for c.6317A>T anywhere in the case materials.
PM1VCEPNot applicable: the ATM VCEP marks PM1 'Not applicable' with the direction not to use it, as benign and pathogenic variants co-occur in the same domains.
Details ▾Hide ▴
Reasoning
The governing specification removes PM1 for ATM entirely. The ClinGen HBOP ATM VCEP specification (cspec v1.6, doc 639508985) records PM1 with applicability 'Not applicable' and the instruction 'Do not use: Benign and pathogenic variants are known to occur within the same domains and germline mutational hotspots are not well defined at this time'; every strength tier (Stand Alone, Very Strong, Strong, Moderate, Supporting) is likewise marked Not applicable with no rule text. This is a class-level, gene-wide prohibition, so it is not rescued by anything specific to NM_000051.4:c.6317A>T / NP_000042.3:p.(Asn2106Ile). As instructed for this group, the case's vcep_materials.json was checked for a domain_tables entry: the key exists but is an empty object ({}), so no extracted ATM critical-domain list is available to test residue 2106 against - and even an alignment with a PI3K-related or FATKIN domain could not override the VCEP's own decision not to use PM1. The variant is a missense substitution at residue 2106 in exon 43 with no hotspot signal in the collected data (CancerHotspots lookup for 'ATM N2106' returned no row; OncoKB reports Unknown Oncogenic Effect with no variant-specific reviewed evidence), which is consistent with the VCEP's stated rationale that germline mutational hotspots are not well defined for ATM. PM1 is therefore not applicable rather than not met: the criterion is outside this VCEP's framework and carries no weight in either direction.
VCEP rule
ATM VCEP v1.6 PM1: applicability 'Not applicable' - 'Do not use: Benign and pathogenic variants are known to occur within the same domains and germline mutational hotspots are not well defined at this time.' Same-version framework in force for this gene; criterion not applicable.
Evidence considered
cspec v1.6 (doc 639508985) PM1 entry: applicability 'Not applicable'; instructionsToUse 'Do not use: Benign and pathogenic variants are known to occur within the same domains and germline mutational hotspots are not well defined at this time'; all five strength tiers (Stand Alone, Very Strong, Strong, Moderate, Supporting) listed as Not applicable with empty rule text.
vcep_materials.json for this case has a domain_tables key whose value is an empty object - no ATM domain table was extracted, so there is no authoritative domain list to test residue 2106 against.
Variant context: c.6317A>T is p.(Asn2106Ile), a missense substitution in coding exon 43 of NM_000051.4 - the kind of change PM1 would normally address, which is why the VCEP's blanket 'do not use' statement is the operative rule.
CancerHotspots lookup for 'ATM N2106' returned no result (no statistically significant hotspot row); OncoKB reports 'Unknown Oncogenic Effect' with is_variant_specific = false - no hotspot or curated-domain evidence is claimed for this residue.
The ATM VCEP PVS1 document notes that the FAT, PI3-K and FATC regions are treated as one contiguous domain (PMID 28508083), but that caveat governs PVS1/null-variant reasoning; it does not reinstate PM1, which this VCEP has disabled.
Still missing
vcep_materials.json was checked for the instructed domain_tables key: the key is present but empty ({}), so there is no extracted per-gene critical-domain/hotspot table for ATM to compare residue 2106 against. This is a statement about the extracted data, not evidence that PM1 is met.
No converted ATM VCEP file at /Users/gs_agent/LYFESCI_HERMES_v9/vcep_db/ATM/fulltext contains a per-variant PM1 entry for c.6317A>T / p.Asn2106Ile / p.N2106I ('PM1' does not appear in the PS1/PVS1/PM3-BP2 documents or the supplementary-table texts); the cspec class-level entry is the governing statement.
PM4N/ANot applicable: the ATM VCEP limits PM4 to stop-loss variants, and p.Asn2106Ile is a single missense substitution in exon 43 with no protein-length change.
Details ▾Hide ▴
Reasoning
The ATM VCEP specification (cspec v1.6, doc 639508985) restricts PM4 explicitly: 'Use for stop-loss variants.' PM4 as used by this VCEP therefore requires a termination-codon (stop-loss) change that elongates the protein; it is not used for the generic missense/in-frame-length-change reading. NM_000051.4:c.6317A>T is p.(Asn2106Ile) - a single amino-acid substitution within exon 43 - which neither removes nor alters a stop codon, and the variant is not a stop-loss allele (no stop-loss annotation appears in the normalization data, and the Sun et al. 2025 Table S1 row classifies it as Missense). It is likewise not an in-frame deletion/insertion: it is a single-nucleotide substitution with no change in protein length. No stop-loss variant at ATM codon 2106 (or any other codon) is under assessment here. PM4 is therefore not applicable on both the VCEP-specific rule (stop-loss only) and the underlying criterion definition (protein length change).
Rule
ATM VCEP v1.6 PM4: 'Use for stop-loss variants.' Variant is a missense substitution, not a stop-loss or length-changing variant; criterion not applicable.
Evidence considered
cspec v1.6 (doc 639508985) PM4 entry: default strength Pathogenic Moderate with the single rule 'Use for stop-loss variants.'
Variant consequence: NM_000051.4:c.6317A>T produces NP_000042.3:p.(Asn2106Ile) - a substitution, not a stop-loss or a length-changing in-frame indel (VariantValidator/Mutalyzer normalization data in the case prefetch).
Sun et al. 2025 (PMID 40580951) Table S1 row for hg38_pos 108317491 records the same change as Variant_consequence = Missense at residue 2106 in exon 43, consistent with a simple amino-acid substitution.
Still missing
No ATM VCEP file in /Users/gs_agent/LYFESCI_HERMES_v9/vcep_db/ATM/fulltext contains a per-variant PM4 entry for c.6317A>T / p.Asn2106Ile / p.N2106I ('PM4' appears in no converted VCEP text); the cspec rule is the governing statement and it is class-based, not variant-specific.
PM5N/ANot applicable: ATM VCEP PM5_Supporting covers only truncating variants upstream of p.Leu3048 and explicitly excludes missense changes such as p.Asn2106Ile.
Details ▾Hide ▴
Reasoning
The ATM VCEP repurposes PM5 away from the classic same-residue missense rule. The cspec v1.6 entry (doc 639508985) carries the original ACMG definition as background but the controlling instructions are: 'Based on location of the most C-terminal known pathogenic variant, p.Arg3047*'; 'Use as PM5_Supporting (not moderate)'; 'Do not use for start-loss variants'; '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 single applicable strength rule is explicitly truncation-based: 'DNA Variants: Apply to all NMD-prone, truncating variants receiving PVS1 (must be at VeryStrong) with premature termination codons upstream of p.Leu3048; Splice Variants: Apply only to observed (not predicted) NMD-prone, truncating variants receiving PVS1(RNA) (must be at VeryStrong with high-quality RNA data) with premature termination codons upstream of p.Leu3048.' NM_000051.4:c.6317A>T is p.(Asn2106Ile), a single missense substitution in exon 43: it is neither a truncating variant nor a splice variant and cannot receive PVS1, so neither branch of the rule is reachable, and the VCEP's explicit 'Do not use for missense changes' instruction applies directly. Independently, the classic same-residue reading also fails on the evidence: the pipeline's PM5 comparator collection found zero candidates (total_candidates_seen = 0, same_residue_candidates = 0), and no alternate change at residue 2106 is reported as (likely) pathogenic in ClinVar (the N2106Y/H/D/T/S/K rows of the Sun et al. 2025 supplement have empty ClinVar_classification fields, while the residue-2106 change under assessment is itself only a 2-star VUS). PM5 is therefore not applicable to this variant, and this determination is a consequence of the governing VCEP's scope, not an assertion that a same-residue comparator was found.
Rule
ATM VCEP v1.6 PM5: PM5_Supporting applies only to NMD-prone truncating variants (DNA) or observed NMD-prone truncating splice variants (RNA data) with premature termination codons upstream of p.Leu3048; 'Do not use for missense changes.' This variant is a missense substitution that cannot receive PVS1, so criterion not applicable.
Evidence considered
cspec v1.6 (doc 639508985) PM5 entry: applicability 'Applicable', default strength Pathogenic Moderate, instructionsToUse 'Based on location of the most C-terminal known pathogenic variant, p.Arg3047*. Use as PM5_Supporting (not moderate). Do not use for start-loss variants. 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 only PM5 strength rule marked Applicable in the v1.6 ruleset is truncation-based: 'DNA Variants: Apply to all NMD-prone, truncating variants receiving PVS1 (must be at VeryStrong) with premature termination codons upstream of p.Leu3048' and the observed-RNA splice counterpart, both requiring PVS1 at Very Strong.
Variant consequence: c.6317A>T -> p.(Asn2106Ile), a missense substitution in exon 43 (NM_000051.4 / NP_000042.3); SpliceAI max delta 0.005 - the variant is not a truncating or splice variant and cannot receive PVS1, so neither PM5 branch is reachable.
output/pm5_candidates.json for this case: pm5_mode 'truncation_cutoff_gene_specific', eligible_for_classic_pm5_search = false, total_candidates_seen = 0, same_residue_candidates = 0, expert_panel_plp_candidates = 0 - no same-residue comparator variants were collected, and the artifact itself flags PM5 as a truncation/non-missense framework rule rather than classic same-residue PM5.
ClinVar (variation 135768, VCV000135768): the residue-2106 change under assessment is Uncertain significance (2-star, multiple submitters, no conflicts) - there is no pathogenic anchor at this residue.
Sun et al. 2025 (PMID 40580951) Table S1 rows for the other residue-2106 substitutions (N2106Y c.6316A>T, N2106T c.6317A>C, N2106S c.6317A>G, N2106K c.6318T>A/G) show empty ClinVar_classification fields - none is an established pathogenic same-residue comparator.
Still missing
No converted ATM VCEP file at /Users/gs_agent/LYFESCI_HERMES_v9/vcep_db/ATM/fulltext contains a per-variant PM5 entry for c.6317A>T / p.Asn2106Ile / p.N2106I; the cspec v1.6 class-level rule is the governing statement and it is consequence-type-based, not variant-specific.
The ClinVar-wide search for every P/LP missense change at ATM codon 2106 was carried out through the pipeline's comparator artifact (output/pm5_candidates.json, 0 candidates) plus the ClinVar_classification column of the Sun et al. 2025 supplement; a direct per-codon ClinVar API query was not independently re-run for this group, so an unreported ultrarare same-residue P/LP submission cannot be excluded with complete certainty. Under the governing VCEP rule this would not change the outcome, since PM5 may not be used for missense changes at all.
PM6VCEPNot applicable: the ATM VCEP v1.6 prohibits PM6, as no informative de novo occurrences are available to calibrate it for ATM disease.
Details ▾Hide ▴
Reasoning
The ClinGen HBOP ATM VCEP specification version 1.6 marks PM6 not applicable and instructs that it not be used for ATM-related autosomal dominant or autosomal recessive disease, for the same reason given for PS2: informative de novo occurrences have not been observed and de novo autosomal-recessive disease is unlikely to be informed by phase. This governing gene-specific specification overrides the generic ACMG/AMP PM6 criterion, and no case-specific assumed de novo occurrence is documented for NM_000051.4:c.6317A>T (NP_000042.3:p.(Asn2106Ile)).
VCEP rule
Do not use for AD or AR disease: informative de novo occurrences have not yet been observed and de novo AR conditions are unlikely to be informed by phase.
Evidence considered
The ClinGen HBOP ATM VCEP specification (version 1.6, retrieved as cspec) lists PM6 with applicability 'Not applicable' and the instruction 'Do not use for AD or AR disease: Informative de novo occurrences have not yet been observed and de novo AR conditions are unlikely to be informed by phase'.
The generic ACMG/AMP 2015 PM6 definition covers an assumed de novo occurrence without confirmation of maternity and paternity; the gene-specific VCEP prohibition takes precedence.
No assumed de novo occurrence or parental testing result is documented for c.6317A>T anywhere in the case materials.
PP2VCEPNot applicable: the ATM VCEP marks PP2 'Not applicable' with the direction not to use it, because ATM has no defined low rate of benign missense variation.
Details ▾Hide ▴
Reasoning
The governing specification disables PP2 for ATM. The ClinGen HBOP ATM VCEP specification (cspec v1.6, doc 639508985) records PP2 with applicability 'Not applicable' and the instruction 'Do not use: ATM does not have a defined low rate of missense benign variation'; no strength tier is populated for it. PP2 as originally defined requires both a low rate of benign missense variation and that missense variants are a common mechanism of disease, and the VCEP has determined the first condition does not hold for ATM - a determination consistent with the data collected here, where the residue-2106 alternates in the Sun et al. 2025 supplement show a wide spread of predicted effects (for example REVEL 0.539 for N2106I, 0.635 for N2106Y, 0.291 for N2106S in that table) rather than a uniformly constrained residue. Because NM_000051.4:c.6317A>T is indeed a missense variant (NP_000042.3:p.(Asn2106Ile), exon 43), this is precisely the variant class the VCEP's prohibition addresses, and no other VCEP file supplies a per-variant PP2 code. PP2 is therefore not applicable rather than not met: it is not part of the ATM framework in either direction.
VCEP rule
ATM VCEP v1.6 PP2: applicability 'Not applicable' - 'Do not use: ATM does not have a defined low rate of missense benign variation.' Criterion not applicable for this gene and this missense variant.
Evidence considered
cspec v1.6 (doc 639508985) PP2 entry: applicability 'Not applicable'; instructionsToUse 'Do not use: ATM does not have a defined low rate of missense benign variation'; all strength tiers listed as Not applicable with empty rule text.
Variant class: NM_000051.4:c.6317A>T yields NP_000042.3:p.(Asn2106Ile) - a missense substitution in exon 43, i.e. the variant class the PP2 prohibition addresses.
Sun et al. 2025 (PMID 40580951) Table S1 rows for residue 2106 show divergent predictor values across substitutions in the same codon (REVEL 0.539 for N2106I, 0.635 for N2106Y, 0.551 for N2106H, 0.498 for N2106D, 0.466 for N2106T, 0.291 for N2106S, 0.5 for N2106K), consistent with the VCEP's position that this gene lacks a defined low rate of benign missense variation; these scores are noted only as the VCEP's stated rationale, not as PP3/BP4 evidence for this group.
Still missing
No converted ATM VCEP file at /Users/gs_agent/LYFESCI_HERMES_v9/vcep_db/ATM/fulltext contains a per-variant PP2 entry for c.6317A>T / p.Asn2106Ile / p.N2106I ('PP2' does not appear in the PS1, PVS1 or PM3/BP2 VCEP documents or in the supplementary-table texts); the cspec v1.6 class-level entry is the governing statement.
PP4VCEPNot applicable: the ATM VCEP excludes PP4 at every strength, and no patient phenotype or family history accompanies this variant.
Details ▾Hide ▴
Reasoning
Governing specification: HBOP ATM VCEP v1.6 (cspec doc 639508985) lists PP4 as 'Not applicable' at every strength. Its instructions exclude the criterion on both inheritance models for ATM: for autosomal dominant disease 'do not use as breast cancer is a disease with multiple genetic etiology (genetic heterogeneity) and there are no features that can readily distinguish hereditary from sporadic causes', and for autosomal recessive disease 'do not use as a separate line of evidence. Such evidence is built into the Ataxia Telangiectasia PM3|BP2 table'. Consistent with this, no patient-level phenotype or family history accompanies the variant in this case: the ClinVar conditions attached to VCV000135768 are only the broad labels 'Ataxia-telangiectasia syndrome', 'Hereditary cancer-predisposing syndrome' and 'not provided', which give no phenotype specificity to weigh. The VCEP has pre-empted PP4 for ATM on both models, so this is not_applicable rather than a negative finding about the variant; the phenotype-specificity credit for a recessive A-T proband is captured through the points-based PM3/BP2 table instead.
VCEP rule
Not applicable per HBOP ATM VCEP v1.6 (PP4 excluded at all strengths; recessive A-T phenotype weight is folded into the PM3/BP2 table).
Evidence considered
cspec (HBOP ATM v1.6) PP4: applicability 'Not applicable' at Stand Alone, Very Strong, Strong, Moderate and Supporting strengths.
VCEP PP4 instruction (autosomal dominant): 'do not use as breast cancer is a disease with multiple genetic etiology (genetic heterogeneity) and there are no features that can readily distinguish hereditary from sporadic causes.'
VCEP PP4 instruction (autosomal recessive): 'do not use as a separate line of evidence. Such evidence is built into the Ataxia Telangiectasia PM3|BP2 table.'
ClinVar conditions for this exact variant are the broad labels Ataxia-telangiectasia syndrome, Hereditary cancer-predisposing syndrome and not provided; no individual case phenotype or family history is recorded in the case evidence.
Still missing
No patient-level phenotype or family history is present in the case record, so even outside the VCEP exclusion there would be no phenotype-specificity data to score.
BS2VCEPNot applicable: the ATM VCEP excludes BS2 because ATM-related disease is incompletely penetrant, and no homozygous carrier appears in gnomAD.
Details ▾Hide ▴
Reasoning
The ATM VCEP v1.6 explicitly designates BS2 as Not Applicable, with the instruction 'Do not use: ATM has incomplete penetrance.' BS2's premise - a fully penetrant, typically early-onset condition with the variant observed in a well-documented healthy adult - cannot be satisfied for ATM-related cancer predisposition, which is incompletely penetrant. As a corroborating observation, no homozygous carrier is present in any captured dataset (0 homozygotes in gnomAD v2.1, v4.1 and v2.1 non-cancer exomes; absent from gnomAD v3.1 non-cancer genomes and gnomAD-Canada v1.0), so there is also no population observation of a healthy homozygous adult that could be weighed. BS2 is recorded as not applicable by specification rather than as evidence against benignity.
VCEP rule
BS2 not applicable per ATM VCEP v1.6: 'Do not use: ATM has incomplete penetrance.'
Evidence considered
ATM VCEP v1.6 BS2 entry: applicability 'Not applicable', instructions 'Do not use: ATM has incomplete penetrance.'
gnomAD homozygote counts: 0 in gnomAD v2.1 (1/251,304 alleles), 0 in gnomAD v4.1 (11/1,611,514 alleles), 0 in gnomAD v2.1 non-cancer exomes (1/236,766); absent from gnomAD v3.1 non-cancer genomes and gnomAD-Canada v1.0, so no homozygous carrier observation is available.
ACMG/AMP 2015 (PMID:25741868) restricts BS2 to fully penetrant early-onset conditions observed in well-documented healthy adults, a premise the specification states ATM does not meet.
BS4VCEPNot applicable: the ATM VCEP v1.6 excludes BS4, because informative lack of co-segregation in A-T families is too rare to carry weight.
Details ▾Hide ▴
Reasoning
The ClinGen HBOP ATM VCEP specification version 1.6 marks BS4 not applicable: for the autosomal-recessive ATM condition, informative instances of lack of co-segregation in A-T families are too rare to be considered for weight at this time, and such observations may instead be considered for BP2 if biallelic unaffected patients are observed in an A-T family. This governing gene-specific specification overrides the generic ACMG/AMP BS4 criterion; the local VCEP guidance for this criterion is provided entirely through BP2. No non-segregation observation in an informative affected family is documented for NM_000051.4:c.6317A>T (NP_000042.3:p.(Asn2106Ile)).
VCEP rule
Not applicable under the ClinGen HBOP ATM VCEP specification version 1.6; lack of co-segregation in A-T families may instead be considered for BP2 when biallelic unaffected patients are observed.
Evidence considered
The ClinGen HBOP ATM VCEP specification (version 1.6, retrieved as cspec) lists BS4 with applicability 'Not applicable' and notes that for the AR condition informative instances of lack of co-segregation in A-T families are too rare to be considered for weight, with such observations potentially routed to BP2 instead.
The generic ACMG/AMP 2015 BS4 definition covers lack of segregation in affected family members; the gene-specific VCEP prohibition takes precedence.
No non-segregation or affected-family genotype data are documented for c.6317A>T in the case materials.
BP1VCEPNot applicable: the ATM VCEP marks BP1 'Not applicable' with the direction not to use it, because missense pathogenic variants are known for ATM.
Details ▾Hide ▴
Reasoning
The governing specification disables BP1 for ATM. The ClinGen HBOP ATM VCEP specification (cspec v1.6, doc 639508985) records BP1 with applicability 'Not applicable' and the instruction 'Do not use: Missense pathogenic variants are known for ATM'; no benign strength tier is populated. BP1 would ordinarily apply to a missense variant in a gene where primarily truncating variants cause disease, so the criterion is directed at exactly this variant class (NM_000051.4:c.6317A>T is NP_000042.3:p.(Asn2106Ile), a missense substitution in exon 43). The VCEP's reason for withholding it is well supported in the collected evidence: protein-truncating and missense ATM alleles both cause disease in the biallelic ataxia-telangiectasia / cancer-predisposition setting, and the PM3/BP2 VCEP attachment and the cspec framework both operate on missense variants as pathogenic-eligible. Correspondingly, no converted VCEP file assigns a per-variant BP1 code to this variant, and nothing in the case evidence indicates a truncating-only mechanism. BP1 is therefore not applicable rather than not met: it is excluded from the ATM framework.
VCEP rule
ATM VCEP v1.6 BP1: applicability 'Not applicable' - 'Do not use: Missense pathogenic variants are known for ATM.' Criterion not applicable for this gene.
Evidence considered
cspec v1.6 (doc 639508985) BP1 entry: applicability 'Not applicable'; instructionsToUse 'Do not use: Missense pathogenic variants are known for ATM'; all strength tiers listed as Not applicable with empty rule text.
Variant class: NM_000051.4:c.6317A>T yields NP_000042.3:p.(Asn2106Ile), a missense substitution in exon 43 - the class of variant BP1 would otherwise be aimed at.
The ATM VCEP framework elsewhere treats missense changes as pathogenic-eligible (PS1 has an explicit missense sub-path at Pathogenic Strong; the PM3/BP2 points table applies regardless of variant class), which is the mechanism behind the VCEP's statement that missense pathogenic variants are known for ATM.
No converted ATM VCEP file mentions BP1, so there is no pre-assigned BP1 code to defer to for this variant.
Still missing
No converted ATM VCEP file at /Users/gs_agent/LYFESCI_HERMES_v9/vcep_db/ATM/fulltext contains a per-variant BP1 entry for c.6317A>T / p.Asn2106Ile / p.N2106I ('BP1' does not appear in the PS1, PVS1 or PM3/BP2 VCEP documents or in the supplementary-table texts); the cspec v1.6 class-level entry is the governing statement.
BP3VCEPNot applicable: the ATM VCEP marks BP3 'Not Applicable' and p.Asn2106Ile is a substitution, not an in-frame indel in a repetitive region.
Details ▾Hide ▴
Reasoning
The ATM VCEP specification (cspec v1.6, doc 639508985) records BP3 with applicability 'Not Applicable' - this VCEP does not use BP3 for ATM at any strength. The generic ACMG/AMP definition of BP3 is also inapplicable on this variant's own terms: BP3 covers an in-frame deletion/insertion in a repetitive region without a known function, whereas NM_000051.4:c.6317A>T is a single-nucleotide substitution, p.(Asn2106Ile), with no insertion or deletion and therefore no repeat-region-length reasoning to apply (the exon-43 coding sequence around residue 2106 is not a low-complexity/repeat insertion site for this variant). There is consequently no VCEP pre-assigned BP3 code and no generic BP3 branch reachable. Note that 'no BP3' here carries no benign weight: BP3 is simply not part of the ATM VCEP framework, and the near-neutral SpliceAI result for this variant is not BP4/BP3 evidence either way.
VCEP rule
ATM VCEP v1.6 BP3: applicability 'Not Applicable'; generic BP3 (in-frame indel in a repetitive region) requires an insertion/deletion, which this substitution is not.
Evidence considered
cspec v1.6 (doc 639508985) BP3 entry: applicability = 'Not Applicable' - no default strength and no rules are specified for ATM.
Variant class: c.6317A>T is a single-nucleotide substitution yielding p.(Asn2106Ile) (VariantValidator/Mutalyzer normalization data), i.e. not an in-frame deletion or insertion, which excludes the generic BP3 scenario of an in-frame indel within a repetitive region without known function.
No converted ATM VCEP file in the gene's VCEP directory mentions BP3, so there is no pre-assigned code to defer to.
Still missing
Searched all converted ATM VCEP texts in /Users/gs_agent/LYFESCI_HERMES_v9/vcep_db/ATM/fulltext for 'BP3': the string occurs in no VCEP file, and the cspec v1.6 entry itself lists BP3 as 'Not Applicable', so no governing per-variant or class-level BP3 rule exists for ATM.
BP5VCEPNot applicable: the ATM VCEP excludes BP5 because ATM is low penetrance and co-occurring pathogenic variants show no phenotype difference.
Details ▾Hide ▴
Reasoning
Governing specification: HBOP ATM VCEP v1.6 (cspec doc 639508985) lists BP5 as 'Not applicable' at every strength, with the instruction 'Do not use: Cases with multiple pathogenic variants have been observed with no noticeable difference in phenotype (e.g. BRCA1 and BRCA2). In addition, ATM has low penetrance and will naturally occur with other pathogenic variants more frequently due to higher tolerance/presence in the general population.' There is also no case in this curation to which the criterion could be applied: the evidence contains only variant-level data (population frequencies, ClinVar assertions, COSMIC somatic counts) and no proband genotype, so there is no affected individual with an alternate molecular basis for disease to evaluate. BP5 is therefore not_applicable by explicit VCEP exclusion, not a negative finding about the variant.
VCEP rule
Not applicable per HBOP ATM VCEP v1.6 (BP5 excluded: do not use, ATM is low penetrance and co-occurring pathogenic variants show no phenotype difference).
Evidence considered
cspec (HBOP ATM v1.6) BP5: applicability 'Not applicable' at Stand Alone, Very Strong, Strong, Moderate and Supporting strengths.
VCEP BP5 instruction: 'Do not use: Cases with multiple pathogenic variants have been observed with no noticeable difference in phenotype (e.g. BRCA1 and BRCA2). In addition, ATM has low penetrance and will naturally occur with other pathogenic variants more frequently due to higher tolerance/presence in the general population.'
No proband genotype, segregation or case-level phenotype is present in the case evidence; the only case-adjacent data are COSMIC somatic observations (2) and ClinVar assertions, neither of which is a case with an alternate molecular basis for disease.
Still missing
No case-level (proband) data are present for this variant, so an alternate-molecular-basis scenario could not be assessed even if the VCEP permitted BP5.
BP7N/ANot applicable: BP7 covers only synonymous and deep intronic variants, and c.6317A>T is a missense change (p.Asn2106Ile).
Details ▾Hide ▴
Reasoning
BP7 is scoped by the ATM VCEP to synonymous substitutions and deep intronic variants (beyond, but not including, +7 at donor and -21 at acceptor sites). c.6317A>T is neither: it is a coding missense change, NP_000042.3:p.(Asn2106Ile), which the VCEP's own supplementary table independently annotates as 'Missense' at exon 43, and it is not an intronic variant at all. BP7 is therefore out of scope on its own terms, upstream of any predictor evidence; no SpliceAI/REVEL/VarSeak/BayesDel value was evaluated for it and the criterion must not be recorded as not_met on the basis of scores.
Evidence considered
ClinGen HBOP ATM VCEP v1.6 (source_registry key 'cspec') BP7 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.'
Variant consequence from case normalization and from Suppl_TableS1_PMID 40580951.xlsx (row c.6317A>T / N2106I, Variant_consequence = Missense), confirming BP7 scope is not met.
No RNA-based BP7(RNA) evidence (observed lack of aberrant splice defect) is present for this variant in the case evidence.
AF 6.8e-06 · 11 / 1,611,514 alleles · 0 hom · FAF 2.9e-06
Remaining individuals3 / 62,362
4.8e-05
European (non-Finnish)8 / 1,178,890
6.8e-06
0 alternate alleles recorded in 8 other groups: Admixed American, European (Finnish), Amish, East Asian, Middle Eastern, South Asian, Ashkenazi Jewish, African/African American.
gnomAD v2.1
AF 4e-06 · 1 / 251,304 alleles · 0 hom
European (non-Finnish)1 / 113,632
8.8e-06
0 alternate alleles recorded in 7 other groups: African/African American, Admixed American, Ashkenazi Jewish, East Asian, European (Finnish), Remaining individuals, South Asian.
Statements
gnomAD v4.1
This variant is present in gnomAD v4.1 (AF= 6.82588e-06; MAF= 0.00068%, 11/1611514 alleles, homozygotes = 0) and has highest observed frequency in the Remaining individuals population (AF= 4.81062e-05; MAF= 0.00481%, 3/62362 alleles, homozygotes = 0); grpmax FAF= 2.92e-06.
gnomAD v2.1
This variant is present in gnomAD v2.1 (AF= 3.97924e-06; MAF= 0.00040%, 1/251304 alleles, homozygotes = 0) and has highest observed frequency in the European (non-Finnish) population (AF= 8.80034e-06; MAF= 0.00088%, 1/113632 alleles, homozygotes = 0).
gnomAD-Canada
Absent from gnomAD-Canada v1.0.
SpliceAI and Pangolin
SpliceAI acceptor lossat -69 bp
0.002
SpliceAI acceptor gainat -109 bp
0.005
SpliceAI donor lossat +30 bp
0.000
SpliceAI donor gainat +454 bp
0.000
Pangolin splice lossmagnitude shown
-0.006
Pangolin splice gain
0.001
Other predictors
REVEL
0.539
BayesDel
-0.029
How the predictions were used
PP3Not met: missense REVEL 0.539 sits in the gray zone, below the ATM VCEP PP3 threshold of REVEL >0.7333.
Details ▾Hide ▴
Reasoning
NM_000051.4:c.6317A>T is a missense change (NP_000042.3:p.(Asn2106Ile)), so the applicable computational path for PP3 is the missense protein-impact path (REVEL), not splicing. The governing specification is ClinGen HBOP ATM VCEP v1.6 (framework_mode=vcep), whose PP3 rule for missense is REVEL > 0.7333. The REVEL score for this variant is 0.539, from the pipeline's local REVEL v1.3 lookup (source_registry key 'revel'), and the same value (0.539) is reported in the VCEP-declared supporting table Suppl_TableS1_PMID 40580951 (row c.6317A>T / N2106I, exon 43, Missense), so the two independent reads agree exactly. 0.539 is well below the VCEP threshold of 0.7333 and also below the generic ClinGen SVI supporting threshold of 0.644 (Pejaver et al. 2022, PMID 36413997), so PP3 is not met at any strength. The variant's SpliceAI result (max delta 0.005-0.01, no predicted impact) was not used here: it belongs to the splice path and says nothing about tolerance of the Asn2106Ile substitution; conversely it does not create BP4 support either. Secondary computational columns in the VCEP table are concordantly non-supportive for a pathogenic call (AlphaMissense 0.2578 - ambiguous/low range, EVE 0.695, BayesDel -0.028517, all treated as corroboration only, never as the primary source), and none of them changes the REVEL-based verdict.
Variant consequence settled first: NM_000051.4:c.6317A>T = NP_000042.3:p.(Asn2106Ile), a missense variant, so the REVEL path applies and the SpliceAI path does not.
Local REVEL v1.3 lookup returned 0.539 for this variant (source_registry key 'revel').
VCEP-declared supporter searched for all three notations with grep -rn over /Users/gs_agent/LYFESCI_HERMES_v9/vcep_db/ATM/fulltext: 'c.6317A>T', 'p.Asn2106Ile', 'p.N2106I'. Only Suppl_TableS1_PMID 40580951.xlsx.txt matched (line 20199; row: hg38 108317491 A>T, c.6317A>T, N2106I, exon_43, Missense, REVEL 0.539, AlphaMissense 0.2578, CADD 21.5, EVE 0.695, GnomAD_all_AF 7.53e-06, ClinVar Uncertain significance, Combined_score -0.754, Classification Functional, Confidence High, DeepATM_predicted No).
That table pre-assigns no PP3 or BP4 code for this variant; it carries only functional-classification and computational-predictor columns, so no code was available to apply directly. Its REVEL column corroborates the local lookup rather than replacing it.
Generic ClinGen SVI REVEL calibration consulted and also not met (supporting PP3 requires >=0.644; Pejaver et al. 2022, PMID 36413997).
SpliceAI result for this variant (max delta 0.005-0.01) was deliberately not used as evidence for this missense variant's PP3, per the one-path rule.
Still missing
The full text of PMID 40580951 (Sun et al. 2025) was not fetched for this case; only the VCEP-declared supplement table Suppl_TableS1_PMID 40580951.xlsx was available and searched. Its per-variant REVEL column (0.539) agrees with the independent local REVEL lookup, so the PP3/BP4 missense verdict does not depend on the unfetched manuscript.
BP4Not met: missense REVEL 0.539 exceeds the ATM VCEP BP4 cutoff of REVEL <=0.249, and the clean SpliceAI score cannot substitute.
Details ▾Hide ▴
Reasoning
Same scope determination as PP3: c.6317A>T is a missense variant (p.(Asn2106Ile)), so benign computational support must come from the missense REVEL path. ClinGen HBOP ATM VCEP v1.6 BP4 rule for missense is REVEL <= 0.249. The measured REVEL is 0.539 (local REVEL v1.3 lookup; identical 0.539 in the VCEP-declared Table S1), which is far above both the VCEP cutoff of 0.249 and the generic ClinGen SVI supporting cutoff of <= 0.29 (Pejaver et al. 2022, PMID 36413997). BP4 is therefore not met - this is a gray-zone REVEL result, which explicitly means BP4 is not met and does not license reaching across to the splice path. The variant's clean SpliceAI result (max delta 0.005-0.01, below the <= 0.1 BP4 splice cutoff) cannot be used as BP4 support because SpliceAI measures splice impact, not missense tolerance; splicing and protein impact are separate mechanisms and the VCEP instructs against applying the wrong type of computational evidence. No VCEP supporting material pre-assigns a BP4 code for this variant.
Local REVEL v1.3 lookup returned 0.539 (source_registry key 'revel'), independent of and identical to the REVEL column value 0.539 in the VCEP-declared Table S1 row for c.6317A>T / N2106I.
Generic ClinGen SVI REVEL calibration also not met (supporting BP4 requires <= 0.29; Pejaver et al. 2022, PMID 36413997).
SpliceAI max delta 0.005-0.01 is below the VCEP splice BP4 cutoff of <= 0.1, but was not applied: the variant is missense, so the splice path is out of scope for this criterion and a clean splice prediction is not BP4 support for a missense change.
No PP3/BP4 code for this variant was found in any VCEP-declared file after grep of 'c.6317A>T', 'p.Asn2106Ile' and 'p.N2106I' across /Users/gs_agent/LYFESCI_HERMES_v9/vcep_db/ATM/fulltext.
Still missing
No RP/BP pre-assigned code for this variant exists in the VCEP-declared Suppl_TableS1_PMID 40580951.xlsx, so BP4 rests entirely on the missense REVEL path, which is indeterminate at 0.539.
BP7N/ANot applicable: BP7 covers only synonymous and deep intronic variants, and c.6317A>T is a missense change (p.Asn2106Ile).
Details ▾Hide ▴
Reasoning
BP7 is scoped by the ATM VCEP to synonymous substitutions and deep intronic variants (beyond, but not including, +7 at donor and -21 at acceptor sites). c.6317A>T is neither: it is a coding missense change, NP_000042.3:p.(Asn2106Ile), which the VCEP's own supplementary table independently annotates as 'Missense' at exon 43, and it is not an intronic variant at all. BP7 is therefore out of scope on its own terms, upstream of any predictor evidence; no SpliceAI/REVEL/VarSeak/BayesDel value was evaluated for it and the criterion must not be recorded as not_met on the basis of scores.
Evidence considered
ClinGen HBOP ATM VCEP v1.6 (source_registry key 'cspec') BP7 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.'
Variant consequence from case normalization and from Suppl_TableS1_PMID 40580951.xlsx (row c.6317A>T / N2106I, Variant_consequence = Missense), confirming BP7 scope is not met.
No RNA-based BP7(RNA) evidence (observed lack of aberrant splice defect) is present for this variant in the case evidence.
SpliceAI predicts no significant splice impact for this variant (max delta score = 0.01). REVEL score = 0.539. BayesDel score = -0.028517.
6 paper records listed. 1 cited in the classification. None of the listed paper records explicitly indicates that the variant is named.
Cited in the classification
Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.
Richards S et al. · 2015 May
The ACMG/AMP 2015 standards and guidelines (Richards et al.) provide the general criteria framework, including the definition of PVS1 as a null variant in a gene where loss of function is a known mechanism, and the definitions of PM4 and BP3 in terms of protein length change and in-frame indels in repetitive regions. The paper reports no ATM-specific data and no variant-level data for c.6317A>T (p.(N2106I)); the gene examples it uses are MYH7, CFTR, NIPBL and APOE. It was consulted in this group only for the generic definitions that the ATM VCEP PVS1 document explicitly adapts.
For example, rule PM3 could be upgraded to strong if there were multiple observations of detection of the variant in trans (on opposite chromosomes) with other pathogenic variants (see section PM3 for further guidance).
Location Main text, paragraph describing use of the criteria and weight flexibility (section introducing Tables 3-5); PM3 and BP2 definitions in Table 3 / Table 4
Context Professional society standards and guidelines document (framework only); no cohort, assay or case series.
EFNS/ENS Consensus on the diagnosis and management of chronic ataxias in adulthood.
van de Warrenburg BP et al. · 2014
EFNS/ENS consensus on chronic ataxias in adulthood. The paper discusses ataxia telangiectasia at the gene/disease level only and does not report the variant c.6317A>T (p.N2106I); it contains no variant-level consequence, functional or case data usable for PVS1, PM4 or BP3.
In Europe and the USA, Friedreich's ataxia is the most common recessive ataxia, unlike in Japan where ataxia with oculomotor apraxia type 1 (AOA1) is the most common. The second most frequent recessive ataxia globally is ataxia telangiectasia, with the exception of North Africa where this is ataxia with vitamin E deficiency
Diagnostic work-up in chronic ataxias - 'Recessive family history' section (paragraph on relative frequencies of recessive ataxias)
A practice guideline from the American College of Medical Genetics and Genomics and the National Society of Genetic Counselors: referral indications for cancer predisposition assessment.
Hampel H et al. · 2015
ACMG/NSGC practice guideline on referral indications for cancer predisposition assessment. The gene ATM is not discussed and neither c.6317A>T nor p.(N2106I) appears; the paper provides no variant-specific evidence for PVS1, PM4 or BP3.
The evidence statements recorded by the interpretation pipeline. They have not been reinterpreted by the page.
gnomAD v4.1
This variant is present in gnomAD v4.1 (AF= 6.82588e-06; MAF= 0.00068%, 11/1611514 alleles, homozygotes = 0) and has highest observed frequency in the Remaining individuals population (AF= 4.81062e-05; MAF= 0.00481%, 3/62362 alleles, homozygotes = 0); grpmax FAF= 2.92e-06.
gnomAD v2.1
This variant is present in gnomAD v2.1 (AF= 3.97924e-06; MAF= 0.00040%, 1/251304 alleles, homozygotes = 0) and has highest observed frequency in the European (non-Finnish) population (AF= 8.80034e-06; MAF= 0.00088%, 1/113632 alleles, homozygotes = 0).
gnomAD-Canada
Absent from gnomAD-Canada v1.0.
ClinVar
This variant has been reported in ClinVar as Uncertain significance (5 clinical laboratories). (ClinVarID = 135768)
In silico
SpliceAI predicts no significant splice impact for this variant (max delta score = 0.01). REVEL score = 0.539. BayesDel score = -0.028517.
Functional
OncoKB did not identify variant-specific reviewed functional evidence for this variant; gene-level curated context is available for reviewer follow-up. ATM, a kinase involved in the DNA damage response, is mutated in various solid and hematologic malignancies.
Somatic
This variant does not lie in a statistically significant hotspot. This variant has previously been reported in somatic cancers (COSMIC; COSV99069667, n = 2 times).
Cancer Hotspots
This variant does not lie in a statistically significant hotspot.
This variant was interpreted by pipeline 8.0.0.
The current version is 11.0.07.
Re-running queues a fresh interpretation; the result replaces this page when complete.
Re-runs are not free. Blank draws on today's public interpretation allowance; with a code, it uses one of that code's uses.