STAG2 c.2431G>T, p.(Glu811Ter)

Reasoning

PM2 is met: the variant is absent (AF = 0) from gnomAD v2.1, gnomAD v4.1 (a large ~800K-allele cohort), and gnomAD-Canada v1.0, satisfying 'absent in large population cohorts' (Richards et al. 2015, PMID 25741868) and well below this lab's non-VCEP PM2 threshold of AF <0.1%. Absence is expected for a pathogenic loss-of-function variant in STAG2, an X-linked gene with strong germline LoF disease context (pvs1_gene_context). Consistent with ClinGen SVI recommendations, PM2 is applied at supporting strength only (PM2_supporting).

Rule

PM2: absent (or AF <0.1% lab non-VCEP threshold) in large population cohorts. Met at supporting strength (PM2_supporting).

Evidence considered

• Variant is absent from gnomAD v2.1 (X-123205071-G-T), gnomAD v4.1 (chrX-124071221-G-T), and gnomAD-Canada v1.0 (search_status 'absent' in all three), satisfying PM2 'absent in large population cohorts' (Richards et al. 2015, PMID 25741868).
• Absence across the gnomAD v4.1 cohort (~800K alleles) and gnomAD v2.1 is far below this lab's non-VCEP PM2 operating threshold of AF <0.1%.
• STAG2 has strong germline loss-of-function disease context (pvs1_gene_context), and extreme rarity in population databases is the expectation for a pathogenic X-linked LoF variant; PM2 is applied at supporting strength only (PM2_supporting) consistent with ClinGen SVI recommendations, rather than the moderate strength listed in the original 2015 framework.

Still missing

• Exact allele counts and per-population breakdown (e.g., hemizygous males vs heterozygous females) were not captured (source_data empty), so absence could not be verified against gnomAD coverage/quality metrics at this site.
• If the governing framework intends PM2 at moderate strength rather than supporting, strength would need to be adjusted; supporting strength is used here per ClinGen SVI guidance.

Sources

Reasoning

PS4 requires the prevalence of this exact variant to be significantly increased in affected individuals versus controls (case-control enrichment). No such data exists for STAG2 c.2431G>T (p.Glu811Ter): the variant is absent from ClinVar (no variation record, 0 submissions), absent from gnomAD v2.1/v4.1 and gnomAD-Canada, and absent from COSMIC. All four retrieved full-text papers (PMID:21852505, PMID:22417201, PMID:24121789, PMID:25010205) were reviewed: they describe gene-level STAG2 truncating/inactivating mutation prevalence in somatic cancers (e.g., 18% of urothelial carcinomas; 21.5% of Ewing sarcoma tumors) but never mention c.2431G>T or p.Glu811Ter, so they cannot support variant-level enrichment. OncoKB lists a variant-specific page for STAG2 E811* (curated 'Likely Oncogenic', Likely Loss-of-function), suggesting the alteration has been observed in somatic tumor sequencing, but no affected-individual counts or germline case-control data are provided, and a somatic curation label without quantified enrichment is not PS4-grade evidence. Generic ACMG/AMP 2015 (PMID 25741868) governs because no ClinGen CSPEC/VCEP or local gene framework exists for STAG2.

Rule

PS4 - The prevalence of the variant in affected individuals is significantly increased compared with controls (ACMG/AMP 2015, Richards et al., PMID 25741868; generic ACMG fallback, no STAG2 VCEP/CSPEC available)

Evidence considered

• Variant c.2431G>T (p.Glu811Ter) is absent from ClinVar: variation_ids empty, 0 submissions, no expert panels (evidence_brief clinvar block).
• Variant absent from gnomAD v2.1, gnomAD v4.1, and gnomAD-Canada v1.0; no control allele frequency exists to support or refute case-control enrichment.
• COSMIC search for 'STAG2 c.2431G>T' returned no record; no somatic recurrence count available.
• OncoKB variant page for STAG2 E811* curates 'Likely Oncogenic' with biological effect 'Likely Loss-of-function' (variant-specific, citations PMID:21852505, 22417201, 24121789, 25010205), but provides no case counts and is a somatic oncology curation, not germline case-control enrichment.
• Full-text review of the four retrieved papers found zero mentions of c.2431G>T, p.Glu811Ter, E811, or residue 811; they report gene-level STAG2 truncating-mutation prevalence in somatic cancers only.
• PS4 rule from Richards et al. 2015 (ACMG/AMP, PMID 25741868) applied under the generic ACMG fallback (generic_acmg_combination_rules): enrichment of the variant in affected individuals vs controls not demonstrated.

Still missing

• No case series or case-control study reporting the exact variant c.2431G>T in affected individuals versus controls
• No proband clinical context or indication for which this variant is being interpreted
• No germline disease-specific prevalence data for this exact variant

Sources

Reasoning

PP3 is not met. No computational evidence supports a deleterious effect for NM_001042749.1:c.2431G>T. Splice path: SpliceAI reports max delta 0.033 (DS_AG=0.033, DS_AL=0.026, DS_DG=0.001, DS_DL=0.020), far below the delta >= 0.2 splice-altering threshold recommended in the SpliceAI publication (Jaganathan et al. 2019, PMID 30661751), so there is no splice-impact support for PP3. Missense path: the variant is a nonsense substitution (NP_001036214.1:p.(Glu811Ter), consequence_class 'nonsense'), so missense-oriented predictors (REVEL/BayesDel/aGVGD) do not apply to this change; REVEL returned no score (lookup failed for chrX:124071221), and the BayesDel score of 0.652534 is treated as uncalibrated because no verified published PP3/BP4 calibration threshold is available to this pipeline (generic_acmg_combination_rules). Per the governing generic fallback, PP3 via the splice path requires SpliceAI max delta > 0.2, which is not met. The variant's protein-truncating consequence is captured by PVS1 (evaluated by another group), not by PP3; awarding PP3 here would double-count that effect. No criterion-supporting prediction exists, so PP3 is not met.

Rule

PP3 (ACMG/AMP 2015, PMID 25741868): Multiple lines of computational evidence support a deleterious effect on the gene or gene product (conservation, evolutionary, splicing impact, etc.).

Evidence considered

• No CSPEC/VCEP exists for STAG2 (cspec found=false; framework_mode=generic_acmg; precedence_rank 3); PP3/BP4 are governed by the generic ACMG/AMP 2015 fallback (Richards et al., PMID 25741868) with the operating calibration defined in generic_acmg_combination_rules: missense variants are assessed with REVEL only; intronic/synonymous/non-canonical-splice-position variants with SpliceAI only (SpliceAI max delta > 0.2 -> PP3 supporting; < 0.1 -> BP4 supporting; [0.1, 0.2] gray zone); BayesDel is never used because no verified published threshold is available.
• SpliceAI Lookup (spliceai) for NM_001042749.1:c.2431G>T reports max delta 0.033 (DS_AG=0.033, DS_AL=0.026, DS_DG=0.001, DS_DL=0.020; displayed summary 'max delta score = 0.03'), far below the delta >= 0.2 high-sensitivity splice-altering threshold recommended by Jaganathan et al. 2019 (PMID 30661751); provides no support for PP3 via the splice path.
• REVEL returned no score for chrX:124071221 (prefetch error: 'No REVEL ZIP found for chrX:124071221'); the ClinGen SVI REVEL calibration (Pejaver et al. 2022, PMID 36413997: REVEL >= 0.932 supports PP3 and <= 0.016 supports BP4 at supporting strength) cannot be applied because no score exists.
• Local BayesDel lookup (bayesdel) returned score 0.652534 for X:123205071 G>T; per generic_acmg_combination_rules, BayesDel has no verified published PP3/BP4 calibration threshold available to this pipeline and is never used for PP3/BP4, so this score is treated as uncalibrated (not_available) regardless of its value.
• The variant is a nonsense substitution (NP_001036214.1:p.(Glu811Ter) per variant normalization; consequence_class 'nonsense'), so missense in-silico predictors (REVEL/BayesDel/aGVGD) are not applicable to this change.

Still missing

• REVEL score unavailable for chrX:124071221 (prefetch error: 'No REVEL ZIP found for chrX:124071221'); the ClinGen SVI REVEL calibration (Pejaver et al. 2022, PMID 36413997) could not be applied, though the missense in-silico path is not applicable to this nonsense variant in any case.
• BayesDel score 0.652534 was returned by the local lookup but is treated as not usable: no verified published PP3/BP4 calibration threshold (named publication and table) is available to this pipeline. If such a calibration is added, this score could be re-evaluated.

Sources

Reasoning

Per group instructions, PP5 applies only for an exact-variant ClinVar expert-panel classification of Pathogenic/Likely pathogenic. The variant is absent from ClinVar entirely: no variation record, 0 submissions, no expert panels, so no exact-variant expert-panel pathogenic classification exists to trigger PP5. The OncoKB 'Likely Oncogenic' label was considered and rejected: it is a somatic oncology curation from a non-expert database, not a ClinVar expert-panel germline classification, and the instructions state non-expert assertions must never trigger PP5.

Rule

PP5 - Reputable source recently reports variant as pathogenic, but the evidence is unavailable to the laboratory (ACMG/AMP 2015, PMID 25741868); per group rule, applies ONLY for an exact-variant ClinVar expert-panel Pathogenic/Likely pathogenic classification

Evidence considered

• ClinVar query for NM_001042749.1:c.2431G>T (multiple search term combinations) returned no variation record; submitter_count 0, expert_panels [] (evidence_brief clinvar block).
• OncoKB labels STAG2 E811* 'Likely Oncogenic' (somatic, Likely Loss-of-function); this is not a ClinVar expert-panel classification and per group rule cannot trigger PP5.
• PP5 rule from Richards et al. 2015 (ACMG/AMP, PMID 25741868) as constrained by this group's rule: only exact-variant ClinVar expert-panel Pathogenic/Likely pathogenic classifications qualify.

Still missing

• No ClinVar record exists for this variant; if a future expert-panel submission appears, PP5 should be revisited

Sources

Reasoning

BA1 requires an allele frequency >5% in a large, ethnically diverse population cohort (Richards et al. 2015, PMID 25741868; reflected in generic_acmg_combination_rules). The variant is absent from gnomAD v2.1 (GRCh37 X-123205071-G-T), gnomAD v4.1 (GRCh38 chrX-124071221-G-T), and gnomAD-Canada v1.0 (search_status 'absent' in all three), i.e. observed allele frequency = 0. This is far below the 5% BA1 threshold and also below this lab's more lenient non-VCEP BA1 operating threshold (AF >1%). BA1 is not met.

Rule

BA1: allele frequency >5% in large population database (Richards et al. 2015); lab non-VCEP operating threshold AF >1%. Not met (AF = 0).

Evidence considered

• BA1 is defined as allele frequency >5% in large population datasets (Richards et al. 2015, PMID 25741868), as reflected in generic_acmg_combination_rules; this lab's non-VCEP BA1 operating threshold is AF >1%, which is more lenient and still not met.
• gnomAD v2.1 (variant X-123205071-G-T, GRCh37), gnomAD v4.1 (variant chrX-124071221-G-T, GRCh38), and gnomAD-Canada v1.0 all record the variant as absent (search_status 'absent'); observed allele frequency is 0, far below any BA1 threshold.
• No threshold artifact explains the absence: this is a single-nucleotide G>T transversion queried at genomic coordinates in both builds, and STAG2 has no known common benign LoF allele at this position.

Still missing

• Exact allele counts, per-population frequencies, and site coverage metrics were not captured in the gnomAD lookup results (source_data empty), so absence was not cross-checked against coverage/quality filters.

Sources

Reasoning

BS1 requires an allele frequency greater than expected for the disorder. STAG2 germline loss-of-function disease is a very rare X-linked condition (pvs1_gene_context establishes germline LoF disease context), so the expected pathogenic allele frequency is far below 1%; this lab's non-VCEP BS1 operating threshold is AF >0.3%. The variant is absent from gnomAD v2.1, gnomAD v4.1, and gnomAD-Canada v1.0 (AF = 0), so the observed frequency does not exceed the expected disorder frequency. BS1 is not met.

Rule

BS1: allele frequency greater than expected for disorder (lab non-VCEP operating threshold AF >0.3%). Not met (AF = 0).

Evidence considered

• BS1 requires allele frequency greater than expected for the disorder; STAG2 germline LoF disease is a very rare X-linked condition (germline disease context per pvs1_gene_context), so the expected pathogenic allele frequency is far below 1%, and this lab's non-VCEP BS1 operating threshold is AF >0.3%.
• The variant is absent from gnomAD v2.1, gnomAD v4.1, and gnomAD-Canada v1.0 (AF = 0); a frequency of 0 cannot exceed the expected disorder allele frequency, so BS1 is not met.
• Penetrance/prevalence considerations: for a highly LoF-intolerant X-linked gene such as STAG2, absence from large population cohorts is consistent with pathogenicity and provides no benign-frequency signal.

Still missing

• A quantitative disorder-specific expected allele frequency for STAG2 germline disease was not available in the bundle; the conclusion rests on the variant being entirely absent rather than on a precise prevalence-based comparison.

Sources

Reasoning

BS2 requires observation of the variant in healthy adult individuals in an appropriate state (heterozygous for dominant, hemizygous for X-linked, homozygous for recessive disorders). The variant is absent from gnomAD v2.1, gnomAD v4.1, and gnomAD-Canada v1.0, and has no ClinVar submissions, so there are no observations of the variant in any healthy control, adult, homozygote, or hemizygote. Absence of the variant is the opposite of what BS2 requires; BS2 is not met.

Rule

BS2: variant observed in healthy adult (homozygous/heterozygous/hemizygous as appropriate). Not met (no observations; variant absent from all population databases).

Evidence considered

• BS2 requires the variant to be observed in healthy adult individuals in an appropriate state (hemizygous males for an X-linked disorder such as STAG2); the variant was not observed in any population cohort (gnomAD v2.1, gnomAD v4.1, and gnomAD-Canada v1.0 all report absence), so no healthy-carrier observations exist.
• ClinVar returns no result for this variant (clinvar: no_result), providing no healthy-individual or phenotype-based observations that could support BS2.

Still missing

• No dedicated healthy-control cohort (e.g., in-house controls or disease-specific control sets) beyond the queried population databases was available in the case bundle.

Sources

Reasoning

BP4 is not met. BP4 requires computational evidence suggesting no impact on the gene or gene product; NM_001042749.1:c.2431G>T is a stop-gain (NP_001036214.1:p.(Glu811Ter)) whose deterministic protein-truncating consequence is itself an impact on the gene product, so the criterion's 'no impact' premise is not satisfied regardless of any splicing prediction. This mirrors the consequence-class gate applied deterministically to BP7 for this case (BP7.json: requires_synonymous not met for a nonsense variant). The only no-impact prediction available is SpliceAI (max delta 0.033, below the 0.2 splice-altering threshold of Jaganathan et al. 2019, PMID 30661751): it is a single splice-mechanism line, it does not constitute 'multiple lines of computational evidence suggest no impact on gene or gene product', and it does not negate the stop-gain effect. The governing generic fallback (generic_acmg_combination_rules) assigns the SpliceAI-only path to intronic/synonymous/non-canonical-splice-position variants; a nonsense substitution is not within that enumerated scope, and BP7 - the designated criterion that absorbs splice no-impact predictions for silent variants - is not applicable here. REVEL returned no score, BayesDel (0.652534) is uncalibrated and unusable, and no conservation score was available, so no independent no-impact line exists. BP4 is therefore not applied, and the single SpliceAI prediction is not double-counted into BP4.

Rule

BP4 (ACMG/AMP 2015, PMID 25741868): Multiple lines of computational evidence suggest no impact on gene or gene product (conservation, evolutionary, splicing impact, etc.).

Evidence considered

• Governing framework: generic ACMG/AMP 2015 fallback (generic_acmg_combination_rules; Richards et al., PMID 25741868) because no STAG2 CSPEC/VCEP exists; BP4 requires multiple lines of computational evidence suggesting no impact on the gene or gene product.
• SpliceAI (spliceai) predicts no splice impact for NM_001042749.1:c.2431G>T (max delta 0.033, below the 0.2 splice-altering threshold of Jaganathan et al. 2019, PMID 30661751); this is a single splice-mechanism line and is the only no-impact prediction available in the case bundle.
• The variant is a stop-gain (NP_001036214.1:p.(Glu811Ter)); the deterministic protein-truncating consequence is an impact on the gene product, so BP4's 'no impact on gene or gene product' conclusion is not satisfied regardless of the splice prediction - the same structural reasoning applied to BP7 in this case (BP7.json: not_applicable, requires_synonymous).
• REVEL returned no score for chrX:124071221 (lookup failed), and BayesDel (0.652534) is uncalibrated and never used for PP3/BP4 (generic_acmg_combination_rules); no independent missense-path no-impact line exists, and the missense path is not applicable to a nonsense variant in any case.
• No nucleotide conservation score was available in the case bundle, so no conservation-based no-impact evidence exists for BP4.

Still missing

• No nucleotide conservation score (e.g., phyloP/GERP) was present in the case bundle; conservation-based no-impact evidence for BP4 could not be assessed.
• REVEL/BayesDel missense in-silico scores were not usable (REVEL lookup failed; BayesDel uncalibrated per generic_acmg_combination_rules), so the missense BP4 path could not contribute an independent no-impact line; for this nonsense variant that path is not applicable in any case.

Sources

Reasoning

Per group instructions, BP6 applies only for an exact-variant ClinVar expert-panel classification of Benign/Likely benign. The variant is absent from ClinVar entirely (no variation record, 0 submissions, no expert panels), so no exact-variant expert-panel benign classification exists to trigger BP6. No other source provides a benign assertion for this variant; OncoKB's 'Likely Oncogenic' label is a somatic, non-expert curation and runs contrary to benign status.

Rule

BP6 - Reputable source recently reports variant as benign, but the evidence is unavailable to the laboratory (ACMG/AMP 2015, PMID 25741868); per group rule, applies ONLY for an exact-variant ClinVar expert-panel Benign/Likely benign classification

Evidence considered

• ClinVar query for NM_001042749.1:c.2431G>T returned no variation record; no benign expert-panel classification exists.
• OncoKB labels STAG2 E811* 'Likely Oncogenic' (non-expert, somatic); it is not a benign assertion and cannot trigger BP6.
• BP6 rule from Richards et al. 2015 (ACMG/AMP, PMID 25741868) as constrained by this group's rule: only exact-variant ClinVar expert-panel Benign/Likely benign classifications qualify.

Still missing

• No ClinVar record exists for this variant; if a future expert-panel submission appears, BP6 should be revisited

Sources

Reasoning

PVS1 could not be assessed because its agent did not produce valid output.

Still missing

• attempt 1: invalid criterion strength: pvs1
• attempt 2: invalid criterion strength: pvs1
• attempt 3: invalid criterion strength: pvs1

Reasoning

PS2 (de novo with confirmed maternity and paternity in a proband with the disease and no family history) cannot be adjudicated: the case bundle contains no proband clinical/pedigree information and no parental genotype or maternity/paternity confirmation data for STAG2 c.2431G>T (p.Glu811Ter). No source in the registry reports a de novo observation of this variant. Absence of family data is a data gap, not evidence against de novo origin.

Rule

PS2: De novo (maternity and paternity confirmed) in a patient with the disease and no family history (Richards et al. 2015, PMID:25741868; generic ACMG fallback).

Evidence considered

• No VCEP/CSPEC exists for STAG2 (cspec found=false; framework_mode=generic_acmg); the governing framework for PS2 is the generic ACMG/AMP 2015 criteria (Richards et al., PMID:25741868), as represented by generic_acmg_combination_rules.
• The case bundle contains no proband, parental, or pedigree data of any kind (review_manifest, case_summary, evidence.json, literature_pass all lack family/segregation content), so the de novo status of c.2431G>T cannot be established or excluded.
• ClinVar contains no record for this variant (evidence_sentence: 'This variant is absent from ClinVar.'), so no external submission documents a de novo occurrence.
• All 4 available full-text publications (PMID:21852505, PMID:22417201, PMID:24121789, PMID:25010205) are somatic cancer genomics studies; full-text search for '2431', 'Glu811', and 'E811' returned no hits, so none reports this variant or any germline de novo observation (per citation rule, papers are only citable when the exact variant is mentioned).
• STAG2 is X-linked (Xq25); for a nonsense variant in an X-linked gene a true de novo finding would be a legitimate PS2 observation, but it requires trio/proband and parental genotype data that are absent here.

Still missing

• No proband genotype/phenotype data in the case
• No parental (maternity/paternity) testing results
• No documentation of family history (affected/unaffected relatives)
• Variant absent from ClinVar, so no externally reported de novo observation is available

Sources

Reasoning

PS3 (Richards et al. 2015, PMID:25741868) requires well-established functional studies demonstrating a damaging effect on the gene or gene product. No CSPEC/VCEP specification exists for STAG2 (cspec found=false, gene_vcep_dir=null), so the generic ACMG/AMP 2015 fallback applies. The evidence bundle contains no validated functional assay performed on this exact variant: OncoKB provides a variant-specific curation for E811* ('biological-effect context: Likely Loss-of-function; curated oncogenicity label: Likely Oncogenic'), but this is a curated classification label from a somatic-oncology knowledge base, not a well-established functional study, and a criterion must not be inferred from another source's final classification. The four ONCOKB-attached publications (PMID:21852505, 22417201, 24121789, 25010205) document gene-level STAG2 loss-of-function biology in cancers (aneuploidy upon STAG2 inactivation; loss of STAG2 expression by IHC with truncating mutations), but none of them mentions c.2431G>T / p.Glu811Ter in the abstract or full text, so under strict citation grounding they cannot serve as variant-specific PS3 evidence. Gene-level LoF mechanism support exists (pvs1_gene_context, literature-only) and is biologically relevant context, but it is not a variant-specific functional assay. BayesDel (0.6525) is in-silico with no verified published threshold and is not functional evidence. PS3 is therefore not assessable from the available evidence; the gap is absence of a well-established variant-specific functional assay.

Rule

generic_acmg_2015_PS3

Evidence considered

• Adjudicated under generic ACMG/AMP 2015 (Richards et al., PMID:25741868) via generic_acmg_combination_rules because no CSPEC/VCEP specification exists for STAG2 (cspec found=false, gene_vcep_dir=null, framework_mode=generic_acmg).
• Searched the full text of all four ONCOKB-attached STAG2 papers (PMID:21852505, PMID:22417201, PMID:24121789, PMID:25010205) for 'c.2431G>T', 'Glu811', 'E811', 'p.Glu811Ter' and '2431': no paper mentions this exact variant, so none can be cited as variant-specific PS3 evidence.
• PMID:21852505 (Solomon et al. 2011, Science) provides gene-level functional evidence that STAG2 inactivation causes aneuploidy: AAV-mediated correction of endogenous STAG2 mutations in glioblastoma cell lines restored STAG2 protein expression and narrowed ploidy peaks; the tested alleles (25-bp insertion in H4 cells; a nonsense mutation in 42MGBA cells) are not c.2431G>T.
• PMID:24121789 (Balbas-Martinez et al. 2013) shows truncating STAG2 mutations lead to loss of the C-terminal epitope and loss of STAG2 expression by IHC in 52/295 urothelial carcinomas, but no variant-level assay of c.2431G>T.
• OncoKB curates the E811* allele as 'Likely Oncogenic' with biological-effect context 'Likely Loss-of-function' (variant-specific page, screenshot screenshots/oncokb.png); this is a curated classification in a somatic-oncology context, not a validated functional assay, and PS3 is not inferred from another source's final classification.
• pvs1_gene_context confirms germline LoF mechanism support for STAG2 (literature-only source, supporting_sources PMID:42409735, 31334757, 32687945, 34529785, 35203280), which supports biological relevance of loss-of-function but is gene-level context, not a variant-specific functional study.
• BayesDel score 0.6525 is an in-silico predictor with no verified published threshold in this pipeline; it is not functional evidence and was not used for PS3.

Still missing

• No well-established functional assay (e.g., sister chromatid cohesion assay, aneuploidy/ploidy analysis, STAG2 protein expression by Western blot/IHC, or cohesin complex assembly) performed specifically on c.2431G>T / p.Glu811Ter.
• Retrieved STAG2 functional literature is gene-level LoF evidence from somatic cancers and does not test this variant; OncoKB biological-effect context is a curated label, not assay-based evidence.

Sources

Reasoning

PM3 (Richards et al. 2015, PMID:25741868) requires, for a recessive disorder, observation of the variant in trans with a pathogenic/likely-pathogenic variant in an affected proband (or homozygosity/hemizygosity consistent with a biallelic model). No such data exist for NM_001042749.1:c.2431G>T (p.Glu811Ter) in STAG2. The case bundle contains no affected-proband observation with a second STAG2 variant, no phase (trans/cis) determination, no biallelic/homozygous/hemizygous genotype observation, and no family segregation data. The variant is absent from ClinVar and is not reported in any of the four bundled publications (PMID 21852505, 22417201, 24121789, 25010205 — all somatic cancer cohorts that do not mention c.2431G>T or p.Glu811Ter). In addition, no CSPEC/VCEP or local gene-specific framework exists for STAG2 (generic ACMG/AMP 2015 fallback governs), and the germline disease inheritance mode for STAG2 is not established in this case (pvs1_gene_context inheritance list is empty), so the recessive-model prerequisite for PM3 cannot be confirmed. Evidence is therefore insufficient; the criterion is not assessed pending phase/biallelic observations in affected individuals.

Rule

PM3 - For recessive disorders, detected in trans with a pathogenic variant (Richards et al. 2015 generic ACMG/AMP framework)

Evidence considered

• Governing framework: no CSPEC/VCEP or local gene-specific specification for STAG2; generic ACMG/AMP 2015 rules (Richards et al. 2015, PMID:25741868) apply, in which PM3 requires observation in trans with a pathogenic variant for recessive disorders.
• ClinVar search for STAG2 c.2431G>T (including transcript, genomic, and gene+varName queries) returned no records, so no prior allelic/phase observations are documented.
• Screened all four bundled publications (PMID 21852505, 22417201, 24121789, 25010205); none report this exact variant or any germline allelic (trans/cis/proband) observation for it. Full text for 25010205 was checked directly (no c.2431/E811/Glu811 mentions); the other three full-text files were empty (0 bytes), so only abstracts were available, and none mention the variant.
• Case bundle contains no proband/pedigree, family, segregation, or phase data anywhere in prefetch.json, evidence.json, case_summary.json, or clinvar_submission_audit.json; PM3 evidence is entirely absent rather than conflicting.
• STAG2 is X-linked (Xq25); pvs1_gene_context lists no established inheritance mode (inheritance list empty), and the germline disease context is literature-only, so a recessive biallelic model cannot be assumed for PM3.

Still missing

• No affected proband(s) with a second STAG2 pathogenic variant observed; compound-heterozygous or homozygous/hemizygous genotype data for c.2431G>T are required.
• No phase (trans/cis) determination is available for c.2431G>T relative to any second variant.
• Germline inheritance mode for STAG2 disease is not established in this case; a recessive (or X-linked recessive) disease model with an allelic requirement must be confirmed before PM3 can be applied.

Sources

Reasoning

PM6 (assumed de novo without confirmation of maternity/paternity) cannot be adjudicated: there is no proband observation of c.2431G>T and no parental testing data of any kind in the case bundle, so an 'assumed de novo' status cannot be inferred. There is no evidence to support or refute a de novo origin.

Rule

PM6: Assumed de novo, without confirmation of maternity and paternity (Richards et al. 2015, PMID:25741868; generic ACMG fallback).

Evidence considered

• No VCEP/CSPEC exists for STAG2; PM6 is evaluated under the generic ACMG/AMP 2015 framework (Richards et al., PMID:25741868) via generic_acmg_combination_rules.
• The case bundle provides no proband or parental genotype data, so an assumed de novo status (PM6) cannot be established for c.2431G>T.
• ClinVar has no entry for this variant, and none of the 4 available papers (PMID:21852505, PMID:22417201, PMID:24121789, PMID:25010205) mentions c.2431G>T (full-text search for '2431', 'Glu811', 'E811' returned no hits), so no externally reported assumed-de novo observation exists.
• Because PS2's required parental confirmation data are also absent, PM6 cannot be inferred from any partial trio information — no partial information exists at all.

Still missing

• No proband genotype data for c.2431G>T
• No parental genotype results (needed to infer de novo origin)
• No ClinVar or literature report of an assumed de novo occurrence for this variant

Sources

Reasoning

PP1 (co-segregation with disease in affected family members) cannot be adjudicated: no family members, affected or unaffected, are genotyped anywhere in the case bundle, and no meiotic/segregation observations are reported for c.2431G>T in ClinVar or the available literature. There are zero meioses to score.

Rule

PP1: Co-segregation with disease in affected family members (Richards et al. 2015, PMID:25741868; generic ACMG fallback).

Evidence considered

• No VCEP/CSPEC exists for STAG2; PP1 is evaluated under the generic ACMG/AMP 2015 framework (Richards et al., PMID:25741868) via generic_acmg_combination_rules.
• The case bundle contains no pedigree, no family member genotypes, and no segregation observations, so there are no informative meioses to count for PP1.
• ClinVar has no entry for this variant, and the 4 available papers (PMID:21852505, PMID:22417201, PMID:24121789, PMID:25010205) are somatic tumor studies that never mention c.2431G>T (full-text search returned no hits), so no published segregation evidence exists for this exact variant.
• For an X-linked gene (STAG2, Xq25), segregation in informative meioses (hemizygous affected males, carrier females) could support PP1 if family data were available, but none is present.

Still missing

• No affected family members tested for the variant
• No unaffected family members tested (needed to count informative meioses)
• No pedigree structure or segregation data in the case
• No ClinVar/literature report of segregation for this variant

Sources

Reasoning

PP4 requires the proband's phenotype or family history to be highly specific for a disease with a single genetic etiology. The case bundle contains no proband phenotype data whatsoever: case_summary.json has no phenotype/clinical section, no family-history or segregation context, and no indication is recorded. Without any phenotypic information, phenotype specificity cannot be evaluated. Gene-level knowledge that STAG2 loss-of-function is associated with cohesinopathy phenotypes and somatic cancer does not substitute for proband-specific phenotype evidence under PP4.

Rule

PP4 - Variant's phenotype or family history is highly specific for a disease with a single genetic etiology (ACMG/AMP 2015, Richards et al., PMID 25741868)

Evidence considered

• Case bundle contains no proband phenotype, clinical summary, or family-history data (case_summary.json has no phenotype section), so PP4 cannot be evaluated.

Still missing

• No proband phenotype, clinical summary, or family history provided in the case bundle
• No disease/indication context for which the variant is being interpreted

Reasoning

BS3 (Richards et al. 2015, PMID:25741868) requires well-established functional studies demonstrating no damaging effect on protein function or splicing. No CSPEC/VCEP specification exists for STAG2, so the generic ACMG/AMP 2015 fallback applies. The evidence bundle contains no functional study of c.2431G>T at all, and nothing demonstrates retained STAG2 function or normal splicing for this allele. The variant is a nonsense change (p.Glu811Ter) predicted to truncate the protein, and the retrieved gene-level literature consistently associates STAG2 truncating variants with loss of protein expression and aneuploidy, providing no support for a benign functional effect. SpliceAI (max delta 0.033) is an in-silico prediction and is not a well-established functional study; it belongs to the in-silico/splicing domain (PP3/BP4 rules), not to BS3. BS3 is therefore not assessable from the available evidence; the gap is the complete absence of functional evidence of a non-damaging effect.

Rule

generic_acmg_2015_BS3

Evidence considered

• Adjudicated under generic ACMG/AMP 2015 (Richards et al., PMID:25741868) via generic_acmg_combination_rules because no CSPEC/VCEP specification exists for STAG2 (cspec found=false, gene_vcep_dir=null).
• Searched the full text of all four ONCOKB-attached STAG2 papers (PMID:21852505, PMID:22417201, PMID:24121789, PMID:25010205) for 'c.2431G>T', 'Glu811', 'E811', 'p.Glu811Ter' and '2431': none mentions this variant, and none reports functional data suggesting a benign effect.
• Gene-level literature associates STAG2 truncating mutations with loss of STAG2 expression by IHC (PMID:24121789: 'Virtually all tumor-derived STAG2 mutations discovered to date are truncating ... which lead to absence of the carboxyl-terminal epitope and therefore loss of expression via IHC') and with aneuploidy upon STAG2 inactivation (PMID:21852505), which runs counter to a benign functional interpretation.
• SpliceAI max delta 0.033 (spliceai source) is an in-silico prediction, not a well-established functional study; it is evaluated under in-silico BP4/PP3 rules by the appropriate group and cannot support BS3.
• OncoKB E811* curation ('Likely Loss-of-function' biological effect context) provides no evidence of retained function and is a classification label rather than a functional assay.
• No ClinVar submissions (clinvar source: variant absent) and no COSMIC records were found for this variant, so no curated functional annotation from those sources is available either.

Still missing

• No functional studies demonstrate absence of a damaging effect on STAG2 function or splicing for c.2431G>T / p.Glu811Ter (e.g., no assay showing normal cohesion, normal protein expression, or preserved splicing).
• No evidence in the record supports retained protein function; the variant is a nonsense allele in a gene whose LoF mechanism is supported, so BS3 cannot be met without dedicated variant-specific functional data.

Sources

Reasoning

BS4 (lack of segregation in affected members of a family) cannot be adjudicated: BS4 requires documented non-segregation, i.e., an affected family member who does not carry the variant. No family members are genotyped in this case, so there is no observation of non-segregation; the absence of family testing is not evidence of non-segregation.

Rule

BS4: Lack of segregation in affected members of a family (Richards et al. 2015, PMID:25741868; generic ACMG fallback).

Evidence considered

• No VCEP/CSPEC exists for STAG2; BS4 is evaluated under the generic ACMG/AMP 2015 framework (Richards et al., PMID:25741868) via generic_acmg_combination_rules.
• The case bundle contains no family member genotypes, so no affected individual can be shown to lack c.2431G>T; BS4 (non-segregation) is therefore unsupported and cannot be scored.
• ClinVar has no record for this variant and none of the 4 available papers (PMID:21852505, PMID:22417201, PMID:24121789, PMID:25010205) mentions c.2431G>T, so no non-segregation observation exists in any source.
• Population absence (gnomAD v2.1/v4.1/Canada) is not BS4 evidence; BS4 specifically requires intra-familial non-segregation, which is absent here.

Still missing

• No affected family members tested (required to demonstrate non-segregation)
• No family pedigree or segregation data in the case
• No ClinVar/literature report of non-segregation for this variant

Sources

Reasoning

BP2 (Richards et al. 2015, PMID:25741868) requires observation of the variant in trans with a pathogenic variant for a fully dominant disease/gene, or observation in cis with a pathogenic variant in any inheritance pattern. Neither condition can be evaluated for NM_001042749.1:c.2431G>T (p.Glu811Ter) in STAG2: no second variant has been observed in cis or in trans with this variant, no phase information exists, and no affected individuals carrying this variant together with another STAG2 allele are reported. The variant is absent from ClinVar, is not reported in the four bundled publications (all somatic cancer cohorts), and no fully dominant germline disease model for STAG2 with a defined allelic requirement is established in this case. Evidence is therefore insufficient; the criterion is not assessed pending familial/phase observations.

Rule

BP2 - Observed in trans with a pathogenic variant for a fully dominant disease/gene, or observed in cis with a pathogenic variant in any inheritance pattern (Richards et al. 2015 generic ACMG/AMP framework)

Evidence considered

• Governing framework: no CSPEC/VCEP or local gene-specific specification for STAG2; generic ACMG/AMP 2015 rules (Richards et al. 2015, PMID:25741868) apply, in which BP2 requires observation in trans with a pathogenic variant for a fully dominant disease/gene or in cis with a pathogenic variant in any inheritance pattern.
• ClinVar search for STAG2 c.2431G>T returned no records, so no documented cis/trans observations with a second pathogenic variant exist.
• Screened all four bundled publications (PMID 21852505, 22417201, 24121789, 25010205); none report this exact variant or any cis/trans allelic observation involving it; full text for 25010205 was checked directly (no c.2431/E811/Glu811 mentions), and the other three full-text files were empty (0 bytes).
• Case bundle contains no proband/pedigree, family, segregation, or phase data anywhere in prefetch.json, evidence.json, case_summary.json, or clinvar_submission_audit.json; BP2 evidence is entirely absent rather than conflicting.
• STAG2 is X-linked (Xq25) and its germline disease context is literature-only with no established inheritance mode in this case; the cis/trans observations required by BP2 cannot be interpreted without family data.

Still missing

• No observation of c.2431G>T in cis with a pathogenic variant (any inheritance pattern) is available.
• No observation of c.2431G>T in trans with a pathogenic variant for a fully dominant disease/gene is available; no fully dominant germline STAG2 disease model is established in this case.
• No family/phase testing data (parental or sibling genotypes) are present in the case bundle.

Sources

Reasoning

BP5 applies when the variant is found in a case with an alternate molecular basis for disease (e.g., a second causal variant or alternative diagnosis identified in the proband). The case bundle contains no proband-level data at all - no genotype context, no alternate-diagnosis findings, no clinical case description - so the presence or absence of an alternate molecular basis cannot be determined.

Rule

BP5 - Variant found in a case with an alternate molecular basis for disease (ACMG/AMP 2015, Richards et al., PMID 25741868)

Evidence considered

• No case-level data exists in the bundle (no proband genotype context, no alternate-diagnosis information), so BP5 cannot be evaluated.

Still missing

• No proband case data (alternate causal variant or alternate diagnosis) available to evaluate an alternate molecular basis