MLH3 c.2390G>A, p.(Arg797His)

Reasoning

Scope and governing framework: no ClinGen CSPEC, VCEP specification or local gene-specific framework exists for MLH3 (case_summary.json: cspec.found = false; review_manifest.json: gene_vcep_dir null; vcep_materials.json: no CSPEC URL and no gene VCEP files; final_classification_framework.json resolves to the generic ACMG/AMP 2015 fallback, framework_mode = generic_acmg), so BA1 was adjudicated under generic ACMG/AMP 2015 (Richards et al. 2015, PMID:25741868), which assigns stand-alone benign support when the allele frequency in a control population exceeds 5%. No specification designates a non-cancer or exome-only population source for this gene, so the all-comers gnomAD v2.1/v4.1 entries are the default sources, and the non-cancer subsets present in the case evidence as well as the exome-only and genome-only splits were reviewed only as sensitivity checks. Frequencies for this exact variant (hg37 chr14:75513969 C>T = hg38 chr14:75047266 C>T): gnomAD v2.1 total AF 0.006369 (1801/282780 alleles, 72 homozygotes), African/African American AF 0.066113 (1649/24942 alleles, 71 homozygotes) with grpmax filtering AF 0.063292; gnomAD v4.1 total AF 0.003509 (5663/1614070 alleles, 174 homozygotes), African/African American AF 0.066275 (4971/75006 alleles, 171 homozygotes) with grpmax filtering AF 0.064736. Both ancestry-specific popmax filtering AF values (0.0633 and 0.0647) and both African-ancestry point frequencies (6.611% and 6.627%) exceed the 5% stand-alone threshold, and they rest on large allele counts (AC 1649 and 4971 over AN 24942 and 75006), so this is neither a small-sample nor a threshold artifact; the 16.7% HGDP:MANDENKA value in the gnomAD v3.1 non-cancer record (7/42 alleles, 0 homozygotes) is a small-sample artifact and was deliberately excluded. The signal reproduces across independent dataset partitions - gnomAD v2.1 exome-only AF 0.004881 (1227/251392) with exome grpmax filtering AF 0.063292 and genome-only AF 0.018287 (574/31388); gnomAD v2.1 non-cancer exomes African AF 0.065627 (1548/23588) with grpmax filtering AF 0.062438; gnomAD v3.1 non-cancer genomes African AF 0.066300 (2724/41086) with grpmax filtering AF 0.064295; gnomAD v4.1 joint grpmax filtering AF 0.064736 with exome grpmax 0.063578 and genome grpmax 0.064533 - so BA1 remains met on an exome-only reading, on a non-cancer-only reading and on genomes alone, ruling out exome/genome pooling or cancer-cohort ascertainment as the explanation. gnomAD-Canada v1.0 (gnomad_canada; 96/18416 alleles, AF 0.005213, 5 homozygotes) matches the global all-comers figure. Supporting context: for a rare, adult-onset, incompletely penetrant hereditary cancer predisposition, an allele at 6.6% in a major ancestry group (and 0.35-0.64% overall) with dozens to hundreds of homozygous carriers is incompatible with a monogenic disease allele, and the homozygote counts sit at or modestly above Hardy-Weinberg expectation (gnomAD v4.1 African 171 observed vs about 165 expected; gnomAD v3.1 non-cancer African 91 vs about 90; gnomAD v2.1 African 71 vs about 55), giving no sign of selection against biallelic carriers. The ClinVar record for this variant was inspected at submission level only (13 submissions, all from ordinary clinical laboratories, no expert panel; clinvar_submission_audit.json shows no submission supplying a frequency criterion) and its aggregate Benign classification was deliberately not used to derive this or any other criterion. BA1 is met at stand-alone strength.

Rule

Generic ACMG/AMP 2015 BA1 (stand-alone benign): control-population allele frequency above 5% (Richards et al. 2015, PMID:25741868). No MLH3 CSPEC/VCEP threshold exists; the all-comers gnomAD v2.1/v4.1 popmax (grpmax) filtering AF was compared.

Evidence considered

• Governing framework: no CSPEC/VCEP or local MLH3 framework exists (case_summary.json cspec.found = false; review_manifest.json gene_vcep_dir null; vcep_materials.json no CSPEC URL and no gene VCEP files; final_classification_framework.json source = generic_acmg_fallback), so the generic ACMG/AMP 2015 framework governs (PMID:25741868).
• BA1 threshold applied verbatim from the supplied generic calibration: allele frequency >= 0.05 = stand-alone benign support (ACMG/AMP 2015 stand-alone allele-frequency criterion, PMID:25741868).
• gnomAD v2.1 (gnomad_v2): total AF 0.006369 (1801/282780 alleles, 72 homozygotes); African/African American AF 0.066113 (1649/24942, 71 homozygotes); grpmax filtering AF 0.063292; exome-only AF 0.004881 (1227/251392, 49 homozygotes) with exome grpmax filtering AF 0.063292; genome-only AF 0.018287 (574/31388, 23 homozygotes).
• gnomAD v4.1 (gnomad_v4): total AF 0.003509 (5663/1614070 alleles, 174 homozygotes); African/African American AF 0.066275 (4971/75006, 171 homozygotes); grpmax filtering AF 0.064736 (joint); exome grpmax filtering AF 0.063578; genome grpmax filtering AF 0.064533.
• Non-cancer and subset sensitivity checks drawn from the same population source records: gnomAD v2.1 non-cancer exomes AF 0.004716 (1117/236856, 40 homozygotes) with African AF 0.065627 (1548/23588, 62 homozygotes) and grpmax filtering AF 0.062438; gnomAD v3.1 non-cancer genomes AF 0.019239 (2845/147874, 92 homozygotes) with African AF 0.066300 (2724/41086, 91 homozygotes) and grpmax filtering AF 0.064295; gnomAD-Canada v1.0 AF 0.005213 (96/18416, 5 homozygotes). An exome-only reading keeps BA1 met (exome grpmax filtering AF 0.0633-0.0636).
• Small-sample/threshold-artifact control: the 16.7% HGDP:MANDENKA frequency in the gnomAD v3.1 non-cancer record is 7/42 alleles with 0 homozygotes and was excluded; the popmax values used rest on AC 1649-4971 over AN 23588-75006.
• Ancestry distribution: the allele is common in African/African American (6.61-6.63%) and Admixed American (0.33% in v2.1) individuals and essentially absent in East Asian individuals (0/19948 in v2.1), so the popmax used is the African-ancestry estimate with the largest sample size.
• Disease context for interpreting that frequency: dominant mismatch-repair carrier prevalence is estimated at about 1 in 300 and hereditary MMR cancer is adult-onset with variable penetrance (PMID:34043773), so a 6.6% allele frequency is far outside the range compatible with a rare pathogenic allele.
• ClinVar VCV000314383 was consulted at submission level only (13 submissions, 12 clinical laboratories, no expert-panel submission; clinvar_submission_audit.json records no frequency criterion in any submission). The aggregate Benign classification was deliberately not used to derive BA1 or any other criterion.
• Combination-rule context: the generic benign combination rules state that BA1 alone yields a Benign classification (generic_acmg_combination_rules, per PMID:25741868), so no second benign criterion is required for this variant.

Still missing

• No MLH3-specific CSPEC/VCEP frequency rule exists, so the generic 5% ACMG/AMP stand-alone threshold was applied rather than a gene-specific derived threshold.
• No MLH3-specific disease prevalence or penetrance source was retrievable, so an expected allele frequency for MLH3-associated hereditary cancer could not be derived independently; this cannot change the verdict because the popmax frequency already exceeds the generic stand-alone threshold by more than an order of magnitude.
• Sherloc (PMID:28492532), which refines population criteria and is cited by one ClinVar submission (SCV000562043) as a generic criteria reference, was not retrievable in full text, so its refinements could not be checked for an alternative BA1 treatment.
• No patient ancestry is recorded in this variant-level case, so BA1 was driven by the highest-ancestry popmax frequency rather than by an ancestry-matched comparison; the AFR point estimates rest on 24942 and 75006 alleles, so this does not affect the call.

Sources

Reasoning

Governing framework: generic ACMG/AMP 2015 (no MLH3 CSPEC/VCEP or local framework; cspec.found = false, gene_vcep_dir null, framework_mode generic_acmg), under which an allele frequency in a control population greater than expected for the disorder is strong support for a benign interpretation (BS1; Richards et al. 2015, PMID:25741868). No MLH3 specification exists, so the generic calibration threshold of AF >= 0.01 was applied, with per-gene derivation preferred in principle but unavailable here (threshold source: Whiffin et al. 2017, PMID:28518168). The same frequency observations that satisfy BA1 also satisfy BS1. Ancestry popmax: gnomAD v2.1 African/African American AF 0.066113 (1649/24942 alleles) with grpmax filtering AF 0.063292; gnomAD v4.1 African/African American AF 0.066275 (4971/75006) with grpmax filtering AF 0.064736 - 6.3-6.5 fold above the 0.01 threshold. Dataset-wide frequencies also exceed it in the datasets where the allele is common: gnomAD v3.1 non-cancer genomes total AF 0.019239 (2845/147874 alleles, 92 homozygotes), gnomAD v2.1 genome-only AF 0.018287 (574/31388), gnomAD v2.1 total AF 0.006369 (1801/282780), gnomAD v4.1 total AF 0.003509 (5663/1614070), gnomAD-Canada v1.0 AF 0.005213 (96/18416); only the exome-only partitions fall marginally below 0.01 (gnomAD v2.1 exome AF 0.004881; gnomAD v2.1 non-cancer exome AF 0.004716), while their grpmax filtering AF remains 0.0624-0.0633. Disease-prevalence context (PMID:34043773): dominant mismatch-repair carrier prevalence is about 1 in 300, so a 6.6% allele frequency is far greater than expected for this disorder class, and the observable frequency signal is reproduced in the non-cancer subsets (gnomAD v2.1 non-cancer exomes African AF 0.065627; gnomAD v3.1 non-cancer genomes African AF 0.066300) and in genomes alone, so it is not an artifact of cancer-ascertained cohorts or of pooling exomes with genomes. Because BS1 and BA1 rest on the same observation, BS1 adds no independent benign weight in this case - under the generic combination rules BA1 alone already produces a Benign classification (generic_acmg_combination_rules) - and is recorded at strong strength for completeness rather than to be counted twice. BS1 is met at strong strength.

Rule

Generic ACMG/AMP 2015 BS1 (strong benign): control-population allele frequency greater than expected for the disorder, applied at the supplied generic threshold AF >= 0.01 (Whiffin et al. 2017, PMID:28518168), using the gnomAD popmax filtering AF because no MLH3 specification exists.

Evidence considered

• Governing framework: generic ACMG/AMP 2015 (no CSPEC/VCEP or local MLH3 framework), under which BS1 is strong benign support when the control-population frequency is greater than expected for the disorder (Richards et al. 2015, PMID:25741868).
• BS1 threshold applied verbatim from the supplied generic calibration: AF >= 0.01 (generic default with per-gene derivation preferred; Whiffin et al. 2017, PMID:28518168).
• gnomAD v2.1 (gnomad_v2): African/African American AF 0.066113 (1649/24942) with grpmax filtering AF 0.063292; total AF 0.006369 (1801/282780, 72 homozygotes); genome-only AF 0.018287; exome-only AF 0.004881 with exome grpmax filtering AF 0.063292.
• gnomAD v4.1 (gnomad_v4): African/African American AF 0.066275 (4971/75006) with grpmax filtering AF 0.064736; total AF 0.003509 (5663/1614070, 174 homozygotes).
• Non-cancer and subset checks: gnomAD v2.1 non-cancer exomes AF 0.004716 (1117/236856) with African AF 0.065627 (1548/23588) and grpmax filtering AF 0.062438; gnomAD v3.1 non-cancer genomes AF 0.019239 (2845/147874) with African AF 0.066300 (2724/41086) and grpmax filtering AF 0.064295; gnomAD-Canada v1.0 AF 0.005213 (96/18416).
• Disease-prevalence context (PMID:34043773): path_MMR carrier prevalence is estimated at about one in 300 and penetrance varies by gene and gender 'from very high to immeasurable', so a 6.6% allele frequency is far greater than expected for a rare dominant hereditary cancer predisposition.
• The 16.7% HGDP:MANDENKA subpopulation value in the gnomAD v3.1 non-cancer record (7/42 alleles, 0 homozygotes) was excluded as a small-sample artifact; the frequencies used rest on large allele counts.
• ClinVar VCV000314383 was inspected at submission level only (13 submissions, no expert-panel submission, no submission supplying a frequency criterion); the aggregate Benign classification was not used to derive BS1.
• Because BS1 rests on the same observation as BA1, BS1 was recorded for completeness and is not counted as a second, independent line of benign evidence (generic_acmg_combination_rules: BA1 alone already gives a Benign classification).

Still missing

• No MLH3-specific BS1 threshold exists (no CSPEC/VCEP and no local framework), so the generic 0.01 threshold was used although per-gene derivation is preferred; a gene-specific threshold would likely be lower and would not change the met verdict.
• No MLH3-specific prevalence or penetrance data were retrievable to derive an expected allele frequency for the disorder independently.
• Sherloc (PMID:28492532) was not retrievable in full text, so any refinement of frequency-based benign criteria it contains could not be evaluated.
• Deriving a race-matched versus popmax threshold was not possible because the case records no patient ancestry.

Sources

Reasoning

BP1 applies to a missense variant in a gene for which primarily truncating variants cause disease, and it is met here. MLH3's established pathogenic spectrum is truncating: the P/LP MLH3 records in ClinVar are nonsense, frameshift and canonical splice alleles (e.g. p.Glu8Ter, p.Ser80Ter, p.Thr618fs, p.Thr968fs, p.Asn1038fs, p.Lys1253fs, p.Arg1320Ter, p.Trp1276Ter, c.3571-1G>A), and among 2,176 MLH3 missense records exactly one reaches P/LP - the initiator-codon change p.Met1Thr - while the remaining 2,175 are uncertain, conflicting, benign or likely benign. The documented germline mechanism for MLH3 is loss of function: a biallelic germline nonsense MLH3 variant underlies polyposis predisposition, and gene-level curation describes inactivating/truncating changes as the pathogenic class for a gene acting as a tumour suppressor. This is reinforced by the missense constraint data - MLH3 carries missense variation at essentially expectation (gnomAD v2.1 oe_mis 0.924, mis_z 0.736), consistent with missense substitutions being a tolerated rather than a disease-causing class of change - and by the mismatch-repair disease framework, which defines disease-causing variants only for MLH1, MSH2, MSH6 and PMS2 and excludes MLH3 (PMID:34043773). BP1's stated caution (care where missense changes have been reported as pathogenic) does not apply beyond the initiator-codon change. BP1 is applied at its default supporting strength.

Rule

Generic ACMG/AMP 2015 BP1 (Richards et al., PMID:25741868): missense variant in a gene where primarily truncating variants cause disease; applied at default supporting weight. No MLH3 CSPEC/VCEP BP1 rule exists for this gene.

Evidence considered

• ClinVar MLH3 pathogenic/likely pathogenic records sampled for this assessment are all protein-truncating or splice-disrupting: c.22G>T (p.Glu8Ter), c.239C>A (p.Ser80Ter), c.980dup (p.Ile328fs), c.1852del (p.Thr618fs), c.2903del (p.Thr968fs), c.3112_3115del (p.Asn1038fs), c.3571-1G>A, c.3759del (p.Lys1253fs), c.3958A>T (p.Arg1320Ter), c.3828G>A (p.Trp1276Ter).
• Missense spectrum: of 2,176 MLH3 missense-annotated ClinVar records, 2,175 are uncertain significance, conflicting, likely benign, benign/likely benign or unclassified; the single P/LP missense-class allele is the initiator-codon change NM_001040108.2:c.2T>C (p.Met1Thr), which is not a classical missense mechanism.
• Gene-level mechanism (pvs1_gene_context): a biallelic germline nonsense variant of MLH3 underlies polyposis predisposition, and MLH3 is curated as a tumour suppressor whose inactivating changes are associated with HNPCC7, endometrial and gastric cancers - i.e. loss of function, not missense substitution, is the documented disease mechanism.
• Missense constraint (gnomAD v2.1, metric per Karczewski et al. 2020, PMID:32461654): oe_mis 0.924, mis_z 0.736, obs_mis 687 versus exp_mis 743.4 - missense changes are tolerated at near-expected rates in this gene.
• Mismatch-repair disease framework: the EHTG/ESCP Mallorca-3 Lynch syndrome guideline (PMID:34043773) states 'LS is caused by inherited malfunction in one of the four MMR genes named MLH1, MSH2, MSH6 and PMS2', with MLH3 absent from the list of genes for which disease-causing variants are defined, so missense changes in MLH3 are not an accepted disease class.
• Criterion definition consulted: Richards et al. 2015 (PMID:25741868) BP1 - 'missense variant in a gene for which primarily truncating variants are known to cause disease', applied at supporting strength with the guideline's caution about genes in which missense variants have been reported as pathogenic.

Still missing

• No ClinGen/VCEP-approved MLH3 gene-disease framework or gene-disease validity assertion was available, so the 'primarily truncating variants cause disease' premise rests on the observed ClinVar MLH3 variant-type spectrum plus gene-level loss-of-function literature and constraint data rather than on a curated specification.
• ClinVar record counts were obtained by a direct esearch/esummary sweep performed for this assessment rather than from a curated gene-level variant-type table.

Sources

Reasoning

Scope: NM_001040108.2:c.2390G>A (MLH3) is a missense variant, NP_001035197.1:p.(Arg797His), so BP4 is in scope and is taken from the missense protein-impact path (REVEL) only. No ClinGen CSPEC, VCEP specification, or local gene-specific framework exists for MLH3, so the generic ACMG/AMP 2015 fallback with the published ClinGen SVI REVEL calibration (Pejaver et al. 2022, Am J Hum Genet, PMID:36413997) applies. The local REVEL v1.3 score is 0.083, which is at or below the <=0.290 supporting and <=0.183 moderate BP4 thresholds, but above the <=0.016 strong threshold; the highest strength permitted is therefore moderate. BP4 is met at moderate strength. The SpliceAI max delta score of 0.00 in the bundle does not contribute to this estimate: for a missense variant splices prediction is not valid support for PP3/BP4, and the methodologically independent agreement of SpliceAI and Pangolin was deliberately not double-counted. BayesDel (-0.564808) was not used because no ACMG-strength BayesDel calibration is established. Per ACMG/AMP 2015 (PMID:25741868) a computational prediction is one line of evidence and must not be the sole basis for a clinical assertion, which is consistent with this being a single criterion at moderate strength.

Rule

SpliceAI/missense split: missense variant -> REVEL path only; BP4 requires REVEL <= 0.290 (supporting), <= 0.183 (moderate), <= 0.016 (strong)

Evidence considered

• Variant consequence: missense, NP_001035197.1:p.(Arg797His) from NM_001040108.2:c.2390G>A, establishing that the REVEL protein-impact path governs BP4 (the SpliceAI path applies only to intronic/synonymous/splice-region variants).
• Local REVEL v1.3 lookup: revel_score = 0.083 at chr14:75047266 C>T (hg38; = 14-75513969-C-T hg37); 0.083 <= 0.183 (moderate) and <= 0.290 (supporting) but > 0.016, so BP4 reaches moderate and not strong.
• Thresholds applied verbatim from the supplied ClinGen SVI REVEL calibration (Pejaver et al. 2022, Am J Hum Genet, PMID:36413997): BP4 supporting <=0.290, moderate <=0.183, strong <=0.016.
• SpliceAI Lookup (Broad) max delta score 0.00 (Pangolin SG 0.002, SL -0.006) recorded as consulted but not counted toward BP4, because a clean splice delta is not benign protein-impact evidence for a missense variant and the same prediction must not be double-counted.
• BayesDel (no-AF, local file) score -0.564808 recorded as consulted and set aside as not_available for BP4 in the absence of an established ACMG-strength BayesDel calibration.
• Framework context: cspec not found for MLH3, vcep_materials.json has no summaries/files/lookup values, final_classification_framework.json found=false (generic_acmg_fallback), so no gene-specific pre-assigned BP4 code existed to apply.
• ACMG/AMP 2015 (PMID:25741868) computational-predictor section: predictors are one line of evidence and should not be used as the sole source of evidence for a clinical assertion.
• paper_extracts.json for the four fetched full texts (PMID:25741868, PMID:24493721, PMID:33516529, PMID:34043773) contains no variant-level computational or functional data on c.2390G>A/p.(R797H); no paper was needed beyond the framework source.

Still missing

• The BP4 strength rests on generic ClinGen SVI REVEL calibration; no MLH3-specific calibration (e.g. a VCEP lookup spreadsheet) exists to confirm or adjust the applicable strength.
• No VCEP pre-assigned code was available for BP4 (no CSPEC and no VCEP materials for MLH3).

Sources

Reasoning

PS1 requires that the variant's own amino-acid change (p.Arg797His) has already been established as pathogenic, regardless of the nucleotide change. No such variant exists. A ClinVar-wide sweep of MLH3 returned only four records at codon 797: p.Arg797His (VCV000314383 - this variant - Benign, 'criteria provided, multiple submitters, no conflicts', 13 submissions from 12 clinical laboratories, zero expert-panel submissions), p.Arg797Cys (VCV000941813, conflicting: 2x uncertain significance, 1x likely benign), p.Arg797Asp (VCV004860221, uncertain significance) and p.Arg797= (VCV001096936, benign/likely benign). None of these is classified pathogenic or likely pathogenic, and no alternative nucleotide change producing p.Arg797His carries any pathogenic assertion. Scaled to the whole gene, of 2,176 ClinVar records annotated as MLH3 missense variants exactly one is P/LP - the initiator-codon change p.Met1Thr - and none is at codon 797. Curated protein annotation (UniProt Q9UHC1) lists the R797H substitution as a natural variant referenced only to dbSNP rs28756991, without the 'in HNPCC7' disease annotation that flags MLH3 variants at positions 24, 499, 624, 647, 817, 981, 1007, 1394 and 1451. With no previously established pathogenic variant sharing this amino-acid change, PS1 is not met.

Rule

Generic ACMG/AMP 2015 PS1 (Richards et al., PMID:25741868): same amino acid change as a previously established pathogenic variant, regardless of nucleotide change. No MLH3-specific CSPEC/VCEP rule exists for this gene (generic_acmg fallback), so the generic definition governs.

Evidence considered

• ClinVar VCV000314383 (NM_001040108.2(MLH3):c.2390G>A, p.Arg797His): Benign, review status 'criteria provided, multiple submitters, no conflicts'; 13 submissions from 12 clinical laboratories with no expert-panel submission and no pathogenic or likely pathogenic assertion for this exact amino-acid change.
• Same-residue/same-change search: ClinVar esearch for 'MLH3[gene] AND Arg797' returned four records - p.Arg797His (this variant, Benign), p.Arg797Cys (c.2389C>T, conflicting classifications: 2x uncertain significance and 1x likely benign), p.Arg797Asp (c.2389_2390delinsGA, uncertain significance) and p.Arg797= (c.2391C>T, benign/likely benign). No other nucleotide change creating p.Arg797His exists, and no pathogenic same-amino-acid-change comparator was found.
• Whole-gene ClinVar sweep (esearch 'MLH3[gene] AND "missense variant"[Molecular consequence]', aggregate germline classifications tallied via esummary): 2,176 missense records - 1,986 uncertain significance, 97 conflicting, 72 likely benign, 13 benign/likely benign, 3 benign, 4 without classification, and exactly 1 likely pathogenic (the initiator-codon change p.Met1Thr). None at codon 797.
• UniProt Q9UHC1 annotates the R797H substitution as a natural variant referenced only to dbSNP rs28756991, without the 'in HNPCC7' disease annotation that flags MLH3 variants at positions 24, 499, 624, 647, 817, 981, 1007, 1394 and 1451.
• Criterion definition consulted: Richards et al. 2015 (PMID:25741868) PS1 - 'same amino acid change as a previously established pathogenic variant regardless of nucleotide change'.

Still missing

• No MLH3-specific germline variant database (e.g. InSiGHT/LOVD-style locus-specific database) was consulted independently of ClinVar; ClinVar plus UniProt Q9UHC1 variant annotation together show no established pathogenic p.Arg797His allele.

Sources

Reasoning

PS4 (Richards et al. 2015, PMID:25741868) requires that the prevalence of the variant in affected individuals be significantly increased compared with the prevalence in controls. No CSPEC/VCEP specification exists for MLH3 (vcep_materials.json: 'no cspec url available'; final_classification_framework.found = false; framework_mode = generic_acmg), so the generic ACMG/AMP 2015 framework governs. That framework defines no numeric odds-ratio or likelihood-ratio threshold for PS4, so there is no quantitative operator written in any governing source for this criterion in this case and no numerical comparison to get wrong in either direction; the criterion can only be met by a study showing significant case enrichment, which does not exist here. Evidence search: the case's literature pass screened all 7 ClinVar-cited PMIDs and the four that could be retrieved in full text (PMID:25741868, PMID:24493721, PMID:33516529, PMID:34043773) never mention c.2390G>A or p.(Arg797His)/p.R797H, so no affected-individual cohort, case series, or case-control comparison containing this variant was identified in the retrieved literature; the variant is absent from gnomAD-Canada (AC=0, AN=0, uninformative). Independent population data point the opposite way: gnomAD v4.1 reports this allele at total AF = 0.003509 (5663/1614070 alleles, 174 homozygotes) with a grpmax filtering AF of 0.0647 in African/African American individuals, and gnomAD v2.1 at AF = 0.006369 (1801/282780 alleles, 72 homozygotes; AFR AF = 0.0661). Frequencies of that magnitude in population controls are incompatible with significant enrichment of this allele among affected individuals in a rare-disease cohort. With no case-enrichment evidence and population frequencies that preclude it, PS4 is not met.

Rule

generic_acmg_2015: PS4 requires significant enrichment of the variant in affected individuals versus controls; no numeric threshold defined by any governing specification, and no enrichment study exists.

Evidence considered

• Generic ACMG/AMP 2015 (PMID:25741868) defines PS4 as significantly increased prevalence of the variant in affected individuals compared with controls, with no numeric threshold; no MLH3-specific CSPEC/VCEP specification was retrieved for this case (vcep_materials.json reports 'no cspec url available' and no gene VCEP directory).
• Literature pass over all 7 ClinVar-cited PMIDs (24905773, 24493721, 33451724, 33516529, 34043773, 25741868, 28492532); full-text extraction was completed for 4 of them and none mentions MLH3 c.2390G>A or p.(Arg797His)/p.R797H, so no affected individual carrying this variant is documented in the retrieved literature.
• gnomAD v4.1: total AF 0.003509 (5663/1614070 alleles), 174 homozygotes, grpmax filtering AF 0.0647 (African/African American, 4971/75006 alleles, 171 homozygotes) - shows the allele is common in population controls rather than enriched in cases.
• gnomAD v2.1: total AF 0.006369 (1801/282780 alleles), 72 homozygotes, African/African American AF 0.0661 (1649/24942 alleles) - concordant control-frequency evidence.
• gnomAD-Canada v1.0 returned the variant absent with AN = 0 (no alleles typed), so it contributes no informative case or control data.
• ClinVar VCV000314383 (ClinVarID 314383) records only laboratory Benign assertions (no case series, no case counts) with no submission reporting affected-individual prevalence or enrichment.

Still missing

• No case-control or case-cohort frequency comparison for MLH3 c.2390G>A was found in any retrieved source; PS4 would need such a study (or a variant-specific case series with matched controls) to be evaluable in the positive direction.

Sources

Reasoning

PM1 requires a mutational hotspot and/or a critical, well-established functional domain without benign variation. No governing gene-specific specification exists for MLH3: there is no CSPEC, no local framework and no VCEP, and vcep_materials.json's domain_tables object is empty for this gene, so there is no authoritative VCEP-approved critical-domain list to test residue 797 against. Independent evidence against PM1 is consistent across three lines. (1) Hotspot: the Cancer Hotspots query for MLH3 R797 ('MLH3 R797') returned no statistically significant hotspot, and the variant is not listed. (2) Domain: curated annotation places all of MLH3's annotated functional domains at the protein termini - the N-terminal ATPase/DNA-mismatch-repair region (InterPro/Pfam entries spanning roughly residues 1-349: SSF55874 2-216, PF13589 24-118, IPR013507/SM01340/PF01119 206-349) and the C-terminal MutL dimerisation/regulatory domain (IPR014790/IPR037198/SM00853/PF08676 approx. residues 1189-1403) - leaving residue 797 in the large central region that carries no annotated domain at all. (3) Benign variation at the residue: PM1's 'without benign variation' requirement is directly contradicted here - p.Arg797His is Benign in ClinVar (12 clinical laboratories, no conflicts) and common in gnomAD (v2.1 AF 0.64%, 1,801/282,780 alleles, 72 homozygotes, popmax 6.61% in African/African American; v4.1 AF 0.35%, 5,663/1,614,070 alleles, 174 homozygotes), p.Arg797= is benign/likely benign, and p.Arg797Cys is only VUS/likely benign. PM1 is therefore not met.

Rule

Generic ACMG/AMP 2015 PM1 (Richards et al., PMID:25741868): mutational hotspot and/or critical well-established functional domain without benign variation; no MLH3 CSPEC/VCEP domain table exists, so generic definition plus independent hotspot and curated-domain annotation govern.

Evidence considered

• No gene-specific framework: final_classification_framework reports found=false with source 'generic_acmg_fallback' for MLH3 (no CSPEC, no VCEP directory, no local framework), and vcep_materials.json contains an empty domain_tables object - so no VCEP-extracted critical functional domain list exists to check residue 797 against.
• Hotspot: the Cancer Hotspots query 'MLH3 R797' returned no row ('This variant does not lie in a statistically significant hotspot'; residue_significant=false, exact_variant_listed=no).
• Domain boundaries (curated annotation, UniProt Q9UHC1, 1,453 aa, and InterPro/Pfam matches for the same entry): N-terminal ATPase/DNA-mismatch-repair region approximately residues 1-349 (SSF55874 2-216; PF13589 24-118; IPR013507/SM01340/PF01119 206-349; TIGR00585 2-269) and C-terminal MutL dimerisation/regulatory domain approximately residues 1189-1403 (IPR014790 1191-1372; IPR042120 1217-1403; IPR042121 1246-1367; SM00853 1191-1372; PF08676 1218-1370). Residue 797 falls between these, in a region with no annotated domain; the only features annotated near the middle of the protein are disordered/low-complexity regions at 624-650 and 933-960, neither of which contains residue 797.
• Benign variation at the codon: ClinVar p.Arg797His (VCV000314383) is Benign with 12 clinical laboratories and no conflicting submissions; p.Arg797= (c.2391C>T) is benign/likely benign; p.Arg797Cys is uncertain significance/likely benign only.
• gnomAD: v2.1 AF 0.00636891 (1,801/282,780 alleles, 72 homozygotes), popmax 6.61134% African/African American (1,649/24,942, 71 homozygotes), grpmax FAF 0.0632924; v4.1 AF 0.00350852 (5,663/1,614,070 alleles, 174 homozygotes), popmax 6.62747% African/African American (4,971/75,006, 171 homozygotes), grpmax FAF 0.064736. A residue carrying this much benign variation cannot satisfy PM1's 'without benign variation' requirement.
• Criterion definition consulted: Richards et al. 2015 (PMID:25741868) PM1 - 'located in a mutational hot spot and/or critical and well-established functional domain ... without benign variation'.

Still missing

• vcep_materials.json domain_tables is empty for MLH3 and no CSPEC/VCEP exists for this gene, so there is no authoritative approved critical-domain list; the residue-to-domain test was performed against UniProt Q9UHC1 / InterPro-Pfam curated domain boundaries instead.
• The only paper triaged to this group's hotspot/domain theme (PMID:33451724, SGO endometrial cancer review part II) contains no MLH3 domain or hotspot content and no variant mention, and no full text was fetched; no MLH3-specific germline hotspot dataset was available, so the hotspot prong rests on the Cancer Hotspots query plus the ClinVar/gnomAD benign-variation pattern.

Sources

Reasoning

Governing criterion: generic ACMG/AMP 2015 PM2 at supporting strength (Richards et al. 2015, PMID:25741868), for which the supplied calibration sets the threshold at allele frequency <= 0.0001 (ClinGen SVI recommendation to downgrade PM2 to supporting). No MLH3 CSPEC/VCEP or local framework exists (cspec.found = false; gene_vcep_dir null; framework_mode generic_acmg), so no gene-specific PM2 threshold or filtering-allele-frequency rule applies and the all-comers gnomAD v2.1/v4.1 entries are the default sources. The variant is common rather than absent: gnomAD v4.1 total AF 0.003508 (5663/1614070 alleles, 174 homozygotes) and gnomAD v2.1 total AF 0.006369 (1801/282780 alleles, 72 homozygotes), about 35-fold and 64-fold above the 0.0001 threshold, with ancestry popmax frequencies of 0.066275 (v4.1 African, 4971/75006) and 0.066113 (v2.1 African, 1649/24942), roughly 660-fold above threshold; the filtering AF popmax, which is the quantity ClinGen-style rules compare, is 0.064736 (v4.1 joint) and 0.063292 (v2.1 exome). Every other dataset is concordant and far above threshold: gnomAD v2.1 non-cancer exomes AF 0.004716 (1117/236856, 40 homozygotes), gnomAD v3.1 non-cancer genomes AF 0.019239 (2845/147874, 92 homozygotes), gnomAD-Canada v1.0 AF 0.005213 (96/18416, 5 homozygotes). The only sub-0.0001 observations are in ancestries where the allele was essentially not seen (gnomAD v2.1 East Asian 0/19948; gnomAD v2.1 Ashkenazi Jewish 1/10368 = 0.0000965); selecting the lowest subpopulation frequency instead of the popmax filtering AF would be a threshold artifact of exactly the kind this criterion must avoid, and the variant is plainly not absent from a large general population or from a control cohort of more than 1000 individuals, which is what PM2 requires. Ancestry matching could not be applied because this variant-level case records no patient ancestry, but PM2 cannot be satisfied under any ancestry assignment: the allele is present at 0.33-0.64% in the all-comers population and at about 6.6% in the largest ancestry group, far above a supporting-level rarity threshold for a disorder that cannot be that common. Disease-context check: for an adult-onset, incompletely penetrant cancer predisposition phenotype with an estimated dominant mismatch-repair carrier prevalence of about 1 in 300 (PMID:34043773), a 6.6% allele frequency sits orders of magnitude outside the range in which a rare pathogenic allele could lie. PM2 is not met.

Rule

Generic ACMG/AMP 2015 PM2 at supporting strength: absent or extremely rare in a large general population or a race-matched control cohort (>1000 individuals), applied at the supplied threshold AF <= 0.0001 (ClinGen SVI downgrade of PM2 to supporting, cited as PMID:25741868).

Evidence considered

• Governing definition: generic ACMG/AMP 2015 PM2 (Richards et al. 2015, PMID:25741868) - absence from a large general population or a control cohort (>1000 individuals) race-matched to the patient, or below the expected carrier frequency for a recessive disease; assessed at supporting strength.
• PM2 threshold applied verbatim from the supplied generic calibration: allele frequency <= 0.0001 (ClinGen SVI recommendation to downgrade PM2 to supporting, cited in the supplied calibration as PMID:25741868).
• gnomAD v4.1 (gnomad_v4): total AF 0.003509 (5663/1614070 alleles, 174 homozygotes); African/African American AF 0.066275 (4971/75006); grpmax filtering AF 0.064736 - about 35-fold (total) to 660-fold (popmax) above the 0.0001 threshold.
• gnomAD v2.1 (gnomad_v2): total AF 0.006369 (1801/282780 alleles, 72 homozygotes); African/African American AF 0.066113 (1649/24942); grpmax filtering AF 0.063292 - about 64-fold (total) to 660-fold (popmax) above threshold.
• Additional concordant datasets: gnomAD v2.1 non-cancer exomes AF 0.004716 (1117/236856, 40 homozygotes) with African AF 0.065627; gnomAD v3.1 non-cancer genomes AF 0.019239 (2845/147874, 92 homozygotes) with African AF 0.066300; gnomAD-Canada v1.0 AF 0.005213 (96/18416, 5 homozygotes).
• Threshold-artifact check: the lowest subpopulation frequencies are East Asian 0/19948 and Ashkenazi Jewish 1/10368 (0.0000965) in gnomAD v2.1 - i.e. absence in ancestries where the allele is not segregating - whereas the relevant comparator is the popmax filtering AF of 0.063-0.065, so no sub-threshold subpopulation reading was used to rescue PM2.
• Ancestry: the case provides no patient ancestry, so no ancestry-matched reading was possible; the allele nonetheless occurs at 0.33-0.64% in the general population and about 6.6% in African/African American individuals, which is inconsistent with a supporting-level rarity criterion under any ancestry assumption.
• Disease-context check (PMID:34043773): dominant mismatch-repair carrier prevalence is about 1 in 300 and the phenotype is adult-onset with variable penetrance, so a 6.6% allele frequency is far above what a rare pathogenic allele for this disorder class could reach.
• ClinVar VCV000314383 was consulted at submission level only (13 submissions, no expert-panel submission, no submission providing a frequency criterion); its aggregate Benign classification was not used to derive PM2.

Still missing

• No patient ancestry is recorded in this variant-level case, so a race-matched control comparison of the kind the ACMG PM2 wording contemplates could not be performed; this does not change the verdict because the allele is common in the general population and in the largest ancestry group.
• No MLH3-specific PM2 threshold or filtering-allele-frequency rule exists (no CSPEC/VCEP), so the generic supporting-level 0.0001 threshold was applied.
• PM2 read strictly requires absence (or a very low expected carrier frequency for recessive disease) from a large general population or a control cohort of more than 1000 individuals, which is unambiguously not the case here.

Sources

Reasoning

Generic ACMG/AMP 2015 governs this gene: vcep_materials.json contains no CSPEC/VCEP and no local MLH3 framework (no cspec url, final_classification_framework.source = generic_acmg_fallback, inheritance field empty), so PM3 keeps its generic 2015 wording - 'For recessive disorders, detected in trans with a pathogenic variant' (PMID:25741868). NM_001040108.2:c.2390G>A predicts NP_001035197.1:p.(Arg797His), a missense substitution, so PM3 can only ever be satisfied by an observed allelic relationship, never by variant-class reasoning. Two things are required and both are absent. (1) A pathogenic MLH3 allele in trans: the ClinVar record for this variant (variation 314383; 13 submissions, 11 submitters, 12 clinical laboratories; germline origin; no expert-panel submission; conditions HNPCC type 7 / endometrial carcinoma) contains no second MLH3 variant, no zygosity and no cis/trans statement for this variant; none of the four full-text papers extracted for the case names c.2390G>A, p.Arg797His or p.R797H at all; and the Lynch-syndrome gene lists in the case's own guideline literature (PMID:34043773) define path_MMR alleles only for MLH1, MSH2, MSH6 and PMS2, so no MLH3 pathogenic partner allele is documented anywhere in this workup. (2) A phase-resolved affected proband: this case supplies no proband, pedigree, parental genotypes or phasing data, so trans status could not be tested directly for any individual. Population data argue against, not for, an in-trans-dependent reading: the allele reaches 0.63689% (1801/282780) with 72 homozygotes in gnomAD v2.1 and 0.35085% (5663/1614070) with 174 homozygotes in gnomAD v4.1 (171 of those in the African/African American population, where AF is 6.62747%), i.e. biallelic carriers are common in unselected adults rather than the profile of an allele whose significance rests on a second pathogenic hit. MLH3 does have a documented biallelic disease mechanism (Olkinuora et al., Genet Med 2019, PMID:30573798, 'Biallelic germline nonsense variant of MLH3 underlies polyposis predisposition'), so the recessive premise is not dismissed as impossible for this gene - but that report concerns a different, truncating allele and does not report c.2390G>A. PM3 is therefore not met on the evidence available, rather than not applicable, because a recognised biallelic MLH3 mechanism exists and an in-trans pathogenic observation, if one were documented, would be scorable.

Rule

PM3 (moderate, generic ACMG/AMP 2015): 'For recessive disorders, detected in trans with a pathogenic variant' (Richards et al. 2015, PMID:25741868); no MLH3-specific CSpec/VCEP exists, and no in-trans observation is documented for c.2390G>A.

Evidence considered

• Governing framework: vcep_materials.json records no CSPEC/VCEP and no local MLH3 framework (cspec found = false, no cspec url, framework source generic_acmg_fallback, scoring model ACMG/AMP 2015, inheritance = []), so the generic ACMG/AMP 2015 PM3 definition in PMID:25741868 was applied verbatim: 'For recessive disorders, detected in trans with a pathogenic variant'.
• Variant class: prefetch.json / pvs1_variant_assessment.json normalize NM_001040108.2:c.2390G>A to NP_001035197.1:p.(Arg797His), consequence_class missense, canonical_splice_consensus false - so PM3 can only be met through an observed allelic relationship and not through variant-type reasoning.
• ClinVar (clinvar, variation 314383): 13 submissions from 11 submitters / 12 clinical laboratories, all classified Benign, germline origin, no expert-panel submission, conditions 'Colorectal cancer, hereditary nonpolyposis, type 7', 'Endometrial carcinoma', 'not provided', 'not specified'. No submission reports a second MLH3 variant, zygosity, or any cis/trans relationship for c.2390G>A; the record therefore contains no pathogenic variant in trans. The aggregate ClinVar classification itself was not used as evidence.
• Population data (gnomad_v2): AF 0.63689% (1801/282780 alleles) with 72 homozygotes overall; African/African American AF 6.61134% (1649/24942) with 71 homozygotes; grpmax FAF 0.0632924.
• Population data (gnomad_v4): AF 0.35085% (5663/1614070 alleles) with 174 homozygotes overall; African/African American AF 6.62747% (4971/75006) with 171 homozygotes; grpmax FAF 0.064736. Homozygotes are observed at population scale, which is not the profile of an allele whose interpretation depends on a pathogenic partner in trans; this observation was used only as context for PM3 and not as a benign criterion here.
• Guideline literature (PMID:34043773): the EHTG/ESCP Mallorca guideline defines path_MMR variants only for MLH1, MSH2, MSH6 and PMS2 and never names MLH3, so no MLH3 pathogenic partner allele is documented in the case's guideline evidence.
• Gene-level inheritance context: MLH3 is the HNPCC7 susceptibility gene per the case gene context, but a biallelic mechanism is documented in the case's own PVS1 literature folder - Olkinuora et al., Genet Med 2019 (PMID:30573798), titled 'Biallelic germline nonsense variant of MLH3 underlies polyposis predisposition' (abstract held at output/pvs1_articles/PMID_30573798.md). That paper concerns a truncating allele, is not in source_registry, and does not mention c.2390G>A, so it supports only the applicability of a recessive reading, not the presence of a pathogenic allele in trans.
• Phase search: every source in this case (ClinVar submission audit, evidence.json source records, the four extracted full texts) was checked for phase, cis/trans, zygosity or compound-heterozygote statements about this variant; none exists. Case input is variant-level only, with no affected proband or family genotypes.

Still missing

• No proband, pedigree, parental genotypes or other phase-resolved data were provided in this case, so the trans configuration of c.2390G>A could not be tested directly in an affected individual.
• No pathogenic MLH3 variant in trans with c.2390G>A is reported in ClinVar, in any extracted full text, or in the guidelines consulted; MLH3 has no ClinGen CSpec/VCEP, so PM3 has no gene-specific strength calibration and was applied at the generic ACMG/AMP 2015 moderate level.
• Gene-level recessive evidence exists only for a different MLH3 allele (biallelic truncating MLH3, PMID:30573798); whether the missense c.2390G>A (p.Arg797His) can contribute to a biallelic genotype remains undetermined.

Sources

Reasoning

PM5 requires a different missense change at the same amino-acid residue (797) to have been determined pathogenic or likely pathogenic. Every variant known at MLH3 codon 797 is benign or uncertain: p.Arg797Cys (c.2389C>T, VCV000941813) has conflicting classifications made up of 2x uncertain significance and 1x likely benign and is not established as pathogenic; p.Arg797Asp (c.2389_2390delinsGA, VCV004860221) is uncertain significance; p.Arg797= (c.2391C>T) is benign/likely benign; and p.Arg797His is this variant itself, classified Benign. A whole-gene scan found only one P/LP missense-class allele anywhere in MLH3 (the initiator-codon change p.Met1Thr), and no pathogenic variant at codon 797. UniProt Q9UHC1 likewise does not flag any codon-797 variant as HNPCC7-associated. Because no different, pathogenic missense change exists at this residue, PM5 is not met.

Rule

Generic ACMG/AMP 2015 PM5 (Richards et al., PMID:25741868): a different missense change at the same residue must already be determined pathogenic or likely pathogenic. No MLH3 CSPEC/VCEP PM5 rule exists for this gene.

Evidence considered

• ClinVar codon-797 inventory (esearch 'MLH3[gene] AND Arg797', four records): p.Arg797His (this variant, Benign), p.Arg797Cys c.2389C>T (VCV000941813, 'Conflicting classifications of pathogenicity', aggregate explanation 'Uncertain significance (2); Likely benign (1)', 3 submitters, no expert panel, and the LB submitter's comment cites population frequency, intact protein function, co-occurrence and lack of segregation), p.Arg797Asp c.2389_2390delinsGA (VCV004860221, Uncertain significance, single submitter), and p.Arg797= c.2391C>T (VCV001096936, Benign/Likely benign, multiple submitters). No pathogenic or likely pathogenic same-residue missense exists.
• Whole-gene ClinVar scan of 2,176 MLH3 missense-annotated records found exactly one P/LP allele, the initiator-codon change NM_001040108.2:c.2T>C (p.Met1Thr); none at codon 797.
• pm5_candidates.json: eligible_for_classic_pm5_search=false, pm5_mode 'unknown', total_candidates_seen 0, same_residue_candidates 0; recommendation level 'not_applicable' because classic same-residue PM5 semantics could not be confirmed from the available artifacts - so the comparator question was resolved directly against ClinVar rather than from that artifact.
• UniProt Q9UHC1 lists variants designated 'in HNPCC7' at positions 24, 499, 624, 647, 817, 981, 1007, 1394 and 1451; codon 797 is not among them (R797H is referenced only to dbSNP rs28756991).
• Criterion definition consulted: Richards et al. 2015 (PMID:25741868) PM5 - 'novel missense change at an amino acid residue where a different missense change determined to be pathogenic has been seen before'.

Still missing

• The automated comparator harvest was skipped - pm5_candidates.json records 0 candidates with reason 'Unable to confirm classic same-residue PM5 semantics safely; skipping automatic candidate harvesting' - so the same-residue sweep was performed manually against ClinVar instead; the R797C record is explicitly 'conflicting', so if it is later reclassified likely pathogenic or pathogenic, PM5 would need re-evaluation.

Sources

Reasoning

PP2 requires both that the gene has a low rate of benign missense variation and that missense variants are a common mechanism of disease. Neither prong holds for MLH3. Constraint: gnomAD v2.1 gene constraint gives mis_z 0.736 and oe_mis 0.924 (687 observed versus 743.4 expected missense variants), i.e. missense variation is close to expectation and the gene is not missense-constrained (constraint metric per Karczewski et al. 2020, Nature, PMID:32461654). Mechanism: missense changes are not an established cause of MLH3 disease - across 2,176 ClinVar MLH3 missense records only one is P/LP (the initiator-codon change p.Met1Thr) while 1,986 are uncertain significance and 85 are benign/likely benign, and the documented germline mechanism is loss of function via truncating alleles (biallelic MLH3 nonsense variants underlying polyposis predisposition). This variant's own population behaviour is congruent: it is a tolerated benign missense present at 0.64% allele frequency in gnomAD v2.1 with 72 homozygotes and classified Benign by 12 clinical laboratories. The canonical mismatch-repair disease framework also defines disease-causing 'path_MMR' variants only for MLH1, MSH2, MSH6 and PMS2, with MLH3 absent (EHTG/ESCP Mallorca-3 guideline, PMID:34043773). PP2 is not met.

Rule

Generic ACMG/AMP 2015 PP2 (Richards et al., PMID:25741868): missense variant in a gene with a low rate of benign missense variation and in which missense variants are a common disease mechanism. No MLH3 CSPEC/VCEP PP2 rule exists for this gene.

Evidence considered

• Missense constraint: gnomAD v2.1 gene constraint for MLH3 - mis_z 0.736, oe_mis 0.924 (obs_mis 687, exp_mis 743.4). The constraint metric is that of Karczewski et al. 2020, Nature (PMID:32461654); MLH3 is not missense-constrained, so a low rate of benign missense variation cannot be asserted.
• Missense mechanism: ClinVar tally of all MLH3 missense-annotated records (n=2,176) - 1,986 uncertain significance, 97 conflicting, 72 likely benign, 13 benign/likely benign, 3 benign, 4 unclassified, and only 1 likely pathogenic (initiator-codon p.Met1Thr). Established pathogenic MLH3 alleles are truncating, so missense is not a common disease mechanism.
• Gene-level mechanism context (pvs1_gene_context): targeted germline literature review identified MLH3 loss-of-function/truncating disease mechanism support, including a biallelic germline nonsense MLH3 variant underlying polyposis predisposition, and curated gene context describes inactivating/truncating changes (not missense) as the pathogenic class; MLH3 is curated as a tumour suppressor acting through loss of function.
• MMR disease framework: the EHTG/ESCP Mallorca-3 Lynch syndrome guideline (PMID:34043773) defines disease-causing path_MMR variants only for the four MMR genes MLH1, MSH2, MSH6 and PMS2 ('LS is caused by inherited malfunction in one of the four MMR genes named MLH1, MSH2, MSH6 and PMS2'), with MLH3 absent - there is no ratified missense-driven MLH3 disease mechanism.
• Variant-level check on the 'low benign missense variation' prong: this codon carries common benign missense variation - p.Arg797His is Benign in ClinVar (12 laboratories) and present in gnomAD v2.1 at 0.64% (72 homozygotes, popmax 6.61% African/African American) and gnomAD v4.1 at 0.35% (174 homozygotes).
• Criterion definition consulted: Richards et al. 2015 (PMID:25741868) PP2 - 'missense variant in a gene that has a low rate of benign missense variation and in which missense variants are a common mechanism of disease'.

Still missing

• MLH3 has no CSPEC/VCEP and no ClinGen gene-disease validity assertion was retrieved for this case, so PP2's gene-level prongs were assessed from gnomAD v2.1 constraint, the ClinVar MLH3 missense classification spectrum, and the MMR guideline gene lists rather than from a curated gene-disease framework; the 2,176-record ClinVar tallies come from a direct esearch/esummary sweep performed for this assessment.

Sources

Reasoning

Scope: NM_001040108.2:c.2390G>A (MLH3) is a missense substitution, NP_001035197.1:p.(Arg797His), so PP3 is in scope and must be taken from the missense protein-impact path (REVEL) only; a splice prediction can never supply PP3 for a missense variant. No ClinGen CSPEC, VCEP specification, or local gene-specific framework exists for MLH3 (framework_complete=false), so the generic ACMG/AMP 2015 fallback with the published ClinGen SVI REVEL calibration applies. The local REVEL v1.3 lookup returns 0.083 for chr14:75047266 C>T (hg38) / 14-75513969-C-T (hg37), which is far below the >=0.644 supporting, >=0.773 moderate, and >=0.932 strong PP3 thresholds (Pejaver et al. 2022, Am J Hum Genet, PMID:36413997). PP3 therefore fails at every strength. A SpliceAI max delta score of 0.00 (DS_AG/DS_AL/DS_DG/DS_DL all 0.00) was present in the evidence bundle, but under the one-path rule it is not admissible support for a missense variant's PP3 and was not counted; a locally computed BayesDel score (-0.564808) likewise contributed nothing because no ACMG-strength BayesDel calibration is established. No publication reports a functional or computational study of this exact variant. PP3 is not met.

Rule

SpliceAI/missense split: missense variant -> REVEL path only; PP3 requires REVEL >= 0.644 (supporting), >= 0.773 (moderate), >= 0.932 (strong)

Evidence considered

• Variant consequence: missense, NP_001035197.1:p.(Arg797His) from NM_001040108.2:c.2390G>A (normalization in the case evidence bundle), so the REVEL missense path governs PP3/BP4.
• Local REVEL v1.3 lookup: revel_score = 0.083 at chr14:75047266 C>T (hg38; = 14-75513969-C-T hg37), identical to the value recorded under the source_registry entry for the local REVEL predictor.
• Thresholds applied verbatim from the supplied ClinGen SVI REVEL calibration (Pejaver et al. 2022, Am J Hum Genet, PMID:36413997): PP3 supporting >=0.644, moderate >=0.773, strong >=0.932; BP4 supporting <=0.290, moderate <=0.183, strong <=0.016.
• SpliceAI Lookup (Broad) for the same variant returned max delta score 0.00 (DS_AG 0.0, DS_AL 0.0, DS_DG 0.0, DS_DL 0.0; Pangolin SG 0.002, SL -0.006). Recorded as consulted but explicitly excluded from PP3/BP4 because the variant is missense and evidence may come from only one predictor path.
• BayesDel (no-AF, local file) score -0.564808 recorded as consulted; no established ACMG-strength BayesDel calibration exists, so it was treated as not_available for PP3/BP4.
• ClinGen CSPEC lookup returned not found for MLH3; vcep_materials.json contains no VCEP summaries, files, or PP3/BP4 lookup values, and final_classification_framework.json shows found=false with source=generic_acmg_fallback, so no pre-assigned gene-specific code was available to apply.
• paper_extracts.json was checked for all four already-fetched full texts (PMID:25741868, PMID:24493721, PMID:33516529, PMID:34043773); none mentions c.2390G>A, p.(R797H), or any computational prediction for this variant, so no paper could be cited for PP3.

Still missing

• No gene-specific (MLH3) predictor calibration or VCEP PP3 rule exists; the generic ClinGen SVI REVEL thresholds were applied.
• No VCEP pre-assigned code was available for PP3 (no CSPEC and no VCEP materials for MLH3).

Sources

Reasoning

PP5 (Richards et al. 2015, PMID:25741868) may only be applied here for an exact-variant ClinVar classification issued by a recognised expert panel asserting Pathogenic or Likely pathogenic. The exact variant record VCV000314383 (ClinVarID 314383; NM_001040108.2(MLH3):c.2390G>A (p.Arg797His)) was audited submission by submission: 13 submissions from 12 submitters, all of them ordinary clinical laboratories (Mayo Clinic Laboratories, Illumina Laboratory Services, GeneDx, ARUP, KCCC/NGS Kuwait Cancer Control Center, Ambry Genetics, Breakthrough Genomics, Rigshospitalet Copenhagen, Sinai Health System, Labcorp/Invitae, Myriad Genetics, Dasa). clinvar_submission_audit.json records is_expert_panel = false for every submission (expert_panel_submissions = 0), and no InSiGHT Variant Interpretation Committee or other MMR-gene expert-panel assertion is present. Every submission asserts Benign; none asserts Pathogenic or Likely pathogenic. The record-level review status is 'criteria provided, multiple submitters, no conflicts' (2 stars, an aggregate of independent laboratory assertions) and not the 3-star expert-panel tier. Under the governing rule, ordinary laboratory submissions, aggregate ClinVar labels, and non-expert assertions must never trigger PP5. PP5 is not met.

Rule

expert_panel_clinvar_rule: PP5 requires an exact-variant ClinVar expert-panel (3-star) Pathogenic/Likely pathogenic classification; laboratory and aggregate assertions are excluded.

Evidence considered

• ClinVar VCV000314383 (ClinVarID 314383): classification Benign, record review status 'criteria provided, multiple submitters, no conflicts' (2 stars), 12 submissions from 11-12 submitters, all clinical laboratories; no expert-panel (3-star) assertion.
• clinvar_submission_audit.json: submissions_total 13, expert_panel_submissions 0, is_expert_panel = false for all submissions (Mayo Clinic Laboratories, Illumina, GeneDx, ARUP, KCCC Kuwait x2, Ambry, Breakthrough Genomics, Rigshospitalet, Sinai Health System, Labcorp/Invitae, Myriad, Dasa).
• Every audited submission asserts Benign; no submission to ClinVar for this variant asserts Pathogenic or Likely pathogenic, so even setting the expert-panel requirement aside there is no pathogenic reputable-source assertion to cite.
• Only PMID:25741868 and PMID:28492532 appear as criteria references in the submissions (generic ACMG/AMP and Sherloc), i.e. generic criteria references, not variant-level pathogenic evidence from an expert panel.
• Governing rule applied: PP5 is applied only for an exact-variant ClinVar expert-panel Pathogenic/Likely pathogenic classification; ordinary laboratory submissions, aggregate ClinVar labels and non-expert assertions are excluded.

Still missing

• No expert-panel (3-star) ClinVar assertion exists for this variant; if an InSiGHT or other recognised expert-panel classification were later issued for c.2390G>A, PP5/BP6 would need reassessment.

Sources

Reasoning

No CSPEC/VCEP or local MLH3 framework exists (vcep_materials.json), so generic ACMG/AMP 2015 BP2 governs: 'Observed in trans with a pathogenic variant for a fully penetrant dominant gene/disorder or observed in cis with a pathogenic variant in any inheritance pattern' (PMID:25741868). Neither branch is satisfied for NM_001040108.2:c.2390G>A. Cis branch: no source in this case records any phase relationship for this variant. The ClinVar record (variation 314383; 13 submissions, 11 submitters, germline origin, no expert-panel submission) reports no cis partner variant and no zygosity, and all four full-text papers extracted for the case (PMID:25741868, PMID:34043773, PMID:33516529, PMID:24493721) are negative for this variant by name - none of them reports c.2390G>A in cis with any pathogenic allele, nor does any other bundle source (gnomAD provides no phase for heterozygotes). Trans-dominant branch: no pathogenic MLH3 allele in trans is reported for this variant anywhere, and the branch's premise is not established for MLH3 - the EHTG/ESCP Mallorca guideline (PMID:34043773) states that Lynch syndrome path_MMR variants are defined only for MLH1, MSH2, MSH6 and PMS2, and the ASCO and SGO reviews (PMID:24493721, PMID:33516529) likewise list only MLH1/MSH2/MSH6/PMS2 (plus EPCAM), with MLH3 absent from every gene list and risk model. Population-scale data cannot substitute for a phase observation: gnomAD v2.1 shows 72 homozygotes and gnomAD v4.1 shows 174 homozygotes (171 African/African American) for this allele, demonstrating biallelic tolerance - that observation belongs to BS2 and was deliberately not counted toward BP2, because BP2 requires a documented pathogenic partner in cis or in trans. Because the variant's phase was never established in any source and neither branch has any supporting observation, BP2 is not met rather than not applicable; a documented cis-with-pathogenic or qualifying trans-with-pathogenic observation would change the result.

Rule

BP2 (supporting, generic ACMG/AMP 2015): 'Observed in trans with a pathogenic variant for a fully penetrant dominant gene/disorder or observed in cis with a pathogenic variant in any inheritance pattern' (Richards et al. 2015, PMID:25741868).

Evidence considered

• Governing framework: vcep_materials.json records no CSPEC/VCEP and no local MLH3 specification, so generic ACMG/AMP 2015 BP2 (PMID:25741868) was applied: 'Observed in trans with a pathogenic variant for a fully penetrant dominant gene/disorder or observed in cis with a pathogenic variant in any inheritance pattern'.
• ClinVar (clinvar, variation 314383): 13 submissions from 11 submitters / 12 clinical laboratories, all Benign, germline origin, no expert-panel submission; no submission states phase, zygosity, or a second MLH3 allele, so the record documents neither a cis nor a trans relationship with a pathogenic variant. Only submission-level fields were consulted; the aggregate classification was not used as evidence.
• Full-text paper extracts (paper_extracts.json, all four fetched papers): variant_mentioned = false for PMID:25741868, PMID:34043773, PMID:33516529 and PMID:24493721 when searched for c.2390G>A, p.(R797H), p.R797H and R797H - no published cis or trans observation exists for this variant in any paper retrieved for this case.
• PMID:34043773 (EHTG/ESCP Mallorca guideline, third edition): 'LS is caused by inherited malfunction in one of the four MMR genes named MLH1, MSH2, MSH6 and PMS2. The malfunctioning variants are referred to as path_MLH1, path_MSH2, path_MSH6 and path_PMS2.' MLH3 is never mentioned, so the fully penetrant dominant path_MMR premise of BP2's trans branch has no documented basis for this gene.
• PMID:33516529 (SGO endometrial cancer review): 'Most hereditary endometrial cancers are due to Lynch Syndrome, an autosomal dominant syndrome caused by germline mutations in mismatch repair genes MLH1, MSH2, MSH6, PMS2, and EPCAM.' MLH3 is absent from the gene lists, so no dominant gene level context is provided for this variant.
• PMID:24493721 (ASCO cancer family-history statement): its gene list for mismatch-repair-deficient colorectal cancer is 'MLH1/MSH2/MSH6/PMS2/EPCAM'; MLH3 appears nowhere in Table 2 or Table 3, so it provides no cis/trans or dominant-gene evidence for this variant.
• Population data used only as context, not as BP2 support: gnomAD v2.1 (gnomad_v2) 1801/282780 alleles (AF 0.63689%) with 72 homozygotes; gnomAD v4.1 (gnomad_v4) 5663/1614070 alleles (AF 0.35085%) with 174 homozygotes (171 in the African/African American population, AF 6.62747%). Homozygous carriers at population scale show the allele is tolerated biallelically, which is the domain of BS2, not BP2.
• Phase search: no proband, pedigree, parental genotypes or long-read phasing data were supplied with this case, and no source among the ClinVar submissions, evidence.json records or extracted full texts reports phasing for this variant, so neither arm of BP2 could be satisfied from the available evidence.

Still missing

• Cis/trans phase between c.2390G>A and any pathogenic MLH3 allele is undocumented in every source consulted (ClinVar submissions, the four extracted full texts, gnomAD).
• No affected-proband, parental or family genotype data were provided in this case, so a cis-with-pathogenic configuration could not be excluded by testing.
• MLH3 has no ClinGen CSpec/VCEP, so BP2 was applied only with the generic ACMG/AMP 2015 definition and no gene-specific calibration.

Sources

Reasoning

BP6 (Richards et al. 2015, PMID:25741868) may only be applied here for an exact-variant ClinVar classification issued by a recognised expert panel asserting Benign or Likely benign. The exact record VCV000314383 (NM_001040108.2(MLH3):c.2390G>A (p.Arg797His)) is indeed labelled Benign and has 12 submissions from independent clinical laboratories, but its record-level review status is 'criteria provided, multiple submitters, no conflicts' = 2 stars, which is an aggregate of ordinary laboratory assertions, not an expert-panel classification; clinvar_submission_audit.json records is_expert_panel = false for every one of the 13 submissions (expert_panel_submissions = 0) and no expert-panel (e.g. InSiGHT Variant Interpretation Committee) assertion exists for this variant. Several of these submissions also lack assertion criteria altogether (SCV000691886 Mayo Clinic Laboratories and SCV007602136 Dasa: 'no assertion criteria provided'), and no submission supplies criteria-level evidence. The Benign consensus in ClinVar is therefore exactly the aggregate, non-expert category that must never trigger BP6, and BP6 is not met.

Rule

expert_panel_clinvar_rule: BP6 requires an exact-variant ClinVar expert-panel (3-star) Benign/Likely benign classification; laboratory and aggregate assertions are excluded.

Evidence considered

• ClinVar VCV000314383 is classified Benign with record review status 'criteria provided, multiple submitters, no conflicts' (2 stars) - an aggregate of laboratory submissions, not an expert-panel classification.
• clinvar_submission_audit.json: 13 submissions from 12 clinical laboratories (Mayo Clinic Laboratories, Illumina, GeneDx, ARUP, KCCC Kuwait x2, Ambry, Breakthrough Genomics, Rigshospitalet, Sinai Health System, Labcorp/Invitae, Myriad, Dasa); expert_panel_submissions = 0 and is_expert_panel = false for every submission.
• Two submissions provided no assertion criteria at all (SCV000691886, SCV007602136), and none of the submissions provided criteria-level evidence for independent evaluation.
• Governing rule applied: BP6 is applied only for an exact-variant ClinVar expert-panel Benign/Likely benign classification; aggregate ClinVar labels and ordinary laboratory submissions must never trigger it.

Still missing

• No expert-panel (3-star) Benign/Likely benign ClinVar assertion exists for this variant, so BP6 currently has no qualifying reputable-source classification to cite.

Sources

Reasoning

PS2 (Richards et al. 2015, PMID:25741868) is met when the variant is observed to arise de novo in a proband with the relevant phenotype and parental identity is confirmed by genotyping (i.e. both parents tested and shown to lack the variant). No proband, no parental genotype and no clinical/phenotype record exists anywhere in this case: prefetch.json contains only variant normalization (NM_001040108.2:c.2390G>A, NP_001035197.1:p.(Arg797His), NC_000014.8:g.75513969C>T) and in-silico lookups, evidence.json contains only database scrapes (ClinVar, gnomAD, OncoKB, COSMIC, hotspots, SpliceAI, REVEL), and a text sweep of every case file for 'de novo', 'proband', 'trio', 'pedigree', 'parental' and 'inheritance' returns nothing beyond the title of a retrieved family-history guideline (PMID:24493721, 'collection and use of a cancer family history'). Parental testing evidence is therefore genuinely absent rather than merely unreported, and no de novo claim exists that could be evaluated or refuted, so PS2 is not assessable from this case. Context that does not substitute for parental testing: the variant is common in gnomAD v2.1 (total AF 0.00636891, 1801/282780 alleles, 72 homozygotes; African/African American AF 0.0661134, 1649/24942 alleles, 71 homozygotes) and gnomAD v4.1 (total AF 0.00350852, 5663/1614070 alleles, 174 homozygotes; African/African American AF 0.0662747), and ClinVar holds 13 submissions, 11 usable, all Benign, with no expert-panel submission (VCV000314383, 'criteria provided, multiple submitters, no conflicts', 2 stars, 12 clinical laboratories). This context makes a pathogenic de novo occurrence biologically implausible, but PS2 is decided by parental genotyping, not by frequency or by another source's classification, so the criterion remains not_assessed rather than met or not_met.

Evidence considered

• Criterion definition: PS2 requires a de novo observation in a proband with confirmed parental identity (Richards et al. 2015, PMID:25741868).
• No proband/parental/trio data anywhere in prefetch.json, evidence.json, case_summary.json, literature_pass.json, review_manifest.json or vcep_materials.json; a keyword sweep for 'de novo', 'proband', 'trio', 'pedigree', 'parental', 'inheritance', 'segregation' found no case-level or family-level content.
• gnomAD v2.1: total AF 0.00636891 (1801/282780 alleles) with 72 homozygotes; African/African American AF 0.0661134 (1649/24942 alleles) with 71 homozygotes.
• gnomAD v4.1: total AF 0.00350852 (5663/1614070 alleles) with 174 homozygotes; African/African American AF 0.0662747 with 171 homozygotes.
• ClinVar VCV000314383: Benign, 13 submissions (11 usable for classification), 12 clinical laboratories, no expert-panel submission, review status 'criteria provided, multiple submitters, no conflicts' (2 stars); criterion leads SCV004015898, SCV005294506, SCV002760757, SCV005880860, SCV005918741 cite PMID:25741868 and SCV000562043 cites PMID:28492532, none of which reports proband or parental genotypes.
• ClinVar clinical laboratories do not supply de novo or parental-testing data in this record, so no submission was treated as a de novo observation.
• Absence of this variant from any retrieved family-based publication, and absence of any MLH3 pedigree in the four full texts reviewed (PMID:25741868, PMID:24493721, PMID:33516529, PMID:34043773).

Still missing

• No proband or index patient is defined anywhere in the case; the record is a variant-level query only.
• No parental genotypes and no trio structure (neither confirmed nor assumed parentage) are available, so de novo vs inherited origin cannot be established.
• No phenotype, indication, or clinical history is recorded, so the 'proband with the relevant phenotype' requirement of PS2 cannot be satisfied even if a de novo event were documented.
• No ClinGen VCEP or MLH3 gene-specific specification exists for this gene (cspec.found=false, gene_vcep_dir=null, framework source generic_acmg_fallback), so generic ACMG/AMP 2015 definitions are the only applicable rules.

Sources

Reasoning

PS3 requires well-established in vitro or in vivo functional studies supportive of a damaging effect on the gene or gene product (Richards et al. 2015, PMID:25741868). NM_001040108.2:c.2390G>A in MLH3 is a missense substitution (NP_001035197.1:p.(Arg797His)), and no functional assay result for this variant was identified in any retrieved source. OncoKB's variant page states that it 'did not identify variant-specific reviewed functional evidence for this variant' and returns 'Unknown Oncogenic Effect' (not reviewed); the COSMIC record COSV104996920 is a single somatic occurrence, not an assay; the available predictor scores (REVEL 0.083; BayesDel -0.564808 with no published ACMG-strength calibration; SpliceAI max delta 0.00) are computational, not functional. The ClinVar record (VCV000314383, Benign, 12 submissions) carries PMID trails only to framework papers (PMID:25741868, PMID:28492532), and none of the germline MLH3 papers retrieved by the targeted literature search (PMID:29568967, PMID:30573798, PMID:31822864, PMID:36833239, PMID:37427096) mentions c.2390G>A or p.Arg797His. MLH3 has no ClinGen VCEP or CSPEC, so the generic ACMG/AMP 2015 framework governs and no gene-specific validated functional assay is defined. Because the criterion depends entirely on an experimental assay result that does not exist in the evidence, PS3 is returned as not assessed rather than not met: absence of a study is not evidence of a damaging effect.

Evidence considered

• OncoKB variant page for MLH3 p.Arg797His returned 'Unknown Oncogenic Effect', not reviewed, with the statement: 'OncoKB did not identify variant-specific reviewed functional evidence for this variant' (oncokb).
• COSMIC record COSV104996920 lists one somatic occurrence of this variant; a somatic recurrence count is not a functional assay (cosmic).
• ClinVar VCV000314383 (Benign, 12 submitters, 2-star review status) carries submission PMID trails that cite only framework papers, PMID:25741868 (ACMG/AMP 2015 standards, Richards et al.) and PMID:28492532 (Sherloc criteria refinement); no submitter cites a functional study of p.Arg797His (clinvar).
• Targeted germline MLH3 literature search (queries including 'MLH3 germline syndrome', 'MLH3 germline developmental disorder', 'MLH3 inherited disease loss of function', 'MLH3 truncating variant germline') returned 15 candidate PMIDs and 5 top-scored papers (PMID:29568967, PMID:30573798, PMID:31822864, PMID:36833239, PMID:37427096); none of these titles or abstracts names MLH3 c.2390G>A or p.Arg797His, and none reports a variant-specific functional assay.
• Available molecular scores are computational rather than functional: REVEL 0.083, SpliceAI max delta 0.00, BayesDel -0.564808 (no published ACMG-strength calibration, therefore not usable for PP3/BP4 and never functional evidence) (revel, spliceai, bayesdel).
• No ClinGen CSPEC, VCEP specification, or local gene-specific MLH3 framework was retrieved for this case, so PS3 is evaluated under the generic ACMG/AMP 2015 criterion definition and no MLH3-specific functional assay calibration exists (cspec, generic_acmg_combination_rules).

Still missing

• No variant-specific functional assay for MLH3 p.(Arg797His) was identified in any retrieved source (OncoKB variant review, COSMIC, ClinVar submission references, or the targeted germline MLH3 literature search); PS3 cannot be scored without a well-established assay result demonstrating a damaging effect.
• MLH3 has no ClinGen VCEP or CSPEC, so no gene-specific validated functional assay or assay-strength calibration is defined for PS3 under the generic ACMG/AMP 2015 fallback.

Sources

Reasoning

PM6 (Richards et al. 2015, PMID:25741868) is met when a variant is assumed to be de novo but parental identity has not been confirmed (e.g. parents unavailable for testing or paternity not established). PM6 therefore still presupposes a proband in whom the variant was observed and a reason to assume it arose de novo. This case contains no proband at all: no clinical record, no phenotype, no indication, no parental samples, and no statement that the variant was observed as a singleton or without parental samples. The only family-related text in the entire case is the title of a retrieved guideline (PMID:24493721, ASCO statement on collection and use of a cancer family history), and the four full texts examined (PMID:25741868, PMID:24493721, PMID:33516529, PMID:34043773) contain no MLH3 proband and do not mention c.2390G>A or p.(R797H) (all four extractions record variant_mentioned=false). PM6 is consequently unassessable rather than failed: there is nothing to assume de novo because there is no affected individual to anchor the assumption. For completeness, the variant's population behaviour argues against any de novo pathogenic assertion: 72 homozygotes in gnomAD v2.1 and 174 in gnomAD v4.1, African/African American AF approximately 6.6% in both datasets, and a ClinVar record with 11 usable Benign submissions and no expert-panel review (VCV000314383). That population evidence belongs to the frequency criteria (BA1/BS1/BS2), not to PM6.

Evidence considered

• Criterion definition: PM6 covers a variant assumed to be de novo without confirmation of parental identity (Richards et al. 2015, PMID:25741868).
• No proband, singleton case observation, phenotype, or family relation is recorded in any case file (prefetch.json, evidence.json, case_summary.json), so the de novo assumption PM6 requires has no basis in the record.
• gnomAD v2.1 homozygotes = 72 (total AF 0.00636891) and gnomAD v4.1 homozygotes = 174 (total AF 0.00350852), which is inconsistent with a pathogenic de novo mechanism; recorded here as context, not as PM6 evidence.
• ClinVar VCV000314383 Benign across 11 usable submissions from 12 clinical laboratories, no expert-panel submission, no submission citing a de novo observation in a proband.

Still missing

• No proband, singleton observation, or index case is recorded, so there is no de novo assumption for PM6 to encode.
• No parental or family data of any kind, even unconfirmed, and no phenotype to establish disease relevance of an assumed de novo event.
• No ClinGen VCEP or gene-specific MLH3 framework exists, so no specification modifies the PM6 threshold or its prerequisites.

Sources

Reasoning

PP1 (Richards et al. 2015, PMID:25741868) is met when the variant co-segregates with disease in multiple affected family members, and the strength of the criterion is set by the number of informative meioses genotyped. This case supplies zero meioses. There is no pedigree, no affected relative, no relative genotype, and no proband from whom a family could be constructed; the case is a variant-level database query with normalization and in-silico lookups only. The publications retrieved for this case add no segregation data: the ASCO expert statement on collection and use of a cancer family history (PMID:24493721) is a methodology/consensus document whose gene lists cover MLH1/MSH2/MSH6/PMS2/EPCAM and omit MLH3 entirely, the SGO endometrial cancer reviews (PMID:33516529) likewise describe Lynch syndrome as caused by MLH1, MSH2, MSH6, PMS2 and EPCAM and never mention MLH3, and the updated Mallorca Lynch syndrome guidelines (PMID:34043773) define pathogenic MMR variants only for MLH1, MSH2, MSH6 and PMS2 with MLH3 absent from the text. None of the four full texts mentions c.2390G>A or p.(R797H), and none reports a family study of MLH3. With no family, no affected relatives and no informative meioses, PP1 cannot be assessed; the gap is the complete absence of pedigree-level evidence, not a negative segregation finding. The variant's high population frequency and homozygote counts are not segregation evidence and are handled by the population-frequency criteria.

Evidence considered

• Criterion definition: PP1 requires co-segregation with disease in multiple affected family members, scaled by the number of informative meioses (Richards et al. 2015, PMID:25741868).
• Keyword sweep of prefetch.json, evidence.json, case_summary.json, literature_pass.json, review_manifest.json and vcep_materials.json found no pedigree, proband, or relative-genotype content; the only 'family history' hit is the title of PMID:24493721.
• PMID:24493721 (ASCO expert statement on cancer family history) was reviewed for family/segregation data: it is a consensus methodology document; its gene lists cover MLH1, MSH2, MSH6, PMS2 and EPCAM and MLH3 is absent, and it reports no MLH3 pedigree.
• PMID:33516529 (SGO endometrial cancer review) and PMID:34043773 (EHTG/ESCP Mallorca Lynch syndrome guidelines) were reviewed for MLH3 segregation data: both restrict Lynch syndrome causation to MLH1, MSH2, MSH6 and PMS2, never mention MLH3, and report no segregation of any MLH3 variant.
• PMID:25741868 was checked for variant-level or family data and contains none (framework document only).
• No segregation data were found for this variant in ClinVar submissions (VCV000314383) or in any source in the registry.

Still missing

• No pedigree or family structure of any size is recorded in the case.
• No genotypes from affected or unaffected relatives, and no proband from whom relatives could be ascertained.
• No phenotype or disease definition is available to score co-segregation even if a family existed.
• No ClinGen VCEP or MLH3-specific framework defines a segregation/LOD threshold for this gene.

Sources

Reasoning

PP4 (Richards et al. 2015, PMID:25741868) requires that the patient's phenotype or family history be highly specific for a disease with a single genetic etiology. This case is a variant-level assessment: neither case_summary.json, evidence.json, prefetch.json nor the evidence bundle records any proband HPO term, clinical diagnosis, age at onset, tumour MMR/MSI status, or family history, so there is no patient phenotype against which phenotype specificity can be judged. Independently, the retrieved gene-level guidelines do not establish a highly specific MLH3-associated phenotype: the EHTG/ESCP Lynch syndrome guideline states that 'LS is caused by inherited malfunction in one of the four MMR genes named MLH1, MSH2, MSH6 and PMS2' (PMID:34043773), the ASCO cancer family-history statement lists only MLH1/MSH2/MSH6/PMS2/EPCAM for MMR-deficient colorectal or endometrial cancer and omits MLH3 entirely (PMID:24493721), and the SGO endometrial cancer review likewise names only MLH1/MSH2/MSH6/PMS2/EPCAM (PMID:33516529). MLH3 is absent from all of these phenotype-defining gene sets. Because the decisive input (the case's own phenotype and family history) is missing rather than negative, PP4 is recorded as not_assessed instead of not_met.

Rule

generic_acmg_2015: PP4 requires a patient phenotype or family history highly specific for a disease with a single genetic etiology; no proband-level phenotype data available in this case.

Evidence considered

• No proband phenotype, HPO term, family history, age at onset or tumour MMR/MSI information appears anywhere in the case files; the case is a variant-level classification of NM_001040108.2:c.2390G>A.
• PMID:34043773 (EHTG/ESCP Mallorca-3 Lynch syndrome guideline) defines path_MMR variants exclusively for MLH1, MSH2, MSH6 and PMS2 and never mentions MLH3, so no highly specific MLH3 phenotype is documented.
• PMID:24493721 (ASCO cancer family history statement) lists MLH1/MSH2/MSH6/PMS2/EPCAM for MMR-deficient colorectal or endometrial cancer and omits MLH3 from all gene lists and risk models.
• PMID:33516529 (SGO endometrial cancer review) states hereditary endometrial cancer is mostly Lynch syndrome from MLH1, MSH2, MSH6, PMS2 and EPCAM; MLH3 is not listed.
• Generic ACMG/AMP 2015 (PMID:25741868) requires the patient's phenotype/family history to be highly specific for a disease with a single genetic etiology - the input this criterion needs is absent.

Still missing

• No proband-level clinical phenotype (HPO terms, diagnosis, age at onset, tumour MMR/MSI status) or family history is present in this case, so phenotype specificity for a single-gene etiology cannot be evaluated.
• A gene-disease validity/phenotype-specificity source for MLH3 (e.g. ClinGen gene-disease validity curation) was not available; retrieved Lynch syndrome guidelines (PMID:34043773, PMID:24493721, PMID:33516529) do not include MLH3 among the established MMR/Lynch genes.

Sources

Reasoning

BS3 requires well-established in vitro or in vivo functional studies showing no damaging effect on the protein or gene product (Richards et al. 2015, PMID:25741868). No functional study of MLH3 p.(Arg797His) (c.2390G>A) was identified: OncoKB returns 'Unknown Oncogenic Effect' with no variant-specific reviewed functional evidence, COSMIC holds only a single somatic occurrence (COSV104996920), and the retrieved germline MLH3 literature (PMID:29568967, PMID:30573798, PMID:31822864, PMID:36833239, PMID:37427096) does not mention this variant. Facts that are sometimes mistaken for BS3 are not assay evidence here: the REVEL score (0.083) and SpliceAI max delta (0.00) are computational predictions assessed under PP3/BP4, and the Benign ClinVar assertions of 12 laboratories (VCV000314383) are classification opinions whose PMID trails cite only framework papers (PMID:25741868, PMID:28492532), so they cannot be used to infer BS3. No MLH3 VCEP or CSPEC exists to define an acceptable MLH3 functional assay. BS3 is therefore not assessed; the gap is the absence of any variant-specific functional assay (for example an MLH1-MLH3 endonuclease or MMR-repair activity assay), not a demonstration that function is preserved.

Evidence considered

• OncoKB variant page for MLH3 p.Arg797His returned 'Unknown Oncogenic Effect', not reviewed, with no variant-specific functional evidence (oncokb).
• COSMIC COSV104996920 records a single somatic occurrence; this is a tumour-recurrence observation, not a functional assay showing preserved protein function (cosmic).
• ClinVar VCV000314383 is classified Benign by 12 submitters (2-star review status); its submission PMID trails cite only PMID:25741868 and PMID:28492532, so the benign assertions are not backed by functional assays and cannot be converted into BS3 evidence (clinvar).
• Targeted germline MLH3 literature search (15 candidate PMIDs; 5 top-scored papers PMID:29568967, PMID:30573798, PMID:31822864, PMID:36833239, PMID:37427096) returned no paper naming MLH3 c.2390G>A / p.Arg797His and no MLH3 functional assay of this residue; these papers address MSH2 splicing, biallelic MLH3 truncating polyposis, deep-intronic MMR sequencing, MSI/splicing biology, and renal cell carcinoma genomics respectively.
• Computational scores are not BS3 evidence: REVEL 0.083 and SpliceAI max delta 0.00 are in-silico predictions, and BayesDel -0.564808 has no published ACMG-strength calibration (revel, spliceai, bayesdel).
• No ClinGen CSPEC, VCEP specification, or local MLH3-specific framework was retrieved, so BS3 is evaluated under the generic ACMG/AMP 2015 definition with no MLH3-calibrated functional assay available (cspec, generic_acmg_combination_rules).

Still missing

• No variant-specific functional assay for MLH3 p.(Arg797His) was identified, in either direction; BS3 requires a well-established assay showing no damaging effect, which is absent from the retrieved evidence.
• MLH3 has no ClinGen VCEP or CSPEC, so no MLH3-specific validated functional assay or assay-strength calibration exists under which a benign functional result could be scored.

Sources

Reasoning

BS4 (Richards et al. 2015, PMID:25741868) is met when there is a documented lack of segregation with disease, i.e. genotype data from an informative family showing affected carriers, affected non-carriers, and/or unaffected carriers that affirmatively contradicts co-segregation. This case has no pedigree and no relative genotypes, so non-segregation can neither be demonstrated nor contradicted; absence of segregation data is not itself evidence of non-segregation and must not be scored as BS4. The retrieved literature adds no family data: the four full texts reviewed (PMID:25741868, PMID:24493721, PMID:33516529, PMID:34043773) do not mention c.2390G>A or p.(R797H) at all, and none contains an MLH3 pedigree, so they cannot support BS4. Two sources supply strong benign evidence for this variant, but of a different type than BS4 and belonging to other criteria: gnomAD shows the variant at total AF 0.00350852 (5663/1614070 alleles) with 174 homozygotes in v4.1 and total AF 0.00636891 with 72 homozygotes in v2.1 (African/African American AF approximately 6.6% in both), and ClinVar VCV000314383 is Benign across 11 usable submissions from 12 clinical laboratories with no expert-panel review. Homozygotes in the general population indicate the variant is tolerated in the homozygous state; that is a population/tolerance argument (BA1/BS2/BS1), not a BS4 non-segregation finding, and no family study is available to which BS4 could be applied. BS4 is therefore not_assessed rather than not_met.

Evidence considered

• Criterion definition: BS4 requires documented lack of segregation of the variant with disease in affected family members (Richards et al. 2015, PMID:25741868).
• No pedigree, relative genotype, or family structure recorded in any case file, so a non-segregation claim cannot be constructed.
• gnomAD v4.1: total AF 0.00350852 (5663/1614070 alleles) with 174 homozygotes and African/African American AF 0.0662747; gnomAD v2.1: total AF 0.00636891 (1801/282780 alleles) with 72 homozygotes and African/African American AF 0.0661134.
• gnomAD-Canada v1.0: AF 0.00521286 (96/18416 alleles) with 5 homozygotes.
• ClinVar VCV000314383: Benign, 11 usable of 13 submissions, 12 clinical laboratories, no expert-panel submission, 'criteria provided, multiple submitters, no conflicts' (2 stars).
• All four full-text publications retrieved for this case were searched for the variant and for MLH3 family data; none mentions c.2390G>A or p.(R797H), so none provides segregation or non-segregation evidence.
• The homozygote observations are recorded as population tolerance context (BA1/BS2 territory) and were not scored as BS4.

Still missing

• No pedigree or relative genotypes, so non-segregation cannot be demonstrated; no family study exists to which BS4 could be applied.
• No phenotype or disease definition for any family member, which a non-segregation claim would require.
• No ClinGen VCEP or MLH3-specific framework exists to specify BS4 handling for this gene.
• Population-level benign evidence for this variant (high allele frequency, 174 gnomAD v4.1 homozygotes, ClinVar Benign) must be adjudicated under BA1/BS2/BS1 in the population group; it is not BS4 evidence.

Sources

Reasoning

BP5 (Richards et al. 2015, PMID:25741868) requires that the variant be found in a case with an alternate molecular basis for disease, such as a co-occurring pathogenic variant in the same or another gene or a second independent genetic diagnosis. No CSPEC/VCEP specification exists for MLH3, so the generic ACMG/AMP 2015 definition governs; that definition contains no numeric likelihood-ratio or other quantitative threshold, so no operator or threshold direction applies to BP5 in this case (there is no source text defining a value or comparison for it here). The case file contains no proband-level genotypic or clinical data at all, and the 13 ClinVar submissions for this variant provide no case-level detail, no co-occurring-variant information and no criteria-level evidence: clinvar_submission_audit.json records evidence_signals.criteria_found as empty for every submission except one PP3 mention (SCV007602136, Dasa) that is unrelated to an alternate molecular basis, and the only criteria references cited are the generic papers PMID:25741868 and PMID:28492532. No reported case in which this variant co-exists with an alternate molecular cause could be located, so BP5 cannot be evaluated and is recorded as not_assessed.

Rule

generic_acmg_2015: BP5 requires the variant to be found in a case with an alternate molecular basis for disease; no case-level data available in this case.

Evidence considered

• Generic ACMG/AMP 2015 (PMID:25741868) defines BP5 as the variant being found in a case with an alternate molecular basis for disease (e.g. a co-occurring pathogenic variant); no numeric threshold is defined, and no MLH3-specific CSPEC/VCEP specification was retrieved.
• No proband-level genotype or clinical information is present in case_summary.json, prefetch.json or evidence.json; the case is a variant-level classification only.
• clinvar_submission_audit.json: 13 submissions, criteria_found empty for every submission except one PP3 mention by Dasa (SCV007602136); no submission reports a co-occurring pathogenic variant or an alternate diagnosis.
• ClinVar VCV000314383 submissions cite only generic criteria sources (PMID:25741868; SCV000562043 cites Sherloc, PMID:28492532), giving no case-level genotype information usable for BP5.
• Somatic COSMIC record COSV104996920 (n = 1) and OncoKB 'Unknown Oncogenic Effect' provide no case-level alternate-molecular-basis information.

Still missing

• No case-level data (co-occurring pathogenic variant, second diagnosis, or family genotype data) are available in this case, so an alternate molecular basis for disease cannot be confirmed or excluded.
• ClinVar submissions for this variant provide no criteria-level detail (evidence_signals.criteria_found empty for all submissions), so no submission can be mined for a documented alternate molecular basis.

Sources