MLH3 c.3280+14A>T, p.?

Reasoning

BS1 (strong benign) is met when the allele frequency in a control or general population is greater than expected for the disorder. No MLH3-specific specification exists (ClinGen CSPEC not found; no gene VCEP directory or user specification; framework resolved to generic ACMG/AMP 2015), and no specification mandates a non-cancer or exome-only source, so the threshold applied is the generic supplied default of AF >= 0.01 and the all-comers GNOMAD_V2_1 / GNOMAD_V4_1 entries are the default source. In gnomAD v2.1 the threshold is crossed on two independent metrics: the East Asian ancestry AF is 0.0111791 (223/19,948 alleles, 3 homozygotes) and the exome grpmax filtering allele frequency is 0.0101016 (the variant's overall grpmax FAF is the same 0.0101016). The v2.1 non-cancer exome subset is concordant and slightly higher (EAS AF 0.0114829, 221/19,246, 3 homozygotes; grpmax FAF 0.0103962), so the excess is not an artifact of cancer-cohort inclusion. This is a frequency far above what is expected for MLH3-related disease: the associated phenotype is a rare, adult-onset, incompletely penetrant cancer predisposition (the current EHTG/ESCP Lynch syndrome guideline defines the disorder as caused by MLH1, MSH2, MSH6 and PMS2 and does not include MLH3), no MLH3 prevalence/penetrance figure was available in the case evidence to derive a gene-calibrated (Whiffin-style) threshold, and the generic 1% default is therefore used rather than a lower gene-specific value. The verdict rests on ancestry-specific frequency, which is the correct frame here: 223 of the 225 v2.1 alleles and 332 of the 354 v4.1 alleles are East Asian, so the all-comers totals (0.000796 and 0.000219) understate the frequency in the subpopulation where the variant segregates. The main caveat is dataset discordance: gnomAD v4.1 does not cross the 1% default (EAS AF 0.0073975, 332/44,880, 2 homozygotes; joint grpmax FAF 0.00674173), so a reviewer applying only v4.1 would not call BS1 on the numeric default alone; the 3 homozygotes and 1.12% East Asian frequency in v2.1, together with the low expected frequency for MLH3-associated disease, nevertheless support the strong-benign call, and the criterion is flagged for human review because of that cross-dataset discordance.

Rule

BS1 strong benign: allele frequency greater than expected for the disorder, evaluated at the generic default of AF >= 0.01 (Whiffin et al. 2017, PMID:28518168; per-gene derivation preferred but no MLH3 specification available).

Evidence considered

• gnomAD v2.1 (all-comers, default source): East Asian AF 0.011179065570483256 (223/19,948 alleles, 3 homozygotes; EAS exome 209/18,392, EAS genome 14/1,556); grpmax FAF 0.010101569999999992 (exome), which is also the variant's overall grpmax FAF; total AF 0.0007959586526011929 (225/282,678, 3 homozygotes).
• gnomAD v2.1 non-cancer exome subset (corroboration): East Asian 221/19,246 (AF 0.01148290553881326, 3 homozygotes); grpmax FAF 0.010396179999999986; total exome AF 0.0008783932161016233 (208/236,796, 3 homozygotes). Concordant with, and slightly above, the all-comers figures, so the frequency excess is not attributable to cancer-cohort inclusion.
• gnomAD v4.1 (all-comers): East Asian 332/44,880 (AF 0.007397504456327986, 2 homozygotes); joint grpmax FAF 0.00674173; total AF 0.0002193343882863045 (354/1,613,974, 2 homozygotes). Below the 1% generic default - the source of the discordance flagged for review.
• Threshold applied verbatim from the supplied published calibration block: BS1 strong benign at allele frequency >= 0.01, generic SVI default with per-gene derivation preferred (Whiffin et al. 2017, PMID:28518168); used because no MLH3-specific specification exists. The ACMG/AMP 2015 criterion definition ('an allele frequency in a control population that is greater than expected for the disorder ... is considered strong support for a benign interpretation for a rare Mendelian disorder (BS1)') is carried by the registry entry keyed to PMID:25741868.
• Disease-context considerations for threshold appropriateness: MLH3-associated disease is an adult-onset, incompletely penetrant cancer predisposition; the EHTG/ESCP Lynch syndrome guideline (PMID:34043773) defines the disorder as caused by inherited malfunction in MLH1, MSH2, MSH6 or PMS2 and never mentions MLH3, so no fully penetrant early-onset MLH3 disorder and no MLH3-specific expected allele frequency are established in the case evidence. No numeric prevalence or penetrance figure was available, so no gene-calibrated threshold was derived.
• Source selection: no ClinGen CSPEC (cspec found = false), no gene VCEP directory or user specification (gene_vcep_dir null), framework mode generic_acmg, so no non-cancer/exome-only source is mandated and the all-comers GNOMAD_V2_1 / GNOMAD_V4_1 entries are the default; the non-cancer subsets were read as corroboration only.
• gnomAD v3.1 non-cancer (genomes only) provides 44/147,966 alleles (AF 0.00029736661101874754, 0 homozygotes) with grpmax FAF 0.006480060000000005; its HGDP:TUJIA 1/18 '5.55556%' figure is a small-sample artifact and was not used.
• gnomAD-Canada v1.0 (HostSeq genomes) query returned search_status 'absent' with error 'timed out' - no usable frequency data.

Still missing

• No MLH3-specific CSPEC, VCEP specification, or local framework was available, and no MLH3 prevalence/penetrance figure exists in the case evidence, so BS1 was evaluated against the generic SVI default (>= 0.01) instead of a Whiffin-style gene-calibrated threshold; a derived threshold for a rare adult-onset cancer predisposition would be lower, which would make BS1 easier to meet, not harder.
• Cross-dataset discordance is unresolved: gnomAD v2.1 crosses the 1% default (EAS 1.1179%; exome grpmax FAF 1.0102%) while gnomAD v4.1 does not (EAS 0.7398%; joint grpmax FAF 0.6742%). No governing specification names which dataset is authoritative, so the criterion is met on v2.1 with needs_human_review = true.
• gnomAD-Canada v1.0 (HostSeq genomes) returned no usable data (lookup timed out), so a third independent cohort could not be used to break the v2.1/v4.1 tie.
• gnomAD v2.1 is an older exome-rich dataset with smaller East Asian sample size (AN 19,948) than v4.1 (AN 44,880); the 1.12% vs 0.74% difference is within the range expected from sampling and differing ancestry composition, but it is not resolvable from the available data.

Sources

Reasoning

NM_001040108.2:c.3280+14A>T (NC_000014.8:g.75513065T>A) is an intronic variant 14 nucleotides into the intron following c.3280, so the applicable computational path is SpliceAI (intronic/splice-region path), not REVEL. SpliceAI reports a max delta score of 0.005, with all four component deltas at or below 0.005 (DS_AG 0.005, DS_AL 0.000, DS_DG 0.002, DS_DL 0.003). This is <= 0.1, the SpliceAI recommended cutoff below which no splice impact is predicted (Jaganathan et al. 2019, Cell, PMID:30661751), which is the calibrated BP4 supporting threshold in the generic ACMG/ClinGen-SVI framework. The only named predicted effect is a weak donor gain 14 nt downstream (DP_DG 14) with delta 0.002, and the nearby canonical donor site is not disrupted (DP_AL 166, DP_AG 227 deltas 0.000 and 0.005). Pangolin is concordantly null (splice gain 0.004, splice loss -0.001) but carries no published ACMG calibration and is not used as a scored basis. The variant is not in the canonical +/-1,2 donor or acceptor dinucleotides, so no intrinsically splice-disrupting class is at issue. No MLH3-specific VCEP specification exists (no CSPEC, no gene VCEP framework, no PP3/BP4 lookup spreadsheet), so the generic calibrated SpliceAI cutoff governs. SpliceAI therefore provides BP4 supporting evidence for a benign splicing effect. The single 0.005 delta does not reach the moderate/strong-benign tiers, which are not defined for SpliceAI in the generic framework, so BP4 is capped at supporting strength.

Rule

BP4 met (supporting): SpliceAI max delta 0.005 <= 0.1 per Jaganathan et al. 2019 (PMID:30661751).

Evidence considered

• SpliceAI Lookup for NM_001040108.2:c.3280+14A>T returned max delta score 0.005 (DS_AG 0.005, DS_AL 0.000, DS_DG 0.002, DS_DL 0.003; DP_DG 14 for the only gain predicted), i.e. no predicted splice impact.
• Applied threshold: SpliceAI BP4 supporting <= 0.1 (SpliceAI recommended cutoff, Jaganathan et al. 2019, Cell, PMID:30661751); the generic framework defines no moderate or strong benign tier for SpliceAI, so strength is supporting.
• Scope: the variant is intronic (14 nt into the intron 3' of c.3280) and outside the canonical +/-1,2 splice sites, so the intronic/splice-region SpliceAI path is the applicable predictor path and no REVEL missense path applies (REVEL score null, not found).
• Pangolin SG 0.004 / SL -0.001 concordantly predict no splice effect; no published ACMG-strength calibration for Pangolin, so it is qualitative context only and is not double-counted.
• No ClinGen CSPEC or gene-specific VCEP framework for MLH3 was available (cspec.found = false, gene_vcep_dir null, vcep_materials.files_found empty, framework mode generic_acmg); no gene-specific PP3/BP4 assignment exists to override the generic SpliceAI cutoff.
• ACMG/AMP 2015 (Richards et al., PMID:25741868) defines BP4 as computational evidence supporting a benign effect, the criterion this SpliceAI result satisfies.

Sources

Reasoning

Scope: NM_001040108.2:c.3280+14A>T (NC_000014.8:g.75513065T>A, MLH3, 14 nt into the intron 3' of the exon ending at c.3280; protein consequence NP_001035197.1:p.?) was assessed for case-control / case-enrichment evidence supporting pathogenicity. PS4 requires that the variant be observed in affected individuals more frequently than expected by chance (case-control enrichment) or be supported by an equivalent, pre-specified case-observation enrichment. No such evidence exists for this variant. The literature retrieved for this case was triaged under a case_observation theme with an explicit PS4 lead (PMID:20301390 Lynch Syndrome GeneReviews chapter, PMID:26389505 and PMID:26389258 PDQ health-professional reviews, PMID:34043773 EHTG/ESCP Mallorca Lynch syndrome guideline), but none of it is variant-specific: the only two papers with retrievable full text never mention MLH3 at all, let alone c.3280+14A>T (PMID:34043773 defines Lynch syndrome exclusively by MLH1, MSH2, MSH6 and PMS2), and the remaining retrieved references have no retrievable text or abstract in this case. No cohort, case series, odds ratio, relative risk, penetrance estimate or case-versus-control count for this exact variant appears anywhere in the assembled evidence. The four ClinVar submissions (SCV000388764 Illumina, SCV005090110 Rigshospitalet, SCV005918745 Sinai Health System, SCV002407211 Labcorp/Invitae) are individual clinical-testing classification assertions with no case or control counts attached, and a database classification label cannot be converted into PS4 by inference. The single COSMIC record (COSV53158903, somatic observation count 1) is a somatic tumour occurrence, not case enrichment for a hereditary cancer phenotype, and is not germline case-control evidence. gnomAD instead shows the variant present in the general population (v4.1 AF 0.000219, 354/1,613,974 alleles, 2 homozygotes; v2.1 AF 0.000796, 225/282678 alleles, 3 homozygotes, including 3 homozygotes in East Asians) - a population-frequency observation, which cannot support a case-enrichment criterion. Numeric-threshold and direction check: generic ACMG/AMP 2015 (PMID:25741868) defines PS4 qualitatively and states no numeric cutoff, and MLH3 has no ClinGen CSPEC (cspec.found = false) and no gene-specific VCEP framework (framework_mode = generic_acmg; vcep_materials.files_found and domain_tables empty; rule_payload null). There is therefore no governing numeric PS4 threshold and no comparison operator (>=, <=, >, <) to read, so no inequality is written or evaluated for this criterion; the verdict rests entirely on the absence of any case-enrichment data rather than on a failed numeric comparison. PS4 is not met.

Rule

PS4 not met: no case-control / case-enrichment evidence exists for the exact variant, and no governing numeric PS4 threshold is defined for MLH3 (no ClinGen CSPEC and no gene-specific VCEP; generic ACMG/AMP 2015, PMID:25741868).

Evidence considered

• ClinVar variation 314378 (this exact variant, NM_001040108.2(MLH3):c.3280+14A>T) carries four germline clinical-testing submissions - Illumina (Likely benign), Center for Genomic Medicine Rigshospitalet (Benign), Sinai Health System (Benign), Labcorp Genetics/Invitae (Benign) - with aggregate review status 'criteria provided, multiple submitters, no conflicts' (2 stars). None reports affected/unaffected carrier counts, so none contributes case-control enrichment.
• ClinVar submission audit for this case records no criterion-level evidence signals from any submitter (criteria_found and signals empty for all four submissions; validated PMIDs only 25741868 and 28492532), so no PS4-bearing data can be recovered from the submission trail.
• PMID:34043773 (EHTG/ESCP Mallorca guideline) full text was read: it defines Lynch syndrome by MLH1/MSH2/MSH6/PMS2 only and never mentions MLH3; it reports no case observations for this variant.
• PMID:20301390 (Lynch Syndrome), PMID:26389505 (Genetics of Colorectal Cancer, PDQ) and PMID:26389258 (Cancer Genetics Risk Assessment and Counseling, PDQ) were cited from the ClinVar record with a PS4/case_observation lead; no full text, abstract or variant-level content is retrievable in this case (publication queue records have empty title and abstract fields), and the case-level variant-mention audit across retrieved text found zero occurrences of the variant.
• PMID:25741868 (ACMG/AMP 2015) was consulted as the defining framework for PS4; it provides no variant-level data and states no numeric PS4 threshold.
• COSMIC record COSV53158903 (somatic count 1) is a single somatic tumour observation and is not case-control enrichment for a hereditary phenotype; COSMIC has no source_registry key and is reported here as descriptive context only.
• gnomAD v4.1: AF 0.000219 (354/1,613,974 alleles, 2 homozygotes; East Asian popmax AF 0.00740, grpmax FAF 0.00674). gnomAD v2.1: AF 0.000796 (225/282678 alleles, 3 homozygotes; East Asian AF 0.011179). These are population observations, not affected-case enrichment, and are noted only to exclude a case-enrichment reading of the frequency data.
• Numeric-threshold direction check: no numeric PS4 threshold exists in the governing specification. Generic ACMG/AMP 2015 (PMID:25741868) defines PS4 qualitatively, and no ClinGen CSPEC or gene-specific VCEP rule exists for MLH3 (framework_mode = generic_acmg, vcep_materials.rule_payload null). With no governing threshold there is no operator to quote for this criterion, so no inequality was evaluated; the not_met verdict is based on absence of case-enrichment evidence.

Still missing

• No case-control or cohort enrichment study for MLH3 c.3280+14A>T is available in this case; a case-control or cohort-based PS4 assessment would require primary affected/unaffected carrier counts for this exact variant.

Sources

Reasoning

PM1 requires the variant to lie in a mutational hotspot and/or a critical, well-established functional domain without benign variation. The vcep_materials.json 'domain_tables' key is an empty object ({}), so there is no authoritative VCEP-derived MLH3 critical-domain list to check the variant's residue against, and no entry exists for any gene: the case has no ClinGen CSPEC (cspec.found = false) and no VCEP directory (gene_vcep_dir null), and the final classification framework is a generic ACMG/AMP fallback with no domain rules. NM_001040108.2:c.3280+14A>T is intronic - 14 nucleotides into the intron following the exon ending at c.3280 (VariantValidator intronic exon context '2i') with protein consequence NP_001035197.1:p.? - so it encodes no residue of any MLH3 protein domain (the MLH3 protein features an N-terminal ATPase/MLH1-interaction region and a C-terminal MutL/endonuclease-like domain) and cannot fall within a critical functional domain. No hotspot evidence is available for the position either: the case hotspot lookup returned found = null with an empty source_data object and no query. COSMIC records this genomic position once in somatic cancer (COSV53158903, n = 1) and ClinVar records the same intronic variant (variation 314378) with no submitter reporting a hotspot or domain location, so there is no germline mutational hotspot at this position. SpliceAI max delta 0.005 (DS_AG 0.005, DS_AL 0.000, DS_DG 0.002, DS_DL 0.003) additionally indicates the substitution does not create a functional splice-region hotspot that could substitute for a protein domain. PM1 is therefore not met.

Rule

PM1 not met: the variant is intronic with no residue (p.?), no authoritative MLH3 domain table exists, no hotspot is established at this position, and SpliceAI shows no functional splice hotspot.

Evidence considered

• vcep_materials.json contains an empty domain_tables object ({}), with no CSPEC ('no cspec url available'), no files_found, no gene_notes and no index_guidance, so there is no VCEP-derived critical functional domain list for MLH3 against which the variant's residue could be checked.
• final_classification_framework.json for MLH3 records found=false, source 'generic_acmg_fallback', framework_complete false and rule_payload null, i.e. no gene-specific PM1 or domain rule was available.
• VariantValidator normalization places NM_001040108.2:c.3280+14A>T in the intron following the exon ending at c.3280 (intronic context; protein consequence NP_001035197.1:p.?), so it encodes no amino-acid residue and cannot lie in a protein functional domain.
• Hotspot annotation in the case evidence is unavailable: the 'hotspots' object has found = null, evidence_sentence = null and an empty source_data object, and the prefetch hotspot query string is empty, so no germline hotspot at this position is established by the supplied evidence.
• ClinVar variation 314378 is this same NM_001040108.2(MLH3):c.3280+14A>T intronic variant, submitted by four clinical laboratories (three Benign, one Likely benign); no submitter places it in a hotspot or functional domain.
• COSMIC lists one somatic occurrence at this position (COSV53158903, n = 1), which is a somatic tumour observation rather than a germline mutational hotspot and carries no residue information.
• SpliceAI scores for this variant (DS_AG 0.005, DS_AL 0.000, DS_DG 0.002, DS_DL 0.003; max_delta 0.005) show no predicted splice-altering effect, so the substitution does not create a functional splice hotspot that could stand in for a domain location.
• PMID:25741868 supplies the generic ACMG/AMP PM1 definition (mutational hotspot and/or critical and well-established functional domain without benign variation) applied here; it contains no MLH3 domain data.

Still missing

• vcep_materials.json 'domain_tables' is empty ({}), so no VCEP-approved MLH3 critical functional domain list exists; PM1 could not be checked residue-by-residue against an authoritative domain table and rests on the variant being intronic (no residue at all).
• No MLH3 protein-domain boundary annotation (Pfam/InterPro-style coordinates) was supplied in the case evidence, so domain membership was assessed only from the variant's non-coding position.

Sources

Reasoning

PM2 applies only when a variant is absent from, or present at extremely low frequency in, a large general population or control cohort; the supplied generic calibration applies it at supporting strength with a default frequency threshold of <= 0.0001 (ClinGen SVI recommendation to downgrade PM2 to supporting, PMID:25741868). No MLH3-specific framework exists (ClinGen CSPEC not found; no gene VCEP directory or user specification; framework resolved to generic ACMG/AMP 2015), and no specification mandates a non-cancer or exome-only source, so the all-comers GNOMAD_V2_1 and GNOMAD_V4_1 entries are the default source. Every applicable metric exceeds 0.0001. gnomAD v2.1 all-comers total AF is 0.000795959 (225/282,678 alleles, 3 homozygotes) - about 8-fold above the threshold - and gnomAD v4.1 total AF is 0.000219334 (354/1,613,974 alleles, 2 homozygotes), about 2.2-fold above it; even the lowest all-comers point estimate therefore fails the threshold. The ancestry-specific frequencies are far higher: East Asian AF 0.0111791 (223/19,948 alleles, 3 homozygotes) in v2.1 and 0.0073975 (332/44,880, 2 homozygotes) in v4.1, i.e. 74- to 112-fold above the threshold. The grpmax filtering allele frequencies (0.0101016 in v2.1, 0.00674173 in v4.1) and the v2.1 non-cancer exome totals (exome AF 0.000878393, 208/236,796 alleles, 3 homozygotes; EAS 0.0114829, 221/19,246, 3 homozygotes) confirm the variant is present at ordinary polymorphic frequency in a cancer-cohort-excluded subset, so the excess is not an artifact of oncology-cohort inclusion. The variant is present with homozygotes in every dataset surveyed, and the ACMG/AMP 2015 requirement that absence be confirmed against adequate read depth is moot because the variant is not absent anywhere; the two gnomAD v4.1 homozygotes and three v2.1 homozygotes further establish that the allele segregates as a common East Asian polymorphism. PM2 is therefore not met, at any strength. The only sub-0.0001-flavoured figure in the case evidence is the HGDP:TUJIA 1/18 (5.6%) entry in gnomAD v3.1 non-cancer, which is a small-sample artifact and does not create a low-frequency argument - that dataset's overall AF is 0.00029736661101874754 (44/147,966 alleles), itself above the PM2 threshold.

Rule

PM2 supporting: variant absent from, or at extremely low frequency in, population databases; default threshold frequency <= 0.0001 for genes without a specific specification (SVI recommendation to apply PM2 at supporting strength; PMID:25741868).

Evidence considered

• gnomAD v2.1 all-comers: total AC 225 / AN 282,678, AF 0.0007959586526011929, total homozygotes 3; exome 211/251,292 (AF 0.0008396606338442927, hom 3); genome 14/31,386 (AF 0.00044605875230994713, hom 0); grpmax FAF 0.010101569999999992; East Asian 223/19,948 (AF 0.011179065570483256, 3 homozygotes).
• gnomAD v4.1 all-comers: total AC 354 / AN 1,613,974, AF 0.0002193343882863045, total homozygotes 2; exome 309/1,461,640 (AF 0.00021140636545250542, hom 2); genome 45/152,334 (AF 0.0002954035212099728, hom 0); joint grpmax FAF 0.00674173; East Asian 332/44,880 (AF 0.007397504456327986, 2 homozygotes).
• gnomAD v2.1 non-cancer exome subset (corroboration only, not mandated by any specification): total exome AC 208 / AN 236,796 (AF 0.0008783932161016233, hom 3); East Asian 221/19,246 (AF 0.01148290553881326, 3 homozygotes); grpmax FAF 0.010396179999999986 - the allele is not rare even after cancer cohorts are excluded.
• gnomAD v3.1 non-cancer (genomes only): total AC 44 / AN 147,966, AF 0.00029736661101874754, 0 homozygotes, grpmax FAF 0.006480060000000005; above the PM2 threshold as well. Its HGDP:TUJIA 1/18 '5.55556%' value is a small-sample artifact and was not used as a frequency estimate.
• Threshold applied verbatim from the supplied published calibration block: PM2 supporting at frequency <= 0.0001 (SVI recommendation to downgrade PM2 to supporting, ClinGen SVI 2020; PMID:25741868), used because no MLH3-specific threshold exists. The ACMG/AMP 2015 PM2 definition - absent from, or below the expected carrier frequency in, a large general population or control cohort (>1000 individuals) race-matched to the patient - is carried by the registry entry keyed to PMID:25741868.
• Source selection: no ClinGen CSPEC (cspec found = false), no gene VCEP directory or user specification (gene_vcep_dir null), framework mode generic_acmg, so no non-cancer/exome-only source is mandated and the all-comers GNOMAD_V2_1 / GNOMAD_V4_1 entries are the default source.
• gnomAD-Canada v1.0 (HostSeq genomes) query returned search_status 'absent' with error 'timed out' - no frequency data returned.
• Literature triage for this case assigned all six ClinVar-linked PMIDs to functional or case-observation themes (PP5/PS4); no population-frequency content or ancestry data for this variant was available from the literature, and the two fetched full texts (PMID:25741868, PMID:34043773) report no MLH3 frequency data.

Still missing

• No MLH3-specific ClinGen CSPEC, gene VCEP specification, or local gene framework exists, so PM2 was assessed with the generic supporting default (<= 0.0001) rather than a gene-calibrated rare-variant threshold.
• gnomAD-Canada v1.0 (HostSeq genomes) returned no usable data (lookup timed out), so one additional independent cohort is missing; the variant is already above the threshold in both datasets that were available.
• No ancestry-matched control threshold exists for MLH3, but ancestry specificity argues further against PM2 here: 223 of 225 v2.1 alleles and 332 of 354 v4.1 alleles are East Asian, where the frequency is 0.74-1.12% rather than rare.
• None of these gaps could change the verdict: the lowest observed all-comers total AF (0.000219334) is still about 2.2-fold above the 0.0001 threshold, and the variant is present with homozygotes in every dataset surveyed.

Sources

Reasoning

NM_001040108.2:c.3280+14A>T (NC_000014.8:g.75513065T>A) lies 14 nucleotides into the intron following the exon ending at c.3280, i.e. an intronic, non-coding, non-synonymous variant. Scope check first: this is not a missense variant (no amino-acid substitution is produced; the protein consequence is p.?), so the REVEL protein-impact path is not the applicable path and REVEL was in any case not available (revel.found = false, score null). The applicable calibrated path is SpliceAI. SpliceAI max delta score = 0.005, with component deltas DS_AG 0.005, DS_AL 0.000, DS_DG 0.002, DS_DL 0.003 and the strongest predicted effect a donor gain 14 nt downstream (DP_DG 14) at a delta of only 0.002. This is far below both the SpliceAI recommended supporting cutoff of >= 0.2 and the high-precision moderate cutoff of >= 0.5 (Jaganathan et al. 2019, Cell, PMID:30661751). No other calibrated splice predictor supports a damaging effect: Pangolin gives splice gain 0.004 and splice loss -0.001, i.e. concordantly null, but no published ACMG-strength calibration is available for Pangolin so it is not used as a scored basis. VarSeak scores are not present in this case. No VPCE or gene-specific specification exists for MLH3 (no ClinGen CSPEC and no gene VCEP framework; vcep_materials lists no PP3/BP4 pre-assignment and no lookup spreadsheet), so the published generic SpliceAI calibration applies directly. No in-silico splice prediction supports a splicing impact, therefore PP3 is not met. This is a genuine negative evaluation of an available calibrated predictor, not a missing-data situation.

Rule

PP3 not met: SpliceAI max delta 0.005 < 0.2 (supporting) and < 0.5 (moderate) per Jaganathan et al. 2019 (PMID:30661751).

Evidence considered

• SpliceAI Lookup for NM_001040108.2:c.3280+14A>T returned max delta score 0.005 (DS_AG 0.005, DS_AL 0.000, DS_DG 0.002, DS_DL 0.003; DP_AG 227, DP_AL 166, DP_DG 14, DP_DL 10); the largest single predicted effect is a donor gain at position +14 with delta 0.002.
• Applied threshold: SpliceAI PP3 supporting >= 0.2 and PP3 moderate >= 0.5 (SpliceAI recommended and high-precision cutoffs, Jaganathan et al. 2019, Cell, PMID:30661751).
• Pangolin scores for the same variant (splice gain 0.004, splice loss -0.001) are concordantly null; no published ACMG-level calibration exists for Pangolin, so it is reported only as qualitative concordance and was not scored.
• Consequence type: c.3280+14A>T is intronic (14 nt into the intron 3' of the exon ending at c.3280); protein consequence is NP_001035197.1:p.? The REVEL missense path is therefore not the applicable path, and REVEL was not available in any case (score null, found false).
• No ClinGen CSPEC was found for MLH3 (cspec.found = false) and no gene-specific VCEP framework or PP3/BP4 lookup spreadsheet is present (gene_vcep_dir null, vcep_materials.files_found empty); the case framework mode is generic_acmg, so the published generic SpliceAI calibration is applied directly.
• ACMG/AMP 2015 (Richards et al., PMID:25741868) defines PP3 as computational evidence supporting a damaging effect and notes that splice-altering predictions are predicted rather than proven effects; no prediction here supports a damaging effect.

Sources

Reasoning

PP5 requires a reputable independent source to report this exact variant as pathogenic; under this case's rule, PP5 is triggered only by an exact-variant ClinGen expert-panel / VCEP classification of Pathogenic or Likely pathogenic, and ordinary laboratory submissions, aggregate ClinVar labels and non-expert assertions must never trigger it. ClinVar variation 314378 (the exact variant, NM_001040108.2(MLH3):c.3280+14A>T) has four germline submissions, all from ordinary clinical laboratories, and zero expert-panel submissions: the API record's expert_panels object is empty, the submission audit sets is_expert_panel = false for all four submissions (SCV000388764 Illumina, SCV005090110 Rigshospitalet, SCV005918745 Sinai Health System, SCV002407211 Labcorp/Invitae) and counts expert_panel_submissions = 0; every submission carries review status 'criteria provided, single submitter', and the aggregate record is 'criteria provided, multiple submitters, no conflicts' (2 stars, not 3-star expert-panel reviewed). No exact-variant expert-panel assertion therefore exists, so the PP5 trigger condition is absent. Independently, the direction of the only ClinVar classification available is opposite to what PP5 needs: the record is Benign/Likely benign (3 laboratories Benign, 1 Likely benign), which could never support a pathogenic criterion even if the submitter tier were elevated. The variant-specific literature trail cited by those submitters (PMID:25741868, PMID:28492532) contains no assertion about this variant - both are methodological/classification-framework papers that never mention MLH3 c.3280+14A>T - and the case-observation themed references (PMID:20301390, PMID:26389505, PMID:26389258, PMID:34043773) contain no MLH3-specific clinical assertion either. PP5 is not met. No numeric threshold or comparison operator is involved in PP5 under either generic ACMG/AMP 2015 or any (absent) MLH3 specification.

Rule

PP5 not met: no exact-variant ClinGen expert-panel (VCEP) classification of Pathogenic/Likely pathogenic exists for c.3280+14A>T (expert_panel_submissions = 0; all four submissions are non-expert laboratory assertions, aggregate label benign).

Evidence considered

• ClinVar variation 314378: expert_panels = {} (empty) in the API record, 4 submissions, 4 submitters, review status 'criteria provided, multiple submitters, no conflicts' (review_stars = 2), last updated 2026-03-08.
• ClinVar submission audit: expert_panel_submissions = 0 and is_expert_panel = false for every submission - SCV000388764 (Illumina, Likely benign), SCV005090110 (Center for Genomic Medicine, Rigshospitalet, Benign), SCV005918745 (Sinai Health System, Benign), SCV002407211 (Labcorp Genetics/Invitae, Benign); all are method = clinical testing, origin germline, trust_tier = single_submitter.
• Aggregate ClinVar label is Benign/Likely benign (3 clinical laboratories Benign, 1 clinical laboratory Likely benign) - the opposite direction from what PP5 requires - and per this case's rule an aggregate label must not be used to trigger PP5 in any case.
• PMID:25741868 (ACMG/AMP 2015) was consulted for the definition of PP5 as a reputable-source assertion of pathogenicity; it contains no variant-level assertion for MLH3 c.3280+14A>T.
• PMID:28492532 (Sherloc) was cited by the Labcorp/Invitae ClinVar submission and was reviewed as the submission's PMID trail; it is a classification-methodology refinement applied to unspecified variants and makes no assertion about MLH3 c.3280+14A>T.
• No numeric threshold applies to PP5: generic ACMG/AMP 2015 (PMID:25741868) defines PP5 qualitatively, and MLH3 has no ClinGen CSPEC and no gene-specific VCEP rule (framework_mode = generic_acmg; vcep_materials contains no PP5 rule), so no comparison operator is available or evaluated for this criterion.

Still missing

• No ClinGen expert-panel / VCEP classification of MLH3 c.3280+14A>T exists in ClinVar; PP5 would need to be re-adjudicated if such an expert-panel assertion is later submitted.
• PP5 here rests on absence of a qualifying expert-panel classification, not on a contradicting expert opinion; the four non-expert submissions lean benign but carry no expert weighting.

Sources

Reasoning

BA1 (stand-alone benign) requires an allele frequency over 5% in a control or general population dataset appropriate to the gene-disease context (ACMG/AMP 2015, PMID:25741868). No MLH3-specific framework exists: ClinGen CSPEC was not found (cspec.found = false), no gene VCEP directory or user-supplied specification is present (gene_vcep_dir null; vcep_materials lists no MLH3 summaries or files), and the case framework resolved to the generic ACMG/AMP 2015 fallback. No specification therefore mandates a non-cancer or exome-only population source, so the all-comers GNOMAD_V2_1 and GNOMAD_V4_1 entries were used as the default source, with the non-cancer subsets read as corroboration only. NM_001040108.2:c.3280+14A>T (NC_000014.8:g.75513065T>A; NC_000014.9:g.75046362T>A) is present in gnomAD v2.1 at total AF 0.000795959 (225/282,678 alleles, 3 homozygotes; exomes 211/251,292, genomes 14/31,386) and in gnomAD v4.1 at total AF 0.000219334 (354/1,613,974 alleles, 2 homozygotes). The highest credible ancestry-specific frequency is East Asian: AF 0.0111791 (223/19,948 alleles, 3 homozygotes) in v2.1 and AF 0.0073975 (332/44,880 alleles, 2 homozygotes) in v4.1, mirrored by the v2.1 non-cancer exome subset (EAS AF 0.0114829, 221/19,246, 3 homozygotes; total non-cancer exome AF 0.000878393, 208/236,796). The corresponding grpmax filtering allele frequencies are 0.0101016 (v2.1) and 0.00674173 (v4.1). Every credible metric is roughly 4.5- to 70-fold below the 5% stand-alone threshold. Threshold artifacts were checked explicitly rather than accepted at face value: gnomAD v3.1 non-cancer reports a 'highest observed frequency' of 5.55556% in HGDP:TUJIA, but that is a single allele in 18 (1/18), a subpopulation point estimate with a binomial confidence interval spanning roughly 0.1-27% and no homozygotes; it is not a usable stand-alone-benign frequency and would in any case be contradicted by the same dataset's grpmax FAF of 0.00648006. The variant is also not absent anywhere it matters: it is present in every gnomAD version surveyed and in the non-cancer (cancer-cohort-excluded) exome subset, so the frequency is not an artifact of oncology-cohort inclusion. Disease context reinforces the direction of the call: MLH3-associated disease is an adult-onset, incompletely penetrant cancer predisposition, so a true stand-alone benign frequency would have to be far above 5%, which is not observed. BA1 is therefore not met, at any strength.

Rule

BA1 stand-alone benign: allele frequency over 5% in a control or general population dataset (ACMG/AMP 2015, PMID:25741868; generic default, no MLH3-specific specification available).

Evidence considered

• gnomAD v2.1 all-comers (GRCh37 14-75513065-T-A): total AC 225 / AN 282,678, total AF 0.0007959586526011929, homozygotes 3; exomes 211/251,292 (AF 0.0008396606338442927, hom 3); genomes 14/31,386 (AF 0.00044605875230994713, hom 0); grpmax FAF 0.010101569999999992 (exome), genome grpmax FAF 0.005439360000000048.
• gnomAD v2.1 ancestry: East Asian 223/19,948 (AF 0.011179065570483256, 3 homozygotes), the only ancestry with observations (AFR 0/24,948; AMR 0/35,436; ASJ 0/10,362); within EAS, KOR 48/3,818 (AF 0.012572, 1 hom) and OEA 161/14,422 (AF 0.011164, 2 hom).
• gnomAD v4.1 all-comers (GRCh38 chr14-75046362-T-A): total AC 354 / AN 1,613,974, total AF 0.0002193343882863045, homozygotes 2; exomes 309/1,461,640 (AF 0.00021140636545250542, hom 2); genomes 45/152,334 (AF 0.0002954035212099728, hom 0); grpmax FAF 0.00674173 (joint; exome 0.006590110000000001, genome 0.006326690000000007); East Asian 332/44,880 (AF 0.007397504456327986, 2 homozygotes), best_subpop = eas.
• gnomAD v2.1 non-cancer exome subset (corroboration only; no specification mandates this source for MLH3): exome AC 208 / AN 236,796 (AF 0.0008783932161016233, hom 3), total including the genome component 222/268,182 (AF 0.0008277960489518312, hom 3); East Asian 221/19,246 (AF 0.01148290553881326, 3 homozygotes); grpmax FAF 0.010396179999999986. The cancer-cohort-excluded subset therefore shows a frequency at least as high as the all-comers total, so the observed frequency is not driven by oncology-cohort inclusion.
• gnomAD v3.1 non-cancer (genomes only): total AC 44 / AN 147,966, AF 0.00029736661101874754, 0 homozygotes, grpmax FAF 0.006480060000000005; the reported 'highest observed frequency' of 0.0555556 (1/18 alleles) is confined to HGDP:TUJIA - a small-sample threshold artifact, not an ancestry-level frequency.
• Threshold applied verbatim from the supplied published calibration block: BA1 stand-alone benign at allele frequency >= 0.05 (ACMG/AMP 2015 stand-alone benign frequency, PMID:25741868), used because no MLH3-specific specification exists; the ACMG/AMP 2015 full text likewise states that a control-population allele frequency 'if over 5% then it is considered as stand-alone support (BA1)'.
• Source selection: no ClinGen CSPEC (cspec found = false), no gene VCEP directory or user specification (gene_vcep_dir null), framework mode generic_acmg, so no non-cancer or exome-only population source is mandated and the all-comers GNOMAD_V2_1 / GNOMAD_V4_1 entries are the default source for BA1.
• gnomAD-Canada v1.0 (HostSeq genomes) query returned search_status 'absent' with error 'timed out' - no data contributed to BA1.
• Literature triage for this case (literature_pass.json / pmid_triage.json) assigned all six ClinVar-linked PMIDs to functional or case-observation themes (PP5/PS4); none supplies MLH3 population-frequency content, and the two fetched full texts (PMID:25741868, PMID:34043773) report no allele-frequency data for this variant.

Still missing

• No MLH3-specific ClinGen CSPEC, gene VCEP specification, or local gene framework exists, so BA1 was evaluated against the generic ACMG/AMP 2015 stand-alone frequency (over 5%) rather than a gene-calibrated stand-alone threshold.
• gnomAD-Canada v1.0 (HostSeq genomes) returned no usable result for this variant (search_status 'absent' with error 'timed out'), so one additional independent genome cohort could not be added to the BA1 frequency estimate.
• The 5.55556% figure reported for HGDP:TUJIA in gnomAD v3.1 non-cancer rests on 1 allele in 18 and cannot be resolved to a credible population estimate (no larger ancestry group in that dataset exceeds grpmax FAF 0.00648006); no dataset reports a frequency at or above 5% on an adequate sample.
• None of these gaps could change the verdict: the largest credible frequency (1.12% East Asian) is still about 4.5-fold below the 5% threshold, and the all-comers totals are 60- to 230-fold below it.

Sources

Reasoning

BS2 requires that the variant be observed in a well-documented healthy adult individual for a disorder in which full penetrance is expected at an early age (recessive homozygous, dominant heterozygous, or X-linked hemizygous) - ACMG/AMP 2015, PMID:25741868. The second and controlling precondition fails for this gene-disease pair. MLH3-associated disease is an adult-onset, incompletely penetrant cancer predisposition: ClinVar conditions recorded for this variant are 'Colorectal cancer' and 'Colorectal cancer, hereditary nonpolyposis, type 7', the gene-level OncoKB entry describes inactivating MLH3 changes in hereditary nonpolyposis colorectal cancer, endometrial and gastric cancers, and the current EHTG/ESCP Lynch syndrome guideline defines the disorder as caused by inherited malfunction in MLH1, MSH2, MSH6 or PMS2, with MLH3 never mentioned. No fully penetrant, early-onset MLH3 disorder is established in the case evidence, so the requirement that full penetrance be expected at an early age is not satisfied. The first precondition is partly, but not decisively, satisfied: the variant is not rare, and homozygotes are observed - 3 homozygotes in gnomAD v2.1 (all in East Asian: EAS 223/19,948, 3 homozygotes), 3 in the v2.1 non-cancer exome subset (208/236,796 alleles; EAS 221/19,246, 3 homozygotes), and 2 in gnomAD v4.1 (EAS 332/44,880, 2 homozygotes) - with a carrier frequency of roughly 1 in 600 individuals in that ancestry. Those homozygotes, however, do not establish 'health' in the required sense: gnomAD participants are not phenotyped for colorectal cancer, the v2.1 non-cancer subset excludes cancer-ascertained cohorts but is not a health screen, and for a late-onset phenotype an adult homozygous carrier may simply be pre-symptomatic, which is precisely why the criterion demands an early-onset, fully penetrant condition. Because the disorder premise is not met, the homozygous observations - while genuine and population-level in nature - cannot be scored as BS2 evidence; they are otherwise accounted for under BS1. BS2 is therefore not met. The call is flagged for human review because a reviewer applying BS2 at reduced strength to gnomAD homozygotes in a cancer-cohort-excluded subset would be recording the same observations, and because no MLH3 penetrance quantification exists in the case evidence to settle the premise quantitatively.

Rule

BS2 strong benign: variant observed in a well-documented healthy adult individual for a recessive (homozygous), dominant (heterozygous), or X-linked (hemizygous) disorder with full penetrance expected at an early age (ACMG/AMP 2015, PMID:25741868).

Evidence considered

• Homozygote observations: gnomAD v2.1 total_hom 3 (all within East Asian, EAS 223/19,948 alleles, 3 homozygotes; EAS_KOR 48/3,818 with 1 homozygote, EAS_OEA 161/14,422 with 2 homozygotes); gnomAD v2.1 non-cancer exomes total_hom 3 (208/236,796 alleles; EAS 221/19,246, 3 homozygotes); gnomAD v4.1 total_hom 2 (EAS 332/44,880, 2 homozygotes). No homozygotes in AFR, AMR, ASJ or the remaining gnomAD v4.1 groups, and none in the gnomAD v3.1 non-cancer genome set.
• Carrier frequency implied by the above: total AF 0.0007959586526011929 in gnomAD v2.1 and 0.0002193343882863045 in gnomAD v4.1, rising to 0.011179065570483256 (East Asian) and 0.01148290553881326 (East Asian, v2.1 non-cancer exomes) - i.e. the variant is a common East Asian polymorphism, not a rare observation.
• Disease premise: the associated conditions are adult-onset cancer predisposition (colorectal cancer; colorectal cancer, hereditary nonpolyposis, type 7). The gene-level OncoKB entry describes MLH3 as a DNA mismatch repair protein with inactivating changes reported in hereditary nonpolyposis colorectal cancer, endometrial and gastric cancers, and states 'MLH3, a DNA mismatch repair protein, is infrequently altered in cancer'; no fully penetrant early-onset phenotype is described.
• EHTG/ESCP (PMID:34043773) defines Lynch syndrome as caused by inherited malfunction in one of the four MMR genes MLH1, MSH2, MSH6 and PMS2; MLH3 is never mentioned, so the disorder in whose context this variant would be judged is not established as an MLH3-caused, fully penetrant, early-onset disease.
• Threshold/premise applied from the supplied published calibration block and the ACMG/AMP 2015 BS2 definition (PMID:25741868, carried by the registry entry for generic ACMG/AMP combination rules): BS2 requires a well-documented healthy adult observation for a recessive (homozygous), dominant (heterozygous), or X-linked (hemizygous) disorder with full penetrance expected at an early age. ACMG/AMP 2015 full text additionally notes that population data are most useful when the Mendelian disease under study is early-onset, which this phenotype is not.
• gnomAD is not a phenotyped healthy-adult screen for colorectal cancer; the v2.1 non-cancer subset excludes cancer-ascertained cohorts but does not document disease-free status at the relevant ages, so the health precondition cannot be positively established even for the observed homozygotes.
• gnomAD-Canada v1.0 (HostSeq genomes) query returned search_status 'absent' with error 'timed out' - no additional homozygous-carrier data available.
• Literature triage for this case assigned all six ClinVar-linked PMIDs to functional or case-observation themes (PP5/PS4); none reports MLH3 carrier, homozygote or penetrance data bearing on BS2, and the only fetched papers (PMID:25741868, PMID:34043773) report no MLH3 content at all.

Still missing

• No MLH3-specific CSPEC, gene VCEP specification, or local framework exists, so BS2 was assessed against the generic ACMG/AMP 2015 definition (PMID:25741868), which requires a disorder with full penetrance expected at an early age; no MLH3-specific penetrance or age-of-onset quantification is present in the case evidence.
• Homozygous carriers are documented (3 in gnomAD v2.1 East Asians, 3 in the v2.1 non-cancer exome subset, 2 in gnomAD v4.1 East Asians) but no age, phenotype, or colorectal-cancer status is available for them, and gnomAD is not a phenotyped healthy-adult screen for this late-onset disease; the health precondition of BS2 therefore cannot be positively established even though the observation exists.
• gnomAD-Canada v1.0 (HostSeq genomes) returned no usable data (lookup timed out), so no additional cohort could be screened for homozygous carriers.
• The verdict turns on the disease-premise test (adult-onset, incompletely penetrant cancer predisposition versus the fully penetrant early-onset disorder BS2 requires), not on a missing datum; a gene-specific penetrance estimate would be needed to overturn it.

Sources

Reasoning

BP6 requires a reputable independent source - under this case's rule, specifically an exact-variant ClinGen expert-panel / VCEP classification - to report this variant as Benign or Likely benign; ordinary laboratory submissions, aggregate ClinVar labels and non-expert assertions must never trigger BP6. ClinVar variation 314378 covers the exact variant NM_001040108.2(MLH3):c.3280+14A>T and has zero expert-panel submissions: the API record's expert_panels object is empty, the submission audit sets is_expert_panel = false for all four submissions (SCV000388764 Illumina - Likely benign; SCV005090110 Rigshospitalet - Benign; SCV005918745 Sinai Health System - Benign; SCV002407211 Labcorp/Invitae - Benign) and counts expert_panel_submissions = 0, and every submission's review status is 'criteria provided, single submitter', aggregating to 'criteria provided, multiple submitters, no conflicts' (2 stars, not expert-panel reviewed). The aggregate Benign/Likely benign label is therefore exactly the kind of non-expert, aggregated assertion that must not trigger BP6, and no exact-variant expert-panel benign classification exists to cite. Independently, no MLH3 gene-specific specification exists (no ClinGen CSPEC, framework_mode = generic_acmg, no BP6 rule or lookup table in vcep_materials), so no gene-level BP6 authorization is available either. The PubMed trail cited by these submitters (PMID:25741868, PMID:28492532) makes no assertion about this variant, and none of the case-observation themed references reports a benign classification of c.3280+14A>T by an expert panel. BP6 is not met on the absence of a qualifying expert-panel source; this is a definitive negative evaluation of the available classification sources, not a data-availability gap.

Rule

BP6 not met: no exact-variant ClinGen expert-panel (VCEP) Benign/Likely benign classification exists for c.3280+14A>T (expert_panel_submissions = 0; the four benign-leaning submissions are non-expert laboratory assertions and the aggregate 2-star label does not qualify).

Evidence considered

• ClinVar variation 314378: expert_panels = {} (empty), 4 submissions, review status 'criteria provided, multiple submitters, no conflicts' (review_stars = 2); the record contains no 3-star expert-panel review.
• ClinVar submission audit: expert_panel_submissions = 0, is_expert_panel = false for all four submissions, all trust_tier = single_submitter, all method = clinical testing, origin germline, review status 'criteria provided, single submitter'.
• Submitter classifications: 3 clinical laboratories Benign and 1 clinical laboratory Likely benign (SCV000388764 Illumina Likely benign; SCV005090110 Rigshospitalet Benign; SCV005918745 Sinai Health System Benign; SCV002407211 Labcorp Genetics/Invitae Benign). These are ordinary laboratory assertions and per this case's rule cannot trigger BP6, individually or as an aggregate label.
• PMID:25741868 (ACMG/AMP 2015) was consulted for the definition of BP6; it contains no variant-level assertion for this variant.
• PMID:28492532 (Sherloc) was reviewed as the PMID trail of the Labcorp/Invitae submission; it is a classification-methodology paper and makes no assertion about MLH3 c.3280+14A>T.
• No ClinGen CSPEC and no gene-specific VCEP framework exists for MLH3 (framework_mode = generic_acmg; vcep_materials.files_found empty, domain_tables empty), so there is no gene-level BP6 authorization and the generic ACMG/AMP 2015 expert-source requirement governs. No numeric threshold applies to BP6.

Still missing

• No ClinGen expert-panel / VCEP classification of MLH3 c.3280+14A>T exists in ClinVar; BP6 would need re-adjudication if an expert panel later submits a benign assertion for this variant.

Sources

Reasoning

PS2 requires a confirmed de novo occurrence of the variant in a proband with the disease, with maternity and paternity confirmed by parental testing. This case is a variant-level evaluation (MLH3 NM_001040108.2:c.3280+14A>T, NC_000014.9:g.75046362T>A, a deep intronic substitution) and contains no proband record, no parental/trio genotype results, and no pedigree of any kind. The only variant-level clinical source, ClinVar variation 314378, holds four germline clinical-testing submissions (three Benign, one Likely benign) with no reported familial, de novo, or segregation information and no submitted criteria (submission_audit evidence_signals empty for all four submissions; no expert-panel submission). The two full-text papers retrieved for this case, the ACMG/AMP 2015 standards (PMID:25741868) and the EHTG/ESCP Lynch syndrome guideline (PMID:34043773), contain no MLH3 variant-level case data and do not mention this variant; PMID:34043773 never mentions MLH3 at all. No ClinGen CSPEC and no VCEP specification exist for MLH3, so the generic ACMG/AMP 2015 rule governs. Because no proband and no parental genotypes are available, a de novo occurrence can be neither established nor excluded, and PS2 cannot be evaluated at any strength.

Rule

PS2 (Pathogenic, Strong) - De novo (both maternity and paternity confirmed) in a patient with the disease and no family history (Richards et al. 2015, PMID:25741868). Generic ACMG/AMP applies: no MLH3 ClinGen CSPEC or VCEP specification is available.

Evidence considered

• ClinVar variation 314378 (NM_001040108.2(MLH3):c.3280+14A>T) reports four germline clinical-testing submissions - three Benign and one Likely benign - each 'criteria provided, single submitter', with zero expert-panel submissions; no submission reports a de novo occurrence, parental testing, or family structure.
• Per-submission audit of ClinVar 314378 shows empty evidence_signals (criteria_found and signals both empty) for all four submissions (SCV000388764 Illumina; SCV005090110 Rigshospitalet; SCV005918745 Sinai Health System; SCV002407211 Labcorp/Invitae), so no submitter-reported PS2/PM6/PP1/BS4 reasoning is recoverable.
• The only validated PMID trails in those ClinVar submissions point to the ACMG/AMP 2015 standards (PMID:25741868) and Sherloc (PMID:28492532) - both general classification frameworks, neither containing variant-level or family data.
• Full-text check of PMID:25741868: no MLH3 content, no variant-level data, no segregation or de novo observations; it supplies only the PS2 definition (de novo in a patient with the disease, both maternity and paternity confirmed).
• Full-text check of PMID:34043773 (EHTG/ESCP Lynch syndrome guideline): MLH3 is never mentioned and no case, segregation, or de novo data applicable to this variant is reported.
• No ClinGen CSPEC was found for MLH3 (cspec lookup found = false), no gene VCEP directory exists, and the classification framework resolved to the generic ACMG/AMP 2015 fallback, so the generic PS2 rule applies.
• Literature_pass for this case indexed PMIDs 20301390, 26389258, 26389505, 34043773 only against PP5/PS4 and contains no segregation or de novo candidate evidence for any of them.

Still missing

• No proband-level genotype and no parental (trio) testing results are present in any case source; a confirmed de novo occurrence requires proband plus both-parent genotypes with maternity/paternity confirmed.
• No pedigree or family-history record is available to establish disease status of the proband or absence of family history.
• No ClinGen CSPEC or VCEP specification exists for MLH3, so no gene-specific PS2 modification (e.g. downgrading) could be applied.

Sources

Reasoning

Framework check first: there is no ClinGen CSPEC for MLH3 (cspec.found = false), no gene-specific VCEP material (gene_vcep_dir null, vcep_materials.files_found empty, no PP3/BP4/PS3 lookup spreadsheet, no index guidance) and no local custom framework, so this case is adjudicated under generic ACMG/AMP 2015 (framework_mode = generic_acmg, final_classification_framework.source = generic_acmg_fallback, precedence_rank 3). Under generic ACMG/AMP, PS3 is met only when a well-established functional study, i.e. a validated assay relevant to the gene's disease mechanism performed on this variant (or an equivalent assayed variant), demonstrates a damaging effect; no gene-specific strength calibration exists here to modify the default strong weighting. Variant scope: NM_001040108.2:c.3280+14A>T (NC_000014.8:g.75513065T>A) is 14 nucleotides into the intron 3' of the exon ending at c.3280, i.e. a deep-intronic, non-coding, non-canonical-splice-site change with protein consequence NP_001035197.1:p.? Any functional evidence capable of supporting PS3 for this variant would therefore have to be an RNA-based splice assay (RT-PCR/fragment analysis of patient or minigene transcripts) or a downstream MMR-activity/MSI/IHC assay; no such assay result exists in the evidence base. ClinVar variation 314378 (exact transcript-HGVS match, 4 submissions, 0 expert-panel) carries no assay data: the three criterion-lead submissions cite only the generic classification frameworks (PMID:25741868 ACMG/AMP 2015 for SCV005090110 and SCV005918745; PMID:28492532 Sherloc for SCV002407211), and the submission audit recovered no criteria-code signals and no functional/laboratory evidence. The only paper fetched under the functional theme is PMID:25741868 (ACMG/AMP 2015) which, as extracted, reports no MLH3 data and names no MLH3 variant; its single relevant statement is that splice-site/splice-region effects of sequence variants are predicted rather than proven and that confirmation of impact requires RNA or protein functional analysis. The other fetched full text, PMID:34043773 (EHTG/ESCP Mallorca guideline), contains no MLH3 content at all - MLH3 is never mentioned - and no intronic-variant assay data. The gene-level MLH3 literature assembled for the loss-of-function mechanism review (PMID:29568967 on an MSH2 splice variant, PMID:30573798 on a biallelic MLH3 nonsense variant, PMID:31822864 on deep-intronic sequencing across MLH3 and related MMR genes, PMID:36833239 on MSI and aberrant pre-mRNA splicing, PMID:37427096 on renal cell carcinoma genomics) was screened and none of these reports a functional assay of c.3280+14A>T; they are gene/mechanism context, not assay evidence for this variant, and they mention the variant nowhere. Finally, the only variant-specific functional-adjacent result in the case is the in-silico SpliceAI prediction (max delta 0.005, donor gain at +14 delta 0.002), which is a computational prediction and is expressly not a functional assay; SpliceAI is also scored separately under PP3/BP4 and is not assay evidence capable of satisfying PS3 (or BS3). Because no validated functional assay exists for this variant, the criterion cannot be evaluated in either direction: PS3 is therefore not_assessed rather than not_met, and the outstanding gap is the absence of an RNA/MMR functional assay, not an absence of a damaging assay result.

Rule

PS3 not assessed: PS3 requires a well-established functional study showing a damaging effect (ACMG/AMP 2015, PMID:25741868), and no validated functional assay (RNA splice assay or MMR activity assay) has been performed or reported for MLH3 c.3280+14A>T.

Evidence considered

• ClinVar variation 314378 matches NM_001040108.2:c.3280+14A>T exactly (transcript HGVS); 4 submissions, 0 expert-panel submissions. No submission reports a functional assay, RNA analysis, MMR activity assay, MSI or IHC result; the submission audit found no criteria-code signals. The three criterion-lead submissions cite only PMID:25741868 (ACMG/AMP 2015, SCV005090110 and SCV005918745) and PMID:28492532 (Sherloc, SCV002407211) as classification frameworks. Recorded only as the absence of functional evidence, not as an inference from ClinVar's classification.
• PMID:25741868 (ACMG/AMP 2015, Richards et al.) is the only source fetched under the functional theme and reports no MLH3 data and no MLH3 variant. It defines PS3 as evidence from a well-established functional study showing a damaging effect, and states that splice-altering effects of sequence variants are predicted rather than proven, with confirmation of impact requiring functional analysis by RNA or protein analysis: 'For splice site variants, the variant may lead to exon skipping, shortening or inclusion of intronic material, due to alternative donor/acceptor site usage or creation of new sites. Although splice site variants are predicted to lead to a null effect, confirmation of impact requires functional analysis by either RNA or protein analysis.'
• PMID:34043773 (EHTG/ESCP Mallorca guideline) was reviewed and contains no MLH3 content and no functional assay data; MLH3 is never mentioned, and all recommendations concern MLH1, MSH2, MSH6 and PMS2.
• Variant scope: c.3280+14A>T lies 14 nt into the intron 3' of the exon ending at c.3280, outside the canonical +/-1,2 donor/acceptor dinucleotides; protein consequence is NP_001035197.1:p.? Any PS3-capable evidence would therefore have to be an RNA splice assay or a downstream MMR functional assay, and neither exists.
• Gene/mechanism literature screened during the loss-of-function review (PMID:29568967, PMID:30573798, PMID:31822864, PMID:36833239, PMID:37427096) contains gene-level MLH3 and MMR-splicing context but no functional assay of c.3280+14A>T; the variant is not mentioned in any of them, so none can support or refute PS3. These PMIDs are not case citation keys and are recorded here only to document that the negative search was performed.
• In-silico result present in the case but not usable as functional evidence: SpliceAI max delta 0.005 (DS_AG 0.005, DS_AL 0.000, DS_DG 0.002, DS_DL 0.003; best predicted effect a donor gain at +14 with delta 0.002). It is a computational prediction, is adjudicated under PP3/BP4, and cannot satisfy PS3, which requires an experimental assay.
• No ClinGen CSPEC was found for MLH3 (cspec.found = false), no gene-specific VCEP directory or files exist (gene_vcep_dir null, vcep_materials.files_found empty and no PP3/BP4 pre-assignment table), and framework_mode is generic_acmg; therefore no gene-specific PS3 assay specification or strength calibration overrides the generic ACMG/AMP 2015 requirement.

Still missing

• No RNA-based splice assay (RT-PCR transcript analysis or minigene) and no MMR-activity/MSI/IHC assay has been reported for MLH3 c.3280+14A>T; PS3 requires such a validated functional study and none exists, so the criterion cannot be evaluated.
• The ClinVar record (variation 314378) has no expert-panel submission and no functional/laboratory evidence in any submission; the criterion-lead submissions cite only generic classification frameworks (PMID:25741868, PMID:28492532).
• MLH3 has no ClinGen CSPEC or gene-specific VCEP framework, so there is no gene-specific PS3 assay list or strength calibration (e.g. a pre-specified validated splice assay or MMR functional assay set) to which this variant could be referred.

Sources

Reasoning

No MLH3-specific VCEP, ClinGen CSPEC, or local gene framework exists for this gene (vcep_materials.json: no CSPEC URL, no framework files, scoring model generic ACMG/AMP 2015), so the generic ACMG/AMP PM3 definition was applied: evidence that the variant is detected in trans with a pathogenic variant in an individual with a recessive disorder (Richards et al. 2015, PMID:25741868). MLH3 does have a documented recessive disease mechanism, so PM3 is applicable in principle: a biallelic germline MLH3 loss-of-function variant has been reported to underlie polyposis predisposition (Olkinuora et al., Genet Med 2019, PMID:30573798), alongside the disputed dominant HNPCC7 association (OMIM 604395, NCBI Gene 27030). The case, however, is a purely variant-level record: the evidence set documents NM_001040108.2:c.3280+14A>T (ClinVar VCV000314378; gnomAD v2.1 225/282,678 alleles, 3 homozygotes, East Asian AF 1.12%; COSMIC COSV53158903), with no affected-proband genotype, no second MLH3 variant, no clinical history, no segregation and no cis/trans phase determination anywhere in the bundle. The four ClinVar submissions (three Benign, one Likely benign, all single submitters, none an expert panel) list the variant but supply no criteria or variant-level allelic content, and the reviewed full texts (PMID:25741868, PMID:34043773) contain no MLH3 allelic data. Population homozygosity in gnomAD reflects unphased population genotypes and cannot establish whether this allele sits in trans with a pathogenic MLH3 variant in an affected individual. PM3 therefore cannot be assessed and no strength or points can be assigned.

Rule

ACMG/AMP 2015 PM3: for recessive disorders, the variant is detected in trans with a pathogenic variant (Richards et al. 2015, PMID:25741868); no gene-specific MLH3 specification exists to modify this criterion.

Evidence considered

• No MLH3-specific VCEP/CSPEC or local framework is available (vcep_materials.json carries no CSPEC URL, no framework files, no domain tables), and the framework artefact records a generic ACMG/AMP 2015 fallback, so the generic PM3 definition was used.
• Generic ACMG/AMP PM3 (PMID:25741868): 'For recessive disorders, detected in trans with a pathogenic variant'; this requires an affected individual in whom the variant is phased in trans with a pathogenic/likely pathogenic allele.
• MLH3 has a documented recessive mechanism relevant to PM3 eligibility: Olkinuora et al. (Genet Med 2019, PMID:30573798) report that a biallelic germline nonsense variant of MLH3 underlies polyposis predisposition, identified by exome sequencing of 40 unexplained FAP/AFAP cases; the paper does not mention c.3280+14A>T.
• The case bundle contains no proband, pedigree, second MLH3 variant, segregation result, or phase information for NM_001040108.2:c.3280+14A>T; the evidence is limited to ClinVar (VCV000314378), gnomAD v2.1/v4, SpliceAI, COSMIC and general Lynch syndrome literature.
• The variant's ClinVar record (variation 314378) has four germline submissions - three Benign and one Likely benign, all 'criteria provided, single submitter', none from an expert panel - with no criterion-level assertions captured; these are classification statements only and were not used to infer PM3.
• gnomAD v2.1 reports 225/282,678 alleles (AF 0.0796%), including 3 homozygotes in the East Asian subset (AF 1.12%); these are unphased population genotypes and do not demonstrate an in-trans configuration with a pathogenic MLH3 allele.
• The two full texts available for this case (PMID:25741868 ACMG/AMP standards; PMID:34043773 EHTG/ESCP Lynch syndrome guideline) report no MLH3 allelic, segregation or case-observation data.

Still missing

• No affected-proband clinical record or genotype is documented for this variant.
• No second pathogenic or likely pathogenic MLH3 allele is documented in the same individual.
• No cis/trans phase determination (family study, long-read phasing, or allele-specific testing) is available.
• The relevant phenotype/inheritance mode for this specific case is not documented (dominant HNPCC7 association vs the recessive biallelic polyposis mechanism, PMID:30573798).
• ClinVar submissions provide no criterion-level content (no criteria statements captured for any of the four submitters, no expert-panel submission).

Sources

Reasoning

PM6 requires an assumed de novo occurrence in a patient with the disease, without confirmation by parental testing, and therefore still requires a proband-level observation of the variant with a clinical phenotype. This case provides no proband, no phenotype record, and no family information. ClinVar 314378 is the only per-variant clinical source and is entirely germline clinical-testing submissions without any de novo statement, phenotype attribution to the variant, or family data; the case-summary phenotype context is limited to generic condition labels (Colorectal cancer; Colorectal cancer, hereditary nonpolyposis, type 7; not specified) attached to the ClinVar record, with no individual-level clinical description. Neither full-text paper (PMID:25741868, PMID:34043773) reports this variant or any MLH3 case observation. Since there is no proband in whom the variant could be assumed de novo, PM6 cannot be met, and its absence of evidence is not evidence against the variant either - it is simply not evaluable.

Rule

PM6 (Pathogenic, Moderate) - Assumed de novo, but without confirmation of paternity and maternity (Richards et al. 2015, PMID:25741868). Generic ACMG/AMP applies: no MLH3 ClinGen CSPEC or VCEP specification is available.

Evidence considered

• ClinVar 314378 carries only germline clinical-testing submissions; none asserts that the variant arose de novo, and none provides parental genotype information, so there is no basis even for an unconfirmed de novo claim.
• Condition labels on the ClinVar record are generic (Colorectal cancer; Colorectal cancer, hereditary nonpolyposis, type 7; not specified) and are not linked to a described proband or to this variant's causality, so phenotype consistency cannot be judged.
• PMID:25741868 defines PM6 as an assumed de novo occurrence without confirmation of parental testing, but supplies no case or variant data for MLH3.
• PMID:34043773 contains no MLH3-specific content and no case observations, so it cannot support an assumed de novo claim for this variant.
• No proband, family, or clinical case file exists in this case directory; the case is a variant-level evaluation with literature and database sources only.

Still missing

• No proband observation of this variant is documented in any source, so the prerequisite de novo assumption cannot be made.
• No individual-level phenotype record is available to support the requirement that the proband has the disease.
• No parental testing status is recorded, so it cannot even be determined whether this is an unsupported de novo claim that would fall under PM6.
• No MLH3 CSPEC/VCEP exists to define a gene-specific PM6 strength or phenotype-specificity requirement.

Sources

Reasoning

PP1 requires co-segregation of the variant with disease across multiple affected family members. No pedigree, no family members, and no genotype-phenotype correlation data exist anywhere in this case. The only per-variant clinical source (ClinVar 314378) records four germline clinical-testing submissions with no family information and no submitter-reported criteria, and the two full-text papers retrieved (PMID:25741868 and PMID:34043773) report no MLH3 segregation data - the ACMG framework paper is methodological only, and the Lynch syndrome guideline never mentions MLH3 and reports no segregation for this variant. Because there are zero meioses available to count, co-segregation cannot be demonstrated and no PP1 strength (supporting, moderate, or strong) can be assigned.

Rule

PP1 (Pathogenic, Supporting) - Co-segregation with disease in multiple affected family members in a gene definitively known to cause the disease; strength may be increased with more meioses (Richards et al. 2015, PMID:25741868). Generic ACMG/AMP applies: no MLH3 ClinGen CSPEC or VCEP specification is available.

Evidence considered

• ClinVar 314378 contains four germline clinical-testing submissions (three Benign, one Likely benign) and no family, pedigree, or segregation statement; submitters provided criteria fields that are empty of signal.
• PMID:25741868 defines PP1 as co-segregation with disease in multiple affected family members but reports no MLH3 or variant-level segregation data.
• Full-text check of PMID:34043773 confirms MLH3 is not mentioned anywhere in the guideline and that no segregation or family data relevant to this variant is present.
• The case directory contains no proband, family, or pedigree file, and literature triage produced no segregation-themed papers with this variant mentioned.
• No ClinGen CSPEC is available for MLH3 and the framework resolved to generic ACMG/AMP 2015, so no gene-specific segregation framework exists to substitute.

Still missing

• No pedigree or family structure is recorded for this variant, so the number of informative meioses is zero and co-segregation cannot be assessed.
• No genotype data for affected or unaffected relatives are available.
• The disease phenotype to be co-segregated is not defined for any individual carrying this variant.
• No MLH3 CSPEC/VCEP exists, so no gene-specific PP1 meiosis thresholds or strength-modification rules could be applied.

Sources

Reasoning

PP4 requires that the individual's (proband's) phenotype or family history be highly specific for a disease with a single genetic etiology, and it is scored against that individual's clinical presentation. This case contains no proband-level clinical information: no indication for testing, no HPO-style phenotype terms, no age of onset, no tumour/MSI/IHC data and no family history for any individual carrying c.3280+14A>T. The only disease context available is the set of submitter-supplied ClinVar condition labels ('Colorectal cancer', 'Colorectal cancer, hereditary nonpolyposis, type 7', 'not specified'), which record the condition a laboratory attached to the variant record; they do not establish phenotype specificity for any tested individual and one submission explicitly labels the condition 'not specified'. Nor can phenotype specificity be imported from the gene: the MLH3 disease association rests on hereditary nonpolyposis colorectal cancer type 7, while the Lynch syndrome guideline in this case (PMID:34043773) defines the phenotype by MLH1/MSH2/MSH6/PMS2 and never mentions MLH3, and no retrieved paper names this variant. Because PP4's input data (a specific patient's phenotype/family history) is genuinely absent from this case rather than affirmatively inconsistent with a specific disease, the criterion is not assessed rather than not met, and no phenotype-specificity claim is made either way. Gap: proband phenotype and family history were not provided for this variant-level review; PP4 can be adjudicated if a phenotype/FHx is supplied.

Evidence considered

• No proband-level clinical data (phenotype, age of onset, tumour testing, or family history) exists anywhere in this case's assembled evidence; the case is a variant-level review with ClinVar laboratory conditions attached to the record.
• ClinVar variation 314378 conditions are 'Colorectal cancer', 'Colorectal cancer, hereditary nonpolyposis, type 7' and 'not specified' - submitter-supplied condition labels, with one submission leaving the condition unspecified; they do not demonstrate a highly specific single-gene phenotype in an individual.
• PP4 is defined in ACMG/AMP 2015 (PMID:25741868) as the patient's phenotype or family history being highly specific for a disease with a single genetic etiology; without patient-level data this input is unavailable.
• PMID:34043773 (EHTG/ESCP Mallorca guideline) full text was read: it restricts the Lynch phenotype to MLH1/MSH2/MSH6/PMS2 and never mentions MLH3, so no MLH3-specific phenotype definition can be taken from the single guideline reference available here.
• No ClinGen CSPEC and no gene-specific VCEP framework exists for MLH3 (framework_mode = generic_acmg), so PP4 is evaluated under generic ACMG/AMP 2015 with no gene-specific phenotype-specificity rule.

Still missing

• No proband-level phenotype, indication for testing, tumour/MSI/IHC data or family history is present in this case, so PP4's phenotype-specificity test cannot be applied; supply the proband's clinical phenotype and family history to adjudicate PP4.
• The ClinVar condition labels for variation 314378 are submitter-supplied disease links (one of them 'not specified') and are not a proband phenotype.

Sources

Reasoning

Framework check first: no ClinGen CSPEC and no gene-specific VCEP or local framework exist for MLH3 (cspec.found = false; gene_vcep_dir null; vcep_materials.files_found empty; framework_mode = generic_acmg with generic_acmg_fallback precedence_rank 3), so BS3 is adjudicated under generic ACMG/AMP 2015. BS3 requires well-established functional studies showing no damaging effect - i.e. a validated assay performed on this variant (or equivalent) that demonstrates normal function, with documented controls and assays relevant to the gene's disease mechanism; PS3/BS3 strength calibration cannot be assigned without such an assay. Variant scope: NM_001040108.2:c.3280+14A>T (NC_000014.8:g.75513065T>A) is a deep-intronic change 14 nt into the intron 3' of the exon ending at c.3280, protein consequence NP_001035197.1:p.?; for a variant of this class the assay that could support BS3 would be an RNA-based splice assay (RT-PCR/fragment analysis or minigene showing normal transcript processing) or a downstream MMR functional readout, and no such assay result exists for this variant anywhere in the evidence base. The ClinVar record (variation 314378, exact transcript-HGVS match, 4 submissions, no expert panel) contains no functional evidence and no criteria-code signals; the criterion-lead submissions cite only generic frameworks (PMID:25741868, PMID:28492532). The two fetched full texts were reviewed: PMID:25741868 (ACMG/AMP 2015) reports no MLH3 data and names no MLH3 variant, and PMID:34043773 (EHTG/ESCP Mallorca guideline) never mentions MLH3 and contains no intronic-variant functional data. Gene-level MLH3/MMR papers screened for the loss-of-function review (PMID:29568967, PMID:30573798, PMID:31822864, PMID:36833239, PMID:37427096) report no assay of this variant. Two lines of indirect evidence point toward a benign functional consequence but neither is a functional study and neither may be used to satisfy BS3: (1) SpliceAI max delta 0.005 with all component deltas <= 0.005 and the canonical donor site intact, i.e. no predicted splice impact - a computational prediction only, adjudicated under BP4; (2) the variant's population profile (gnomAD v2.1 total AF 0.000796, East Asian AF 0.0112 with 3 homozygotes) - population data, adjudicated under BA1/BS1. Reading either of these as BS3 would substitute prediction or frequency for experimental evidence and is expressly not done. BS3 therefore cannot be evaluated: it is not_assessed, with the missing validated functional assay (and the absence of any MLH3-specific BS3 assay specification) as the identified gap.

Rule

BS3 not assessed: BS3 requires well-established functional studies showing no damaging effect (ACMG/AMP 2015, PMID:25741868), and no validated functional assay has been performed or reported for MLH3 c.3280+14A>T.

Evidence considered

• PMID:25741868 (ACMG/AMP 2015, Richards et al.) defines BS3 as evidence from well-established functional studies showing no damaging effect and states that splice-altering effects are predicted rather than proven, with confirmation requiring functional analysis by RNA or protein analysis: 'Although splice site variants are predicted to lead to a null effect, confirmation of impact requires functional analysis by either RNA or protein analysis.' It contains no MLH3 data and names no MLH3 variant.
• ClinVar variation 314378 (exact transcript-HGVS match, 4 submissions, 0 expert-panel submissions) reports no functional assay, no RNA analysis, no MMR activity assay, no MSI and no IHC data; the submission audit recovered no criteria-code signals. The criterion-lead submissions cite only generic classification frameworks (PMID:25741868 for SCV005090110 and SCV005918745; PMID:28492532 for SCV002407211). Recorded as the absence of functional evidence; no criterion is inferred from ClinVar's classification.
• PMID:34043773 (EHTG/ESCP Mallorca guideline) was reviewed and never mentions MLH3 and contains no functional assay of any intronic MMR variant.
• SpliceAI result used only as qualitative context and expressly NOT as BS3 evidence: max delta 0.005 (DS_AG 0.005, DS_AL 0.000, DS_DG 0.002, DS_DL 0.003; only predicted effect a donor gain at +14, delta 0.002; canonical donor site not disrupted). It is a computational prediction whose calibrated use is PP3/BP4, and a prediction of no splice impact is not a functional study showing no damaging effect.
• Population data (gnomAD v2.1: total AF 0.000796, 225/282678 alleles, 3 homozygotes; East Asian AF 0.0112, grpmax FAF 0.0101) is frequency evidence, not functional evidence, and cannot be used to meet BS3; it is handled under BA1/BS1 in the population group.
• Gene/mechanism literature screened during the loss-of-function review (PMID:29568967, PMID:30573798, PMID:31822864, PMID:36833239, PMID:37427096) reports no functional assay of c.3280+14A>T; the variant is not mentioned in any of them. These PMIDs are not case citation keys and are listed only to document the negative search.
• No ClinGen CSPEC for MLH3 (cspec.found = false), no gene-specific VCEP directory or files (gene_vcep_dir null, vcep_materials.files_found empty), framework_mode = generic_acmg; there is no MLH3-specific BS3 assay list or strength calibration to apply, so the generic ACMG/AMP 2015 requirement for a well-established functional study governs.

Still missing

• No validated functional study demonstrating normal function (RNA-based splice assay or MMR activity/MSI/IHC readout) has been performed for MLH3 c.3280+14A>T, so BS3 cannot be evaluated despite the benign-leaning prediction and population data.
• MLH3 has no ClinGen CSPEC or gene-specific VCEP framework defining which functional assays are considered well-established for BS3 or how BS3 strength should be calibrated in this gene.
• Indirect evidence that would need experimental confirmation: SpliceAI predicts no splice impact (max delta 0.005) and the variant is common in East Asians (AF 0.0112, 3 homozygotes), but no assay has verified normal transcript processing or MMR function.

Sources

Reasoning

BS4 requires lack of segregation in affected members of a family - that is, at least one affected family member who does not carry the variant. This is a family-level observation and no family data of any kind exists in this case. ClinVar 314378, the only per-variant clinical source, reports only germline clinical-testing submissions with no family or segregation content and no submitter criteria; the retrieved full texts (PMID:25741868, PMID:34043773) contain no MLH3 family data. The variant's population-level benign signal (gnomAD grpmax FAF 0.0101 with three homozygotes in gnomAD v2.1; popmax 0.0112 in East Asian) is relevant to BS1/BA1, not to BS4, which is defined by within-family non-segregation and cannot be inferred from allele frequency or from other submitters' final Benign classifications. With no affected relatives genotyped, non-segregation can be neither demonstrated nor refuted, so BS4 is not evaluable.

Rule

BS4 (Benign, Strong) - Lack of segregation in affected members of a family (Richards et al. 2015, PMID:25741868). Generic ACMG/AMP applies: no MLH3 ClinGen CSPEC or VCEP specification is available.

Evidence considered

• No proband or family file exists in this case, and no source in the bundle reports a genotyped affected relative, so the BS4 requirement of an affected family member lacking the variant is unmet for lack of data rather than by negative observation.
• ClinVar 314378 documents only germline clinical-testing submissions with no segregation or family statement; its Benign/Likely benign submissions were not used as evidence of non-segregation.
• gnomAD v2.1 reports the variant at grpmax FAF 0.0101 (225/282678 alleles, 3 homozygotes; East Asian AF 0.0112, 3 homozygotes) and gnomAD v4.1 at grpmax FAF 0.00674 (354/1613974 alleles, 2 homozygotes) - these are population-frequency observations relevant to BS1/BA1 and do not address BS4.
• Full-text check of PMID:25741868 confirms it is the methodological ACMG/AMP framework with no MLH3 or segregation data; PMID:34043773 never mentions MLH3.
• No ClinGen CSPEC for MLH3 and no VCEP material were available, so generic ACMG/AMP 2015 governs and no gene-specific BS4 rule exists.

Still missing

• No family genotype data are available, so the presence of one or more affected relatives lacking the variant cannot be established.
• No pedigree exists that would allow non-segregation to be scored or strength-weighted.
• The variant's elevated population frequency (gnomAD grpmax FAF 0.0101; East Asian 0.0112) supports BS1/BA1 considerations only and cannot substitute for BS4, which requires within-family non-segregation.
• No MLH3 CSPEC/VCEP exists to define gene-specific BS4 handling.

Sources

Reasoning

No MLH3-specific VCEP, ClinGen CSPEC, or local gene framework exists (vcep_materials.json has no CSPEC URL and no framework files; generic ACMG/AMP 2015 fallback recorded), so the generic ACMG/AMP BP2 definition 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 (Richards et al. 2015, PMID:25741868). Both clauses require allelic observation against a known pathogenic comparator allele, and the case contains no such comparator: the bundle documents only the single variant NM_001040108.2:c.3280+14A>T with ClinVar records (variation 314378; four germline submissions, three Benign and one Likely benign, all single submitters, none expert panel, no criterion content captured), gnomAD frequencies (v2.1 225/282,678 alleles; 3 homozygotes; East Asian AF 1.12%; grpmax FAF 0.0101), COSMIC COSV53158903, and general Lynch syndrome literature. No second MLH3/PMS2/MLH1-class comparator variant, no proband genotype, no pedigree, no segregation and no cis/trans phase result exist, so neither the trans clause nor the cis clause can be evaluated. The first clause is additionally not established for this gene: MLH3 is not one of the four genes (MLH1, MSH2, MSH6, PMS2) with an accepted fully penetrant dominant mechanism in the EHTG/ESCP Lynch syndrome guideline (PMID:34043773), and its gene-disease relationship remains disputed, with a recessive biallelic polyposis mechanism reported (PMID:30573798). BP2 therefore cannot be assessed and no strength can be assigned. The variant's ClinVar benign classification was not used to infer this criterion.

Rule

ACMG/AMP 2015 BP2: 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); no gene-specific MLH3 specification exists to modify this criterion.

Evidence considered

• No MLH3-specific VCEP/CSPEC or local framework is available (vcep_materials.json: no CSPEC URL, no framework files), and the framework artefact records a generic ACMG/AMP 2015 fallback, so the generic BP2 definition was used.
• Generic ACMG/AMP BP2 (PMID:25741868): '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'; both clauses require a known pathogenic comparator allele plus phase.
• The case bundle contains no pathogenic comparator variant, no proband or pedigree genotype, no segregation result, and no cis/trans phase information for NM_001040108.2:c.3280+14A>T.
• ClinVar variation 314378 records four germline submissions (three Benign, one Likely benign; all 'criteria provided, single submitter'; no expert-panel submission) with no criterion-level assertions captured, and the gnomAD data (v2.1: 225/282,678 alleles, 3 homozygotes, East Asian AF 1.12%, grpmax FAF 0.0101) are unphased population genotypes - neither source documents cis or trans configuration with a pathogenic variant.
• MLH3 is not included among the four genes with an established fully penetrant dominant Lynch syndrome mechanism (MLH1, MSH2, MSH6, PMS2) in the EHTG/ESCP guideline, which never mentions MLH3 (PMID:34043773); the gene instead has a reported recessive biallelic mechanism (Olkinuora et al., Genet Med 2019, PMID:30573798 - biallelic germline MLH3 nonsense variant underlying polyposis predisposition), and that paper does not mention c.3280+14A>T.
• The ClinVar Benign/Likely benign classification was noted but not used to infer BP2.

Still missing

• No pathogenic or likely pathogenic comparator variant is documented in cis or in trans with this allele.
• No cis/trans phase determination (family study, allele-specific assay, or long-read phasing) is available.
• No affected-proband genotyping context that could support or refute an allelic observation is documented.
• The fully penetrant dominant gene premise is not established for MLH3 (disputed HNPCC7 association; recessive biallelic mechanism reported, PMID:30573798).

Sources

Reasoning

BP5 (ACMG/AMP 2015, PMID:25741868) requires that the variant have been found in a patient who has an alternate molecular basis for the disease, i.e. case-level evidence that a different variant or gene explains the patient's phenotype. This case contains no case-level data at all: there is no proband, no co-tested genes or second variant, no family information and no case report naming c.3280+14A>T. The ClinVar submissions for variation 314378 are classification assertions without case descriptions, and none of them records an alternate molecular basis (the submission audit's evidence signals are empty for all four and two submissions carry no condition label). The retrieved literature is gene-level/guideline-level only - the Lynch syndrome guideline (PMID:34043773) covers MLH1/MSH2/MSH6/PMS2 surveillance and never mentions MLH3 - and the case-level variant-mention audit across the retrieved corpus found zero occurrences of this variant. Threshold/direction check as required for this group: some VCEP specifications define BP5 (or a sibling criterion in this group) with a likelihood-ratio-style numeric value that strengthens as it goes DOWN, so met would require value <= threshold; for MLH3 no such specification exists - there is no ClinGen CSPEC (cspec.found = false) and no gene-specific VCEP framework (framework_mode = generic_acmg; vcep_materials.files_found and domain_tables empty, rule_payload null, no BP5 lookup table) - so there is no governing numeric BP5 threshold and no operator to quote or substitute numbers into. No inequality can therefore be evaluated, and the criterion cannot be scored on a numeric basis. Because BP5's required input (a case with an alternate molecular basis) is missing rather than affirmatively disproven, BP5 is recorded as not assessed, and the gap is a case-level one; needs_human_review is marked because a case-level finding would flip this criterion to met at supporting strength.

Evidence considered

• BP5 is defined in ACMG/AMP 2015 (PMID:25741868) as the variant being found in a case with an alternate molecular basis for disease; the criterion therefore requires case-level data.
• This case contains no proband, family, co-tested gene or co-occurring variant information for c.3280+14A>T; the review is variant-level only.
• ClinVar variation 314378 submissions (Illumina, Rigshospitalet, Sinai Health System, Labcorp/Invitae) are clinical-testing classification assertions with empty evidence_signals and criteria_found; two carry no condition label at all, and none documents an alternate molecular basis.
• PMID:34043773 (EHTG/ESCP Mallorca guideline) full text was read: it defines Lynch syndrome by MLH1/MSH2/MSH6/PMS2 and reports no MLH3 case; no case with an alternate molecular basis involving this variant is described.
• Numeric-threshold direction check: no governing numeric BP5 threshold exists for MLH3 - generic ACMG/AMP 2015 defines BP5 qualitatively, and MLH3 has no ClinGen CSPEC and no gene-specific VCEP rule (framework_mode = generic_acmg; vcep_materials.rule_payload null, domain_tables empty). No operator (>=, <=, >, <) and no threshold value are therefore available to substitute into an inequality for BP5, and no likelihood-ratio style BP5 definition is present in the governing materials.

Still missing

• No proband-level data on an alternate molecular basis (co-occurring pathogenic variant in another gene, or a second explanatory diagnosis) exists in this case; BP5 cannot be scored without case-level information.
• ClinVar submissions for this variant carry no case descriptions and empty evidence signals, so no alternate-molecular-basis signal can be recovered from the submission trail.

Sources