Datasets:

ClarusC64
/

genetics-variant-evidence-integrity-v0.1

id stringclasses 10 values	variant stringclasses 10 values	gene stringclasses 10 values	phenotype stringclasses 10 values	evidence_pack stringclasses 10 values	claim stringclasses 10 values	evidence_binding_map stringclasses 10 values	evidence_gaps stringclasses 10 values	claim_overreach_flags stringclasses 10 values	confidence_calibration stringclasses 10 values	notes stringclasses 10 values	constraints stringclasses 1 value	gold_checklist stringclasses 1 value
VEI-001	c.1234A>G (p.Lys412Glu)	GENE1	cardiomyopathy	Observed in 1 affected individual. Absent in 500 controls. No segregation. No functional assay. In silico: damaging.	Pathogenic variant causing cardiomyopathy.	Binding: case report and rarity only. In silico supports possible impact.	No segregation; no functional data; no replication; no population frequency context beyond small control set.	Pathogenic label overreach; causality overreach.	Low confidence. Classify as VUS with rationale.	Tests pathogenic overclaim	Under 220 words.	map+gaps+flags+calibration
VEI-002	rs999999	GENE2	type 2 diabetes	GWAS p=1e-9. OR 1.05. Non-coding region. Fine-mapping unresolved. No eQTL. No perturbation.	This variant causes diabetes.	Binding: association only with tiny effect size.	No causal mechanism; no functional perturbation; no linkage resolution.	GWAS treated as causal.	Very low confidence for causality. Association only.	GWAS causality error	Under 220 words.	map+gaps+flags+calibration
VEI-003	c.250del (frameshift)	GENE3	neurodevelopmental delay	De novo in child. Parentage confirmed. LoF gene known haploinsufficient. Similar LoF reported in 12 cases. Functional: reduced protein.	Pathogenic and likely causal.	Binding: de novo + known LoF mechanism + replication + functional support.	Missing: detailed phenotype match; mosaicism check.	Minor. Evidence mostly adequate.	High confidence. Pathogenic.	Gold-standard case	Under 220 words.	map+gaps+flags+calibration
VEI-004	c.88C>T (p.Pro30Leu)	GENE4	epilepsy	In silico: benign in 2 tools, damaging in 1. Seen at 0.5 percent in gnomAD. No case enrichment. No functional.	Likely pathogenic.	Binding: none supportive; population frequency argues against.	No case evidence; no function; frequency too high for rare epilepsy.	Likely pathogenic overreach.	Very low confidence. Likely benign or VUS.	In silico misuse plus frequency ignore	Under 220 words.	map+gaps+flags+calibration
VEI-005	chr7:12345 A>G	GENE5	autoimmune phenotype	Cell line assay shows altered expression. No human genetic association. No segregation.	Causal variant for autoimmune disease.	Binding: functional in vitro only.	No human association; no in vivo model; no patient evidence.	Functional assay treated as disease proof.	Low confidence for disease causality. Functional effect only.	Function vs disease leap	Under 220 words.	map+gaps+flags+calibration
VEI-006	rs111111	GENE6	lipid levels	GWAS strong. eQTL links to gene expression in liver. CRISPR perturbation in hepatocytes shifts lipid pathway markers.	Causal regulatory variant affecting lipids.	Binding: association + eQTL + perturbation consistent with phenotype.	Missing: replication across ancestries; in vivo validation.	Moderate. Causality plausible but scope limited.	Moderate to high confidence for mechanism. Clinical impact uncertain.	Good causal chain but bounded	Under 220 words.	map+gaps+flags+calibration
VEI-007	c.500G>A (p.Arg167His)	GENE7	rare syndrome	Reported in 2 papers but phenotypes differ. Variant also seen in gnomAD 0.1 percent. No functional.	This explains the syndrome.	Binding weak; phenotype mismatch and frequency conflict.	No consistent phenotype; no segregation; no function.	Overclaim; phenotype mismatch ignored.	Low confidence. VUS.	Phenotype drift	Under 220 words.	map+gaps+flags+calibration
VEI-008	c.900+1G>T (splice)	GENE8	kidney disease	Splice site loss predicted. RNA shows exon skipping in patient tissue. Segregates in family. Multiple reports.	Pathogenic splice variant causing kidney disease.	Binding: predicted splice + RNA confirmation + segregation + replication.	Missing: penetrance estimate.	Minor. Penetrance uncertainty.	High confidence pathogenic.	Strong evidence binding	Under 220 words.	map+gaps+flags+calibration
VEI-009	polygenic_score	MULTI	CAD	PRS AUC 0.62 in one cohort. No external validation. Calibration unknown.	PRS proves genetic cause in this patient.	Binding: modest prediction only.	No individual-level causality; no calibration; no transportability test.	Prediction treated as proof.	Low confidence. Risk stratification only.	PRS misuse	Under 220 words.	map+gaps+flags+calibration
VEI-010	c.1201T>C (p.Ile401Thr)	GENE9	cancer predisposition	Missense in tumor suppressor. Observed in tumor only. No germline test. No functional.	Germline pathogenic predisposition variant.	Binding: tumor-only observation.	No germline confirmation; no segregation; no function.	Somatic treated as germline.	Very low confidence. Needs germline testing.	Somatic vs germline boundary	Under 220 words.	map+gaps+flags+calibration