🧬 CFTR Compound Heterozygosity

Overview

A clinical variant-interpretation cohort across four recessive disease genes (CFTR, PAH, HBB, GBA1), three oligogenic disease pairs, and a VUS-reclassification candidate, built so the diagnosis depends on phasing rather than any single variant. 19 ClinVar / gnomAD-annotated variants across 10 genes plus one real 1000 Genomes genome (NA12878).

Data Structure

Metric	Value
Disease genes	CFTR, PAH, HBB, GBA1 (essential, recessive) plus three oligogenic pairs: LDLR+PCSK9, KCNQ1+KCNH2, BMPR2+ACVRL1
Cohort	12 controlled patients (P1–P12, incl. P12 a VUS-reclassification candidate) + 1 real 1000 Genomes genome (NA12878)
Variants	19 ClinVar / gnomAD-annotated variants across 10 genes, plus 8 real NA12878 calls (GRCh38)
Reference knowledge	gnomAD, ClinVar, Reactome, HPO, pulled live
Real phasing	NA12878's parent-of-origin from its CEU trio (NA12891/NA12892), by Mendelian transmission

Causal Reasoning Model

Three reasoning mechanisms (two that diagnose, one that flags for review):

1. Compound heterozygosity in trans: two pathogenic variants in the same essential gene, one maternal and one paternal, disable both copies and cause recessive disease. The same mechanism covers cystic fibrosis (CFTR), phenylketonuria (PAH), β-thalassemia (HBB), and Gaucher disease (GBA1). 2. Oligogenic convergence: pathogenic variants in two different genes that act on one shared Reactome pathway combine to raise risk. The same rule diagnoses three real disease pairs, each on its own shared pathway: LDLR+PCSK9 (familial hypercholesterolemia), KCNQ1+KCNH2 (long-QT syndrome), BMPR2+ACVRL1 (pulmonary arterial hypertension). 3. VUS reclassification (P12): when one variant is pathogenic and the other is a variant of uncertain significance (VUS), in trans in an essential gene, the engine does not confirm disease (no over-call) and does not dismiss it — it flags the patient for VUS reclassification review. This mirrors the clinician's real daily decision.

The Phasing Test

The signature case: the same two CFTR variants cause disease when inherited on opposite parental copies (in trans) but not when they sit on the same copy (in cis). A per-variant filter cannot tell these apart; phasing decides it. For NA12878, the phase is derived from its real CEU trio rather than assumed.

Evidence Provenance

Every disease label is traceable to its public source: per-variant ClinVar significance and review stars, Reactome pathway membership, and CFTR2 corroboration cited by reference. CFTR variant pairs additionally carry real gnomAD population co-occurrence (in-cis vs in-trans), which corroborates the phasing rather than asserting it.