Systematic overestimation using in silico-driven curation in a population database caused reevaluation of the genetic prevalence of Alport Syndrome.
Abstract
INTRODUCTION A landmark study by Gibson and coworkers in 2021 estimated that one in 106 individuals carry a predicted pathogenic COL4A3 or COL4A4 variant, suggesting autosomal dominant Alport syndrome is massively underdiagnosed. However, pathogenicity of all triple-helical Glycine substitutions was inferred using in silico tools, which may overestimate prevalence.
Methods
COL4A3, COL4A4, and COL4A5 variants were extracted from population database gnomADv4.1.1 (about 800,000 individuals, including UK Biobank) and from a clinical exome cohort of 29,055 patients (Eurofins Biomnis), stratified into a kidney disease group (KIDNEY) of 8663 individuals and 20,392 individuals of a non-kidney control group (NON-KIDNEY). Three curation strategies were compared: a ClinVar-dependent model, an in silico-driven model adapted from Gibson, and an Expert Strategy combining stringent allele frequency thresholds, high-confidence loss-of-function filtering, and with ClinVar validation for missense variants. Estimates were validated against the control group and replicated in the non-UKB Biobank subset.
Results
Applied to gnomADv4.1.1, the Gibson methodology yielded one in 221, a 52% reduction from the original one in 106 (gnomADv2). The Expert Strategy estimated one in 283, validated in the NON-KIDNEY group (one in 295; concordance ratio 0.96), whereas Gibson inflated the control rate by 91% (one in 154 vs. one in 295). Glycine substitutions (COL4A3 p.Gly1277Ser odds ratio 0.95; COL4A4 p.Gly999Glu odds ratio 1.01) were misclassified by all in silico tools. Expert Strategy variants were found in one in 44 KIDNEY patients (odds ratio 6.76). Enrichment was unchanged by high-confidence loss-of-function restriction.
Conclusions
Autosomal dominant Alport syndrome affects one in 283 individuals, one to two orders of magnitude above classical clinical estimates (one in 5,000 to one in 53,000), confirming underdiagnosis, but substantially lower than the Gibson estimate. These findings argue for downgrading the PM2 rarity criterion in collagenopathies. Systematic COL4A3/COL4A4 screening in chronic kidney disease is justified by a diagnostic yield of one in 44.