Systematic overestimation using in silico-driven curation in a population database caused reevaluation of the genetic prevalence of Alport Syndrome

COL4A3, COL4A4

Heterozygous COL4A3/COL4A4 variants; overestimated pathogenicity of triple-helical glycine substitutions by in silico tools

A 2021 study estimated that one in 106 individuals carries a predicted pathogenic COL4A3 or COL4A4 variant, inferring the pathogenicity of all triple-helical glycine substitutions with in silico tools. The authors compare three curation strategies (ClinVar-dependent, in silico adapted from that study, and an Expert Strategy) in gnomAD v4.1.1, about 800,000 individuals, and in 29,055 clinical exomes split into 8,663 kidney patients and 20,392 controls. Applied to gnomAD v4.1.1, the in silico method yields one in 221; the Expert Strategy estimates one in 283, validated in the control group (one in 295, concordance ratio 0.96), whereas the in silico method inflates the control rate by 91%. Glycine substitutions such as COL4A3 p.Gly1277Ser (OR 0.95) and COL4A4 p.Gly999Glu (OR 1.01) are misclassified by all in silico tools, and Expert Strategy variants are found in one in 44 kidney patients (OR 6.76).

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

A useful paper for day-to-day interpretation: a triple-helical glycine substitution in COL4A3/COL4A4 is not pathogenic by default, and the PM2 rarity criterion carries less weight in genes where causal variants are common in the population. The revised prevalence remains one to two orders of magnitude above classical clinical estimates, confirming underdiagnosis. The Expert Strategy partly relies on ClinVar for missense variants, with the submission biases this entails.

Analysis by Dr Thibaut Benquey