Benchmarking AlphaMissense against ClinVar for Diagnostic Interpretation of Missense Variants in Inherited Retinal Diseases

TL;DR

This study evaluates how well AlphaMissense predicts harmful genetic changes in inherited retinal diseases by comparing it to ClinVar and analyzing its accuracy and limitations.

Contribution

The study benchmarks AlphaMissense for inherited retinal diseases and identifies its performance limitations in hypomorphic variants.

Findings

AlphaMissense achieved 94.1% specificity and 79.4% sensitivity in predicting pathogenicity in IRD genes.

It underperformed in predicting hypomorphic variants, especially in ABCA4-associated Stargardt disease.

Population-based metrics helped identify false negatives, and long-read sequencing improved variant classification.

Abstract

AlphaMissense is a newer deep learning–based variant predictor that evaluates the structural consequences of missense variants, the most common pathogenic variant type in inherited retinal diseases (IRDs). This study evaluates the diagnostic utility of AlphaMissense in IRDs by assessing its concordance with ClinVar annotations and exploring how other variant-level metrics may refine its predictions.

A cross-sectional benchmarking study using public variant resources, with a single illustrative clinical case.

Missense variants from 107 IRD genes; 1 patient case undergoing long-read sequencing.

Pathogenicity scores from AlphaMissense were extracted from 128 248 variants present in both IRD genes and the Genome Aggregation Database. Among these, 4204 had definitive ClinVar classifications and were used to calculate AlphaMissense specificity, sensitivity, and false discovery rate (FDR). Population-based metrics, including allele frequency, homozygote count, and Combined Annotation Dependent Depletion score, were analyzed to identify salient features that would be associated with discordance. Long-read sequencing was carried out in a monoallelic ABCA4 patient with late-onset macular dystrophy for phased variant analysis.

Concordance between AlphaMissense predictions and ClinVar annotations was used to calculate sensitivity, specificity, and FDR. Variant-level metrics between discordant variants. Case-based reclassification of hypomorphic variants with long-read sequencing.

AlphaMissense achieved a specificity of 94.1% and sensitivity of 79.4% in IRD genes, with specificity reaching 100% in ABCA4, USH2A, RPGR, and PRPH2, which are 4 of the most common IRD genes. The FDR was 9.6%. AlphaMissense underperformed in predicting hypomorphic variants, particularly in ABCA4-associated Stargardt disease. Variant-level metrics were effective in identifying false negatives. In a clinical case, phased variant analysis identified a potential hypomorphic ABCA4 variant predicted as benign by AlphaMissense.

AlphaMissense demonstrates high specificity for pathogenicity prediction in IRD-associated genes; however, its reduced sensitivity, as seen in hypomorphic variants, suggests a need to incorporate population and functional metrics scores, which may improve classification accuracy, especially as long-read sequencing enables phased variant analysis.

The authors have no proprietary or commercial interest in any materials discussed in this article.

Figures