Comparative Analysis of In-Silico Tools in Identifying Pathogenic Variants in Dominant Inherited Retinal Diseases

Abstract

Purpose : Inherited retinal diseases (IRDs) are genetic eye conditions caused by variants that disrupt retinal function, resulting in blindness. Despite progress in identifying genes associated with IRDs, improvements are necessary for the correct classification of rare autosomal dominant (AD) disorders. AD diseases are highly heterogenous, with causal variants being restricted to specific amino acid changes within certain protein domains, making AD conditions difficult to classify. We aim to determine the top-performing in-silico tools on predicting the pathogenicity of AD IRD variants and apply them on a cohort of patients with undiagnosed IRDs.

Methods : Extracting annotated pathogenic and benign variants from ClinVar, we systematically benchmarked 39 variant classifiers for genes associated with IRDs, split by inheritance pattern. Using area-under-the-curve (AUC) analysis, we determined the top-performing classifiers and defined thresholds for variants that were likely-pathogenic, likely-benign, or variants of unknown significance. We evaluated the top-performing classifiers using whole genome sequencing data of a cohort of >1,000 patients affected with IRDs of unknown etiology.

Results : Benchmark analysis identified MutScore, ClinPred, BayesDel, MetaRNN, REVEL, and VEST4 as the top-performing classifiers for AD IRD variants. MutScore achieved the highest accuracy in differentiating likely-pathogenic and likely-benign variants within AD genes, yielding an AUC of 0.969. A filtered benchmark of gain-of-function and dominant negative variants resulted in BayesDel achieving the highest performance with an AUC of 0.997. Five patients with variants in NR2E3, RHO, GUCA1A, and GUCY2D were confirmed to have dominantly inherited disease based on pedigree, phenotype, and segregation analysis. We identified two novel variants in GUCA1A (c.428T>A, p.Ile143Thr) and RHO (c.631C>G, p.His211Asp) in two patients.

Conclusions : Our findings support the effectiveness of using a multi-classifier approach consisting of new missense classifiers to identify pathogenic variants in patients affected with AD IRDs. The top-performing models significantly outperformed tools traditionally used in clinical practice, such as PolyPhen2, SIFT, and CADD. Our results provide a foundation for improved genetic diagnosis for patients with IRDs, a crucial step in guiding patient care in the era of gene therapy.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.