Systematic approach to analyzing protein structure predicts mutations associated with inherited retinal disease

Abstract

Purpose : Mechanistic understanding of the genetics of inherited retinal disease (IRD) facilitates development of therapeutics. Structure-based network analysis (SBNA), a tool we previously published, identifies which amino acids are structurally critical and thus immutable. SBNA has yielded insights into the consequence of genetic variation on viral protein structure, however this approach has never been applied to human proteins, which are subject to a greater repertoire of possible biological functions and interactions that may influence amino acid importance. We sought to validate SBNA in well-studied human proteins and apply it to test whether a subset of IRD-associated mutations can be explained by their impact on structural topology.

Methods : We augmented the existing SBNA pipeline with homology modeling, which allows for the prediction of human structure from animal experiments and permits saturated sampling of possible structural configurations of a single protein. We compute amino acid SBNA scores and average results across multiple models. We used published well-studied human protein saturation mutagenesis data to validate this approach, then applied it to IRD-associated genes identified among patients at Massachusetts Eye and Ear. IRD SBNA scores were then compared with clinical data from the ClinVar and GnomAD databases.

Results : SBNA scores for 4 well-studied human proteins correlate strongly with empirical functional scores from saturation mutagenesis experiments (Fig. 1A, Spearman correlation p<0.001 for all) and with known variant clinical significance (Fig. 1B). Following validation, we performed SBNA on 21 IRD-associated genes with available structural data, 14 of which had sufficient clinical data available. SBNA scores for known pathogenic variants consistently exceeded scores for benign variants and variants of uncertain significance (VUS) (Fig. 2), though statistical comparisons for specific genes were somewhat limited by dataset power.

Conclusions : Validation with in vitro experiments and known clinical disease phenotypes suggests that SBNA can be meaningfully applied to human proteins, and it may be helpful in predicting which mutations will cause inherited retinal disease.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.

 

A. Correlation between SBNA and functional scores within saturation mutagenesis datasets. B. SBNA scores grouped by clinical effect.

 

SBNA scores of IRD-associated genes grouped by clinical effect.