Structure-based network analysis identifies pathogenic variants in patients with inherited retinal disease

Abstract

Purpose : Mechanistic understanding of the genetics of inherited retinal disease (IRD) in individual patients facilitates diagnostic answers for patients and development of therapeutics. Structure-based network analysis (SBNA), a tool we previously published, identifies which amino acids are structurally critical and thus immutable. We previously presented the successful application of SBNA to IRD-associated genes. We sought to further apply SBNA to identify the putative genetic basis of IRD in patients without a known causative genetic variant.

Methods : We computed SBNA scores for 47 IRD-associated proteins. These were combined with BLOSUM62 scores to create a modified SBNA score (mSBNA) that does not require training on prelabeled datasets. We then calculated the mSBNA score for genetic variants from 455 patients at Mass. Eye and Ear with phenotypic evidence of IRD. Of these, 63 lacked known pathogenic variants, and mSBNA scores were used to identify which variants of uncertain significance (VUSs) could be causative of disease.

Results : For 357 patients with known pathogenic variants, mSBNA scores were examined to determine whether they correctly predicted pathogenicity. mSBNA scores correctly predicted variant pathogenicity in 96.1% of patients (Fig. 1A-B). Of the 63 patients with only VUSs, mSBNA scores identified the putative genetic basis of disease in 40 patients (23 unique variants) (Fig. 1C). Inheritance patterns included autosomal recessive (n = 15; in combination with a known pathogenic variant or a second VUS with a high mSBNA score), autosomal dominant (n = 2), and X-linked recessive (n = 5). For an additional 19 patients, the mSBNA scores contributed towards identifying a possible but not completely solved genetic cause, such as only one heterozygous VUS receiving a strong score in a gene typically requiring a homozygous or compound heterozygous pattern to cause disease.

Conclusions : mSBNA can be meaningfully applied to IRD-associated genes, yielding insight into the putative genetic basis of disease for patients without known causative genetic variants. Importantly, mSBNA does not rely on a training step with prelabeled data and thus may be broadly applicable.

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

 

(A) Categorization of results from a dataset of possibly solving patient variants, further subdivided into patients with known putative genetic causes of disease (B) and those with only VUSs in known IRD-associated genes (C).