Machine Learning Identifies Common Risk Variants and Implicates Abnormal Vision Physiology in ASD

Abstract

Genomic technology advancements have facilitated associations between genetic variants and disease risk. Rare deleterious variants can independently induce disease, while common variants collectively enhance susceptibility with minimal individual effects. The nuanced nature of common variant consequences attenuates their identification. Autism spectrum disorders (ASD) are hereditary neurodevelopmental disorders. Atypical eye-gaze responses frequently occur in ASD; however, this phenotype has been overlooked in genetic studies. Using WizWhy, an interpretable machine learning tool, to analyze the Autism Sequencing Consortium (ASC) data, 210 common variants across 177 genes were associated with ASD risk. This association is supported by significant overlap with the SFARI gene database and relevant gene ontologies. Individuals with a higher variant burden are at increased ASD risk. Notably, 52 genes were linked to abnormal eye physiology, underscoring the role of this pathway in ASD etiology. Focusing on specific hub genes as potential pharmacological targets may benefit patients with ASD.