End-to-end deep learning methods for genetic risk prediction of schizophrenia

Abstract

Schizophrenia is a highly heritable psychiatric disorder with a complex genetic basis. While recent GWAS and whole exome sequencing studies have identified numerous risk loci, existing clinical prediction models rely primarily on linear assumptions, limiting their ability to capture complex, nonlinear genetic effects such as epistasis. In this study, we apply end-to-end genome interpretation neural network models to predict schizophrenia risk using whole exome sequencing data in a cohort of 6,135 cases and 6,245 controls. We show that nonlinear neural networks significantly outperform conventional additive models when sufficient sample size is available. These findings support the fact that high-order genetic interactions between alleles and variants should be considered by clinical and quantitative genetics models. To investigate the decision process our models follow, we integrate explainable AI techniques and biological priors into our models, using them to identify predictive genes and pathways. Our approach recovers both known schizophrenia risk genes and recommends BASP1 as a potential understudied schizophrenia gene involved in neuronal development. Existing prediction models for schizophrenia often fail to capture nonlinear genetic effects, like epistasis. Here, the authors show that explainable neural networks improve genetic prediction of schizophrenia by capturing genetic interactions.