Hybrid deep learning method to identify key genes in autism spectrum disorder

Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder with a strong genetic component. This research aims to identify key genes associated with autism spectrum disorder using a hybrid deep learning approach. To achieve this, a protein-protein interaction network is constructedand analyzed through a graph convolutional network, which extracts features based on gene interactions. Logistic regression is then employed to predict potential key regulatorgenes using probability scores derived from these features. To evaluate the infection ability of these potential key regulator genes, a susceptible-infected (SI) model, is performed, which reveals the higher infection ability for the genes identified by the proposed method, highlighting its effectiveness in pinpointing key genetic factors associated with ASD. The performance of the proposed method is compared with centrality methods, showing significantly improved results. Identified key genes are further compared with the SFARI gene database and the Evaluation of Autism Gene Link Evidence (EAGLE) framework, revealing commongenes that are strongly associated with ASD. This reinforces the validity of the method in identifying key regulator genes. The proposed method aligns with advancements in therapeutic systems, diagnostics, and neural engineering, providing a robust framework for ASD research and other neurodevelopmental disorders.