Inference of Autism-Related Genes by Integrating Protein-Protein Interactions and miRNA-Target Interactions

Autism spectrum disorders (ASD) are a group of conditions characterized by impairments in social interaction and presence of repetitive behavior. These complex neurological diseases are among the fastest growing developmental disorders and cause varying degrees of lifelong disabilities. There have been a lot of ongoing research to unravel the pathogenic mechanism of autism. Computational methods have come to the scene as a promising approach to aid the physicians in studying autism. In this paper, we present an efficient method to predict autism-related candidate genes (autism genes in short) by integrating protein interaction network and miRNA-target interaction network. We combine the two networks by a new technique relying on shortest path calculation. To demonstrate the high performance of our method, we run several experiments on three different PPI networks extracted from the BioGRID database, the HINT database, and the HPRD database. Three supervised learning algorithms were employed, i.e., the Bayesian network and the random tree and the random forest. Among them, the random forest method performs better in terms of precision, recall, and F-measure. It shows that the random forest algorithm is potential to infer autism genes. Carrying out the experiments with five different lengths of the shortest paths in the PPI networks, the results show the advantage of the method in studying autism genes based on the large scale network. In conclusion, the proposed method is beneficial in deciphering the pathogenic mechanism of autism.