Inferring Disease-Associated Piwi-Interacting RNAs via Graph Attention Networks

Abstract

Piwi proteins and Piwi-Interacting RNAs (piRNAs) are commonly detected in human cancers. However, it is time-consuming and costly to detect piRNA-disease associations (PDAs) by traditional experimental methods. In this study, we present a computational method GAPDA to identify potential and biologically significant PDAs based on graph attention network. Specifically, we combined piRNA sequence information, disease semantic similarity, and piRNA-disease association network to construct a new attribute network. Then, the network embedding in node-level is learned via the attention-based graph neural network. Finally, potential piRNA-disease associations are scored.To be our knowledge, this is the first time that the attention-based Graph Neural Networks is introduced to the field of ncRNA-related association prediction. In the experiment, the proposed GAPDA method achieved AUC of 0.9038 using five-fold cross-validation. The experimental results show that the GAPDA approach ensures the prospect of the graph neural network on such problems and can be an excellent supplement for future biomedical research.