Association Prediction of lncRNAs and Diseases Using Multiview Graph Convolution Neural Network

Wei Zhang Yifu Zeng Xiaowen Xiang Bihai Zhao Sai Hu Limiao Li Xiaoyu Zhu Lei Wang ^*

• Changsha University, Changsha, China

The functionality of long noncoding RNAs (lncRNAs) is complex, which facilitates their interaction with a wide range of biological macromolecules, including microRNAs (miRNAs), proteins, genes, and cyclic RNAs, to regulate various physiological processes. These macromolecules are interconnected within a sophisticated regulatory network that is implicated in numerous diseases. The aim of the association prediction methodology, which leverages network analysis and deep learning, is to extract topological structural features from the network and to elucidate intricate associations within the regulatory relationship network, therefore enhancing the reliability of predictions. In this research, we introduce a graph convolutional network (GCN)-based approach to predict the disease relationships of lncRNAs, utilizing macromolecule-disease regulatory networks. Recognizing the collaborative role of miRNAs and lncRNAs in gene expression regulation, we developed a lncRNA-miRNA disease network. Additionally, to mitigate the challenges posed by isolated diseases and lncRNAs, we incorporate disease semantics and the similarity network of lncRNAs into a heterogeneous network framework. We present MVIGCN, an association prediction method for lncRNA-disease relationships that integrates multiview data sources and employs deep learning techniques with attention mechanisms.