GCNPCA: miRNA-Disease Associations Prediction Algorithm Based on Graph Convolutional Neural Networks

A growing number of studies have confirmed the important role of microRNAs (miRNAs) in human diseases and the aberrant expression of miRNAs affects the onset and progression of human diseases. The discovery of disease-associated miRNAs as new biomarkers promote the progress of disease pathology and clinical medicine. However, only a small proportion of miRNA-disease correlations have been validated by biological experiments. And identifying miRNA-disease associations through biological experiments is both expensive and inefficient. Therefore, it is important to develop efficient and highly accurate computational methods to predict miRNA-disease associations. A miRNA-disease associations prediction algorithm based on Graph Convolutional neural Networks and Principal Component Analysis (GCNPCA) is proposed in this paper. Specifically, the deep topological structure information is extracted from the heterogeneous network composed of miRNA and disease nodes by a Graph Convolutional neural Network (GCN) with an additional attention mechanism. The internal attribute information of the nodes is obtained by the Principal Component Analysis (PCA). Then, the topological structure information and the node attribute information are combined to construct comprehensive feature descriptors. Finally, the Random Forest (RF) is used to train and classify these feature descriptors. In the five-fold cross-validation experiment, the AUC and AUPR for the GCNPCA algorithm are 0.983 and 0.988 respectively.

1 Introduction

Micrornas (miRNAs) are small endogenous RNAs that can regulate gene expression post-transcriptionally[1]. MiRNAs consist of a set of genomically encoded non-coding RNAs of approximately 20 to 23 nucleotides in length that direct the silencing complex (RISC) to degrade mRNA or block its translation by base-pairing with the target gene mRNA[2]. Their tissue specificity and temporal sequence, which determine tissue and cellular functional specificity, suggest that miRNAs play multiple roles in the regulation of cell growth and developmental processes. Furthermore, a growing number of recent studies have shown that aberrant miRNA expression often accompanies the development and progression of human diseases. For example, the miR-338-3p has been shown to exacerbate intervertebral disc degeneration progression by directly targeting SIRT6, a negative regulator of the MAPK/ERK pathway. The miR-338-3p is validated as a marker factor for IDD in a large number of samples, and miR-338-3p is obtained to promote cell proliferation and enhance matrix synthesis by in vitro experiments [3]. The miR-128-1 is a key metabolic regulator in mammals. In a mouse model of metabolic disease, antisense targeting and genetic ablation of miR-128-1 resulted in increased energy expenditure and ameliorated metabolic disease. The increased energy expenditure improves high-fat diet-induced obesity and significantly improves glucose tolerance [4]. Therefore, the use of appropriate experimental or computational methods to explore the relationship between miRNAs and diseases can enable miRNAs to become disease suppressors or biomarkers. The studies can help medical researchers to gain insight into the pathological mechanisms of various complex diseases from a molecular perspective and develop appropriate drugs.