LncRNA-disease association prediction based on neighborhood information aggregation in neural network

Recent studies have demonstrated that IncRNAs play pivotal roles in various biological processes. Some computational methods have been developed to infer IncRNA-disease associations. However, the experimental identification is time-consuming. In this paper, we introduce a method named NNHLDA, which is based on neighborhood information aggregation in neural network. NNHLDA outperforms previous methods at IncRNA-disease association prediction in heterogeneous network. To evaluate our method, we conduct several experiments. In leave-one-out cross-validation (LOOCV) experiments, our NNHLDA method performed better than current state-of-the-art approach. Furthermore, we extracted top 100 IncRNA-disease associations identified by our method and conducted case studies on gastric cancer. The predictions have been confirmed by verified experimental results. Therefore, it is anticipated that NNHLDA could be a useful tool for biomedical researches.