Predicting Long Non-coding RNA-disease Associations using Multiple Features and Deep Learning

Tóm tắt

Various long non-coding RNAs have been shown to play crucial roles in different biological processes including cell cycle control, transcription, translation, epigenetic regu-lation, splicing, differentiation, immune response and so forth in the human body. Discovering lncRNA-disease associations promotes the awareness of human complex disease at molec-ular level and support the diagnosis, treatment and preven-tion of complex diseases. It is costly, laboratory and time-consuming to discover and verify lncRNA-disease associations by biological experiments. Therefore, it is crucial to develop a computational method to predict lncRNA-disease associations to save time and resources. In this paper, we proposed a new method to predict lncRNA-disease associations using multiple features and deep learning. Our method uses a weighted K-nearest known neighbors algorithm as a pre-processing step to eliminate the impact of sparsity data problem. And it combines the linear and non-linear features extracted by singular value decomposition and deep learning techniques, respectively, to obtain better prediction performance. Our proposed method achieves a decisive performance with the best AUC and AUPR values of 0.9702 and 0.8814, respectively, under LOOCV experiments. It is superior to other state-of-the-art SDLDA and NCPLDA methods in both AUC and AUPR evaluation metrics. It could be considered as a powerful tool to predict lncRNA-disease associations.