The Study Of Prediction Methods On S-Palmitoylation Sites And Long Non-coding RNAs Based On Deep Learning

Deep learning is a general term for a class of machine learning methods characterized by multi-layer neural networks.In recent years,deep learning has been widely used in the fields of image recognition and natural language processing.Deep learning is suitable for machine learning in high sample size,high dimensional problem scenarios.In the construction of sequence-oriented bioinformatical models,deep learning methods can reduce information loss from feature extraction process by primary coding of sequences.Based on the convolutional neural network,we established a predictive tool,DpPalm,for S-Palmitoylation sites recognition with peptide sequences.We encode peptide sequences with a AAIndex1 combination in DpPalm.The AUC based on prediction result of the independent test set with DpPalm reached 0.866.Furthermore,based on the Bi LSTM-CNN hybrid network,we established a predictive tool DpLNC for lncRNA recognition with nucleic acid sequence.The AUC based on prediction result of the independent test set with Dp LNC reached 0.822.We have also established a comprehensive lncRNA database called LIVE based entirely on low-throughput experiments.The new lncRNA dataset included in LIVE obtained a prediction sensitivity of 81.8% with DpLNC.The data archived in LIVE involves 572 lncRNAs,including 1342 annotation entries from 1045 articles.This paper is a typical application of deep learning in informatical prediction based on biological sequences.The research in this paper not only has provided an effective tool for the recognition of S-Palmitoylation modification sites,but also established a new method for the recognition of lncRNAs.