The Recognition of Protein Methylation Sites Based on CNN and Bi-LSTM Models

Methylation is a protein Post-Translational Modification (PTM) that regulates cell function ,which can provide guidance and help for research in the fields of gene regulation and disease prediction. At present, there are some problems in the research of methylation site recognition, such as few labeled datasets, insufficient positive sample data and low recognition accuracy of methylation research. In order to solve these problems, this paper proposes a protein methylation site recognition method based on Convolutional Neural Network (CNN) and Bi-directional Long Short-Term Memory (Bi-LSTM) model. Our model is divided into two branches. The CNN branch uses a dense connection method to make the feature information of each layer fully transmitted and shared. The stacked Bi-LSTM branches obtain bidirectional long-term dependencies in the sequence, and then the two branches perform feature fusion for methylation recognition. Experiments show that the Accuracy ( ACC), F1 Score ( F1 score) and Matthews Correlation Coefficient (MCC) obtained by using our model to identify methylation sites are 0. 851 9, 0. 849 4 and 0. 728 4, respectively. Compared with other methods, the model has better performance.