Prediction of Protein-Protein Interactions through Deep Learning Based on Sequence Feature Extraction and Interaction Network

Protein-protein interaction (PPI) is an important molecular process in the cell, which is vital to the function of the cell in the biochemical process. This study focuses on human protein. It uses protein information and the relationship of protein interaction network structure to predict PPI. Deep neural network model is implemented to realize PPI prediction. Through five-fold cross-validation, a high performance in the prediction accuracy is produced. The accuracy rate on the test set is 92.45%. To further evaluate the performance of this method, we compared it with other machine learning algorithms. The experimental results show that the method based on neural network is significantly better than the others on the same dataset. It also shows a superior performance compared to previous predictors in this field on PPI prediction.