Interpretable Prediction of Protein-Ligand Interaction by Convolutional Neural Network
Fan Hu, Jiaxin Jiang, Peng Yin · 2019
Evaluation of protein-ligand interaction is a crucial step in the process of drug discovery. Recently, several methods based on deep learning have gained impressive binary classification performance on protein-ligand binding prediction. However, lack of three-dimensional complex data still limits the accuracy and robustness of evaluation of protein-ligand binding affinity, as well as the prediction of their binding sites. In this paper, we propose a novel convolutional neural network based method for estimating the binding affinity between protein and ligand using only 1D sequence data. Even with the same amount of sample size, this model outperforms other structure-dependent traditional and machine learning based methods in terms of both binary classification and regression task. Furthermore, we use this model to identify the key amino acid residues of protein that are vital for binding interaction, which provides biological interpretation.