A Channel-Attention Convolutional Neural Network for Protein-DNA Binding Residue Prediction

Accurately identifying protein-DNA binding residues is important for protein function annotation and drug design. Previous studies have used classic machine learning algorithms like support vector machines (SVMs) and random forest (RF) to predict protein-DNA binding residues. However, as new machine learning techniques are being developed, the prediction performance could be further improved. In this paper, a convolutional neural network (CNN) based framework is proposed for protein-DNA binding residues prediction, named Se-Residual-Inception network, which includes stacked Inception blocks with short-cut connection and channel-attention mechanism. We examine the performance of proposed Se-Residual-Inception network on the benchmark dataset and independent testing set, and achieve the area under the ROC curve (AUC) values of 0.882 and 0.848, respectively. The prediction performance indicates that our proposed method can precisely identify the DNA binding residues in protein sequences, which can provide assistive directions for wet-lab experiments.