Dual Channel based Convolutional Neural Network based Protein–Ligand Interaction Prediction for Drug Design

Nowadays, the drug discovery and drug design is developing to yield safe drugs constantly based on time, labor efficient way and cost. In drug discovery, the binding affinity and interaction strength among ligand and target protein are the main factors used for prediction. However, the existing Deep Learning (DL) based prediction approaches failed to predict the binding affinity between ligand and protein precisely. To overcome this problem, a Dual Channel based Convolutional Neural Network (DC-CNN) method is proposed for predicting binding affinity and interaction between protein and ligand for drug design. The proposed DC-CNN model which consists of two blocks that enhanced the prediction by capturing relationship (interaction) between protein and ligand. This acquired interaction data enhanced the drug design and also improved the prediction results. Experimental results of the proposed DC-CNN model attained Pearson Correlation Coefficient (PCC) of 0.872 which is higher than the existing prediction approaches such as Hybrid Neural Network (HNN) and Graph Attention Network (GAN).