Prediction of Protein-Ligand Binding Affinity Based on LSTM and Attention Mechanism

The prediction of protein-ligand binding affinity is a challenging task in drug repositioning regression. Deep learning methods can effectively predict the binding affinity of protein-ligand interactions, reducing the time and cost of drug discovery. Therefore, a deep convolutional neural network model (DLLSA) based on Long Short-Term Memory (LSTM) modules and attention mechanism modules has been proposed. The model consists of a convolutional network with embedded LSTM and spatial attention modules, where the LSTM module is designed for the long sequence information of protein-ligand contact features, and the spatial attention module aggregates local information of contact features. The model was trained using the PDBbind (v.2020) dataset and validated on the CASF-2013 and CASF-2016 datasets, achieving a Pearson correlation coefficient improvement of 0.6% and 3% respectively compared to the PLEC model, with experimental results significantly outperforming other related methods.