PLANET: A Multi-Objective Graph Neural Network Model for Protein-Ligand Binding Affinity Prediction

Abstract

Predicting protein-ligand binding affinity is a central issue in drug design. Various deep learning models have been developed in recent years to tackle this issue, but many of them merely focus on reproducing the binding affinity of known binders. In this study, we have developed a graph neural network model called PLANET (Protein-Ligand Affinity prediction NETwork). This model takes the graph-represented 3D structure of the binding pocket on the target protein and the 2D chemical structure of the ligand molecule as input, and it was trained through a multi-objective process with three related tasks, including deriving the protein-ligand binding affinity, protein-ligand contact map, and intra-ligand distance matrix. To serve those tasks, a large number of decoy non-binders were selected and added to the standard PDBbind data set. When tested on the CASF-2016 benchmark, PLANET exhibited a scoring power comparable to other deep learning models that rely on 3D protein-ligand complex structures as input. It also showed notably better performance in virtual screening trials on the DUD-E and LIT-PCBA benchmark. In particular, PLANET achieved comparable accuracy on LIT-PCBA as the conventional docking program Glide. However, it only took less than 1% of the computation time required by Glide to finish the same job because it did not perform exhaustive conformational sampling. In summary, PLANET exhibited a decent performance in binding affinity prediction as well as virtual screening, which makes it potentially useful for drug discovery in practice.