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

Xiangying Zhang,Haotian Gao,Haojie Wang,Zhihang Chen,Zhe Zhang,Xinchong Chen,Yan Li,Yifei Qi,Renxiao Wang
DOI: https://doi.org/10.1021/acs.jcim.3c00253
IF: 6.162
2023-06-15
Journal of Chemical Information and Modeling
Abstract:Predicting protein-ligand binding affinity is a central issue in drug design. Various deep learning models have been published in recent years, where many of them rely on 3D protein-ligand complex structures as input and tend to focus on the single task of reproducing binding affinity. 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. It was trained through a multi-objective process with three related tasks, including deriving the protein-ligand binding affinity, protein-ligand contact map, and ligand distance matrix. Besides the protein-ligand complexes with known binding affinity data retrieved from the PDBbind database, a large number of non-binder decoys were also added to the training data for deriving the final model of PLANET. When tested on the CASF-2016 benchmark, PLANET exhibited a scoring power comparable to the best result yielded by other deep learning models as well as a reasonable ranking power and docking power. In virtual screening trials conducted on the DUD-E benchmark, PLANET's performance was notably better than several deep learning and machine learning models. As on the LIT-PCBA benchmark, PLANET achieved comparable accuracy as the conventional docking program Glide, but it only spent less than 1% of Glide's computation time to finish the same job because PLANET did not need exhaustive conformational sampling. Considering the decent accuracy and efficiency of PLANET in binding affinity prediction, it may become a useful tool for conducting large-scale virtual screening.
chemistry, multidisciplinary, medicinal,computer science, interdisciplinary applications, information systems
What problem does this paper attempt to address?
The paper aims to address the key issue of protein-ligand binding affinity prediction in drug design. Specifically, the research team developed a multi-objective graph neural network model called PLANET (Protein-Ligand Affinity Prediction Network) to improve the accuracy, ranking ability, and docking ability of binding affinity prediction. The input of the PLANET model includes the three-dimensional structural representation of the target protein binding pocket and the two-dimensional chemical structure of the ligand molecule. During the training process, the model is optimized through multi-objective learning to simultaneously perform three tasks: predicting protein-ligand binding affinity, protein-ligand contact graph, and ligand distance matrix. In addition to extracting protein-ligand complexes with known binding affinity data from the PDBbind database, a large number of non-binder decoys are also included to further optimize the final model. In the CASF-2016 benchmark test, PLANET demonstrated scoring ability comparable to the best deep learning models, as well as reasonable ranking ability and docking ability. In virtual screening experiments conducted on the DUD-E benchmark, PLANET outperformed several deep learning and machine learning models. In the LIT-PCBA benchmark, PLANET achieved accuracy comparable to the traditional docking program Glide, but consumed less than 1% of Glide's computation time because PLANET does not require thorough conformational sampling. In conclusion, considering the accuracy and efficiency of PLANET in binding affinity prediction, it could be a useful tool for large-scale virtual screening.