SS-GNN: A Simple-Structured Graph Neural Network for Affinity Prediction

Shuke Zhang,Yanzhao Jin,Tianmeng Liu,Qi Wang,Zhaohui Zhang,Shuliang Zhao,Bo Shan
DOI: https://doi.org/10.1021/acsomega.3c00085
IF: 4.1
2023-06-15
ACS Omega
Abstract:Efficient and effective drug-target binding affinity (DTBA) prediction is a challenging task due to the limited computational resources in practical applications and is a crucial basis for drug screening. Inspired by the good representation ability of graph neural networks (GNNs), we propose a simple-structured GNN model named SS-GNN to accurately predict DTBA. By constructing a single undirected graph based on a distance threshold to represent protein-ligand interactions, the scale of the graph data is greatly reduced. Moreover, ignoring covalent bonds in the protein further reduces the computational cost of the model. The graph neural network-multilayer perceptron (GNN-MLP) module takes the latent feature extraction of atoms and edges in the graph as two mutually independent processes. We also develop an edge-based atom-pair feature aggregation method to represent complex interactions and a graph pooling-based method to predict the binding affinity of the complex. We achieve state-of-the-art prediction performance using a simple model (with only 0.6 M parameters) without introducing complicated geometric feature descriptions. SS-GNN achieves Pearson's <i>R</i><sub><i>p</i></sub> = 0.853 on the PDBbind v2016 core set, outperforming state-of-the-art GNN-based methods by 5.2%. Moreover, the simplified model structure and concise data processing procedure improve the prediction efficiency of the model. For a typical protein-ligand complex, affinity prediction takes only 0.2 ms. All codes are freely accessible at https://github.com/xianyuco/SS-GNN.
chemistry, multidisciplinary
What problem does this paper attempt to address?
### Problems the Paper Attempts to Solve The main goal of this paper is to develop a simple and efficient Drug-Target Binding Affinity (DTBA) prediction model to address the challenges of traditional methods in real-world applications with limited computational resources. Specifically: 1. **Simplifying Model Structure**: - A simplified graph neural network model named SS-GNN is proposed for accurate DTBA prediction. By constructing a single undirected graph based on distance thresholds to represent protein-ligand interactions, the scale of graph data is significantly reduced. 2. **Improving Computational Efficiency**: - Ignoring covalent bonds in proteins further reduces the computational cost of the model. The Graph Neural Network and Multi-Layer Perceptron (GNN-MLP) module is used to independently extract latent features of atoms and edges in the graph. An edge-based atom pair feature aggregation method and a graph pooling-based affinity prediction method are developed. 3. **Efficient Prediction**: - Using a simple model structure (containing only 0.6M parameters), advanced prediction performance is achieved without introducing complex geometric feature descriptions. For example, on the PDBbind v2016 core set, SS-GNN achieved a Pearson correlation coefficient Rp=0.853, which is 5.2% higher than the state-of-the-art GNN methods. Additionally, the simplified model structure and streamlined data processing significantly improve prediction efficiency, requiring only 0.2 milliseconds for affinity prediction of a typical protein-ligand complex. Through these improvements, SS-GNN not only demonstrates excellent prediction performance but also offers higher computational efficiency, making it suitable for large-scale engineering applications.