Graph Neural Network-Based Prediction Framework for Protein-Ligand Binding Affinity: A Case Study on Pediatric Gastrointestinal Disease Targets

Abstract

Accurate prediction of protein-ligand binding affinity is a fundamental step in drug and vaccine development, particularly for pediatric gastrointestinal diseases such as peptic ulcers, Crohn's disease, and ulcerative colitis. Traditional computational methods, including molecular docking and physics-based simulations, often suffer from limited accuracy and high computational costs. To address these limitations, this study proposes a prediction framework based on Graph Neural Networks (GNNs), which naturally represent the structural and relational characteristics of protein-ligand complexes. Using publicly available datasets derived from PDBbind and BindingDB, a subset of protein targets highly relevant to pediatric gastrointestinal disorders was curated. Protein-ligand complexes were preprocessed to construct heterogeneous molecular graphs, with atoms as nodes and bonds or intermolecular interactions as edges. Multiple GNN architectures-including Graph Convolutional Networks (GCN), Graph Attention Networks (GAT), and Graph Isomorphism Networks (GIN)-were compared to evaluate prediction performance. Experimental results demonstrate that the GIN-based model achieved the best performance, with a mean squared error (MSE) of 2.05 and a Mean Absolute Error (MAE) of 1.05, outperforming traditional baselines such as RNN-based methods. These findings highlight the potential of graph-based deep learning approaches for accelerating drug discovery in pediatric gastroenterology by providing accurate, scalable, and generalizable predictions of binding affinity.