ABSTRACT:The accurate prediction of drug–protein interactions (DPIs) is a fundamental task in drug discovery, enabling the identification of novel therapeutic targets, drug repurposing opportunities, and reducing the time and cost of wet-lab experiments. Traditional computational approaches, such as molecular docking and similarity-based methods, often face limitations in handling complex molecular structures and integrating heterogeneous biological data. Graph Neural Networks (GNNs) have emerged as a powerful deep learning paradigm that naturally models the structural and relational information inherent in chemical compounds and proteins. This literature review surveys recent advancements in GNN-based DPI prediction, including architectures such as Graph Convolutional Networks (GCNs), Graph Attention Networks (GATs), message-passing neural networks (MPNNs), and heterogeneous GNNs. We compare key models, analyze their strengths and limitations, and explore their integration with structural biology tools such as AlphaFold. Special emphasis is given to ensemble learning approaches, particularly the method proposed by Liu et al. [1], which combines protein sequences and drug fingerprints. Finally, we present a hybrid network architecture that integrates GNN-based embeddings with ensemble learning for improved DPI prediction performance.

KEYWORDS:Graph Neural Networks, Drug–Protein Interaction, Drug Discovery, Deep Learning, Computational Biology, Ensemble Learning.

LITERATURE REVIEW: GRAPH NEURAL NETWORK–BASED METHODS FOR DRUG–PROTEIN INTERACTION PREDICTION

MS. MAMTA PANNASE, PROF. SUSHIL BHISE