A Structure-Aware Multimodal Framework for Drug–Target Interaction Prediction via Heterogeneous Graph Learning

DeepGCL: Multi-View Graph Contrastive Learning for Enhanced Drug-Target Binding Affinity Prediction Through Protein Pocket-Drug Interaction Modeling.

Drug-target binding affinity (DTA) prediction plays a critical role in drug discovery. Although recent approaches have demonstrated the importance of incorporating protein pocket information, they still struggle to jointly model local atomic-level interactions and global sequence contexts, as well as capture their intricate cross-modal dependencies, within a unified framework for accurate prediction. To address this limitation, we present DeepGCL, a novel multi-modal framework that leverages multi-view graph contrastive learning to capture latent representations of pocket-drug interactions and their underlying molecular determinants. The DeepGCL integrates protein sequences, pocket-drug interaction graphs, and drug molecular graphs within a unified framework to learn complementary molecular representations. By combining sequence encoding with multi-view graph contrastive learning through augmented view generation, the DeepGCL effectively captures contextual protein information and structural interactions between pockets and drugs. This enables accurate characterization of spatial and chemical relationships between binding partners. Comprehensive evaluation on multiple benchmarks shows that DeepGCL consistently outperforms state-of-the-art methods across standard metrics. Ablation studies further confirm that the protein feature module and graph contrastive learning module both make significant and complementary contributions to performance. These findings underscore the effectiveness of multi-view learning paradigms in capturing the multifaceted nature of drug-target interactions and highlight the importance of integrating diverse molecular representations for accurate binding affinity prediction. Our code is available at https://github.com/domtryto/DeepGCL.

MAGNETIC: Multilayer Attention and Graph Neural Network Diffusion for Effective Drug-Target Interaction Prediction.

Drug target interaction (DTI) prediction is a critical task in drug discovery, as it has the potential to accelerate the identification of promising drug-target interactions, thus leading to more efficient and focused wet lab validations. Our central premise is that modeling DTI as a multilayer attributed heterogeneous graph and learning layer-specific representations with diffusion-based graph neural networks (GNNs) improves predictive performance over conventional single-view graph learning and matrix factorization approaches. To test this premise, we pro pose Multilayer Attention and Graph Neural nETwork dIffusion for effeCtive Drug-Target Interaction prediction (MAG NETIC). MAGNETIC is a graph auto-encoder framework that represents drugs and targets in a multiplex network. It further incorporates meta-path-derived drug-drug and target-target similarity relations. MAGNETIC trains a separate Adaptive Graph Diffusion Network (AGDN) per layer to preserve layer-specific topology and attributes, then combines embeddings through learnable layer weights to produce a unified representation for DTI inference. The model is optimized with a reconstruction objective tailored to the sparse and imbalanced DTI setting. We evaluate MAGNETIC against strong graph neural network and matrix factorization baselines across multiple experimental scenarios. Results reveal that MAGNETIC consistently out performs baselines on both the area under the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPRC). The largest gains are observed in AUPRC, indicating improved identification and ranking of true interactions under class imbalance. Finally, a real-world validation study on previously undiscovered DTIs provides additional evidence of the model's practical utility for novel interaction discovery.

MKGR: Multimodal Knowledge-Graph Representation Learning for Cold-Start Protein-Protein Interaction Prediction

Accurate protein-protein interaction (PPI) prediction is central to functional genomics, disease mechanism discovery, and drug development. A difficult setting arises when candidate interactions include proteins that have no observed PPI edges during training, where models relying on network topology alone often lose useful context. This paper presents \method, a multimodal representation framework for cold-start PPI prediction. \method\ combines region-aware protein sequence encoding with four protein-centered biomedical knowledge graphs, including protein-drug, protein-disease, protein-miRNA, and protein-lncRNA associations. The sequence branch extracts contextual representations from structurally informed sequence regions, while graph attention encoders learn modality-specific protein embeddings from sparse biomedical associations. A bridge reconstruction objective regularizes graph learning by recovering shared protein-entity associations, and a pair-level gating module adaptively integrates sequence and graph evidence for each candidate protein pair. Experiments on two benchmark datasets under novel-old and novel-novel cold-start settings show that \method\ consistently outperforms competitive sequence, network, and knowledge-graph baselines across ACC, F1, AUC, AUPR, and MCC.

GraphTransDTI: A novel hybrid framework combining graph transformer and CNN-BiLSTM for enhanced Drug-Protein Interaction prediction

Drug-protein interaction (DTI) prediction is a pivotal step in the drug discovery and repurposing process, helping to minimize experimental costs and time. However, existing deep learning methods often face limitations in simultaneously capturing the spatial structure of drug molecules and the deep contextual correlation with protein sequences. To address this issue, we propose GraphTransDTI, a synergistic hybrid framework that integrates a Graph Transformer to represent drug graph structures, a CNN-BiLSTM network to encode protein sequence context, and a Cross-Attention mechanism to model cross-domain interactions. Comprehensive experiments on two benchmark datasets, KIBA and Davis, across three rigorous scenarios: random splits, cold drug splits, and cold target splits demonstrate that GraphTransDTI achieves competitive performance compared to current state-of-the-art baseline models. Our findings confirm that the strategic combination of graph structural information and sequential attention mechanisms significantly enhances prediction accuracy and robustness in cold-start scenarios, offering a reliable and well-validated approach for high-precision virtual drug screening systems.

GraESM-FuseDTA: adaptive gated multimodal fusion of graph neural networks and protein language models for robust drug-target affinity prediction.

Multiview feature fusion-based graph representation model for drug-drug interaction prediction