Aug 2026· IEEE transactions on computational biology and bioinformatics· Vol PP· 0 citations
Medicine
TL;DR
SGTL-DDA is proposed, a novel graph transformer framework designed to incorporate structural information and domain-specific knowledge from heterogeneous biological information networks (HBINs) that successfully identifies both known therapeutics and novel repositioning candidates, supported by molecular docking results and literature evidence.
Abstract
Accurately predicting drug-disease associations (DDAs) is essential for accelerating the discovery of novel therapeutics. Graph representation learning-based computational models have become increasingly popular for this task due to their efficiency and cost-effectiveness. However, existing approaches often suffer from structural inductive biases and a limited ability to capture the rich heterogeneous context of biomedical molecules, which constrains their capacity to learn expressive drug and disease representations. To address this issue, we propose SGTL-DDA, a novel graph transformer framework designed to incorporate structural information and domain-specific knowledge from heterogeneous biological information networks (HBINs). SGTL-DDA integrates a meta-path-guided sampling strategy with a multi-level attention mechanism, enabling the model to jointly learn from both structural dependencies and attribute semantics in an end-to-end manner. Extensive experiments on two benchmark datasets demonstrate that SGTL-DDA consistently outperforms state-of-the-art methods in terms of Accuracy, F1-score, and AUC under a ten-fold cross-validation scheme. Furthermore, case studies on Alzheimer's disease and breast cancer confirm the predictive capability of SGTL-DDA, as it successfully identifies both known therapeutics and novel repositioning candidates, supported by molecular docking results and literature evidence.
Drug repurposing represents a cost-effective strategy to identify novel therapeutic applications for existing pharmaceuticals, circumventing the protracted timelines of traditional drug discovery. While knowledge graph (KG) based methods excel at integrating heterogeneous biomedical data, they often struggle to harmonize high-level domain knowledge with fine-grained molecular mechanisms. We propose KGDDA, a multimodal framework designed for drug-disease association prediction that synergistically integrates KGs with medical ontologies. By leveraging an attention-driven fusion mechanism, KGDDA dynamically merges contextual topological embeddings with ontology-derived priors, enabling the adaptive capture of intricate drug-disease interactions. Extensive evaluations on two benchmark datasets demonstrate that KGDDA consistently outperforms state-of-the-art baselines in both predictive accuracy and generalization. Furthermore, case studies on head and neck cancer and small cell lung cancer validate KGDDA's ability to provide actionable mechanistic insights, highlighting its potential to accelerate therapeutic discovery and precision medicine.
Qichang Zhao, Qiao Ling, Muhammad Habibulla Alamin et al.· IEEE transactions on computa...· 0 citations
VitaGraph is presented, a comprehensive multi-purpose biological knowledge graph built by integrating and refining multiple public datasets and enabling benchmarking of graph-based models and offering the opportunity to tackle tasks such as drug repurposing, PPI prediction, and side-effect prediction, among others.
Francesco Madeddu, Lucia Testa, Gianluca De Carlo et al.· Scientific Data· 0 citations
This work proposes 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.
Vang V. Le, Mai Thi Anh Nhu, Pham Truong Viet Thong· PLoS ONE· 0 citations
Drug–drug interactions (DDIs) are modulated not only by structural relations between drugs but are also profoundly influenced by underlying biological mechanisms. Most existing methods fail to adequately consider biologically interpretable signals at multiple biological hierarchies, limiting their ability to model the mechanistic complexity of drug–drug interactions. In this paper, we propose a heterogeneous semantic-aware framework (HSAF-DDI) that integrates molecular motif, protein sequence, and knowledge graph representations for DDI prediction. This framework learns local functional semantic representations of molecular motifs, fine-grained target protein semantics that capture deep biological characteristics, and higher-order associations encoded in knowledge graphs. We design a hierarchical adaptive fusion module that facilitates robust fusion and adaptive representation learning over multisource heterogeneous information. In addition, we introduce a contrastive learning mechanism with adversarial negatives and perturbations to improve the discrimination of DDI types and enhance the discriminability and robustness of learned representations. Experiments demonstrate that HSAF-DDI achieves superior overall performance compared to state-of-the-art methods, indicating the critical role of fine-grained biological features in improving the DDI prediction performance.
Xiaoli Lin, Siyuan Zhang, Bo Li et al.· Journal of Chemical Informat...· 0 citations
Abstract Motivation Polypharmacy is increasingly common in clinical practice, and the sheer number of possible drug combinations makes manual interaction screening impractical. Early computational approaches relied on chemical similarity metrics and rule-based systems, while subsequent machine learning and deep learning methods improved predictive power but continued to treat drugs as isolated entities, missing the broader biological context that governs interaction behaviour. Graph Neural Network (GNN) based methods address this by modeling drugs alongside proteins, diseases, and side effects in a shared relational graph, but tend to fall short on sparsely represented long-tail interaction classes due to the severe class imbalance that characterizes real-world biomedical interaction data. Results We construct a large-scale heterogeneous biomedical knowledge graph—26 408 nodes across five entity types and 1 679 387 edges across six relation types—and benchmark MLP, GCN, HGT, and RGCN for 105- class DDI prediction. RGCN achieves the strongest overall performance (Macro F1: 0.694, Recall: 0.720), with relation-specific weight matrices proving the critical factor in heterogeneous DDI modelling. Among class imbalance strategies tested, Tail-Aware Focal Loss outperforms standard cross-entropy by 3.1% on Macro F1 and 7.4% on Recall, striking a better balance between class performance than either weighted cross-entropy or weighted random sampling.
Varshini Venkatesh, Varsha G, Dhannya S. M.· Bioinformatics· 0 citations
Accurate prediction of drug-target interactions is pivotal for accelerating drug discovery and drug repurposing. However, existing advanced methods often fail to effectively characterize the many-to-many interactions between drugs and targets. Furthermore, they struggle to fully mine the structural features of drugs and proteins. To address these limitations, we propose IHLO-DTI, a novel prediction model based on an improved hypergraph neural network and Laplacian matrix optimization. First, we construct drug, protein, and drug−protein pair hypergraphs, where shared-drug and shared-target relationships are used to characterize multi-target activity and shared-target regulation. We then optimize hyperedge weights using a Laplacian matrix to enhance biologically meaningful high-order associations and suppress potential noise. Second, we use simplified graph convolution and graph convolutional network to extract global and local features, enabling efficient modeling of multi-level semantic information for drugs and targets. Next, we introduce a cross-attention mechanism and a dynamic gating module to perform fine-grained fusion of multi-channel features, improving the representation of cross-modal information interactions. Finally, we jointly train the model with contrastive learning and cross-entropy loss to enhance the consistency and discriminability of the learned representations. IHLO-DTI achieves mean AUROC and AUPR values of 0.9777 and 0.9716, respectively, on two public datasets. IHLO-DTI can effectively capture high-order many-to-many interactions between drugs and targets, improving prediction accuracy and robustness. It provides a more reliable computational tool for clinical drug screening, repurposing, and precision medicine research.
Guolongwei Dai, Tao Luo, Dandan Li et al.· ACS Synthetic Biology· 0 citations