MVII-GCL is proposed, a novel multitask prediction framework built upon multiview information integration and graph contrastive learning that supports robustness against data sparsity and noise and its ability to discover potential associations.
Abstract
As a part of drug discovery, the exploration of circRNA-drug sensitivity associations (CRAs), drug-disease associations (RDAs) has been an important means of understanding disease mechanisms and proposing new treatment strategies. Due to the time-consuming and labor-intensive nature of biological experiments, computational methods have become effective solutions to this problem. However, existing computational methods for inferring CRAs, RDAs typically operate in isolation, failing to exploit the rich interplay across circRNAs, drugs, and diseases. In this paper, we propose MVII-GCL, a novel multitask prediction framework built upon multiview information integration and graph contrastive learning. MVII-GCL first constructs a multiview network encompassing attribute, topology, and association perspectives, explicitly tailored to capture the unique characteristics of each domain. To map multiview information into a unified latent space, our multiview attention encoder module employs domain-specific encoders to learn specialized representations, followed by a Top-k sparse attention mechanism. To further bolster robustness against data sparsity and noise, a symmetric graph contrastive learning objective is seamlessly integrated as a regularization term within the multitask learning paradigm. 5-fold cross-validation demonstrates our model's strong performance, achieving AUC, AUPR, and F1 scores of 0.9764, 0.9750, and 0.9234 in drug-disease association prediction, and 0.9301, 0.9339, and 0.8609 in circRNA-drug sensitivity prediction.Furthermore, case studies further support the model's ability to discover potential associations.
A multi-task prediction framework, capable of simultaneously predicting drug-disease, drug-protein, and disease-protein associations, is proposed, named MTP-DDA, demonstrating its effectiveness and robustness.
Ming-li Cui, Cui-Na Jiao, Daohui Ge et al.· 0 citations
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.
Bowei Zhao, Hui Zhao, Yu-an Huang et al.· IEEE transactions on computa...· 0 citations
IDEAL (Interpretability-Driven Evolvable Attentive Learning for Microbe-Drug Association) is proposed, a multi-view framework that integrates drug network topological attributes, BERT-encoded drug semantics, drug fingerprints, microbe genome sequence attributes, BERT-encoded microbe semantics, and microbe metabolic pathway attributes.
MRGBMDAT, a multi-relational graph encoder network with bilinear fusion for miRNA-disease association type prediction significantly outperforms five state-of-the-art methods across multiple evaluation metrics, exhibiting strong discriminative power and generalization capability.
Y. Sun, Wenjing Su, Siqi Zhu et al.· Bioinformatics· 0 citations
Noncoding RNAs (ncRNAs) are critical regulators of drug response and disease progression, making accurate prediction of ncRNA-drug resistance associations a key task in pharmacogenomics and precision medicine. However, current methods largely rely on global neighborhood aggregation, which treats node contexts as homogeneous and overlooks fine-grained structural and semantic heterogeneity. Moreover, they often model ncRNAs and drugs as interchangeable nodes, disregarding their biological distinctions and asymmetric interactions, and failing to effectively integrate modality-specific and cross-modal features. To overcome these limitations, we propose HDBI, a higher-order dynamic disentangled framework for predicting ncRNA-drug resistance associations. HDBI integrates multiview hypergraph learning, disentangled representation modeling, and bidirectional cross-modal updating to capture heterogeneous topological and semantic patterns within ncRNA and drug spaces while preserving modality-specific characteristics and enabling cross-modal information exchange. Extensive experiments on two benchmark data sets demonstrate that HDBI consistently outperforms state-of-the-art methods. Case studies on 5-FU and Docetaxel further support the biological relevance of the predictions, with 22/30 and 21/30 top-ranked ncRNAs supported by PubMed evidence, respectively. Functional enrichment and molecular docking analyses further linked these predictions to drug-relevant pathways and structurally plausible regulatory interactions. These findings suggest that HDBI provides an effective and interpretable framework for prioritizing ncRNA-mediated drug resistance associations and guiding downstream mechanistic investigation.
Tiyao Liu, Shudong Wang, Baoming Feng et al.· Journal of Chemical Informat...· 0 citations
Drive by the rapid envolution of deep-learning techniques, a large body of biological experiments bas has uncovered extensive associations between microRNAs (miRNAs) and complex human diseases, hig- hlighting the pivotal roles of miRNAs in pathogenesis. Elucidating these associations is essential for understanding disease mechanisms and developing preventive strategies. Traditional wet-lab validati- on, however, is notoriously labor- and resource-intensive, creating an urgent demand for efficient computational tools that can prioritize the most promising miRNA–disease candidates. Existing predictors predominantly rely on a single category of handcrafted features, thereby overlooking the complementary information embedded in multiple, heterogeneous data sources. Although a few recent attempts integrate diverse features, they usually exploit only a limited subset and fail to capture the intricate, non-linear relationships among them. To address these limitations, we propose MFCAMNet, a Multi-Feature fusion and Cross-Self-Attention model for MiRNA–Disease association prediction. Firstly, we construct multiple similarity matrices and employ two independent autoencoders with multi-source feature attention to obtain deep features of miRNA and disease to extract the inherent relationships between multiple features. Secondly, the proposed model employs element-level addition, element-level multiplication, and concatenation operations to generate miRNA-disease pair features with rich information. Finally, we use the encoder structure of the transformer to fuse the three deep features and predict all potential miRNA disease associations. We conducted comprehensive evaluations on the public HMDD v2.0 and HMDD v3.2 benchmark datasets. MFCAMNet achieved average AUCs of 0.9455 and 0.9420 under 5-fold and 10-fold cross-validation on HMDD v2.0, respectively, and an AUC of 0.9578 under 5-fold cross-validation on HMDD v3.2, outperforming state-of-the-art competitors. Case studies on breast, esophageal, and lung cancers further corroborate the reliability and practical utility of the proposed method.