RGCNMDA: Pathway-Bridged Relational Graph Learning with Adaptive Multi-View Fusion for Human miRNA-Disease Association Prediction.

MRGBMDAT: a multi-relational graph encoder network with bilinear fusion for miRNA-disease association type prediction

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.

RLNSF-MDA: reliability-guided graph-regularized matrix factorization for immune-related miRNA–disease association prediction

Introduction MicroRNAs (miRNAs) regulate gene expression and are closely linked to the onset and progression of immune‑related diseases. Experimental discovery of disease‑associated miRNAs remains costly and time‑consuming, motivating computational prioritization. However, existing matrix‑completion and deep graph‑learning methods either underuse local biological neighborhood evidence or require complex multi‑view neural architectures. Methods In this study, we present RLNSF‑MDA, a reliability‑guided graph‑regularized logistic matrix factorization framework for predicting potential miRNA–disease associations, with an emphasis on immune disease analysis. The method integrates disease semantic similarity; miRNA functional, semantic, and sequence similarities; and training‑only Gaussian interaction profile similarities through data‑driven reliability weights. It then combines multi‑scale neighborhood evidence, diffusion scores, low‑rank reconstruction features, and contrastive graph‑regularized latent factors. Results On the HMDD v3.2 benchmark containing 788 miRNAs and 374 diseases, RLNSF‑MDA achieved an average accuracy of 0.8525, an AUC of 0.9194, and an AUPR of 0.9055 in five‑fold cross‑validation, and an average accuracy of 0.8569, an AUC of 0.9266, and an AUPR of 0.9197 in ten‑fold cross‑validation. Ablation experiments showed that the full model outperformed reduced feature combinations and training strategies, supporting the contribution of reliability‑guided fusion, side‑score construction, graph regularization, and contrastive ranking. Evidence from immune‑related case studies in dbDEMC V2.0 and miR2Disease further covered lupus nephritis, lymphoma, and leukemia, with 45, 47, and 47 confirmed miRNAs, respectively. Discussion These results suggest that RLNSF‑MDA provides an effective and interpretable framework for prioritizing candidate miRNAs associated with immune‑related diseases.

Computational Method Using Attribute-Aware Message Passing and Graph Convolutional Network for Potential miRNA–Disease Association Prediction

MicroRNA (miRNA) dysregulation is a crucial pathogenic factor that extensively participates in the occurrence and progression of various human diseases, especially cancers. Identifying unknown miRNA–disease connections is essential for understanding disease pathogenesis and improving clinical treatment strategies. Traditional biological experiments are often expensive and technically restricted, so computational prediction has become a widely used auxiliary research tool. In this study, we develop a novel predictive model called Attribute-Aware Message Passing Graph Convolutional Network (AAMPGCN) to identify potential miRNA–disease associations. The advantage of AAMPGCN lies in integrating miRNA and disease attribute information into the message-passing process: it partitions the miRNA–disease heterogeneous graph that incorporates miRNA functional similarity, disease semantic similarity, and Gaussian interaction kernel similarity into attribute-homogeneous subgraphs, while restricting high-order message propagation within each subgraph. This mechanism effectively filters cross-attribute noise, preserves the discriminability of miRNA and disease embeddings during deep convolution, and is thus well-adapted to miRNA–disease heterogeneous networks. The AAMPGCN prioritizes miRNA and disease attributes, aggregating messages specifically among nodes with similar attribute characteristics that are relevant to miRNA–disease interactions. Experimental results show that the AAMPGCN model achieves AUC and AUPR values of 94.06 and 93.52 on the HMDD2.0 dataset, which outperforms existing methods. The proposed AAMPGCN provides a new and effective method for miRNA–disease association prediction, and also offers theoretical support for the research on disease molecular mechanisms and the screening of clinical therapeutic targets.

A Hybrid Graph Convolutional Network and XGBoost Framework for miRNA–Disease Association Prediction

GCN-XGB is a novel hybrid computational framework that integrates a two-layer Graph Convolutional Network with Extreme Gradient Boosting to improve the accuracy of miRNA-disease association prediction and is suggested to be a powerful and reliable tool for identifying potential disease-related miRNAs and prioritizing candidates for experimental validation.

Heterogeneous dual-channel and interpretable graph representation learning with global virtual nodes for microRNA-mediated drug sensitivity prediction.

Dual-channel learning framework for microbe-disease association prediction via subgraph enhanced dynamic weight–calibrated graph convolutional and cross-pooling attention networks

Identifying potential microbe–disease associations (MDAs) is vital for elucidating disease pathogenesis and advancing precision medicine. Existing methods primarily learn features from heterogeneous microbe–disease graphs, but often rely solely on global topology for feature propagation, ignoring neighborhood subgraph density, centrality, and edge-weight heterogeneity. The loss of such structural information further exacerbates distributional shifts of microbe–disease feature representations, making it difficult for static concatenation or average fusion to effectively bridge the semantic gaps between modalities or accurately capture the contextual dependencies between node pairs. To address these challenges, we propose STDCAMDA, a dual-channel learning framework for MDA prediction. For structural enhancement, we design a subgraph topology module that fuses global and local topological information via multidimensional edge weights and node-gating mechanisms, thereby encoding rare structural signals while suppressing noise. In feature learning, we adopt a dual-channel strategy: embedding dynamic weight correction into a graph convolutional network for adaptive adjacency calibration and building a cross-pooling attention network to mitigate modality distribution shifts and capture cross-modal dependencies. Finally, we introduce two strategies: a dynamically weighted fusion classifier that integrates dual-channel features and uses a multi-layer perceptron for prediction, and a subgraph-aware negative sampling strategy that selects hard negative samples. Experiments on the Disbiome and HMDAD datasets demonstrate that STDCAMDA outperforms seven existing MDA prediction models, with statistical significance tests confirming the reliability of these improvements and cold-start evaluations validating its robustness and generalization capability. Practically, STDCAMDA provides an effective computational framework for prioritizing candidate disease-related microbes and supporting downstream biomedical validation.