Lei Li

Background: MicroRNAs (miRNAs) regulate biological processes and are associated with human diseases. Computational prediction can prioritize candidate associations for experimental investigation, but many existing methods emphasize nonlinear representations and may underuse complementary linear information and interactions between miRNA and disease features. This study aimed to integrate these complementary representations for miRNA–disease association prediction. Methods: We developed Singular Value Decomposition and Graph Autoencoder for miRNA–Disease Association prediction (SGMDA). Singular value decomposition extracted linear representations from the association matrix, while graph autoencoders learned nonlinear representations from integrated miRNA and disease similarity graphs. Contrastive learning aligned the two representation spaces, and a highway network modeled interactions between the fused miRNA and disease features. A random forest classifier generated association scores. Evaluation included pair-wise five-fold cross-validation, global leave-one-out cross-validation, miRNA-wise and disease-wise cross-validation, ablation experiments, and case studies of lung and prostate neoplasms. Results: The reported comparisons showed competitive predictive performance across the evaluated settings, with lower performance under entity-wise cross-validation than under pair-wise cross-validation. Ablation experiments supported the complementary contributions of linear and nonlinear features, contrastive alignment, and highway-based feature interactions. Many highly ranked case-study candidates were supported by at least one external database. Temporal validation using different HMDD releases provided additional retrospective evidence for prioritizing subsequently recorded associations. Conclusions: Combining linear and nonlinear representations with contrastive alignment and feature-interaction modeling provides a useful framework for prioritizing potential miRNA–disease associations. Database support does not establish biological causality or constitute strict independent validation, and the predicted associations require further experimental investigation.

Synergistic drug combinations have shown promising results in treating cancer cell lines by enhancing therapeutic efficacy and minimizing adverse reactions. The effects of a drug vary across cell lines, and cell lines respond differently to various drugs during treatment. Recently, many AI-based techniques have been developed for predicting synergistic drug combinations. However, existing computational models have not addressed this phenomenon, neglecting the refinement of features for the same drug and cell line in different scenarios. In this work, we propose a feature refinement deep learning framework, termed FRSynergy, to identify synergistic drug combinations. It can guide the refinement of drug and cell line features in different scenarios by capturing relationships among diverse drug-drug-cell line triplet features and learning feature contextual information. The heterogeneous graph attention network is employed to acquire topological information-based original features for drugs and cell lines from sampled sub-graphs. Then, the feature refinement network is designed by combining attention mechanism and context information, which can learn context-aware feature representations for each drug and cell line feature in diverse drug-drug-cell line triplet contexts. Extensive experiments affirm the strong performance of FRSynergy in predicting synergistic drug combinations and, more importantly, demonstrate the effectiveness of feature refinement network in synergistic drug combination prediction.

Synergistic drug combination prediction tasks based on the computational models have been widely studied and applied in the cancer field. However, most of models only consider the interactions between drug pairs and specific cell lines, without taking into account the multiple biological relationships of drug-drug and cell line-cell line that also largely affect synergistic mechanisms. To this end, here we propose a multi-modal deep learning framework, termed MDNNSyn, which adequately applies multi-source information and trains multi-modal features to infer potential synergistic drug combinations. MDNNSyn extracts topology modality features by implementing the multi-layer hypergraph neural network on drug synergy hypergraph and constructs semantic modality features through similarity strategy. A multi-modal fusion network layer with gated neural network is then employed for synergy score prediction. MDNNSyn is compared to five classic and state-of-the-art prediction methods on DrugCombDB and Oncology-Screen datasets. The model achieves area under the curve (AUC) scores of 0.8682 and 0.9013 on two datasets, an improvement of 3.70% and 2.71% over the second-best model. Case study indicates that MDNNSyn is capable of detecting potential synergistic drug combinations.