mDADGAN: Predicting miRNA-Drug Response Associations Using lncRNAs and a Diffusion-Based Generative Adversarial...

Area of Science:

• Biochemistry and Molecular Biology
• Bioinformatics and Computational Biology
• Genomics and Genetics

Background:

• MicroRNAs (miRNAs) are crucial regulators of anticancer drug response, influencing sensitivity and resistance.
• Accurate prediction of miRNA-drug response associations is vital for understanding mechanisms and guiding targeted therapies.
• Existing prediction methods struggle with complex biological interactions, data sparsity, and class imbalance.

Purpose of the Study:

• To develop an advanced method for predicting microRNA-drug response associations.
• To address limitations of existing prediction techniques, particularly data sparsity and class imbalance.
• To enhance the accuracy and reliability of identifying potential miRNA-drug interactions for therapeutic guidance.

Main Methods:

• Constructed a diffusion-based generative adversarial network (mDADGAN) for miRNA-drug response prediction.
• Integrated lncRNA into a miRNA-drug interaction network and employed heterogeneous graph convolution to extract latent features.
• Applied diffusion-based perturbation for denoising and cost-sensitive learning to handle class imbalance during adversarial training.

Main Results:

• mDADGAN significantly outperformed six existing methods across multiple metrics (AUC, AUPR, ACC, Recall, F1-score) on four datasets.
• Integrating lncRNA information demonstrably improved the accuracy of miRNA-drug response association prediction.
• The method successfully identified known associations and predicted novel, evidence-supported candidate associations.

Conclusions:

• mDADGAN effectively predicts microRNA-drug response associations, offering a robust approach to overcome existing prediction challenges.
• The integration of lncRNA data enhances predictive performance, highlighting its importance in biological networks.
• The developed method provides valuable guidance for future biological experiments, potentially increasing efficiency and reducing costs in drug discovery.