DNI-MDCAP

DNI-MDCAP predicts causal microRNA (miRNA)-disease associations by integrating miRNA similarity metrics, deep graph embedding learning-based network imputation, and a semi-supervised learning framework to prioritize causal miRNA–disease relationships.

Key Features:

• miRNA similarity metrics: Incorporates additional miRNA similarity metrics to improve identification of biologically related miRNAs.
• Deep graph embedding network imputation: Applies deep graph embedding learning-based network imputation to infer missing links in miRNA-disease interaction networks.
• Semi-supervised learning: Employs a semi-supervised learning framework combining labeled and unlabeled data to enhance predictive accuracy.
• Performance evaluation: Assessed by tenfold cross-validation (AUROC 0.896) and independent testing (AUROC 0.889).
• Comparative benchmarking: Outperforms existing models such as MDCAP and LE-MDCAP, reporting an AUROC of 0.870 for distinguishing causal from non-causal associations.
• Statistical validation: Uses the Wilcoxon test to show significantly higher prediction scores for causal versus non-causal associations.
• Literature validation: Predictions include associations such as diabetic nephropathies and hsa-miR-193a that have been validated in recent literature.

Scientific Applications:

• Causal miRNA-disease association prediction: Prioritizes causal relationships between miRNAs and diseases.
• Distinguishing causal from non-causal associations: Enables discrimination between causal and non-causal miRNA-disease links.
• Disease mechanism and therapeutic target insights: Provides insights into disease mechanisms and potential therapeutic targets via prioritized causal associations.

Methodology:

Uses additional miRNA similarity metrics, deep graph embedding learning-based network imputation, and a semi-supervised learning framework; evaluated by tenfold cross-validation and independent testing with AUROC reporting and statistical comparison via the Wilcoxon test.