LRWRHLDA

LRWRHLDA predicts associations between long non-coding RNAs (lncRNAs) and human diseases by applying a Laplace normalized random walk with restart on integrated heterogeneous networks.

Key Features:

• Network-Based Framework: Constructs four isomorphic similarity networks: lncRNA similarity, disease similarity, gene similarity, and miRNA similarity.
• Heterogeneous Networks Integration: Integrates six heterogeneous association networks: lncRNA-disease, lncRNA-gene, lncRNA-miRNA, disease-gene, disease-miRNA, and gene-miRNA.
• Laplace Normalized Random Walk with Restart: Applies the Laplace normalized random walk with restart algorithm on the global multi-layer network to predict lncRNA-disease links.
• Performance Evaluation: Validated by ten-fold cross-validation with an Area Under the Curve (AUC) of 0.98402.
• Prediction of Isolated lncRNA Associations: Identifies disease-related lncRNAs that lack previously known disease associations.

Scientific Applications:

• Disease Research and Biomarker Discovery: Predicts candidate lncRNA biomarkers for cancers including colorectal cancer, lung adenocarcinoma, stomach cancer, and breast cancer, with predictions corroborated by independent studies.
• Understanding Disease Mechanisms: Maps lncRNA interactions with genes and miRNAs to provide insights into molecular mechanisms underlying disease progression.

Methodology:

Construct similarity networks for lncRNAs, diseases, genes, and miRNAs; integrate six heterogeneous association networks (lncRNA-disease, lncRNA-gene, lncRNA-miRNA, disease-gene, disease-miRNA, gene-miRNA); apply the Laplace normalized random walk with restart algorithm on the global network; validate predictions by ten-fold cross-validation and verify against independent studies.