NCPHLDA

NCPHLDA predicts potential associations between long non-coding RNAs (lncRNAs) and human diseases by applying network consistency projection across integrated lncRNA cosine similarity, disease cosine similarity, and known lncRNA-disease association networks.

Key Features:

• Network Integration: NCPHLDA integrates a lncRNA cosine similarity network, a disease cosine similarity network, and a known lncRNA-disease association network.
• Parameterless and Negative Sample-Free: Operates without parameters or negative training samples.
• Predictive Capability for Uncharacterized lncRNAs: Can predict associations for lncRNAs with no previously known disease links.
• Network Consistency Projection: Infers associations by projecting relationships across the integrated networks using a network consistency framework.
• Predictive Performance: Reported AUC values are 0.9273 in leave-one-out cross-validation and 0.9179 ± 0.0043 in five-fold cross-validation.

Scientific Applications:

• Disease mechanism exploration: Understanding disease mechanisms by exploring lncRNA involvement.
• Biomarker and clinical inference: Facilitating early diagnosis, treatment, and prognosis through predicted lncRNA–disease associations.
• Experimental prioritization: Prioritizing lncRNA–disease candidates to reduce time and financial costs of experimental validation.
• Case studies: Validated examples include breast cancer, cervical cancer, and hepatocellular carcinoma.

Methodology:

NCPHLDA applies network consistency projection by integrating the lncRNA cosine similarity network, the disease cosine similarity network, and the known lncRNA-disease association network, based on the assumption that functionally similar lncRNAs are likely associated with phenotypically similar diseases and vice versa.