DSCMF

DSCMF predicts associations between long non-coding RNAs (lncRNAs) and human diseases using dual sparse collaborative matrix factorization to support biomarker discovery and studies of disease mechanisms.

Key Features:

• Dual sparse collaborative matrix factorization: Enhances traditional collaborative matrix factorization by incorporating two specific improvements for lncRNA–disease association prediction.
• L2,1-norm regularization: Introduces L2,1-norm regularization to increase model sparsity and capture meaningful lncRNA–disease association patterns.
• Gaussian interaction profile kernel: Integrates a Gaussian interaction profile kernel to compute network similarity between lncRNA and disease profiles.
• Evaluation (ten-fold cross-validation): Performance is assessed by ten-fold cross-validation using the Area Under the Curve (AUC) metric (reported AUC = 0.8523).

Scientific Applications:

• Predicting lncRNA–disease associations: Prioritizes candidate lncRNA–disease links for further experimental validation.
• Biomarker identification: Assists identification of potential lncRNA biomarkers for disease diagnosis and prognosis.
• Understanding molecular mechanisms: Reveals lncRNA–disease relationships to support studies of disease mechanisms.
• Cancer-focused analysis (prostate, breast, ovarian, colorectal): Applied to analyses involving prostate, breast, ovarian, and colorectal cancers to investigate lncRNA roles.

Methodology:

Applies dual sparse collaborative matrix factorization with L2,1-norm regularization and a Gaussian interaction profile kernel, with performance evaluated by ten-fold cross-validation using AUC (reported 0.8523).