HEGANLDA

HEGANLDA predicts associations between long non-coding RNAs (lncRNAs) and human diseases by embedding nodes from an lncRNA–miRNA–disease heterogeneous network and applying supervised classification.

Key Features:

• Methodology: Integrates LDVCHN (LncRNA-Disease Vector Calculation Heterogeneous Networks) with HeGAN (Heterogeneous Embedding Generative Adversarial Networks) to map nodes in an lncRNA–miRNA–disease heterogeneous network into low-dimensional vectors.
• Predictive Model: Trains an XGBoost (eXtreme Gradient Boosting) classifier on the low-dimensional vectors to predict lncRNA-disease associations.
• Performance Evaluation: Evaluated by 10-fold cross-validation with an area under the ROC curve (AUC) of 0.983 and compared against five other methods.
• Robustness and Effectiveness: Validated through case studies and robustness tests confirming prediction effectiveness and reliability.

Scientific Applications:

• lncRNA–disease association discovery: Predicts candidate lncRNA-disease associations for experimental follow-up.
• Prioritization for validation: Ranks lncRNAs for targeted experimental validation in studies of disease mechanisms.
• Integrative heterogeneous-network analysis: Enables analysis combining lncRNA, miRNA, and disease interactions to study regulatory roles of lncRNAs in complex diseases.

Methodology:

Applies LDVCHN and HeGAN to embed nodes of an lncRNA–miRNA–disease heterogeneous network into low-dimensional vectors, trains an XGBoost classifier on these vectors, and evaluates performance using 10-fold cross-validation (reported AUC = 0.983); additional validation used case studies and robustness tests.