IMPMD

IMPMD predicts potential associations between microRNAs (miRNAs) and diseases to prioritize candidate disease-associated miRNAs for diagnostic biomarker discovery.

Key Features:

• Enhanced similarity measures: For miRNAs, IMPMD combines functional similarity, Gaussian similarity, and Jaccard similarity, and for diseases it integrates semantic similarity, Gaussian similarity, and Jaccard similarity.
• Feature construction and selection: Features are constructed from the enhanced similarity measures and a cumulative score is calculated to select robust features indicative of potential associations.
• Machine learning integration: General linear regression is used to assign weights to three classifiers—Support Vector Machine (SVM), K-Nearest Neighbor (KNN), and Logistic Regression—for final prediction.
• Performance evaluation: Model performance was assessed by 10-fold cross-validation with an Area Under the Curve (AUC) of 0.9386 and validated in case studies on lung cancer and breast cancer.

Scientific Applications:

• miRNA–disease association prediction: Prediction of novel miRNA-disease associations using integrated similarity measures and weighted classifier outputs.
• Biomarker discovery: Prioritization of candidate miRNA biomarkers for disease diagnosis and molecular mechanism studies.
• Case study validation: Application to lung cancer and breast cancer with experimental validation of 49/50 and 50/50 top predicted miRNAs, respectively.

Methodology:

IMPMD combines functional, Gaussian, and Jaccard similarities for miRNAs and semantic, Gaussian, and Jaccard similarities for diseases, constructs features from these similarities, computes a cumulative score for feature selection, uses general linear regression to weight SVM, KNN, and Logistic Regression, and evaluates performance by 10-fold cross-validation (AUC 0.9386).