RNACOREX

RNACOREX builds Bayesian Network-based classification models using miRNA-mRNA post-transcriptional networks to identify disease-associated coregulatory interactions and classify new observations of miRNA and mRNA quantifications.

Key Features:

• Curated Database Integration: Combines structural information from curated miRNA-mRNA interaction databases with expression data analysis.
• Conditional Mutual Information Inference: Uses conditional mutual information to infer reliable sets of miRNA-mRNA interactions from expression data.
• Conditional Linear Gaussian Classifiers: Builds probabilistic models based on CLG classifiers that enable prediction on new samples and validation of inferred networks.
• Graph-based Interpretability: Provides interpretability through a graph-based modeling framework that represents post-transcriptional coregulatory networks.
• Cross-cancer Interaction Analysis: Identifies miRNA-mRNA interactions that consistently appear across different cancer types.

Scientific Applications:

• Cancer Subtype Classification: Classifies patients according to survival time across multiple cancer types using TCGA data.
• Disease-associated Network Discovery: Identifies reliable sets of miRNA-mRNA interactions associated with specific diseases from high-throughput expression data.
• Post-transcriptional Regulation Study: Models combined activity of miRNA-mRNA interactions that regulate specific genetic behaviors.
• Biomarker Validation: Validates inferred networks through probabilistic classification performance on unseen observations.

Methodology:

RNACOREX integrates curated interaction databases with expression data analysis, applies conditional mutual information to infer miRNA-mRNA interactions, and models these interactions using Conditional Linear Gaussian classifiers within a Bayesian Network framework.