MWMM

MWMM clusters microRNA (miRNA) and messenger RNA (mRNA) expression data from The Cancer Genome Atlas (TCGA) to identify groups of interacting RNAs potentially implicated in cancer-related pathways.

Key Features:

• Graph theory algorithms: Applies the Hungarian algorithm and the blossom algorithm to a bipartite graph of miRNA–mRNA pairs to partition nodes based on edge weights derived from expression correlations.
• Edge-weight calculation: Quantifies edge weights by integrating tumor (T_CC) and normal (N_CC) correlation coefficients, evaluating six weight formulas and selecting integrated mean value weights as the optimal representation.
• Clustering criteria: Defines clusters by mathematical criteria requiring denser internal than external connections and by biological criteria where intra-cluster Gene Ontology (GO) term similarity exceeds inter-cluster GO term similarity.
• Application and versatility: Developed using breast invasive carcinoma (BRCA) TCGA data and demonstrated applicable across various cancer types, with superior identification of clusters exhibiting high GO term similarity compared with other algorithms.
• Research applications: Clusters miRNAs and mRNAs affected by common causal factors in cancer to support discovery of candidate biomarkers and inform precision medicine strategies.

Scientific Applications:

• Biomarker discovery: Identifies miRNA–mRNA clusters that serve as candidate biomarkers for cancer research.
• Functional interpretation: Detects clusters with high intra-cluster GO term similarity to support biological validation of grouped RNAs.
• TCGA-based cancer analysis: Applies to TCGA datasets, including BRCA, to cluster miRNA–mRNA interactions across cancer types.
• Comparative algorithm assessment: Enables comparison with other clustering algorithms, showing improved detection of clusters with strong GO term similarity.

Methodology:

Compute correlation coefficients for tumor and normal samples (T_CC and N_CC), quantify edge weights between miRNA–mRNA pairs using six candidate weight formulas (selecting integrated mean value weights), construct a bipartite graph, and apply the Hungarian and blossom algorithms to determine clusters that meet the stated mathematical and GO-based biological criteria.