An efficient framework to identify key miRNA-mRNA regulatory modules in cancer

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Micro-RNAs (miRNAs) are key regulators of gene expression, influencing processes like RNA silencing and post-transcriptional control.
• Alterations in miRNA activity are implicated in the development and progression of human cancers.
• Accurate identification of miRNA-mRNA regulatory modules is crucial for understanding cancer biology.

Purpose of the Study:

• To develop a high-performance computational method for identifying cancer-specific miRNA-mRNA regulatory modules.
• To model complex miRNA-mRNA interactions using matched expression profiles from a large tumor cohort.
• To ensure biological relevance and predictive performance of the identified modules.

Main Methods:

• A two-step framework was employed, beginning with estimating a regulatory matrix from miRNA and mRNA expression profiles.
• A unified regularized factor regression (RFR) model was developed to estimate the number of modules and extract them by decomposing the regulatory matrix.
• The RFR model groups correlated miRNAs and mRNAs, controlling sparsity for interpretable and predictive results.

Main Results:

• The method was applied to over 9000 primary tumors across 32 TCGA cancer types.
• Identified miRNA-mRNA modules showed significant enrichment in Hallmark, PID, and KEGG pathways.
• Validation through literature and the miRTarBase database confirmed the biological relevance of the findings.

Conclusions:

• The proposed two-step RFR framework provides a robust and interpretable method for identifying cancer-specific miRNA-mRNA regulatory modules.
• The findings highlight the biological significance of these modules in cancer, offering potential insights for further research.
• The computational framework and associated scripts are publicly available for reproducibility and application.