Identifying Tissue- and Cohort-Specific RNA Regulatory Modules in Cancer Cells Using Multitask Learning

Area of Science:

• Genomics
• Bioinformatics
• Cancer Biology

Background:

• MicroRNA (miRNA) alterations are key drivers in human cancer development and progression.
• miRNAs regulate gene expression by interacting with messenger RNAs (mRNAs).
• Understanding miRNA-mRNA interactions is vital for deciphering tumor heterogeneity and enabling precise cancer diagnosis and treatment.

Purpose of the Study:

• To develop a novel computational method for identifying tissue- and cohort-specific miRNA-mRNA regulatory modules.
• To analyze expression profiles of tumor tissues to uncover these regulatory relationships.
• To support precision medicine by defining tumor-specific molecular signatures.

Main Methods:

• A multitask learning sparse regularized factor regression (MSRFR) method was established.
• MSRFR simultaneously models sparse miRNA-mRNA relationships and extracts tissue- and cohort-specific modules.
• The model was validated on multiple cancer cohorts from different tissues (blood, kidney, lung).

Main Results:

• MSRFR effectively identifies cancer-related miRNAs within cohort-specific modules.
• The method successfully distinguishes between tissue-specific and cohort-specific regulatory modules.
• Tissue-specific information was extracted from diverse cancer cohorts, highlighting MSRFR's versatility.

Conclusions:

• The MSRFR model accurately determines miRNA-mRNA regulatory modules specific to cancer type and tissue.
• This approach enhances the understanding of tumor biology and heterogeneity.
• MSRFR provides a valuable tool for precision medicine, enabling the definition of tumor-specific miRNA-mRNA signatures for improved diagnostics and therapeutics.