MORE interpretable multi-omic regulatory networks to characterise phenotypes.

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Understanding phenotype-specific regulatory mechanisms is key to deciphering disease molecular underpinnings.
• Current methods for constructing multi-omic regulatory networks (MO-RN) are limited in integrating diverse omics data, prior knowledge, and inferring phenotype-specific networks.

Purpose of the Study:

• To introduce MORE (Multi-Omics REgulation), a novel R package designed for inferring multi-modal regulatory networks.
• To provide a flexible tool capable of handling any number and type of omics layers, optionally incorporating prior biological knowledge.

Main Methods:

• MORE utilizes advanced regression-based models and variable selection techniques to identify significant regulatory relationships.
• The package supports integration of diverse omics modalities and prior biological information.
• Network visualization, differential network analysis, and functional enrichment analyses are included for biological interpretation.

Main Results:

• MORE demonstrated superior performance in accuracy, model fit, and computational efficiency compared to existing state-of-the-art tools on simulated datasets.
• Application to an ovarian cancer multi-omic dataset revealed distinct regulatory patterns specific to tumor subtypes and survival outcomes.

Conclusions:

• MORE addresses limitations in existing MO-RN inference methods, offering a valuable resource for studying complex regulatory systems.
• The tool's ability to construct accurate and interpretable phenotype-specific regulatory networks facilitates research into molecular interactions and regulatory mechanisms.