MetaSage: Machine Learning-Based Prioritization of Metabolic Regulators from Multi-Omics Data.

Area of Science:

• Biochemistry
• Bioinformatics
• Oncology

Background:

• Metabolic dysregulation is a key feature of cancer.
• The precise mechanisms governing cancer metabolism are not fully understood.

Purpose of the Study:

• To systematically investigate metabolic regulation across diverse cancer types.
• To develop a computational framework for identifying key regulators of cancer metabolites.

Main Methods:

• Developed MetaSage, an XGBoost-based machine learning pipeline integrating knowledge graphs and multi-omics data.
• Analyzed harmonized data from 15 cohorts across 11 cancer types.
• Utilized Shapley values to determine regulatory features for predictable metabolites.

Main Results:

• Identified 442 variable metabolites and significant gene and upstream metabolite regulators.
• Discovered 1,146 gene features and 363 precursor metabolites as crucial regulators.
• Network analysis revealed conserved and cancer-specific metabolic regulatory patterns.

Conclusions:

• MetaSage provides a robust framework for discovering metabolite regulation in cancer.
• Findings offer potential for developing novel therapeutic strategies targeting cancer metabolism.