Metab8D: a metabolic regulome network from multiomics and machine learning

Area of Science:

• Cancer research
• Systems biology
• Metabolomics

Background:

• Understanding the metabolome's regulation is crucial for cancer research.
• Multiomic data offers a comprehensive view of cellular processes.

Purpose of the Study:

• To explore multiomic regulation of the metabolome using machine learning.
• To predict metabolomic variation across diverse cancer cell lines.

Main Methods:

• Utilized machine learning to analyze matched omics data (genomics, epigenomics, transcriptomics, proteomics) from ~1000 cancer cell lines.
• Investigated eight biomolecular classes including DNA methylation, RNA splice variants, and phosphoproteomics.
• Reconstructed multiomic interaction subnetworks for predictable metabolites.

Main Results:

• The metabolome is strongly associated with the transcriptome, with RNAs as top predictors.
• Peripheral metabolites correlate with enzyme levels, while central metabolites require complex pathway predictors.
• YAP1 signaling was identified as a major predictor across multiple omic layers.

Conclusions:

• Multiomic data, particularly transcriptomics, can predict metabolomic variation in cancer.
• Predictive features highlight potential synthetic-lethal interactions and combination therapies.
• Prioritized features for advanced metabolomics assays like single-cell and spatial.