Whole-body-level metabolic flux quantitation by machine learning

Project Summary Systemic diseases such as diabetes mellitus and metabolic syndrome affect multiple organs of the body. While the human body is naturally capable of self-healing, it faces an increasing challenge as multiple components of the systems of the human body go awry. Metabolism is a dynamic network of biochemical reactions that support cell proliferation and biosynthesis. On the

whole-body level, metabolic networks of individual tissues and organs are connected by the circulatory system and interfaced with the digestive and excretory systems. Our ability to cure systemic diseases relies on a quantitative understanding of whole-body metabolism, which requires comprehensive measurement of its dynamic states. However,

challenges arise from the lack of our ability to quantify metabolic fluxes (i.e., rates at which pathways are utilized) on a systems level. Metabolic fluxes are a direct readout for the dynamic state of metabolism but intangible deduced quantities that result from the catalytic interaction between metabolites and enzymes according to the kinetic and thermodynamic laws. Metabolic

flux analysis (MFA) framework allows quantitation of metabolic fluxes by imposing mass balances on all isotopologues resulting from stable isotope tracing experiments. As carbons form the molecular backbone, 13C-labeled substrates are extensively employed. The overarching aim of this project is to facilitate the measurement of metabolic fluxes on

muti-tissue and whole-body levels by tracing multiple isotope tracers. Knowledge of metabolic fluxes offers dual benefits of laying a solid foundation for understanding and controlling metabolism. To effectively achieve this computationally intensive goal, our teams at UCLA and Stevens will combine deep learning with analytical, stable isotope tracing, and simulation

techniques. Using multilayer neural networks, we will develop deep learning models that predict metabolic fluxes from the isotope labeling patterns of metabolites. With the augmented flux determination capability, we will impart quantitative systems-level knowledge of metabolism in individual and across tissues in co-cultures and animals.