Data-Driven Mechanistic Modeling of Untargeted Metabolome Data

Mathematical modeling can aid in understanding cell metabolism by allowing the construction of mechanistic metabolic models. The cell metabolism can be described as a biochemical reaction network. However, due to the reaction network size, identifying pathways used by the cells is challenging. Enumeration of all the possible pathways is challenging due to the number of combinations increasing exponentially with the reaction network size. Column generation (CG) is a data-driven approach for iteratively identifying a set of metabolic pathways. CG allows a reaction network of up to genome scale. In this chapter, we describe in detail how to carry out CG modeling for untargeted extracellular and intracellular metabolomic data. Metabolomics data are often scarce, and the metabolite concentrations from untargeted metabolomics are relative, posing challenges for modeling. We have demonstrated how to deal with these challenges and identify important metabolites and metabolic routes from untargeted metabolomic data [1]. Steps from initial key metabolite identification to reaction network modification, model construction, and model interpretation are explained. This allows the reader to apply the method to their own data and identify important metabolic pathways, where an interesting application can be to identify feed medium modifications to obtain better production for their mammalian cell-based bioprocess.