Redgem and Lumpgem: Systematic Approaches for Reduction of Genome-Scale Metabolic Models

Genome-scale metabolic reconstructions (GEMs) are invaluable resources for understanding and redesigning cellular networks as they encapsulate all the known biochemical processes occurring in organisms. Complexity of these large metabolic networks often hinders their utility in various practical applications. This difficulty sparked the interest for models of reduced sizes that are consistent with the genome-scale models. Currently available core models are reduced ad hoc with different aims and criteria, and up to date, there are no systematic reduction methods in the literature.

In this work, we have developed two methods, redGEM and lumpGEM, for constructing reduced metabolic models from compartmentalized or non-compartmentalized GEMs. redGEM uses as inputs: (i) definition of the metabolic subsystems/pathways that are of interest for a selected physiology; (ii) information about the extracellular nutrients and possible by-products; and (iii) any available physiological data from metabolomics, fluxomics, proteomics and transcriptomics. Firstly, redGEM employs a directed graph search method to find the pairwise connections (reactions) between the selected subsystems/pathways to account for all possible carbon flow routes. These subsystems (reactions and metabolites), along with the identified connections define the core network for the reduced model.  Then, we apply lumpGEM to construct subnetwork(s) and corresponding lumped reaction(s) that are capable of synthesizing all biomass building blocks from this core network. redGEM/lumpGEM allows the generation of different sized models with the following criteria: (i) different degree of connections between subsystems/pathways of interest, and (ii) alternative lumped reactions for the same biomass building block. The result of this procedure is reduced core model that are consistent with the original GEM in terms of flux profiles and essential reactions and genes. We used these two methods to generate reduced models for E. coli, S. cerevisiae and human metabolic model Recon 2 under aerobic/anaerobic conditions and generated GEM-compatible core models. We also compared the 3 organisms and highlighted the differences and similarities of their core carbon metabolism.