In silico exploration of the metabolic solution space.

摘要

This work developed methods in constraint-based modeling to understand, interpret, and ultimately predict cellular behavior. The first detailed, physiologic interpretation of extreme pathways for a complete, elementally balanced metabolic system, the human red blood cell, was carried out. The 39 red blood cell extreme pathways were classified into nine physiologically meaningful groups. Next, the alpha-spectrum method was developed in which the extreme pathways that can and can not participate in the reconstruction of a steady state metabolic solution are identified. The alpha-spectrum method was used to interpret systems-level regulation of the red cell extreme pathways as a function of NADPH and NADH stressed states simulated using a dynamic model. The alpha-spectrum analysis was then applied to the 2441 extreme pathways of core Escherichia coli metabolism and experimentally measured fluxomic data were incorporated as additional constraints. Singular value decomposition of a condition-specific extreme pathway matrix (identified using the alpha-spectrum) illustrated the ability of the first mode to predict the network's nominal internal flux distributions. The results showed the utility of the alpha-spectrum method to greatly reduce the size of the metabolic solution space by eliminating up to 80% of the network's extreme pathways. The size of the metabolic solution space was determined using various methods. For sample networks, the volume was calculated exactly using methods of simplicial subdivision and signed decomposition. For larger metabolic systems, these algorithms are computationally intractable so Monte Carlo sampling was implemented to determine the relative size and shape of the steady state solution space. Reducing the v max for each reaction in the network by 50% caused a reduction in the size of the solution space that ranged from 7--96% depending on the reaction. Finally, measurements of kinetic parameters associated with single nucleotide polymorphisms (SNPs) of glucose-6-phosphate dehydrogenase and pyruvate kinase in 46 anemic patients were incorporated into a dynamic model of the human red blood cell. This study was a pioneering effort in which the in silico model of metabolism was used to qualitatively predict patients' phenotypes (pathophysiological states) from their genotype based on an analysis of the NADP/NADPH under nominal and stressed states.