(626b) Optimizing the Biomass Equation of a Genome-Scale Model Using Simple Measurements Obtained By Raman Spectroscopy
The utility of genome-scale metabolic flux modeling has been demonstrated over the past decade. However, it has also been shown fairly recently that the chemical composition of the cell (i.e., the biomass equation) exerts major influence on the flux results returned by flux balance analysis (FBA). Studies have also shown that flux solutions returned by 13C-labeling experiments and metabolic flux analysis (13C-MFA) do not necessarily agree with FBA results. Here, we present some of the first research designed to improve FBA predictions to match measured flux values obtained by 13C-MFA. We hypothesized that more accurate representations of the cell chemical composition in the biomass equation of a genome-scale metabolic flux model would produce flux predictions that agree with measured values. We have tested this hypothesis with two approaches. The first involves the use of Raman spectroscopy to obtain critical cell composition data that can be used in the biomass equation of a genome-scale model. Raman spectroscopy is advantageous for this application because (i) data can be obtained in near real-time, (ii) minimal sample preparation is required, and (iii) a single Raman spectra contains a snapshot of the entire chemical composition of a single cell or entire culture. The second approach involves non-linear optimization of the biomass equation by minimizing uptake/secretion processes not observed in the laboratory but required by the model to satisfy the mass balance. Using these techniques, we successfully modeled the growth of the cellulose utilizer Clostridium cellulolyticum ATCC 35319 through several distinct growth phases using a newly developed genome-scale model for this organism. We have also reconciled FBA predictions with 13C-MFA measurements in Escherichia coli K12. This research demonstrates the discrepancies that arise when using approximated biomass equations in genome-scale modeling. The new tools presented here are applicable to all genome-scale models and provide researchers with a means of minimizing the bias in flux predictions arising from sub-optimal biomass equations.