Thermodynamic and Probabilistic Metabolic Control Analysis of Riboflavin (Vitamin B2) Biosynthesis in Bacteria

Abstract

In this study, we applied a coupled in silico thermodynamic and probabilistic metabolic control analysis methodology to investigate the control mechanisms of the commercially relevant riboflavin biosynthetic pathway in bacteria. Under the investigated steady-state conditions, we found that several enzyme reactions of the pathway operate far from thermodynamic equilibrium (transformed Gibbs energies of reaction below about -17 kJ mol(-1)). Using the obtained thermodynamic information and applying enzyme elasticity sampling, we calculated the distributions of the scaled concentration control coefficients (CCCs) and scaled flux control coefficients (FCCs). From the statistical analysis of the calculated distributions, we inferred that the control over the riboflavin producing flux is shared among several enzyme activities and mostly resides in the initial reactions of the pathway. More precisely, the guanosine triphosphate (GTP) cyclohydrolase II activity, and therefore the bifunctional RibA protein of Bacillus subtilis because it catalyzes this activity, appears to mainly control the riboflavin producing flux (mean FCCs = 0.45 and 0.55, respectively). The GTP cyclohydrolase II activity and RibA also exert a high positive control over the riboflavin concentration (mean CCCs = 2.43 and 2.91, respectively). This prediction is consistent with previous findings for microbial riboflavin overproducing strains.

Uncertainty reduction in biochemical kinetic models: Enforcing desired model properties

“…For this type of studies, the ABC rejection algorithm [89] can be used as follows. First, the prior distribution of kinetic parameters is generated using the ORACLE framework or any other method that uses Monte Carlo sampling of uncertain parameters for constructing populations of kinetic models [3–5, 8, 9, 11, 20–28]. The corresponding control coefficients are next computed, and the parameter classification algorithm is then used to discard parameter vectors from the prior that gave rise to ambiguous control over analyzed quantities.…”

“…In biological systems, large uncertainties and partial experimental data commonly result in a population instead of in a unique set of parameter values that could describe the experimental observations. Such population of parameter sets is typically computed using Monte Carlo sampling techniques [3–5, 8, 9, 11, 20–28]. However, the problem is when certain properties differ among models in a model population.…”

“…For this purpose, we extended the capabilities of iSCHRUNK ( i n S ilico Approach to Ch aracterization and R eduction of Un certainty in the K inetic Models), a recently introduced machine learning approach that characterizes uncertainties in parameters of kinetic models, and identifies accurate and narrow ranges of parameters that can describe a studied physiological state [17]. In iSCHRUNK, machine learning is combined with methods that generate populations of kinetic models [3–5, 8, 9, 11, 20–28] to data-mine the integrated data and observed physiology together with the kinetic parameters. The extended iSCHRUNK workflow is amenable for identifying parameters that give rise to a wide variety of properties of metabolic responses.…”

iSCHRUNK – In Silico Approach to Characterization and Reduction of Uncertainty in the Kinetic Models of Genome-scale Metabolic Networks

“…This inspired the development of new modeling frameworks that exploit the sets of additional thermodynamic and physicochemical 6 constraints and integrate available data coming from several levels to reduce the space of admissible parameter values (Chakrabarti et al, 2013;Jamshidi and Palsson, 2010;Miskovic and Hatzimanikatis, 2010;Miskovic and Hatzimanikatis, 2011;Soh et al, 2012;Tran et al, 2008;Wang et al, 2004;Wang and Hatzimanikatis, 2006a;Wang and Hatzimanikatis, 2006b). Some of these approaches use Monte Carlo sampling techniques to extract populations of parameter sets capable of reproducing the observed physiology (Birkenmeier et al, 2015a;Birkenmeier et al, 2015b;Chakrabarti et al, 2013;Miskovic and Hatzimanikatis, 2010;Murabito et al, 2014;Soh et al, 2012;Tran et al, 2008;Wang et al, 2004;Wang and Hatzimanikatis, 2006a;Wang and Hatzimanikatis, 2006b). However, the sheer size of the admissible space that spans through the spaces of kinetic parameters, metabolite concentrations and metabolic fluxes along with the intrinsic nonlinearities of enzyme kinetics require tailored formulations and efficient parameter estimation techniques that are scalable and that can ultimately provide a detailed description of the metabolism.…”

Control Theory Concepts for Modeling Uncertainty in Enzyme Kinetics of Biochemical Networks

“…42, we obtain for the denominator 1 − a b Õ Ö × Ø , and the terms of the numerators are provided in Table 2. 36,37,[49][50][51][52][53][54][55][56][57][58][59][60][61][62] .…”