Section: New Results

Analysis of dynamic metabolomics data

An important step in the study of intracellular metabolism is the quantification of growth rates as well as uptake and excretion rates of metabolites in growing cellular populations. Traditional approaches are based on steady-state experiments, where time-invariant growth rates and exchange fluxes are measured in different experimental conditions. Technological advances in metabolomics have made it possible to monitor the concentration of extracellular metabolites over time, thus paving the way for the study of metabolism in transient conditions. Recovering time-varying exchange and growth rates from time-lapse metabolimics data is a key aspect of this challenge.

We have investigated the reconstruction of exchange reaction and growth rates from time-lapse measurements of external metabolite concentrations and population growth. In particular we have focused on the case of exhaustion of specific substrates, entailing sudden metabolic reorganization of the cell such as diauxie shifts in E. coli. Such discontinuities in the metabolic dynamics make data analysis and rate recosntruction particularly challenging but also information-rich. We have developed a Bayesian method that explicitly accounts for these sudden changes and the correlated adaptation of growth in order to accurately estimate time-varying exchange reaction and growth rates, and tested the method on real data from batch and fed-batch cultures of E. coli and L. lactis obtained at INRA/INSA Toulouse. The method is based on a time-inhomogeneous Gaussian process characterization of the rate dynamics, and Kalman smoothing techniques for the solution of the regularized estimation problem. Method and results were presented at the joint 2017 ISMB-ECCB conference, and published in the corresponding special issue of Bioinformatics [17]. The software implementing the method in Matlab is available at https://team.inria.fr/ibis/rate-estimation-software/, and has also been used for the data analysis in another joint publication with INRA/INSA Toulouse [20]. Further developments of the method are under consideration.