Model-based inference of enzyme inhibitions from perturbation-induced metabolic dynamics

In microbes, metabolism plays a key role in the first rapid adaptation to sudden challenges such as nutrient limitations or toxic compounds. While metabolomics enables the profiling of stimuli-induced metabolic changes, computational approaches that can interpret these data to mechanistically explain how perturbations propagate through metabolism to produce the observed changes are lagging behind. Here, we developed a computational framework, called Inference from Metabolic Fingerprints (IMF), to model the immediate dynamic response to a metabolic perturbation and systematically infer its entry point (i.e. enzymatic target). IMF assumes small perturbations and linearizes the nonlinear dynamics around a reference steady state. This allows IMF to scale with large metabolic networks and bypass missing kinetic parameters by allowing for fast and efficient ensemble sampling. We apply IMF to a model of central metabolism in Escherichia coli. Using in-silico and experimental data, we demonstrate the ability to infer the target of metabolic perturbations in spite of unknown kinetic parameters and incomplete metabolic data. Hence, we show that IMF is an effective approach for designing, analyzing and interpreting time-resolved metabolomics.