A high-throughput metabolomics method to predict high concentration cytotoxicity of drugs from low concentration profiles

A major source of drug attrition in pharmacological development is drug toxicity, which eventually manifests itself in detrimental physiological effects. These effects can be assessed in large sample cohorts, but generating rich sets of output variables that are necessary to predict toxicity from lower drug dosages is problematic. Currently the throughput of methods that enable multi-parametric cellular readouts over many drugs and large ranges of concentrations is limited. Since metabolism is at the core of drug toxicity, we develop here a high-throughput intracellular metabolomics platform for relative measurement of 50-100 targeted metabolites by flow injection-tandem mass spectrometry. Specifically we focused on central metabolism of the yeast Saccharomyces cerevisiae because potential cytotoxic effects of drugs can be expected to affect this ubiquitous core network. By machine learning based on intracellular metabolite responses to 41 drugs that were administered at seven concentrations over three orders of magnitude, we demonstrate prediction of cytotoxicity in yeast from intracellular metabolome patterns obtained at much lower drug concentrations that exert no physiological toxicity. Furthermore, the ¹³C-determined intracellular response of metabolic fluxes to drug treatment demonstrates the functional performance of the network to be rather robust, until growth was compromised. Thus we provide evidence that phenotypic robustness to drug challenges is achieved by a flexible make-up of the metabolome.