Despite rapid technological progress, the discovery of novel antibiotics has been stalled for the past 50 years. To combat the growing burden of antibiotic resistance, innovative drug discovery paradigms are required to improve and expedite the antibiotic discovery process. A crucial bottleneck in drug discovery is the identification of compounds’ Mode of Action (MoA). To address this problem we developed a rapid and systematic metabolome profiling strategy to classify the MoA of bioactive compounds. In contrast to existing methods based on phenotypic drug profiling, mostly on the basis of growth assays, we exploit here the intracellular response of about 1000 metabolites as a truly multiparametric readout of the cellular response. The specific advance over existing metabolic platforms is a faster throughput of 1-2 orders of magnitude, allowing our combined MS-based metabolomics and computational workflow to scale with the size of typical compound libraries. I will present how an entropy based measure of similarity can be used to predict antibiotic MoAs from large compendia of metabolome profiles, and the results obtained from analyzing an open access set of ~200 novel anti-tuberculosis compounds with unknown MoAs.