Untargeted screening of drug-like compounds using high-throughput metabolomics uncovers inhibitors of key biological pathways

Metabolomics has demonstrated immense potential in elucidating mechanisms of action and off-target effects of known drugs. Here, we propose that metabolomic screening of cellular response to treatment with drug-like compounds can uncover clinically relevant inhibitors – driven by the extensive catalog of metabolic pathways implicated in human diseases. We developed a high-throughput LC–MS metabolomics pipeline paired with a computational Graph Neural Network for target deconvolution, and applied it to screen a library of 1,020 drug-like compounds in cultured cells in vitro. The screen revealed four drug-like compounds targeting key disease pathways, including inhibitors of purine and pyrimidine biosynthesis and an inducer of glutathione metabolism; validated via comprehensive metabolic assessments and growth rescue experiments. In a muscle atrophy model, driven by redox imbalance, the glutathione metabolism inducer reversed atrophy markers and restored protein content. Our scalable pipeline can be readily expanded to screen thousands of compounds, paving the way for the discovery of clinically relevant metabolic inhibitors. Metabolomics is widely used in drug development to reveal mechanisms of action. Here, the authors identify drug-like compounds with potent enzyme engagement by screening 1,020 compounds of unknown function, allowing them to discover nucleotide synthesis inhibitors and an atrophy reversing glutathione activator.