Serendipity is to thank for several ground-breaking discoveries and treatments, such as Alexander Fleming’s accidental discovery of penicillin.1 However, most drug breakthroughs stem from systematic investigations designed to identify compounds targeting specific molecules or pathways. But these strategies are not without problems.
Low-throughput and target-specific methods for characterizing lead compounds can result in a time-consuming and expensive drug development process.2 To overcome this, researchers have increasingly turned to high-throughput assays.
Now, using a high-throughput method to examine the metabolic landscape of cells, researchers assessed the effects of more than 1,500 compounds and discovered new modes of action for drugs that are already approved and in use.3 The results, published in Nature Biotechnology, highlight how high-throughput metabolomics can facilitate drug target research, providing a complementary platform to increase efficiency and cut costs in drug discovery.
Researchers have previously profiled the transcriptomes and proteomes of cells to map drug effects. “[But] a protein can interact with hundreds of other proteins, so it’s difficult to predict what would be the final outcome of a drug treatment based just on the knowledge of the target” said Mattia Zampieri, a systems pharmacologist at the University of Basel and a study coauthor. Understanding the drug’s effect on metabolism can complement other molecular readouts and reveal the cellular changes following the drug’s binding to the target, he explained.
Zampieri and his team cultured lung cancer cell lines and treated them with one of 1,520 compounds, most of which are approved by the US Food and Drug Administration. Mass spectrometry revealed that a majority of the drugs triggered significant changes in at least one metabolite. They observed similar results in ovarian and breast cancer cell lines.
To estimate how far drug-induced changes occur from their intended metabolic targets, the researchers systematically treated cancer cells with around 350 drugs that target different enzymes, selected from the pool of 1,520. Then, using a metabolic map of biochemical reactions in the cell, they checked the proximity of the changes with the targeted enzymes.
“You can imagine [the metabolic map] as a city map, where each road is an enzymatic reaction that you can potentially target with a drug,” said Zampieri. The team analyzed how far away a “traffic jam,” symbolizing metabolite buildup due to enzyme inhibition, occurred upon blocking a particular “road,” he explained.
They found that different drugs induced changes at different distances from the target, offering deeper insights into their modes of action. While immunology and oncology drugs triggered changes up to two steps away from their enzymatic targets, some cardiovascular and endocrinology-related drugs caused reactions farther away.
“This can help to understand what could be secondary drug effects or potentially side effects of the drug,” noted Zampieri. “And this also [creates an] opportunity to use already approved drugs potentially in completely different therapeutic areas than the one that they are typically used.”
Zampieri and his team compared metabolic profiles between drugs with similar targets and found that they induced similar metabolic changes. They used this information to predict the modes of action for drugs based solely on their metabolic changes.
The researchers then tested these metabolome-based predictions on drugs with unexpected modes of action. Data collected from high-throughput in vitro assays supported their predictions, underscoring the ability of metabolic profiling to uncover drug actions.
Finally, the researchers investigated whether transcriptomes or proteomes could similarly predict drug modes of action. They found that certain profiling methods were better at predicting the modes of action for different drug classes than others, suggesting that there is no one-size-fits-all approach for making these predictions. For instance, transcriptomes better predicted the modes of action of DNA alkylating drugs, while metabolic signatures were effective in predicting the action of selective serotonin uptake inhibitors.
“I don't think that metabolomics alone is able to solve the entire understanding drug mode of action, [which] is a very complex problem,” said Zampieri. “Metabolomics is one tool that we add to the arsenal of tools that we can use.”
“The overall experimental design is very strong,” said Jessica Ewald, a computational biologist and an incoming research faculty at the European Bioinformatics Institute, who was not involved in the study. “[This] is a really good example of how a new data type could give more information on a certain class of compounds that may not be as well detected by the other data types.”
However, Ewald noted that the cell lines tested are not typically considered metabolically active, and it would be interesting to test this framework in liver cells. “The most exciting thing is this paper really laid the groundwork to show that their method is reliable and that it produces results that are generally trustworthy and consistent with profiling field.”
Zampieri and his team hope to conduct similar studies in systems like organoids that more closely capture the tissue physiology. But overall, he believes that the tool they have developed can enable people to accelerate the drug discovery process.
- Ban TA. The role of serendipity in drug discovery. Dialogues Clin Neurosci. 2006;8(3):335-344.
- Sun D, et al. Why 90% of clinical drug development fails and how to improve it? Acta Pharm Sin B. 2022;12(7):3049-3062.
- Schuhknecht L, et al. A human metabolic map of pharmacological perturbations reveals drug modes of action. Nat Biotechnol. 2025
