Metabolites are molecules that arise as intermediate or end products of an organism’s metabolism. In fungi and plants in particular, they exhibit an enormous structural diversity, which offers significant potential for the discovery of new active substances. However, searching for suitable compounds among thousands of metabolites from a single species can take many years. To accelerate this process, Prof. Dr Sebastian B?cker from Friedrich Schiller University Jena is developing methods in artificial intelligence and machine learning to predict whether an unknown small molecule possesses bioactive properties. The research project "BindingShadows" is funded by the European Research Council (ERC) through an ERC Advanced Grant. The Jena-based bioinformatician will receive approximately three million euros over the next five years, 500,000 euros of which are earmarked for new computing equipment.

“I sincerely congratulate Professor Sebastian B?cker on being awarded an ERC Advanced Grant,” says Prof. Dr Thomas Pertsch, Vice-President for Research and Innovation at Friedrich Schiller University Jena. “This outstanding success exemplifies the scientific excellence of the University of Jena. It demonstrates how internationally recognized cutting-edge research in bioinformatics can contribute to the identification of new bioactive molecules with the help of artificial intelligence—a crucial step on the path to future active substances.”

To identify the metabolites present in a sample, mass spectrometry is typically used. Tandem mass spectrometry fragments a molecule into its components, from whose masses structural information can be inferred. By comparing spectra, individual molecules can be structurally elucidated. “However, with our new method, we’re not aiming to identify the metabolites definitively,” explains Sebastian B?cker. “Many of the structures are completely unknown to humankind. What we’re interested in is whether they exhibit bioactivity—whether they can kill bacteria or fungi, or perhaps even fight cancer cells. We’re filtering for structures that are worth taking a closer look at.”

New structural representation combined with molecular fingerprinting

To this end, Prof. B?cker’s team will also analyze spectra using a newly developed fragmentation method to obtain more information about the molecule. “Our first goal is to find a new representation—a new form of depicting molecular structures—that contains more information and is interpretable by machine learning models,” says the Jena bioinformatician. “We combine this with the molecular fingerprint predicted using our molecule search engine CSI:FingerID, which we developed ten years ago.” All of this information enables the prediction of various types of bioactivity. This process is fundamentally similar to the prediction of substance classes for small molecules—a method called CANOPUS, which B?cker and his team published in the prestigious journal ?Nature Biotechnology“ in 2021.

New compounds and new cellular targets

On one hand, the University of Jena team plans to use the new AI-based method to search existing spectral databases of small molecules, discover promising metabolites, and recommend them for further analysis. These databases contain data on thousands of undiscovered small molecules waiting to be analyzed. On the other hand, the researchers in Jena aim to integrate the new tool into their well-established SIRIUS platform so users worldwide can quickly and easily derive information about bioactivity from their spectra. SIRIUS and related methods are used by thousands of researchers globally, and the group’s servers have already processed over a billion queries for annotating small molecules. “The search for new active compounds can be dramatically expanded and accelerated,” says Sebastian B?cker. “Processes that take months in the lab can be reduced to just minutes. Our methods will help discover many interesting and previously unknown molecules—and with a bit of luck, one or more of them might become the medicines of the future.”

ERC Advanced Grants

The ERC Advanced Grant is the most highly endowed European research funding for individuals and is also the most prestigious European award for outstanding scientists. It is awarded to established, active researchers with an exceptional scientific track record. Evaluation is based on the ten years preceding the application.