A new artificial intelligence tool named AIMe (AI Molecule Explorer) has been developed to help scientists identify small molecules in the human body and its gut microbiome. These molecules play a key role in processes such as immunity and metabolism. AIMe uses a type of AI called neuro-symbolic AI to predict, organize, and search the mass spectra of over 100 million known small organic molecules. This creates a vast, searchable map of chemical structures, which can help researchers identify unknown compounds more efficiently. The tool was developed by Frank Schroeder of the Boyce Thompson Institute and Cornell University, along with Carla Gomes, director of Cornell's AI for Science Institute. The tool is available online, and its source code will be published after the study is released, with a preprint already posted on bioRxiv. Mass spectrometry is a widely used technique for identifying small molecules in fields like toxicology and food science. It works by breaking down a compound into fragments and recording their masses, creating a unique "fingerprint" known as a tandem mass spectrum (MS2 spectrum). Traditionally, scientists compare these fingerprints to known libraries of spectra or analyze the patterns manually. However, these libraries are limited, covering less than 1% of known compounds, and manual analysis is slow and labor-intensive. This makes it difficult to identify many unknown compounds quickly. AIMe takes a different approach by using computational methods to predict mass spectra and organize them into a searchable database called MS2KOSMOS. This database includes over 800 million predicted spectra, covering nearly all known small organic molecules in PubChem, the world's largest public chemical database. This is a massive increase compared to existing experimental libraries. At the heart of AIMe is a model called DeepMS2Reasoner, which simulates how molecules break apart in a mass spectrometer. It uses both chemical rules and machine learning to predict fragmentation patterns and generate interpretable results in chemical terms. To test AIMe's capabilities, researchers applied it to a dataset comparing the metabolomes of germ-free mice (those without gut bacteria) and mice with normal gut microbiota. They found thousands of chemical differences between the two groups, but most could not be identified using traditional methods. Using AIMe, the team analyzed the most abundant unknown compounds and found matches or related structures for about a third of them. For others, the tool helped identify molecular neighborhoods—groups of related compounds that might share similar properties. Two specific compounds were identified as unusual polyamine derivatives, a class of molecules important in biology. Using AIMe's predictions, the researchers were able to confirm their structures through synthesis, uncovering a previously unknown ring-shaped polyamine in human samples. This highlights how AI can accelerate the discovery of new compounds and improve understanding of biological processes.