Researchers at the Boyce Thompson Institute and Cornell University have unveiled AIMe, a neuro-symbolic AI tool designed to predict, organize, and search mass spectra for over 100 million small organic molecules. The technology expands searchable chemical space a thousandfold to solve a long-standing biomedical bottleneck.
Biomedical science has long wrestled with an invisible blind spot: more than 80 percent of small molecules detected in a typical biological sample cannot be matched to any known chemical structure. This persistent gap has stalled discoveries linking the gut microbiome, immunity, and metabolism. To tackle the problem, a collaborative team led by Frank Schroeder at the Boyce Thompson Institute alongside Carla Gomes at Cornell University developed an artificial intelligence tool known as AI Molecule Explorer, or AIMe.
How AIMe Predicts and Maps Chemical Space
Mass spectrometry serves as the workhorse for small molecule identification across toxicology and food analysis. Instruments fragment a compound and record the masses of the resulting pieces to create a tandem mass spectrum, or MS2 spectrum, which acts as a molecular fingerprint. Traditionally, researchers compare unknown spectra against sparse reference libraries or interpret fragmentation patterns manually. Available experimental libraries cover fewer than 1 percent of known compounds, and resolving a single unknown structure can take days to months of iterative analysis.
AIMe bypasses this reliance on sparse experimental libraries by predicting spectra computationally. The system organizes these predictions into a searchable resource called MS2KOSMOS, generating more than 800 million predicted spectra that cover essentially all known small organic molecules in PubChem. At the heart of the platform sits a specialized model.
“At the core of AIMe is DeepMS2Reasoner, a model that simulates how molecules fragment inside a mass spectrometer, It builds fragmentation pathways step by step, using symbolic chemical rules to enumerate physically plausible fragmentation steps and a neural network to assign likelihoods to each step. The result is a predicted spectrum and an annotated map of how a molecule came apart-a feature that makes AIMe’s outputs interpretable in chemical terms, not just computationally useful.”
Carla Gomes, Professor of Computing and Information Science at Cornell University
Tracing Unknown Compounds from Mouse Guts to Human Biology
To test the platform in a real biological setting, the research team applied the tool to a comparative metabolomics dataset from mice. The experiment contrasted germ-free animals raised without gut microbiota against mice possessing a normal complement of gut bacteria.
Broader Implications for Metabolomics and Disease Research
The expansion of searchable chemical space arrives alongside a growing recognition across the medical community that dysregulated metabolites drive complex conditions. Advanced chemo-proteomics and AI-driven structural modeling are increasingly applied to understand how small molecules interact with cellular machinery. While the AIMe framework maps mass spectra to unearth hidden microbial compounds, parallel efforts in metabolomics are tackling the fraction of the metabolites that exist in our bodies that remain unmeasured in human biology.
By turning traditional discovery models around—focusing on native molecular interactions rather than screening random chemical libraries—researchers are establishing logical starting points for therapeutic development. The convergence of neuro-symbolic artificial intelligence and mass spectrometry transforms unannotated spectral data into actionable chemical leads, setting a new baseline for how the scientific community maps the hidden universe of small molecules.
