Cornell's AIMe uses AI to predict mass spectra for 100 million small molecules
Researchers at the Boyce Thompson Institute and Cornell built AIMe, a neuro-symbolic AI tool that maps the mass spectra of more than 100 million known small organic molecules; the work is still a preprint.
Scientists at the Boyce Thompson Institute and Cornell University have built AIMe (AI Molecule Explorer), a tool that uses neuro-symbolic artificial intelligence to predict, organise and search mass spectra for the whole catalogue of known small organic molecules — more than 100 million compounds.
The problem
Mass spectrometry experiments in metabolomics produce huge numbers of signals that cannot be matched to any known molecule, because reference spectra exist for only a small fraction of chemicals. AIMe generates predicted spectra so that those unknown peaks can be compared against a vast searchable map of chemical space.
Early results
- Applied to more than 7 million spectral clusters in the GNPS public database, AIMe returned putative annotations for about 2.69 million, compared with roughly 416,000 annotated in an earlier effort using the same similarity threshold — a more than sixfold increase.
- The project is led by Frank Schroeder (BTI/Cornell chemistry) and Carla Gomes (Cornell AI for Science Institute).
Caveats
The study has been posted as a bioRxiv preprint and has not yet completed peer review. Annotations are putative predictions that still need experimental confirmation. AIMe is available via Cornell’s website; source code is to be released on publication.
Also available in: Саха тылаРусскийEspañol中文Portuguêsالعربية