The abstract is read at the publisher; the summary is JClub's.
Spec2Function infers biological functions directly from untargeted MS2 spectra, achieving higher similarity to reference functional descriptions across metabolite- and pathway-level benchmarks.
- Why it matters: Most LC–MS/MS features lack confident structural annotations, creating “functional dark matter” that limits downstream pathway and functional interpretation.
- What they did: The dual-encoder Spec2Biotext contrastively embeds MS2 spectra with curated multi-aspect functional text, then uses retrieval-augmented generation over curated databases and biomedical literature.
- The result: In a human liver cirrhosis cohort it recapitulated known metabolic alterations and found additional literature-supported functional modules, and in a 30-organ macaque aging atlas it revealed organ-specific and previously unreported functional modules supported by independent human and mouse datasets.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.