AI redefines mass spectroscopy

Chemists and computer scientists from the CNRS, Paris-Saclay University and Paris Sciences & Lettres University have developed MS2Decide, an automated chemical prospecting tool based on machine learning. Designed to cross-reference results provided by mass spectroscopy software, which sometimes may rely on databases that contain inconsistencies, this programme detects previously unidentified molecules in natural extracts. Presented in the journal Chemistry - Methods, this technology relies on the reliability and degree of agreement between different results and classifies molecules according to a score that assesses the probability that they are already known and catalogued. This open-access programme can be customised according to the specific expertise of each laboratory.