Deep learning prediction of electrospray ionization tandem mass spectra of chemically derived molecules.

ABSTRACT: Chemical derivatization is a powerful strategy to enhance sensitivity and selectivity of liquid chromatography-mass spectrometry for non-targeted analysis of chemicals in complex mixtures. However, it remains impossible to obtain large sets of reference spectra for chemically derived molecules (CDMs), representing a major barrier in real-world applications. Herein, we describe a deep learning approach that enables accurate prediction of electrospray ionization tandem mass spectra for CDMs (DeepCDM). DeepCDM is established by transfer learning from a generic spectrum predicting model using a small set of experimentally acquired tandem mass spectra of CDMs, which converts a generic model with low predictability for CDMs into a specialized model with high predictability. We demonstrate DeepCD

SUBMITTER: Chen B 

PROVIDER: S-EPMC11436754 | biostudies-literature | 2024 Sep

REPOSITORIES: biostudies-literature