A Decade of Computational Mass Spectrometry from Reference Spectra to Deep Learning

Area of Science:

• Analytical Chemistry
• Computational Chemistry
• Mass Spectrometry

Background:

• Computational methods are increasingly vital in analytical chemistry, complementing traditional data evaluation.
• Computational mass spectrometry (CompMS) applies computational techniques to mass spectrometry data analysis.

Purpose of the Study:

• This review discusses advances in CompMS for small molecule chemistry.
• Focus areas include spectral libraries, spectrum prediction, and tentative structure identification.

Main Methods:

• Automatic spectrum curation for expanding open spectral libraries.
• Spectrum and molecular fingerprint prediction using classical and deep learning.
• De novo structure generation from fragment spectra using deep learning.

Main Results:

• Open spectral libraries are expanding, benefiting compound annotation and machine learning.
• Various machine learning and deep learning methods enhance spectrum and fingerprint prediction.
• De novo structure generation from fragment spectra is an emerging research area.

Conclusions:

• CompMS is crucial for modern small molecule analysis.
• Advances in spectral libraries, prediction, and de novo generation are transforming the field.
• Tools like RMassBank and MSNovelist exemplify progress in CompMS.