Matchms cleaned mass spectral library

Combination of the MS2 spectra in the GNPS and MSnLib library, cleaned by matchms and additional filters. They have been cleaned and harmonized to be machine learning training ready. The filters are similar to MassSpecGym, but here we included both ionmodes and all adducts. The cleaning i includes metadata harmonization, like adding smiles and adduct. Also verification of possible smiles + adduct + parent mass. And basic checks to check if enough of the MS2 peaks can be explained by the formula. For details see the notebook. The GNPS spectra and MSnLib were downloaded on 22-07-2025. If you use this data source, please cite: gnps Brungs C, Schmid R, Heuckeroth S, Mazumdar A, Drexler M, Šácha P, et al. Efficient generation of open multi-stage fragmentation mass spectral libraries. ChemRxiv. 2025; doi:10.26434/chemrxiv-2024-l1tqh-v3 This content is a preprint and has not been peer-reviewed. de Jonge, N. F., Hecht, H., Strobel, M., Wang, M., van der Hooft, J. J., & Huber, F. (2024). Reproducible MS/MS library cleaning pipeline in matchms. Journal of Cheminformatics, 16(1), 88.