Training split MS2Deepscore

The training data split into training, validation and test set used for training the ms2deepscore model found at: https://doi.org/10.5281/zenodo.13897744
The spectra are a combination of GNPS spectra, MassBank, Mona and a library created by Corinna Brungs (see MSnLib Mass spectral libraries (.mgf and .json) (zenodo.org)) preprossessed using matchms to harmonize metadata. If you use this for academic work, please make sure to cite all relevant work.

The data was split by selecting 1/20th of the unique inchikeys for the training set and the validation set. So no compound in the training data appears in the test or train set. For the rest the split is random.