Calculated similarity matrices for UniqueInchikey dataset
Florian G. Huber · Zenodo (CERN European Organization for Nuclear Research) · 2020
Contains numpy matrices for: Cosine score - all-vs-all matrix (tolerance=0.005, 12,797 x 12,797 scores) Cosine score - number of matching peaks (12,797 x 12,797 values) Modified cosine score - all-vs-all matrix (tolerance=0.005, 12,797 x 12,797 scores) Modified cosine score - number of matching peaks (12,797 x 12,797 values) "Tanimoto scores" - Jaccard indeces build from rdkit daylight fingerprints (2048bits, 12,797 x 12,797 scores Notebooks of how those scores are derived can be found here: https://github.com/iomega/spec2vec_gnps_data_analysis/tree/master/notebooks