Skip to main content
. 2019 Oct 12;36(4):1213–1218. doi: 10.1093/bioinformatics/btz736

Fig. 2.

Fig. 2.

Performance evaluations when removing the test spectra from training dataset. The accuracy and the sensitivity of SF-Matching, and the consensus method are compared against random prediction and two established methods on (a) the CASMI 2016 dataset, (b) the CASMI 2017 dataset and (c) the EMBL-MCF dataset. The number in the parentheses indicates the total number of molecules which are contained in the various chemical databases; on right of the bar the number of correctly identified molecules is shown. Details on the ranking can be found in Supplementary Tables