C-13 NMR and electron ionization mass spectrometric data-activity relationship model of estrogen receptor binding

Two Spectroscopic Data-Activity Relationship (SDAR) models based on C-13 nu clear magnetic resonance (NMR) and electron ionization mass spectra (EI MS) data were developed for 108 compounds whose relative binding affinities (R BA) to the estrogen receptor are known. The C-13 NMR and EI MS data were us ed as spectrometric digital fingerprints to reflect the electronic and stru ctural characteristics of the compounds. Both SDAR models segregated the 10 8 compounds into 20 strong, 15 medium, and 73 weak relative binding classif ications. The first SDAR model, based on C-13 NMR data alone, gave a leave- one-out (LOO) cross-validation of 75.0%. The second SDAR model, based on a composite of C-13 NMR and EI MS data, gave a LOO cross-validation of 82.4%. Many of the misidentifications from the cross-validations were between med ium and weak classifications, where there were fewer specific spectrometric characteristics to identify the relationship of spectra to estrogen recept or binding. Real and predicted C-13 NMR chemical shifts were used to test t he predictive behavior of both SDAR models. The ease of use and speed of SD AR modeling may facilitate their use with other toxicological endpoints.