Scientific Publications by FDA Staff

Environ Toxicol Chem 2017 Mar;36(3):823-30

Rigorous 3D-SDAR modeling strategy for ToxCast estrogen receptor data classification, validation and feature extraction.

Slavov SH, Beger RD

The estrogenic potential (expressed as a score composite of 18 HTS bioassays) of 1528 compounds from the ToxCast database was modeled by a three-dimensional spectral data-activity relationship approach (3D-SDAR). Due to a lack of 17 O NMR simulation software, the most informative carbon-carbon 3D-SDAR fingerprints were augmented with indicator variables representing oxygen atoms from carbonyl and carboxamide, ester, sulfonyl, nitro, aliphatic hydroxyl and phenolic hydroxyl groups. To evaluate the true predictive performance of our model EPA provided us with a blind test set consisting of 2008 compounds. Of these, 543 had available literature data - their binding affinity served to estimate the external classification accuracy of the developed model: predictive accuracy of 0.62, sensitivity of 0.71 and specificity of 0.53 were obtained. Compared to alternative modeling techniques, our model displayed very little reduction in performance between the modeling and the prediction set. A 3D-SDAR mapping technique allowed identification of structural features essential for estrogenicity: i) the presence of a phenolic OH group or cyclohexenone ii) a second aromatic or phenolic ring at a distance of 6 A to 8A from the oxygen of the first phenol ring, iii) the presence of a methyl group approximately 6A away from the centroid of a phenol ring and iv) a carbonyl group in close proximity ( approximately 4 A measured to the centroid) to one of the phenol rings.