A novel computational technology derived from gene structure has been developed for screening, selecting, and designing pharmaceutical candidates. Pharmacophores, or three-dimensional molecular blueprints, were created by docking known active structures into specific sites in partially unwound DNA. The pharmacophores are composites of the van der Waals surfaces and hydrogen bonding functional groups of active molecules. Once created, molecules can be inserted into the pharmacophores and degree of fit quantitated by the volume of the molecule that fits within the composite surface and the magnitude of electrostatic interactions with charged atoms on the pharmacophore. Here, we describe endocrine pharmacophores and in particular the estrogen pharmacophore derived by docking active ligands into partially unwound DNA. Fit of candidate structures into the estrogen pharmacophore correlated with estrogenic (uterotropic) activity. For example, the super active estrogens moxestrol and 11β-acetoxyestradiol fit better within the site than estradiol. Bisphenol A, a putative endocrine disrupter with suspected estrogenic activity, was a poor fit in the pharmacophore. Consistent with this prediction, bisphenol A was recently shown to lack uterotropic activity. The capacity of the endocrine pharmacophores to predict certain nontarget activities was demonstrated by using the antiandrogen cyproterone acetate that did not fit the estrogen or thyroid pharmacophores but fit partially into the progestin and glucocorticoid pharmacophores. Cyproterone acetate has been reported to have weak progestational and glucocorticoid activities. The pharmacophores provide for the first time a multidimensional computational method that can simultaneously predict multiple activities of diverse molecular structures.