Studies in computational chemistry: Molecular models for prediction of the estrogenic activity of raloxifene analogues and phenolic xenoestrogens; computer code for predicting the toxicity of small-molecule compounds

Several 2-D and 3D-QSAR models based on CoMFA (Comparative Molecular Field Analysis) techniques have been developed that possess both excellent predictive ability (high r_cv2) and exceptional statistical validity (high r2) for two sets of compounds that interact with the estrogen receptor site. The first models were developed using values of relative binding affinity (RBA) obtained from published data for 76 raloxifene analogs (J. of Med. Chem. 40:146–167, 1997) for which the predictability of a test set having low relative binding affinity has been examined. The second models were developed using values of activity based upon the logarithm of the inverse of the 50% molar β-galactosidase activity from the Saccharomyces cerevisiae-based Lac-Z reporter obtained from published data produced by Shultz et al. (Environmental Toxicology and Chemistry 19:2637–2642). This data set contained 36 phenols of various substituent size for which the predictability of several test sets having relatively low activity were also examined.; Additionally, a computer application was developed to simplify a manual process for determining toxicity of various compounds. This application was developed in Visual Basic and implemented both Quicksort and Binary Search Tree (BST) algorithms to parse an input text file and produce appropriate output for further analysis.