10.1021/ci700401p.s004 Aleksandr V. Marenich Aleksandr V. Marenich Pei-Han Yong Pei-Han Yong Isaac B. Bersuker Isaac B. Bersuker James E. Boggs James E. Boggs Quantitative Antidiabetic Activity Prediction for the Class of Guanidino- and Aminoguanidinopropionic Acid Analogs Based on Electron-Conformational Studies American Chemical Society 2008 dozen compounds antidiabetic activity model descriptors guanidine group Quantitative Antidiabetic Activity Prediction 120 compounds oxygen atoms aminoguanidinopropionic acid analogs Pha flexibility carboxyl group training 154 compounds topological characteristics 2008-03-24 00:00:00 Journal contribution https://acs.figshare.com/articles/journal_contribution/Quantitative_Antidiabetic_Activity_Prediction_for_the_Class_of_Guanidino_and_Aminoguanidinopropionic_Acid_Analogs_Based_on_Electron_Conformational_Studies/2949394 The electron-conformational method has been employed to reveal the pharmacophore (Pha) and to predict antidiabetic activity, studying 154 compounds in the class of guanidino- and aminoguanidinopropionic acid analogs. The derived Pha consists of four sites with certain electronic and topological characteristics which are represented by two oxygen atoms of the carboxyl group and two nitrogens of the guanidine group but may be substituted with any other atoms that have the same electronic and geometric features. The Pha flexibility and the influence of out-of-Pha features are described by only three model descriptors that predict the experimental activities quantitatively within experimental uncertainty for a training set of 120 compounds. The quality of the derived Pha and the corresponding quantitative model of activity has been validated (and deemed acceptable) by cross-validation including many-fold cross-validations within the training set and against an independent test set of 34 additional analogs with known experimental activities out of the training set. At last, several dozen compounds never tested experimentally have been screened theoretically using this model, and statistically significant hypoglycemic activities for a few of them are predicted.