prvi kongres hrvatskih znanstvenika iz domovine i inozemstva - HATZ

PRVI KONGRES HRVATSKIH ZNANSTVENIKA IZ DOMOVINE I INOZEMSTVAZagreb - Vukovar, 15. - 19. studenog 2004.P - 14NEW CLASSIFICATION PROCEDURE FOR BIOLOGICALLY ACTIVECOMPOUNDS BASED ON SIMILARITY OF THEIR MOLECULARINTERACTIONS AND logP PREDICTIONBranimir Bertoša, Sanja Tomić, Biserka Kojić-ProdićRuñer Bošković Institute, Zagreb, Croatia.The activity of biological compounds is dependent on both specific binding to targetreceptors and ADME properties (Absorption, Distribution, Metabolism, Excretion). Achallenge to prediction of biological activity is to consider both these types ofcontributions simultaneously in deriving quantitative models. Previously we havedeveloped a method for classification of (biologically active) compounds based onsimilarity of their molecular interaction fields. Herein we will present our new, morecomplex model that combines molecular interaction field analysis with the logP prediction.The method is conformation dependent and because of this elucidation of the biologicallyactive conformation of a molecule is possible. The method is tested on a set of abouthundred auxin related compounds. In the first run the auxin related molecules wereclassified on the basis of similarity of their interaction fields, only. Further on the influenceof logP value to biological activity was taken into account. This resulted with theclassification improvement and indicates that the method is especially efficient in caseswhere biological activity of compounds is correlated with their transport through cellmembranes. The majority of the analyzed compounds were classified in accord with theexperimental data available.Presented on :HUMBOLTD RESEARCH CONFERENCE ON COMPUTATIONAL CHEMISTRYVeliko Turnovo, Bulgaria, 30. 10. - 03. 11. 2002Principal Investigator : Dr.sc. Biserka Kojić-ProdićProject No. 009803616