Mendeleev Communications Mendeleev Commun., 2007, 17, 90–91 – 90 – QSAR modeling on the basis of 3D descriptors representing the electrostatic molecular surface (ambergris fragrances) Igor V. Svitanko,* a,b,† Dmitry A. Devetyarov, c Dmitry E. Tcheboukov, b Maksim S. Dolmat, a,‡ Alexey M. Zakharov, c Svetlana S. Grigor’eva, c Viktoriya T. Chichua, c Lyudmila A. Ponomareva b and Mikhail I. Kumskov b a Higher Chemical College, Russian Academy of Sciences, 125047 Moscow, Russian Federation. E-mail: svitanko@mail.ru b N. D. Zelinsky Institute of Organic Chemistry, Russian Academy of Sciences, 119991 Moscow, Russian Federation. Fax: +7 495 135 5328 c Department of Mechanics and Mathematics, M. V. Lomonosov Moscow State University, 119992 Moscow, Russian Federation DOI: 10.1016/j.mencom.2007.03.012 A 3D-QSAR approach based on the electrostatic surface of molecules was used for the ambergris odour, and it showed a cross validation coefficient of 0.8. Previously, we have constructed the models of a musk odour 1 and a model and new structures of the bicycloureas of psycho- tropic activity 2 (which were successfully synthesised). The method proposed allows us to take into account the spatial electrostatic complementary character of two or several mole- cules or a molecule and a receptor (long-range interaction). For this purpose, it is necessary to calculate the electrostatic field created by the molecule and then supplement the model thus obtained by structural complementarity (short-range interaction). The sample consisted of 50 compounds (37 active and 13 inactive, Figure 1) represented by 3D molecular graphs. 3 For every molecule, we considered that the 3D coordinates of nodes (atoms) and their quantitative characteristics (partial atomic charges) are known. Geometry optimization and charge calculations were per- formed by Gaussian03. The molecule was represented by a molecular surface (the distances from atoms were equal to a van der Waals radius), which was ‘colourised’ by a local physi- cal property (LPP), e.g., charge, lipophilicity and the ability to release or accept an electron (donor–acceptor factors) (Figure 2). For every molecule in the test set, we constructed its triangulated molecular surface with excluded solvent using MSMS. 4 We picked critical points on molecular surfaces using Connolly’s methodology 5 to describe local knobs and holes. Finally, we calculated simple electrostatic (Coulombic) potentials of every critical point by adding up the effects of electrostatic fields created by individual atoms. We can now reformulate the QSAR problem as follows: every object (molecule) in the test set is represented by a set of N critical points (x i , y i , z i , S i , Q i ), i = 1, ..., N, where (x i , y i , z i ) are the 3D coordinates of a critical point, S i describes the shape (0 for a hole and 1 for a knob), and Q i is the electrostatic potential. Our aim was to construct 3D structural descriptors representing the critical points, their pairs and triplets and to check their value for the QSAR modeling. We first performed a cluster analysis of the electrostatic potentials of all critical points in all molecules to segment out O R 1 R 2 R 3 O R 1 R 1 O O O O R 1 O O R 1 O O R 1 O R 1 R 1 R 1 R 1 = H, Me; R 2 = H, OH; R 3 = Me, OMe O OMe O O R 1 C(O)Me O O R 1 Figure 1 Examples of compounds in the test sample. † Vice-chairman of the Higher Chemical College (HCC) of the RAS. ‡ A former student of the HCC RAS (1995–2000). 4 3 2 1 0 –1 –2 –3 –4 5 0 –5 –6 –4 –2 0 2 4 6 x axis y axis z axis Figure 2 Molecular surface and critical points (see colour version on the cover of this issue).