4-DIMENSIONAL QUANTITATIVE STRUCTURE-ACTIVITY RELATIONSHIP ANALYSIS OF A SERIES OF INTERPHENYLENE 7-OXABICYCLOHEPTANE OXAZOLE THROMBOXANE A(2) RECEPTOR ANTAGONISTS

A series of 39 (a training set of 29 and a test set of 10) interphenyl ene 7-oxabicyclo[2.2.1]heptane oxazole thromboxane A(2) (TXA(2)) recep tor antagonists were studied using four-dimensional quantitative struc ture-activity relationship (4D-QSAR) analysis. Two thousand conformati ons of each analogue were sampled to generate a conformational energy profile (CEP) from a molecular dynamic simulation (MDS) of 100 000 tra jectory states. Each conformation was placed in a grid cell lattice fo r each of six trial alignments. Cubic grid cell sizes of 1 and 2 Angst rom were considered. The frequency of occupation of each grid cell was computed for each of seven types of pharmcacophoric group classes of atoms of each compound. These grid cell occupancy descriptors (GCODs) were then used as independent variables in constructing three-dimensio nal (3D)-QSAR models after data reduction. The types of data reduction included doing no reducing; reduction based on individual GCOD correl ation with activity, and reduction from minimum variance constraints o ver the GCOD population. The 3D-QSAR models were generated and evaluat ed by a scheme that combines a genetic algorithm (GA) optimization wit h partial least squares (PLS) regression. The 3D-QSAR models were eval uated by cross-validation using the leave-one-out technique. The cross -validated correlation coefficient, Q(2), ranged from 0.27 to 0.86. Th e models are not from chance correlation because a scrambled data set Was generated and evaluated (Q(2) = 0.25-0.37). A composite 3D-QSAR mo del was constructed using the best models derived from GCODs of both 1 and 2 Angstrom grid cell size lattices. The 3D-QSAR models provide de tailed 3D pharmacophore requirements in terms of atom types and corres ponding locations needed for high TXA(2) inhibition activity. Specific sites in space that should not be occupied by an active inhibitor are also specified. The GCOD measures for the compounds in the training s et permit reference points regarding which pharmacophore sites can pro vide the largest boosts in inhibition activity relative to the existin g analogues.