Pharmacophore identification and quantitative bioactivity prediction using the electron-conformational method

A review of the Electron-Conformational (EC) method of pharmacophore (Pha) identification and quantitative bioactivity prediction in drug design and toxicology is presented, which includes the latest advances and improvements of the method as a whole and details of its realization with illustration of results. In the first part devoted to Pha identification the data of conformational analysis and electronic structure calculation of each of the molecules in the training set are used to construct EC matrices of congruity (ECMC) that include atomic interaction indices as diagonal elements, and bond orders and interatomic distances as off-diagonal elements. Multiple comparisons of the ECMC's of the active compounds between themselves and with those of inactive compounds allows one to separate a relatively small number of matrix elements that within certain tolerances are common to all the ECMC's of the active compounds, while not present in the same combination in the inactive compounds. This is the EC submatrix of activity that represent the Pha, while the tolerances characterize the Pha flexibilities. Distinguished from QSAR approaches, the Pha is obtained here by computational (non-statistical) methods only. The second part of the problem, quantitative activity prediction, is based on using the Pha flexibilities together with the anti-Pha shielding and other auxiliary groups influence in a parameterization and regression analysis procedure that allows for quantitative prediction. An original approach is suggested that side steps the multi-conformational implications. This post-Pha problem is similar to a QSAR approach with special physically transparent descriptors that allow one to avoid chance correlations in the regression procedure. Illustration of the method is given for several drug design problems in which, where sufficiently accurate experimental data are available, the identification of the Pha as a necessary condition of activity is almost 100% correct, while quantitative activity prediction is near to the accuracy of the experimental data, 80% - 90%.