Multiple-conformation and protonation-state representation in 4D-QSAR: Theneurokinin-1 receptor system

Using a 4D-QSAR approach (software Quasar) allowing for multiple-conformati on, orientation, and protonation-state ligand representation as well as for the simulation of local induced-fit phenomena, we have validated a family of receptor surrogates for the neurokinin-1 (NK-1) receptor system. The evo lution was based on a population of 500 receptor models and simulated durin g 40 000 crossover steps, corresponding to 80 generations. It yielded a cro ss-validated r(2) of 0.887 for the 50 ligands of the training set (represen ted by a total of 218 conformers and protomers) and a predictive r(2) of 0. 834 for the 15 ligands of the test set (70 conformers and protomers). A ser ies of five "scramble tests" (with an average predictive r(2) of -0.438) de monstrates the sensitivity of the surrogate toward the biological data, for which it should establish a QSAR. On the basis of this model, the activiti es of 12 new compounds - four of which have been synthesized and tested in the meantime - are predicted. For most of the NK-1 antagonists, the genetic algorithm selected a single entity - out of the up to 12 conformers or pro tomers - to preferably bind to the receptor surrogate. Moreover, the evolut ion converged at an identical protonation scheme for all NK-1 antagonists. This indicates that 4D-QSAR techniques may, indeed, reduce the bias associa ted with the choice of the bioactive conformation as each ligand molecule c an be represented by an ensemble of conformations, orientations, and proton ation states.