Multipole corrections to account for structure and proximity effects in group contribution methods: Octanol-water partition coefficients

We propose a new method to correct for structure and proximity effects that are not generally accounted for in group contribution models. Molecular or bital calculations for isolated molecules are performed to obtain the net c harge and dipole moment on each group within the molecule. These group mult ipoles, which vary as a result of the presence of proximate groups, are use d as the basis for corrections to simple group contribution methods. We hav e applied this method to our recently developed GCSKOW model to predict the octanol-water partition coefficients (K-OW) for complex molecules. We stud ied this simple, theoretically based method with a total of 450 solutes, in cluding five pharmaceutical compounds. The root-mean-square (RMS) deviation in the log(10) K-OW calculated from the GCSKOW model with these multipole corrections is found to be 0.18 (which leads to a mean error of 52% in K-OW ), compared with a RMS deviation of 0.44 (177%) when no correction is made. This simpler approach is more accurate than the KOW-UNIFAC model (RMS of 0 .28 or an error of 92% in K-OW for 344 compounds) and is comparable to the ClogP model (0.18 or 52%), which heretofore had been considered to be the m ost accurate predictive method for octanol-water partition coefficients.