Partial order ranking-based QSAr's: estimation of solubilities and octanol-water partitioning.

Partial order ranking appears as an attractive alternative to conventional Quantitative Structure Activity Relationships (QSAR) methods, the latter typically relying on the application of statistical methods. The method seems attractive as a priori knowledge of specific functional relationships is not required. In the present study, it is demonstrated that QSAR models based on a partial order ranking approach can be used satisfactorily to predict solubilities and octanol-water partitioning for a selection of organic compounds exhibiting different structural and electronic characteristics. The uncertainty is validated using well-established LSER descriptors. Two requirements to the model with regard to precision prevail, i.e., the model must be able to rank the single compounds in the basis set correctly compared to the experimental data, and the model should be based on a basis set of compounds large enough to secure a satisfactorily fine-meshed net, taking the number of descriptors into account. In the present study, the model was able to rank 318 out of 319 comparisons correctly in the case of solubilities. The corresponding figures for the octanol-water partitioning were 407 out of 408. The precision and the uncertainties of the method which, were found closely related to the mutual interplay between the number of compounds and the number of descriptors is discussed in terms of the number of descriptors and compounds involved. The limitations of the method are discussed.