A validation study for the estimation of aqueous solubility from n-octanol/water partition coefficients.

Physico-chemical properties of chemicals are important data for exposure analysis. They can be estimated from structural information and/or via property-property relationships (PPRs). Often, such PPRs are restricted to distinct chemical classes. A concept has been developed in which molecules are characterized according to the presence or non-presence of typical structural elements (e.g., carbonyl functions, rings, aromatics). All structural elements are represented by binary digits (1 or 0 according to presence or non-presence). Using this tuple of digits, the estimation of aqueous solubility from the n-octanol/water partition coefficient is validated. According to different goodness criteria the PPRs are ranked by applying the graph-theoretical concept of HASSE diagrams. The best are analyzed further. Results indicate that some specific sets of molecules with the OH function should be excluded to keep the PPRs sufficiently accurate. Consequently, new PPRs especially for OH compounds are examined