Predicting water solubility of congeners: chloronaphthalenes--a case study

Since the important physicochemical data for chloronaphtalenes (PCNs) are still scarce, we have predicted water solubility (logS) of all 75 congeners with the Quantitative Structure-Property Relationship (QSPR) scheme. The values of logS, predicted by the most efficient model, varied from 0.01 to 1660 microg dm(-3) (2.85 x 10(-11)-1.02 x 10(-5) mol dm(-3)), depending on the number of chlorine atoms present in the molecule and the substitution pattern. We found that the main factor determining relative differences in solubility between the congeners is the solvent accessible volume related to the cavitation process occurring in the solvent. The results are presented as a case study of QSPR modeling for those Persistent Organic Pollutants (POPs) that exist as families of congeners. By investigating the impact of (i) the way of the molecular descriptors' calculation, (ii) the size of applied database and (iii) chemometric method of modeling (Multiple Linear Regression, MLR, and/or Partial Least Squares regression, PLS) on the quality of the models we proposed general recommendations for dealing with congeners. We found that the combination of the B3LYP functional with 6-311++G(d,p) basis set was the most optimal technique of the molecular descriptors' calculation for congeners when comparing with semi-empirical PM3, ab initio Hartee-Fock (HF), and Møller-Pleset 2 (MP2) method carried out with different-size basis sets. Moreover, the model developed with a larger and more general database that includes chloronaphthalenes, polychlorinated dibezno-p-dioxins, furans and biphenyls predicted the values of logS for PCNs noticeable worse than the model calibrated only on PCNs. In the later case it was possible to obtain satisfactory results by employing even the simplest MLR method and only one molecular descriptor. The values of logS were also calculated with the WSKOWIN and COSMO-RS models as the reference techniques and then compared to our results.