Modeling physical and toxicity endpoints of alkyl (1-phenylsulfonyl) cycloalkane-carboxylates using the Ordered Predictors Selection (OPS) for variable selection and descriptors derived with SMILES

Among the methods of variable selection for Quantitative Structure-Property Relationship (QSPR) studies, one of the currently available alternatives is the Ordered Predictors Selection (OPS). Using this algorithm and descriptors obtained using only Simplified Molecular Input Line Entry System (SMILES) strings in the free web server Parameter Client, a QSPR study with a data set of 28 alkyl (1-phenylsulfonyl) cycloalkane-carboxylates and six different endpoints of environmental importance were developed and compared with other works. The comparison with models previously published was performed only with the internal validation, and four of the six new models proved to be superior. However, the six new models also presented high quality for external predictions, were robust and showed no chance correlation. The predicted endpoints of the six models were within the applicability domain. Thus, it can be concluded that the OPS algorithm was able to generate QSA(P)R models with high statistical quality for predicting of physicochemical and toxicological endpoints, thus showing its potential for development of predictive models of environmental interest. (C) 2012 Elsevier B.V. All rights reserved.