ATOMBASED QUADRATIC INDICES TO PREDICT AQUATIC TOXICITY OF BENZENE

Atom-Based Quadratic Indices to Predict Aquatic Toxicity of Benzene Derivatives to Tetrahymena pyriformis

Atom-Based Quadratic Indices to Predict Aquatic Toxicity of Benzene Derivatives to Tetrahymena pyriformis

Juan A. Castillo-Garit,^a,b,c,^* Concepción Abad,^c Yovani Marrero-Ponce,^b,d Francisco Torrens,^d Jeanette Escobar,^a and Amparo Torreblanca^e

^aApplied Chemistry Research Center, Faculty of Chemistry-Pharmacy, Central University of Las Villas, Santa Clara, 54830, Villa Clara, Cuba.

^bUnit of Computer-Aided Molecular “Biosilico” Discovery and Bioinformatic Research (CAMD-BIR Unit), Department of Pharmacy, Faculty of Chemistry-Pharmacy, Central University of Las Villas, Santa Clara, 54830, Villa Clara, Cuba.

^cDepartament de Bioquímica i Biologia Molecular, Universitat de València, E-46100 Burjassot, Spain.

^dInstitut Universitari de Ciència Molecular, Universitat de València, Edifici d'Instituts de Paterna, P. O. Box 22085, 46071 Valencia, Spain

^eDepartament de Biologia Funcional i Antropologia Fisica, Universitat de València, E-46100 Burjassot, Spain

Abstract

The atom-based quadratic indices are applied to develop quantitative structure-toxicity relationship (QSTR) models for the prediction of aquatic toxicity. Ours models agree with the OECD principles for the validation, for regulatory purposes, of QSAR models. The used dataset, consisting of 392 benzene derivatives for which toxicity data to the ciliate Tetrahymena pyriformis were available (defined endpoint), is divided into training and test sets using cluster analysis. We obtain (with an unambiguous algorithm) multiple linear regression models statistically significant (R² = 0.807 and s = 0.334, R²= 0.817 and s = 0.321, for non-stochastic and stochastic quadratic indices, respectively). The models were internally validated using leave-one-out cross-validation, bootstrapping as well as Y-scrambling experiments. In addition, a validation through an external test set is performed, which yields significant values of R²_pred of 0.754 and 0.760, showing these results that our models have appropriate measures of goodness-of-fit, robustness and predictivity. Also, we define a domain of applicability for our models. The comparison with other approaches indicates a good behaviour of our models in the prediction the aquatic toxicity of benzene’s derivatives. The obtained results demonstrated that, the non-stochastic and stochastic quadratic indices can provide an interesting alternative to the costly and time-consuming experiments currently used for determining toxicity.

Keywords: Atom-based non-stochastic and stochastic linear index, multiple linear regressions, QSAR, Tetrahymena pyriformis, Program TOMOCOMD-CARDD.

Tags: aquatic toxicity, the aquatic, aquatic, predict, quadratic, indices, toxicity, benzene, atombased