QSPR Study on Non-Ionic Organic Pesticides

DUCHOWICZ, P. R.; PÉREZ-GONZÁLEZ, M.; TEIJEIRA, M.; HELGUERA A. M.; CASTRO, E. A.

Eighth J.J. Giambiagi Winter School Part A, "Clusters, Molecules, Biomolecules and Materials"

Workshop; Eighth J.J. Giambiagi Winter School Part A, "Clusters, Molecules, Biomolecules and Materials"; 2006

In the present work we apply the Quantitative Structure-Property Relationships (QSPR) theory[1] to predict soil sorption coefficients (KOC) for 185 non-ionic organic heterogeneous pesticides usually employed in Agriculture. This sort of partition coefficient plays an important role in describing the pollution potential of pesticides when undergoing leaching once applied to the soil, and represents a measure of the retaining of a chemical by the organic matter of soils and sediments through a wide variety of intermolecular interactions[2]. The different formulations of the QSPR theory suggest mathematical models for estimating relevant physicochemical properties, especially when these can not be experimentally determined for some reason. The QSPR studies rely on the assumption that the structure of a compound determinates the physicochemical properties it manifests. The molecular structure is therefore translated into numerical variables with physical interpretation, known as molecular descriptors, through mathematical formulae obtained from several theories, such as Chemical Graph Theory, Quantum Mechanics, etc. There exist more than a thousands of molecular descriptors available in the literature, and one usually face the problem of selecting the most representatives for the property under consideration. In order to relate the experimental KOC with the molecular structure, many standard statistical techniques can be employed. Recently, our research group has elaborated a very useful algorithm based, on linear regressions called “Replacement Method”[3, 4] which enables to find the best molecular descriptors among a thousand of them. This technique is applied on the present data set, and the best model obtained is properly validated by means of the Cross-Validation technique and also with an external test set of 22 structurally-related pesticides, in order to assess its true predictive performance. In addition, the results are compared with those obtained with an alternative algorithm, the Genetic Algorithm Approach[5]. The model reported aims to easily obtain numerous, reliable, and comparable KOC for application in risk assessment. 1.Hansch, C., Leo,A.(1995). Exploring QSAR Fundamentals and Applications in Chemistry and Biology. Am. Chem. Soc., Washington, D. C. 2.Jury, W. A. (1986) In: Henn, S. C., Melancon, S. M. (ed) Vadose Zone Modeling of Organic Pollutants . Lewis Publisher p 177. 3.Duchowicz P. R., Castro, E. A., Fernández, F. M., González, M. P. (2005) Chem. Phys. Lett. 412: 376. 4.Duchowicz P. R., Castro, E. A., Fernández, F. M. (2006) MATCH Commun. Math. Comput. Chem. 55: 179. 5.So, S. S., Karplus, M. (1996) J. Med. Chem. 39: 1521.