35.

Ribeiro F. A. L., Sabino L. C., Ferreira M.M.C., “QSPR Models to Predict the Organic Carbon-Water Partition Coefficient (Koc) and Bioconcentration Factor (BCF) for Polycyclic Aromatic Hydrocarbons (PAHs)”.Caxambu, MG, Brazil, 11-16/11/2001: 1º Simpósio Brasileiro em Química Medicinal, QSAR e Modelagem Molecular: Novas Estratégias em Planejamento Racional de Fármacos [1st Brazilian Symposium on Medicinal Chemistry, QSAR and Molecular Modeling: New Approaches in Drug Design]. Poster APM17. Section: 4. Applications of Multivariate QSAR.

Chemometrics

Name
Ribeiro, F. A. L.; Sabino, L. C. ; Ferreira, M. M. C

Institution/Company
Universidade Estadual de Campinas

Country
Brasil

First Abstract Title
QSPR Models to Predict the Organic Carbon-Water Partition Coefficient (Koc) and Biococentration Factor (BCF) for
Polycyclic Aromatic Hydrocarbons (PAH)

First Abstract
      The polycyclic aromatic hydrocarbons (PAHs) have been, for a long time, a focus of great attention by the scientific
      community due to their impact on public health and the environment. Some of these compounds such as
      benz[a]anthracene, chrysene, dibenz[a,h]anthracene are known for their toxicity and carcinogenic actions. They are
      usually introduced in the environment as a result of anthropogenic activities which have increased dramatically in the
      last 20 years. It is well established that the fate of PAHs in the environment is primarily controlled by their
      physicochemical properties. These compounds are quite involatile, highly hydrophobic and resistant to chemical
      reaction and as a result, they show tendency to accumulate in biota, soils sediments and are also highly dispersed by
      the atmosphere. Consequently, meaningful health and environmental assessment requires reliable experimental data.
      Unfortunately, very often these information cannot be found in the literature, mainly due to experimental difficulties. In
      order to solve these deficiencies, the use of QSPR models have become popular in the scientific community. These
      models use molecular descriptors in an attempt to predict the physicochemical and biological activities for those
      chemicals with no experimental value. Two of these properties are subject of this work: 1- the organic carbon-water
      partition coefficient (Log Koc) which is the most used parameter to indicate soil mobility in water-soil systems
      (compounds with higher values of of Log Koc tend to be less mobile), and 2- the bioconcentration factor (BCF) which
      describes bioconcentration in terms of the ratio of the concentration of a chemical present in an aquatic organism to
      the concentration in the surrounding environment. A set of 70 PAHs were studied, and QSPR models constructed for
      the properties mentioned above. The experimental data was taken from literature (1). The models were constructed by
      partial least square methods (PLS) in autoscaled data and leave-one-out cross validation was the method used for
      validation. The molecular descriptors were calculated by semi-empirical method AM1 implemented in Spartan and
      WHIM-3D software. The molecular descriptors which best correlated with Log Koc were molecular volume, molecular
      weight, Randic conectivity index and Wiener Index (2, 3, 4). For BCF they were molecular area, molecular volume,
      molecular weight, Randic conectivity index and Wiener Index. The results for PLS model are shown Table 1.

      Table 1. PLS results for Log, Koc and BCF properties:
      --------------------------------------------------------------------
      Property   LatentVariable   SEV %   Variance   R2
      Log Koc          2               0.1400    99.9612 0.9220
      Log BCF         2               0.2000    99.7131   0.9202
      --------------------------------------------------------------------

      The crossvalidated residuals (experimental - estimated values) for Log Koc model are below 4 %, and for Log BCF model
      are below 10 %. The models obtained were used to predict these properties for those compounds in which the experimental
      measurements were not performed yet. Ten molecules were used to model the Log Koc, and the soil sorption was predicted
      for the other 60 molecules. Log BCF was modeled using a set of 13 molecules and the bioconcentration factor values for o
      other 57 molecules were predicted.

      Acknowledgments. This work was supported by FAPESP, CAPES and FAEP. We thank Dr. Lucicleide R. Cirino and
      Thaís F. Parreira, for assistance with the theoretical calculations and valuable suggestions. Also CENAPAD, for
      computational support.

      References.
      1-Mackay et alli, Illustrated Handbook of Physical-Chemical Properties and Environmental Fate for Organic
      Chemicals, Lewis, 1998.
      2- Randic, J. Am. Chem. Soc., 1975, 97, 6609.
      3- Wiener, J. Am. Chem. Soc., 1947, 69, 17. et alli, Chemosphere, 1996, 33, 71.
      4-Todeschini & Gramatica, Quanti Struc-Act. Relat., 1997, 16, 113.