Estimation of Soil Partition Coefficients of Pesticides from the Chemical Structure

For pesticides, the distribution coefficient in the soil and the soil sorption coefficient play a major role in determining the polluting potential of the chemicals. A non-empirical model is developed for the estimation of soil sorption coefficients, which is based solely on chemical structural data. Two methods of parameter estimation are utilized: multiple linear regression, and neural networks based on radial basis functions. All input variables are derived from the connection tables of the chemical structures, which are tabulated. The final model, which is presented, is a linear model, obtained by multiple linear regression, and comprises 11 input variables.