Abstract
Adequate multivariate structure‐activity relationships (SARs) were calculated by partial least‐squares (PLS) predicting the complete uptake and elimination curves for 63 PCDFs (tetra to hexa chlorinated congeners) in adipose tissue for guinea pig, hamster, mouse and rat. The SARs for uptake and elimination for these 63 congeners were calculated from a experimental determined data set which contained 16 congeners. The models were tested by selecting a training set (12 congeners) from which predictions were computed for a validation set (4 congeners). This validation set was selected from the 16 congeners by means of a principal component analyses (PCA) of 13 physico‐chemical properties of the PCDFs. Models using only the half‐life (t½) for elimination showed less predictive capability than modelling the complete uptake and elimination curves.
Loading plots from the PLS analyses revealed that different physico‐chemical properties were responsible for uptake and elimination. The ionisation potential, the lowest unoccupied molecular orbital, the octanol/water partition coefficient, the number of lateral chlorine atoms and the UV‐maximum wavelength are important variables in modelling elimination. The dipole moment, GC‐retention times and the number of non‐lateral chlorine atoms are more influential in the uptake and distribution steps. The molecular weight and the total number of chlorine atoms play a role in both processes.
Key words: