USE OF PRINCIPAL COMPONENT ANALYSIS AND CLUSTER-ANALYSIS IN QUANTITATIVE STRUCTURE-ACTIVITY-RELATIONSHIPS - A COMPARATIVE-STUDY

The inhibitory effect of some new benzothiazole derivatives on Bacillu s subtilis, Aspergillus niger, Helminthosporium sativum and Fusarium g raminearum was determined. Hydrophobicity and specific hydrophobic sur face area of the pesticides were measured by reversed-phase thin-layer chromatography in neutral, acidic, alkaline and salt-containing envir onments. To find the relationship between the physicochemical paramete rs and biological activity, stepwise regression analysis, principal co mponent analysis, nonlinear mapping technique and cluster analysis wer e applied. Good linear correlations were found between the biological activity, steric parameter and calculated hydrophobicity value of comp ounds proving the significant impact of the above physicochemical para meters on the inhibitory effect. Nonlinear mapping technique cannot be used on the original data matrix, but it can be successfully applied for the principal component loadings and variables. Cluster analysis r equires the lowest computational time and the results are commensurabl e with those of principal component analysis. Both methods indicate th at the pH and salt concentration have a negligible impact on the hydro phobicity and specific hydrophobic surface area of pesticides, whereas the biological activity and the measured physicochemical parameters f orm two different clusters suggesting the poor correlation between the two groups of variables.