Application of fuzzy clustering to data dealing with phytotoxicity

There is a lack of knowledge in the scope of phytotoxicity because the concentration of chemicals in the environment and their uptake by vascular plants is quite difficult to control. Fuzzy clustering, a clustering method which derives the degree of membership to each cluster, was applied to analyze TOXSCREEN, a dataset of chemicals tested with soybean and barley under identical conditions. The clustered variables were the partition coefficients octanol-water (KOW) and atmosphere-water (KAW), the molecular weight (MW) and effective concentration regarding 50% of the experiment population (EC50). The clustering of only physical parameters yielded in a good separation of aromatics and non-aromatics, but cluster decomposition in the sense of phytotoxicity was rather poor. Adding EC50 resulted in the good decomposition of chemicals with different toxicity.