Modelling and Prediction of Toxicity of Environmental Pollutants

Abstract. This paper describes the problem of modelling toxicity of environmental pollutants using molecular descriptors from a systems theoretical viewpoint. It is shown that current toxicity modelling problems systematically incorporate very high levels of noise a priori. By means of a set of individual and combined models self-organised by KnowledgeMiner from a high-dimensional molecular descriptor data set calculated within the DEMETRA project we suggest a way how results interpretation and final decision making can effectively take into account the huge uncertainty of toxicity models.

• Type: Conference
• Year: 2004
• Where: KELSI
• Authors: Frank Lemke, Johann-Adolf Müller, Emilio Benfenati
• Keywords: Informatics, KELSI 2004, Molecular Descriptor, Toxicity Modelling Problems, Toxicity Models