The main target of this project is the identification of metabolite patterns, which can be used to develop a diagnostic test for pre-diabetic metabolic traits. Therefore, statistics, mathematical modelling and pattern recognition are used to analyze mass spectrometry data samples of the TULIP (TUebingen Lifestyle Intervention Program) cohort. The mathematical modelling and classification is based on “differentially expressed” mass-profiles of individuals, who show metabolic traits, i.e. who are at high risk to develop type 2 Diabetes mellitus. Machine learning methods are applied to infer a mathematical method, which is able to classify "pre-diabetic" versus "non diabetic" probands from their mass-profile-patterns only. The next step involves the application of Evolutionary Algorithms to extract an optimal metabolite pattern, which on the one hand is able to classify with high sensitivity and specificity and, on the other hand, has minimal complexity.

Contact: Holger Franken, Raum A312, Tel. (07071) 29-78970, franken at informatik.uni-tuebingen.de