Slideshow Image

Leibniz Institute of Plant Genetics and Crop Plant Research, Departament of Molecular Genetics, Data Inspection Research Group. Gatersleben, Germany.


Environmental and biomedical sciences share major properties that push utilization of state-of-the-art analysis tools from computational intelligence and machine learning research. Common properties are large numbers of data attributes, attribute correlations caused by spatio-temporal dependencies, and the presence of uncertainty induced by noise and missing values. We plan to combine the learning metrics principle with genetic algorithms for a uniform approach to retrieve relevant subsets of data attributes, and to minimalistic models of associations between different data sets. This way, robust mappings of physico-chemical substance properties to octanol-water partitioning coefficients shall be generated, which are essential building blocks in models of compound transport, e.g. PCB, in the environment. Using the same computational framework, modelling and understanding metabolic processes shall be enhanced by mapping gene- and protein-expression levels to phenotypic properties - a first essential step towards systems biology of stress tolerance in crop plants. The joint development of the new computational methods will allow addressing these and upcoming analytical challenges.