Bioinformatics & Mass SpectrometryWe're creating a Bioinformatics and Metabolomics platform, which includes databases and applications for a data-driven approach to analysis and annotation of metabolomics experiments. This includes pattern recognition, database technology and modern software development approaches. Metabolomics PlatformThe platform includes two ArMet and mzData compiant databases and preprocessing and alignment of mass spectrometry raw data.
For the description of metabolomics experiments several standards (MiaMet, SMRS) have been proposed in the community. Coming from developments in the proteomics area, two exchange formats (mzData and mzXML) are the basis of our archival system. For the analysis of several measurements the raw data has to be processed and aligned to compare the signal intensities. We use and improve several programs for this task. MzData Eclipse Editor Advanced Clustering of Metabolite DataDiscovering comon features and differences between metabolic states is done using datamining technologies such as hierarchical or density based clustering and visualisation methods for high-dimensional data.
Metabolite data can be interpreted as phenotypic feature and used to distinguish different metabolic states of a system or to identify ecotypes. Different (hierarchical, density based and other) clustering algorithms implemented in the R statistics software are used with several distance measurements, e.g. to separate 9 Arabidopsis lines or discriminate between ler wildtype and a tt4 mutant. Network Reconstruction with Bayesian NetworksBayesian Networks are graphical models that can be used for modelling of regulatory networks and visualisation of biological processes.
Bayesian Networks are graphical models used to model and visualise biological processes. A Bayesian Networks consists of a directed acyclic graph representing metabolites with edges for their relationships and a conditional probability distribution. The Sparse Candidate Algorithm (SCA) has been implemented in Java to efficiently find a network that models the measured data. Networks have been trained on several datasets, and some Hypotheses on effects on the pathway of a tt4 Mutant have been recognized. |
Research
|
