IBaI Institute |
Past events |
2nd Industrial Conference on Data Mining ICDM´2002
13. 06. - 15. 06. 2002 Leipzig/Germany
Automatically Generation of Gene-Nets
Dr. A. Glass and Prof. Lothar Gierl, Institut für Medizinische Informatik und Biometrie, Universität Rostock/Germany
Most of diseases to be led back on several gene defects which stand in a complex connection. Clearly this connection from expressed genes (proteins) can be represented as a genetic network which encloses poss. thousands from genes. There is a functional relation f-rel (gene-n, gene-m) between two genes in each case with gene-i, i Î G for all genes which are involved directly or indirectly in the disease events.
In the project GENSYS (sponsored by the Germany Ministry of Education and Science BMBF) we develop system based on case (Case-Based-Reasoning) with which incremental modelling of genetic networks is possible. Every genetic network, i.e. a functional connection of expressed proteins with regard to disease, state of disease, tissue, cell, cell type should be considered case within the human genome. Similar cases show similar genetic networks. Every identified case facilitates finding other genetic networks, i.e. cases. The separate genetic networks, must become indexed been suitable to find similar genetic networks rapidly or to integrate new genetic networks into the knowledge base. Because inconsistency and incompleteness are an utstanding feature of genetic networks on the base of the incremental knowledge progress about the proteome, the case base must be settled constantly.
Therefore, within the Retrieval-Reuse-Revise-Retain-Cycle of systems based on case is the Revise-phase in which the knowledge basis is checked and is revised poss., particularly importantly. Moreover, e.g., Tverskys contrast model are available from own works and the international state of the research a row of tested colleges of technology. We develop methods with which the extensive knowledge to be expected during the next years about the function is extracted by genes (1) automatically from the gene data banks in the Internet, (2) gene-expression-data from experiments with cDNA-Chip-technology and 2-D-gel-electrophoresis provided by research project "BMBF-Leitprojekt Proteom-Analyse des Menschen" are processed, (3) data from (1) and (2) into the case base of the genetic networks are integrated, this is checked (4) sequentially on inconsistencies or is corrected and (5) the genetic networks visualized become.
With it a tree of phenotypic classes and phenotypes of gene-nets with the diagnosis and therapy appears self-learning from genomic knowledge crucially are improved and Target-genes for medicaments can be discovered.
We have started first studies to the classification of genetic networks by means of Neural ARTnets which should give explanation about the biological variability of classes of selected autoimmune diseases. To prototypes to the extraction of knowledge about genes from the Internet and to the 3-D-visualization of genetic networks are tested at present. In addition, we have implemented filter methods of the assumption of cDNA-Chip-data and have tested.