Dynamic network reconstruction from gene expression data applied to immune response during bacterial infection

• Published in: Bioinformatics 2005-04, Vol.21 (8), p.1626-1634
• Main Authors: Guthke, Reinhard, Möller, Ulrich, Hoffmann, Martin, Thies, Frank, Töpfer, Susanne
• Format: Article
• Language: English
• Subjects: Algorithms; Biological and medical sciences; Computer Simulation; Cytokines - immunology; Escherichia coli Infections - immunology; Fundamental and applied biological sciences. Psychology; Gene Expression Profiling - methods; Gene Expression Regulation - immunology; General aspects; HLA-D Antigens - immunology; Humans; Mathematics in biology. Statistical analysis. Models. Metrology. Data processing in biology (general aspects); Models, Immunological; Oligonucleotide Array Sequence Analysis - methods; Signal Transduction - immunology
• ISSN: 1367-4803; 1460-2059; 1367-4811
• DOI: 10.1093/bioinformatics/bti226

Motivation: The immune response to bacterial infection represents a complex network of dynamic gene and protein interactions. We present an optimized reverse engineering strategy aimed at a reconstruction of this kind of interaction networks. The proposed approach is based on both microarray data and available biological knowledge. Results: The main kinetics of the immune response were identified by fuzzy clustering of gene expression profiles (time series). The number of clusters was optimized using various evaluation criteria. For each cluster a representative gene with a high fuzzy-membership was chosen in accordance with available physiological knowledge. Then hypothetical network structures were identified by seeking systems of ordinary differential equations, whose simulated kinetics could fit the gene expression profiles of the cluster-representative genes. For the construction of hypothetical network structures singular value decomposition (SVD) based methods and a newly introduced heuristic Network Generation Method here were compared. It turned out that the proposed novel method could find sparser networks and gave better fits to the experimental data. Contact: [e-mail]