Adapted Boolean network models for extracellular matrix formation
- Equal contributors
1 Systems Biology/Bioinformatics, Leibniz Institute for Natural Product Research and Infection Biology – Hans Knöll Institute, Beutenbergstr. 11a, 07745 Jena, Germany
2 Institute of Algebra, Technische Universität Dresden, Zellescher Weg 12-14, 01062 Dresden, Germany
3 Experimental Rheumatology Unit, Department of Orthopaedics, University Hospital Jena, Friedrich Schiller University Jena, Klosterlausnitzer Str. 81, 07607 Eisenberg, Germany
4 Institute of Clinical Chemistry, Hannover Medical School, Carl-Neuberg-Str. 1, 30625 Hannover, Germany
5 Proteome Center Rostock, University of Rostock, Schillingallee 69, 18055 Rostock, Germany
6 Genome Analysis, Leibniz Institute for Age Research – Fritz Lipmann Institute, Beutenbergstr.11, 07745 Jena, Germany
BMC Systems Biology 2009, 3:77 doi:10.1186/1752-0509-3-77Published: 21 July 2009
Due to the rapid data accumulation on pathogenesis and progression of chronic inflammation, there is an increasing demand for approaches to analyse the underlying regulatory networks. For example, rheumatoid arthritis (RA) is a chronic inflammatory disease, characterised by joint destruction and perpetuated by activated synovial fibroblasts (SFB). These abnormally express and/or secrete pro-inflammatory cytokines, collagens causing joint fibrosis, or tissue-degrading enzymes resulting in destruction of the extra-cellular matrix (ECM).
We applied three methods to analyse ECM regulation: data discretisation to filter out noise and to reduce complexity, Boolean network construction to implement logic relationships, and formal concept analysis (FCA) for the formation of minimal, but complete rule sets from the data.
First, we extracted literature information to develop an interaction network containing 18 genes representing ECM formation and destruction. Subsequently, we constructed an asynchronous Boolean network with biologically plausible time intervals for mRNA and protein production, secretion, and inactivation. Experimental gene expression data was obtained from SFB stimulated by TGFβ1 or by TNFα and discretised thereafter. The Boolean functions of the initial network were improved iteratively by the comparison of the simulation runs to the experimental data and by exploitation of expert knowledge. This resulted in adapted networks for both cytokine stimulation conditions.
The simulations were further analysed by the attribute exploration algorithm of FCA, integrating the observed time series in a fine-tuned and automated manner. The resulting temporal rules yielded new contributions to controversially discussed aspects of fibroblast biology (e.g., considerable expression of TNF and MMP9 by fibroblasts stimulation) and corroborated previously known facts (e.g., co-expression of collagens and MMPs after TNFα stimulation), but also revealed some discrepancies to literature knowledge (e.g., MMP1 expression in the absence of FOS).
The newly developed method successfully and iteratively integrated expert knowledge at different steps, resulting in a promising solution for the in-depth understanding of regulatory pathways in disease dynamics. The knowledge base containing all the temporal rules may be queried to predict the functional consequences of observed or hypothetical gene expression disturbances. Furthermore, new hypotheses about gene relations were derived which await further experimental validation.