Transcript PowerPoint-Präsentation

Identification of geneexpression networks in
different immunological states
Marc Bonin1, Jekaterina Kokatjuhha1, Florian Heyl1, Karsten Mans1, Andreas Grützkau2, Biljana Smiljanovic1, Till Sörensen1, Thomas Häupl1
1
Department of Rheumatology and Clinical Immunology, Charité University Hospital, Berlin, Germany,
2German Arthritis Research Center, Berlin, Germany
Initially, correlation matrices were determined for each individual stimulation
condition and its control. Stepwise combination of the three different conditions for
calculation of correlation coefficients revealed a reduction of the correlation network
and a reduction of overlap between the networks. This indicates increasing functional
specificity of the identified candidates.
All of the typical previously published IFN related genes were identified and thus
confirmed our strategy. In a similar way, cell type specific co-expression networks
were determined. Additional filtering for high signal intensity provides candidates for
sensitive detection of the function related patterns even in highly diluted conditions.
These marker panels are currently tested for detection and quantification of
functional signatures in biopsies of inflamed tissue.
GeneChip HG-U133 Plus 2.0 transcriptomes from highly purified blood cell types
(granulocytes, monocytes, CD4+ and CD8+ T-cell, B-cells, NK-cells) as well as from
monocyte stimulation with LPS, TNF and type 1 IFN were selected from the BioRetis
database (www.bioretis.de). Correlations of expression between all probesets were
calculated to filter for co-regulation. Correlation matrices of selected genes were
calculated, clustered and displayed in heat maps (Fig. 2). The database and the
correlation-algorithm will be provided on our homepage http://www.charitebioinformatik.de. (Fig. 1)
Figure 2
CD14 IFNa2a
CD14
Material and Methods:
A
B
CD14 Ctl.
Knowledge about gene networks is of great importance for analysis of transcriptome
data. However, current tools mainly rely on information about direct molecular
interactions between proteins, which is not directly connected to expression levels.
These differences between transcriptome based perception of biological information
and tools for network analysis are the main reason for difficulties in functional
interpretation. Therefore, we started to use transcriptome data of biologically welldefined states to define functional markers and signatures as tools for future analysis.
CD14 IFNa2a
CD14
Results:
CD14 Ctl.
Background and Objective:
Figure 1
a. File-Upload (drag &drop)
a. Qualitycontrol (I. Sample quality (Quality control plot, RNA degradation
plot) II. Array comparability (Chip images, Boxplots, MA plots, Histograms,
Pos/Neg position plot, Scatterplots, RLE & NUSE-Plots) III. Hybridization
quality (Background intensity plot, Pos/Neg distribution plot) IV. Array
correlation (Correlation plot, Cluster dendrograms, PCA))
a. Create a correlation-group (select the CEL-Files and type in a name)
c. Start correlation
a. E-Mail Notification (Including: Name of the correlation-group, number of
probesets, the threshold, number of chips, number of correlations,
statistical information, download-links and the correlation matrix)
correlation matrix
relative correlation matrix
relative
expression
expression
Relative expression values and correlationon matrix
A.
Pearson correlation of all probesets in CD14+ monocytes, "CD14_Ctl" controls before and
CD14+ after IFN-α2a stimulation revealed 1808 correlation pairs with correlations coeffizient
≥0.99 or ≤-0.99. These consisted of 869 genes. The heatmap of the hierarchical clustering
illustrates positive correlation in red and negative correlation in green.
B.
Filtering for IFN-specific genes and exclusion of "false" positive correlations reduced the 1808
probesets to 543 and the 869 genes to 151. Of all genes, only a single one was identified to be
negatively correlated (suppressed by IFN-α2a stimulation), while all others were induced by
IFN-α2a.
Figure 3
% overlap
b. Select an individual Filter to reduce the dataset or include a list with an
individual geneset.
Conclusions:
Correlating transcription between genes in well-defined biological states identifies
function-related markers and signatures. Depending on the type of function,
appropriate conditions have to be selected.
Comparison of correlation networks in different combinations of functional profiles (sample
sets). Displayed is the percentage of those gene pairs, which were identified in the CD14 IFN
sample set by correlation with R>0.99 and which correlate in other sample sets by R>0.9.
Contacts:
Marc Bonin
Department of Rheumatology
and Clinical Immunology
Charité University Hospital
Charitéplatz 1
D-10117 Berlin Germany
Tel: +49(0) 30 450 513 296
Fax: +49(0) 30 450 513 968
E-Mail: [email protected]
Web: www.charite-bioinformatik.de
www.charite-bioinformatik.de