Guiyuan Lei - iNotes - Gaussian Graphical Models for Microarray Data

The procedure for constructing decomposable Gaussian Graph from microarry data:

• Pre-process affymetrix microarray data using BioConductor

Use BioConductor to pre-process the affymetrix microarray data.

At the directory when you put your .CEL files, start R by type;
```
R
```
Load the CDF library for hgu133plus2
```
library("hgu133plus2cdf",lib.loc="/data/BioConductor/")
```
Because the hgu133plus2cdf is not defaultly installed when install BioConductor, I install hgu133plus2cdf by hand. First download hgu133plus2cdf_1.43.tar.gz from website http://bioconductor.org/data/cdfenvs/repos/html/, store it in one directory. Go to this directory, use the following command to install the cdf library. "/data/BioConductor" is the directory where BioConductor is installed.
```
R CMD INSTALL -l "/data/BioConductor" "hgu133plus2cdf_1.4.3.tar.gz"
```
```
library(affy)
```
```
immortalised<-ReadAffy()
```
```
eset<-justRMA()
```
```
write.exprs(eset,file="mydata.txt")
```

• Select gene expression for sub gene list

Select gene expression by given gene IDs (the outfile of justRMA in BioConductor consists of gene IDs). compg:~/GaussianModels/MH-d/pre_process_data/select_genes_expression/readGenName.pl

• Center the data to Standard Normal

Center the data to N(0,1), compg:~/GaussianModels/MH-d/pre_process_data/StandardNormalized.cpp

• Construct Gaussian Graphical models using MH-d

Generate empty graph file empty.dat for specific number of genes, compg:~/GaussianModels/MH-d/generate_empty_graph.C;
Modifiy the data file and graph file, number of genes, number of samples and so on in the input file "infile";
In file metropolis.C, set the interval of iteration for recording output networks:
```
if((k>2000000)&&(k%1000==0)){output graphs}
```
```
make metropolis
```

./metropolis < infile > Immortalised_networks.txt &

Select top networks and calculate edge frequency. compg:~/GaussianModels/MH-d/calculate_edge_frequency.C

• Graph Visionlization

Input files: GeneNames.txt, immortalised_network.txt (final network) and edge_frequency.txt, output file, the .dot file can be used by GraphViz compg:~/GaussianModels/MH-d/graphdrawing/generate_dot.C Then install the software ZGRviewer (Zoomable Visual Transformation Machine) to view the network dynamicly (zoom in and out, search for genes etc). GraphViz should be installed when using ZGRviewer. Download ZGRviewer, unzip it, modify the file run.sh by setting the directory of ZGRviewer and the location of java.

ZGRV_HOME=/home/ngl9/GraphDrawing/zgrviewer

/usr/local/java/j2sdk1.4.2/bin/java - cp ...

./run.sh

Go to preference, and set the directories of dot executable (like /usr/bin/dot) and neato executable file (/usr/bin/neato), etc. Load the .dot file, you can view the graph now. Search for specific gene and view its neighbors, nice software. But a better graph visionlization software is Cytoscape.

Last modified:
9 June, 2006