- Pathway and Network Analysis 99
then subjected to the hypergeometric test on overrepresentation in KEGG
pathways for those upregulated genes in ALL1/AF4 samples (or called
downregulated genes in BCR/ABL samples):
hgOver<- hyperGTest(params)
df.up<- summary(hgOver)
head(df.up)
You can also test the downregulated genes in ALL1/AF4 samples (or called
upregulated genes in BCR/ABL samples) under the same settings. You can
create another KEGGHyperGParams object or simply adjust the parameter
geneIds in the KEGGHyperGParams object params and test as follows:
geneIds(params)<-dn
hgOver<- hyperGTest(params)
df.dn<- summary(hgOver)
head(df.dn)
Or you can test the union of both upregulated and downregulated genes like
this:
geneIds(params)<- c(up, dn)
hgOver<- hyperGTest(params)
df.both<- summary(hgOver)
head(df.both)
From these pathway analyses, you can observe that KEGG terms like “Cell
adhesion molecules (CAMs)” or “Antigen processing and presentation” are
enriched for downregulated genes in ALL1/AF4 samples (or upregulated
genes in BCR/ABL samples), suggesting that BCR/ABL-rearranged B cell
leukemia may be more immunoreactive than ALL1/AF4-rearranged B cell
leukemia.
- Creating gene correlation networks
To create gene−gene correlation networks for these BCR/ABL and
ALL1/AF4 leukemia samples, you can use Pearson correlation coefficients
(PCCs) and DE genes derived from the SAM analysis. You can utilize
the function compCorrGraph() in R package GOstats to compute PCCs of
genes. Additionally, you can specifically take the DE genes with fold change
greater than or equal to 4 into consideration. In order to be able to plot the
graph object returned by compCorrGraph(), you have to load the package
Rgraphviz:
library(Rgraphviz)