[bigmw]

Thanks for the good words on GAGE. Let us know if you have more comments/questions.

I have done a quick test with GOseq, but I must admit I like GAGE better after first glance. Easy to follow, nice manual, nice plots, lots of results and possibilities. It really facilitates further analysis I think.

But Im going to give GOseq another go for sure!

[sindrle]

I have not done a direct comparison yet, but I will in the future.

[entrez]

Thanks for the input. I agree on your feeling on these two packages. Do you see differences in the results using GAGE and GOseq?

[sindrle]

I have done a quick test with GOseq, but I must admit I like GAGE better after first glance. Easy to follow, nice manual, nice plots, lots of results and possibilities. It really facilitates further analysis I think.

But Im going to give GOseq another go for sure!

[entrez]

Have you compared GAGE vs other gene set analysis methods like GOseq etc?

Thank you!


Another question, why is it that the absolute expression levels of genes is not important when doing GAGE/GOseq/DAVID etc.?

If you have a new cell type/tissue or even a species, how do you go about doing GAGE if you do NO COMPARISONS, only you want to get familiar with this new transcriptome?
You only have set of genes that expressed (including CPM/FPKM values), and you also know which genes thats not expressed.

Can you use this information? Basically, you just use one condition, and the expression values of the genes expressed.

[bigmw]

Having that said, you can always do GAGE (or other pathway analysis) on absolute expression levels. GAGE does work on single sample or condition (single-column matrices or vectors). But very likely, you will always see some housing keep pathways/groups, like cell growth, protein synthesis and energy metabolism etc, more expressed than others.

[bigmw]

In pathway analysis or gene set analysis, you may still use absolute expression levels rather than relative expression levels. However, two things make absolute expression analyses less desirable.

First, it is very hard to quantify absolute expression level. The expression measurements are affected by multiple factors other than the actual expression level or abundance, including transcript sequences, length, GC contents etc. These factors are mostly gene/transcript specific, normalization methods don’t model all these factor well.
Second, absolute expression levels are not very biologically relevant even if estimated accurately. Different genes, groups or pathways have very different working abundance ranges and dynamics. Therefore, it is less meaningful to compare the absolute expression levels between genes or pathways directly.

On the opposite, relative expression levels can be accurately estimated and are more comparable between genes because all gene specific factors are cancelled out in such measurements, i.e. a gene's expression relative to its own reference level.

[sindrle]

Thank you!


Another question, why is it that the absolute expression levels of genes is not important when doing GAGE/GOseq/DAVID etc.?

If you have a new cell type/tissue or even a species, how do you go about doing GAGE if you do NO COMPARISONS, only you want to get familiar with this new transcriptome?
You only have set of genes that expressed (including CPM/FPKM values), and you also know which genes thats not expressed.

Can you use this information? Basically, you just use one condition, and the expression values of the genes expressed.

[bigmw]

Forgot that sigGeneSet function has been updated to give users more control on the margin and font size. sigGeneSet calls a internal function heatmap2 to generate the heatmaps. So check the argument for this function
args(gage:::heatmap2)
The argument two relevant arguments here are margins and cexRow, which control the margins for column/row names and row name font size, you may do something like:
kegg.sig<-sigGeneSet(cnts.kegg.p,outname="~/RNAseq/13_Acute-Changes/14_GAGE_native_A1A2/A1A2All/A1A2All.kegg",pdf.size = c(7,12), margins = c(5,10))

Another question... How to alter the heatmaps text-part size (KEGG pathway names)?
The names are all capped, its unreadable.. The "pdf.size" - option only regualtes the graphic part.

###################################################
### significant.genesets
###################################################
kegg.sig<-sigGeneSet(cnts.kegg.p, outname="~/RNAseq/13_Acute-Changes/14_GAGE_native_A1A2/A1A2All/A1A2All.kegg",pdf.size = c(7,12))

[bigmw]

Why don’t you go beyond the preview step and actually open/load the file, you will see that the full graph extends beyond the artboard area.

You can do the following things:
1 Save a png file
File -> Save for Microsoft office

2 Edit artboard size and save another copy of new pdf:
Object -> Artboard -> Fit to Artwork Bounds->..

[sindrle]

Hm.
Ok, Ill try Adobe Illustrator, I tried Preview, but there they were capped as well...

I ended up editing it in Excel, cropping of the names and copy them in again from the "significant-gs.txt", then saving as new PDF.

[bigmw]

Someone asked a similar question here:

https://stat.ethz.ch/pipermail/bioconductor/2012-March/044427.html

Another question... How to alter the heatmaps text-part size (KEGG pathway names)?
The names are all capped, its unreadable.. The "pdf.size" - option only regualtes the graphic part.

###################################################
### significant.genesets
###################################################
kegg.sig<-sigGeneSet(cnts.kegg.p, outname="~/RNAseq/13_Acute-Changes/14_GAGE_native_A1A2/A1A2All/A1A2All.kegg",pdf.size = c(7,12))

[bigmw]

Well, you can check the first attached image in post 47. Obviously you do get different colors on pathview graphs. I think you might have checked the pathview output in the wrong directory:
kegg.dir = "~/RNAseq/13_Acute-Changes/13_GAGE_native_A1A2/A1A2pT2D/Pathview"

This is the directory storing original KEGG pathway graphs (and xml data). you should check in your current working directory for pathview output:
getwd()

Pasted wrong sessioninfo..

# > sessionInfo()
# R version 3.0.3 (2014-03-06)
# Platform: x86_64-apple-darwin10.8.0 (64-bit)
...

Pathview works, but I dont get colors or up/down regulated genes...

[shriram]

Shriram,
Here is the original graph from official KEGG website:

http://www.genome.jp/kegg-bin/show_pathway?sce00051+YCL040W

With the GLK1 gene node marked by red box. Hover your mouse over there, do you see 3 budding yeast genes in this nodes: YCL040W (GLK1), YFR053C (HXK1), YGL253W (HXK2)? You may click on this node, and check the details of these genes. Again, the nodes in pathview graphs are labelled with the most representative gene name rather than all mapped gene names for clarity.

That's great, it makes full sense now, thank you very much for your help.
Shriram

[bigmw]

Shriram,
Here is the original graph from official KEGG website:

http://www.genome.jp/kegg-bin/show_pathway?sce00051+YCL040W

With the GLK1 gene node marked by red box. Hover your mouse over there, do you see 3 budding yeast genes in this nodes: YCL040W (GLK1), YFR053C (HXK1), YGL253W (HXK2)? You may click on this node, and check the details of these genes. Again, the nodes in pathview graphs are labelled with the most representative gene name rather than all mapped gene names for clarity.

pv.out.list <- sapply(enriched_pathways, function(pid) pathview(gene.data = gene_fc, pathway.id = pid, species = "sce", gene.idtype="KEGG", same.layer = F, kegg.native = T, node.sum="median"))

Data in pathview:
pv.out.list[1]
GLK1 -0.35 0.000 0.620 -1.118 -0.900

# original data supplied for pathview
gene_fc[1,]
GLK 0.14 -1.6 0.62 -1.1 -0.37

I have attached the resultant pathway image.
Original image sce00051.png shows genes specific[in green] to yeast.

I am wondering why pathview data differ for GLK when GLK is the only gene on that node.

Thanks,
Shriram