Remove All Genes Associated With A Kegg Pathway From Analysis

Question

[Mike Dewar]

I'm interested in the change in gene expression in T-Cells during the immune response. One way you can distinguish different functions of T-Cells is by which surface proteins they express at different times along the way. The trouble is that these surface proteins are such excellent discriminators that we use a small number of them to sort the cells out into different phenotypes. Hence, when I come along and apply my fancy differential gene expression tool to find out what's changing during the immune response all I find is a whole ton of other surface receptors that change with the phenotype.

This has a couple of problems: the first is that it's kind of boring - all I've discovered is that lots of surface receptors change during the immune response, which I already know. The second is that my bio colleagues are less interested in surface proteins as they're not very well conserved between organisms.

So I was thinking of removing all of the genes that are expressed in a couple of KEGG pathways (the cytokine-cytokine signalling and T-Cell receptor pathways) before I start my analysis and proceed. Two questions:

1) Is this a bad idea for some reason that I haven't realised? Are there more appropriate ways of filtering lists of genes before analysing expression?

2) Assuming it's an awesome, well-motivated-by-the-biology-and-the-data idea, are there any tools, preferably in bioconductor, that can give me a simple yes/no answer to the question: is this gene in either of these pathways?

[Mike Dewar]

Oh, I think question 2 has been answered already - if this is the case I'll just check it works and remove this bit from my original question. Sorry for missing it before asking!

[Will]

I gave you some suggestions for question #2

Answer 1

[Will]

If you're looking to remove genes that are "uninteresting" then I would suggest using Gene Ontology instead of Kegg Pathways. It tends to be more "complete" of a list then KEGG gene lists. I don't use bioconductor on a regular basis so I don't know of any library that will do this ... but I'm sure there must be something.

And removing genes before doing an analysis is a good way to help out your p-values since your background is smaller and more focused.

This is a pretty common thing to do in some GWAS analysis techniques. They remove SNPs that are not in coding regions, or near a gene, since any positive results will be "difficult to interpret".

In answer to question #2:

While I love parsing XML as much as the next programmer there's a much easier way to get all of the genes in a pathway. You can download the gene-ids (Entrez-IDS) for each pathway from their FTP site. Not sure which organism you're dealing with but each pathway has a .list file which lists all of the genes in the respective pathways.

[Mike Dewar]

thanks will! Useful to know about the .list files...

[Boboppie]

Hi will, can this gene list be fetched via KEGG API/web services?

[Will]

@Fengyuan: It probably can be retrieved from the KEGG SOAP but I find their design to be "unintuitive" (at least to me). I've always found that a simple wget of the FTP site gives me the results with less hassle but YMMV.

Answer 2

[Fred Fleche]

[?]

[?]

[?]

<?php
$myGeneId = "786";
if (in_array($myGeneId, $cytokinesCytokinesSignaling)) {
  //This gene belongs to the Pathway so do ......
}
else {
  //This gene does not belong to the Pathway so do ......
}

?>