Rohan raised a valid point about attributing read counts to pathway. What can be done is to use DESeq or edgeR to estimate a log2 fold change for each gene.

Use these log2 fold changes, you can perform a geneset enrichment analysis by considering the log 2 fold changes within a group of genes, to see whether there is evidence of up or down regulation as a group. Below are two links, one is an introduction to gsea and the other a package I use every now and then.

http://master.bioconductor.org/help/course-materials/2009/SeattleApr09/gsea/

http://bioconductor.org/packages/release/bioc/html/gage.html

If you would really want to use a glm approach, then what I can think of now is to use the log 2 fold changes as a response, and introduce the pathway information as variables. 

It all depends on the question you need to answer.. The glm approach can take a bit more time to tweak out.

Thanks for the answers. I am familiar with DE analysis and GSEA as well. I am thinking about gain some power though combine them directly.

Hi Eszter,

Best approach is to run GSEA on your normalized counts rather at the level of fold change. GSEA also provides an option to run enrichment on your own pathway of interest. http://software.broadinstitute.org/gsea/index.jsp