Pathway analysis of Differentially expressed genes

[himanshu04]

Dear all,
I am new to the field of RNA-seq analysis. I used Tophat and the Cufflinks/Cuffdiff to get the differentially expressed genes.I had two conditions a wild type and tumor. I wanted to now do the Pathway analysis and determine which genes are up and down regulated for the PI3K/AKT/mTor pathways in the tumor . What should be the approach from here?. Also , I have the cuff diff output but I was not sure which one of the files would I be using for the Pathway analysis.
I am naive to this field and so I apologize for the basic nature of the questions?.
Any help will be much appreciated.
Thanks,
Himanshu Sharma.

[phoss]

Hi Himanshu,
We've done similar work ourselves.
What we did was map each differentially-expressed gene onto their respective Enzyme Commission accession (EC). Doing so was quite helpful because as you know, KEGG uses KO and/or EC accessions in their pathways. KEGG also has mTOR signaling; you noted you're interested in that pathway.
You could then use tools such as KeggAtlas, DAVID, Paice, SubPathwayMiner... just to name a few. These tools interface your dataset with KEGG pathways.
If the tools above do not do what you're looking for, you may want to build your own custom solution via the KEGG web-service API... it's well documented and available for many popular programming languages.
Hope this was helpful.

[himanshu04]

Dear phoss,
Thanks a lot for the reply . It is very helpful indeed. Although I wanted to know how did you map each DE to their Enzyme commission accession?. I know I need EC as it is compatible with KEGG and GO gives me Uniprot Id's. So what exactly did you use ?. It will be really greatful If you could give me some leads.
Thanks,
Himanshu Sharma.

[phoss]

Hi Himanshu,
Glad it was helpful.
We obtained accession-to-EC mappings via biomart. If however such mappings do not exist for your model, you could use uniprot by running uniprot BLAST against your DEGs. This is handy because a good-number of their accessions have ECs. EBI and GOA are good resources too.

In-case you're wondering, we had good results with DAVD and the KEGG api (both java and python). We've used / had good results with KEGGAnnotator, KEGGanim, paice and DAVID. PathRender in R is pretty-neat too. I actually developed paice ~2yrs ago but it's always good to try other tools and use the one best-suited for the job. We developed paice to help with gene-family visualization since some ECs occur in multiple copies but yet have different expression values.

KEGG tools have been exhaustively studied so you'll have access to many tools / resources. Have you checked out other databases such as Metacyc or Reactome?

[himanshu04]

Dear Phoss,
Thanks again for your help. It is really useful. Now I had a few questions if you dont mind. I have approximately 1200 genes which are significant and my model is mouse (mus musculus). So is there a way to get the EC of all 1200 genes together because it will be very long and tedious to do it one by one. Also, if you could guide me a bit to obtain EC from the gene name via Biomart.
Thanks again for your help. I really appreciate it .
Thanks,
Himanshu Sharma.

[phoss]

Hi Himanshu,
I personally have not worked with M. musculus but I recall a well-known mouse resource to be 'Mouse Genome Informatics' (MGI).
You could use MGI Biomart (under Analysis Tools) to mine-out GO mappings for your DEGs and then map such GO accessions against GO->EC mappings from http://www.geneontology.org/external2go/ec2go
Please correct me if I'm wrong, but I did not see any EC retrieval-option @ MGI-BioMart. If you have GO accessions, you could easily cross-link it with EC accessions nonetheless.
All the above info is pretty-much enclosed in the EBI-GOA page: http://www.ebi.ac.uk/GOA/downloads.html

[chknbio]

I am working with chicken RNAseq dataset. I completed my analysis with cuffdiff and have a list of DE genes. How can I do about finding if these genes are involved in a pathway? Can I use the diff_gene file from cuffdiff for pathway analysis?

[Charitra]

hi Phoss
Thank you for your guidance towards pathway analysis of my deferentially expressed RNA-seq genes. However, as a beginner to pathway analysis, I would like to know details with protocol. Can you please give me some link of study material so I can study and follow the same.

I need to learn every steps of the same.
Thanks again.

[jp.]

anybody got answer ? I am also waiting ?

[bigmw]

If you know R, GAGE/Pathview workflow in Bioconductor can do the pathway analysis with your data. It works for both RNA-seq and microarray data. No mapping to EC is needed, and no pre-selection or filtering of genes either.

Here is an example workflow: https://stat.ethz.ch/pipermail/bioco...ly/054021.html
The packages are available at:

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

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

Some example graphic outputs here:

http://pathview.r-forge.r-project.org/

[vishnuamaram]

Hi charitra & jp,

Once you have your list of genes either up or down regulated obtained from either RNA seq or microarray,
you can go to DAVID
and put your list of genes in Start Analysis options,
select the option usually official gene symbol & proceed for the analysis.

Good luck

[jp.]

Dear bigmw and vishnuamaram
Thank you for your kind guidance. However, I am not able to use both of the programs (pathview and DAVID). For the pathview, I followed the tutorial successfuly. But not with my sample.
My main problem is that I have RNA-seq cuff_data_diff_gene.txt or original files. I really can not upload into either programs.
Any help ?

[bigmw]

You haven’t really followed the GAGE/pathview workflow for RNA-seq data, where you don’t really need Cufflinks/Cuffdiff, just map the raw reads using tophat:

https://stat.ethz.ch/pipermail/bioconductor/2013-July/054021.html

I would suggest you go through that first. You can’t input arbitrary data and ask the programs to figure it out for you. You may also want to read the GAGE/pathview documentation first.
I am not familiar with Cufflinks/Cuffdiff and its output. I assume it is a list of significant genes with certain p-value cutoff. To be able to use GAGE/pathview, you will need:
1. a list of all genes (usually thousands or tens of thousands entries) with its differential expression score, like fold change, t-statstics etc.
2 read that list into R as a vector or a 1-column matrix named with Gene IDs (Entrez Gene, Gene symbol etc). You may want to check functions “read.delim” or “read.table” for how to do this.
3 make sure you know enough R

You may want to to collaborate with a bioinformatician if you are not sure how to do these.