GO categorization

Question

[m.habib]

Hi all,

I am working on a non-model organism. I finished annotation of the transcriptome using trinotate and now I want to do gene set enrichment analysis.

• How to do GSEA for non-model organisms? I saw many posts and it doesn't seem to solve the problem?
• How to categorize GO terms into general functions such as immune response, reproduction, development, ....etc?

Answer 1

[geneontologyhelp]

We have this in our FAQ. PANTHER version 18.0, which powers the enrichment analysis on our homepage, has 143 species loaded and can work on an additional 1454 genomes from the Reference Proteome project. If you don't/can't add your genome to the Reference Proteome project, you can also score your genome against the PANTHER library. See more in our Nature Protocols paper, Box 2.

For the second question, you can use a tool like GO Term Mapper or GO Term Finder. I believe both of these allow you to upload custom GAFs, and here are the specs for GAF 2.2.

• The Term Mapper will map back up to terms in the specified GO Subset (GO Slim). These predetermined bins can be customised by using your own Subset, or for most organisms the Generic GO Subsets will suffice. You might explore the other subsets and use the SGD Subset if you have an organism more closely related to Saccharomyces cerevisiae, for example. This tool is probably what you are looking for to get a summary of GO terms.
• The Term Finder is not limited to predetermined bins but will just show significant shared GO terms. This tool might be helpful if you find the bins in Term Mapper are too general to be informative.

Answer 2

[Istvan Albert]

If you have an organism that lacks the GO to gene mapping then you have to find a way to assign these terms to your genes.

Basically your first step of the analysis needs to be annotating your genome with functions. Many people use reciprocal blast with the closest species.

A paper on reciprocal blast and how it works

• Quickly Finding Orthologs as Reciprocal Best Hits with BLAT, LAST, and UBLAST: How Much Do We Miss?

here is a program that can do it

• https://github.com/cboursnell/crb-blast

and many other alternatives

• https://github.com/cboursnell/crb-blast
• https://github.com/masaomi/rbh

[m.habib]

Thanks Istvan for your reply. I think these programs will do the same as Trinotate. Trinotate searches against numerous databases and generate an annotation report. I now have that annotation report and I extracted the GO terms in a two columns file with the first for the genes ids (TRINITY_DN100006_c0_g1, TRINITY_DN100020_c0_g1, .....) and the second column for the GO ids (GO:0000054,GO:0000055,GO:0000056,GO:0000122,GO:0000776,GO:0003674,GO:0003676,GO:0003723,GO:0005049,GO:0005215,....) with some genes having so many GO ids.

I will do DGE and I would like to perform GSEA for significantly differentialy expressed genes and to categorize the GO terms into general functions such as immunity and reproduction and so on.

I also want to assign the genes to the same general functions!