How Can I Compare Enriched/Depleted Go Terms Between Genelists?

Question

[Aaron Statham]

So I'm analysing human ChIP-seq data, and have made genelists of genes with enrichment at their promoters in two samples - each sample has about 3000 genes enriched, with ~1500 in common. I wanted to do some gene ontology analysis to see what kinds of enriched genes I am getting, and as a GO newbie I used DAVID which is a bit clunky but OK.

So now I have a heap of tables downloaded from DAVID of the enriched GO terms (for BP & MF 1,2,3 and FAT) in each cell line - is there a framework available for comparing the two samples enrichments robustly? I'd like to explore which GO terms overlap and which differ.

Additionally, I have a second ChIP-seq set of data for these two samples, so in the future I want to split each genelist into whether this second factor is present as well, and see if that stratifies out any particular biological groupings of genes.

Any help/pointers appreciated! I've only sunk a bit of time into DAVID so I'm more than happy to try something else (if it takes refseq mrna IDs) - I'm quite intimate with the NGS side of R/Bioconductor but haven't looked at any of its GO packages.

Thanks!

Answer 1

[Giovanni M Dall'Olio]

You can use babelomics, in particular their FatiGO tool

edit: have also a look at the list of third party Tools on the GeneOntology website.

Answer 2

[Ian]

I am very 'fond' of GREAT http://great.stanford.edu/ at the moment, but have historically mostly used DAVID. GREAT's principle advantage is that it associates genes (using one of three different models) to your binding region coordinates. There is also a new function that allows you to download all information at the same time, rather than the laborious saving of each table individually.

I can see nothing wrong on performing gene ontology analysis on the gene lists you describe.

Answer 3

[Khader Shameer]

FatiGo suggested by Giovanni will be a quick solution, in addition:

You may perform the GO enrichment analysis of the two sets separately and then compare the GO terms of the two sets either using visualization tools (for example REViGO) or based on comparison based on semantic similarity Ontologizer. You may use one of the standard GO annotation based enrichment calculation library for the comparison of enriched terms (topGO, GOStats).

Make sure that the tool/server that you use will help to define your own background data or already provide the platform that you are using for your analysis.

The following questions will be useful references:

• Tools For Visualizing Overlap Between Go Terms
• Tools To Find Gene Ontology Term Enrichment

Answer 4

[Chris Evelo]

You could try Go_Elite, it has the advantage that it can do pruning of the GO tree. Meaning it can first calculate the significance of the higher, more detailed classes and then, in case these are found to be significant, it ca remove these from the remainder of the lower, larger level class it occurs in. This prevents that you will find classes like "metabolism" to be different without knowing what really occurs.

Edit: the GO-Elite paper has now been published. It is at: http://dx.doi.org/10.1093/bioinformatics/bts366

[Chris Evelo]

Couldn't edit my own (old) post. Wanted to add that a GO-Elite paper has now been published. It is at: http://dx.doi.org/10.1093/bioinformatics/bts366

[Giovanni M Dall'Olio]

Hi Chris, it seems that you were logged with another account (chirs.evelo lower case, instead of Chris Evelo).

[Leonor Palmeira]

You might be using a different address in your 'chris.evelo' account. You could merge 'chris.evelo' into 'Chris Evelo' here.

[Chris Evelo]

Thanks! I have requested that now.