Best Way To Visualize Geneontology Enrichments For Multiple Datasets?

Question

[enricoferrero]

Hi,

I'm looking for ways/methods/tools/ideas to visualize Gene Ontology terms enrichment in multiple datasets.

I have ~20 datasets (these are originally clusters of proteins obtained from a big protein interaction network), which show differences in the GO biological_process enrichment and I would like to find a nice way to show them all at the same time in one figure/slide.

I am aware of tools to visualise GO enrichments for single datasets, like GOrilla, Ontologizer and BiNGO but they're not very suitable to visualise and compare multiple datasets.

Thanks for your help!

Answer 1

[seidel]

Use a heat map whereby you have go terms as the rows, experiments as the columns, and the colors represent the degree of enrichment of the term in that data set. You can build it yourself, or I think the GeneAnswers package in R has something to this effect. It may help to represent the p-values of enrichment as a transformation whereby you represent the, as -10*log(P-value).

[enricoferrero]

Thanks, that's exactly how I was doing it and I can't think of any better method. So far I've just used some Perl-fu and R's heatmap function and the results are not too disappointing. I'll look into GeneAnswers tough, looks interesting.

Can you please expand a bit on why I should use the -10*log(P-value) instead that the p-value itself?

Thanks!

[Irsan]

I guess: p-values range from 0-1 and you are probalby interested in the ones between 0-0.5. When you do a continuous color mapping on p-values, it will be hard to get a nice visual. When you transform p-values with -10*log(P-value) the range will be 0 - infinitely high. When you map colors to those, its easier to get a nice visual.

Answer 2

[enricoferrero]

Just found about clusterProfiler in Bioconductor which looks exactly what I'm after!