R: Heatmap Of Rna-Seq Count Table With Functional Gene Clustering

Question

[mtw]

I would like to prepare a heatmap of differentially expressed genes from an RNA-seq experiment with some sort of functional gene clustering in R. Starting from standard read counts (as obtained by htseq-count), the DESeq manual illustrates how to create a heatmap for the most differentially expressed genes, however I would like to have a clustering based on GO terms.

Is there any R package available?

Thanks, Michael

[Pavel Senin]

Is that boils down to some sort of metrics based on the product of GO group assignment score (similarity) and observed expression variance? It is quite interesting to see!

[Damian Kao]

I am not sure if any R packages are available for this. Closest service I can think of is Revigo: http://revigo.irb.hr/

It probably isn't what you want, but they provide some resources on how you can cluster GO terms based on semantic distance calculated from the GO graph.

[Dave Bridges]

Depending on how you annotate your DEseq results, you can add GO annotations (for example using biomaRt), then filter those rows for whichever GO id you are interested in. After you have the filtered dataset, the heatmap should be the same as for differentially expressed genes.

Answer 1

[vj]

Try using http://bioconductor.org/packages/2.12/bioc/html/GOSemSim.html The function mgeneSim may be what you are looking for.

Answer 2

[Wrf]

maybe check out edgeR. if you have numbers already, i think the image() command can generate heatmap graphs.