Difference between go enrichment and go semantic similarity

Question

[joram]

Well ...

I'm a little bit confused, about this two methods.

If I have a set of genes, go enrichment analysis will give me the statistically over-represented terms for those set. However, I can create a script (computationally expensive) to measure the semantic similarity of a set of genes and all go terms, one by one, for a given go branch (MF,BP or CC).

What's the difference?

Is there a paper talking about this question?

Answer 1

[svlachavas]

Dear Joram,

generally go semantic similarity and go enrichment analysis have different scopes-essentially, go semantic similarity analysis on some genes/proteins you have(for instance your deg genes), will ""cluster" your genes into different clusters based on their functional enrichment similarities(which is defined by the semantic metric you will use, i.e. resnik). However, it can be useful to reduce and "cut" redundant and "general" GO terms found enriched in your GO enrichment analysis, based on the measured similarities among the GO terms. Although this vignette is from an R package, it might be useful:

http://www.bioconductor.org/packages/release/bioc/vignettes/GOSemSim/inst/doc/GOSemSim.pdf

Best,

Efstathios