How Do You Align Or Compare Two Networks Content-Wise?

Question

[Michi]

Hello BioStars

Imagine the following scenario: you have a few networks of connected genes. Let's say for 20 different tissues. To make it easier without direction.

How would you compare them to access the similarity of the network? So I am interested in comparing the content and not network properties like "shortest path" etc... I found some Network Alignment tools, but they are (almost?) exclusively for homology alignments, which is not the case in my scenario.

So again the resulting question is: How do you align or compare two networks content-wise? Also note, that in the end I am interested to be able to make a statistical statement about the similarity / alignment between two or more given networks.

Thanks in advance for any help!

[Qdjm]

Hi dmeu, could you please clarify a bit further what you mean by content? Are the networks over the same gene set and you want to compare the connectivity patterns?

Answer 1

[Pawel Szczesny]

If I understand the question correctly, you want to do enrichment analysis of these sets of genes. There are several tools for that (I guess more than a hundred). As for me, for simple sets I use FatiGO (maps genes to Gene Ontology, Kegg and Reactome and returns statistical significance of differences) but I'm keen to try more sophisticated DAVID (which includes literature and semantic analyses). The latter has a nice feature of reading wide range of identifiers.

[Michi]

Hi, thanks for your answer! So you say I should make an enrichment analysis with the pairs of connected genes in the network? I was thinking about that..but stumbled upon the question of what would be the background for making the enrichment analysis? All connections possible between all existing genes? In the end I just want to know the similarity between two networks independently of further variables...

[Pawel Szczesny]

You have two networks composed of two sets of genes. So forget the network, but focus on two lists of genes. Both sets of genes represent certain functional annotations and you check whether these functional annotations differ significantly (using any of the mentioned services). If they don't differ, you get no differences. If they differ, you will get information on the differences (for example one network may have more genes related to glycolysis). In case you have a single set of genes, the second set is the rest of genes in the organism/cell/tissue.

[Michi]

OK, I see your point. I will surely do that (also). But still I am looking for a direct comparison method.

Thanks a lot freesci