Clustering Gene Regulatory Networks

Question

[Diana]

Does anyone have any suggestions as to which clustering technique is best for producing gene regulatory networks from expression data and why is it best?

Answer 1

[Sudeep]

Have you tried WGCNA?

Answer 2

[Manu Prestat]

First, I think that you want to infer genetic regulatory networks from expression data (that is actually not "clustering"). If I'm not wrong, you can start with these articles:

• de Jong. Modeling and simulation of genetic regulatory systems: a literature review. J Comput Biol (2002) vol. 9 (1) pp. 67-103
• Werhli et al. Comparative evaluation of reverse engineering gene regulatory networks with relevance networks, graphical gaussian models and bayesian networks. Bioinformatics (2006) vol. 22 (20) pp. 2523-31

Most of these methods are not so trivial and do not come from the same fields/"phylosophies". Dig into one of them (are your data kinetics, i.e. do you need a dynamic/static method?...)

[Diana]

Thanks Manu, I didn't realize its inferring rather than clustering. We want to identify groups of genes from expression assays that have a similar expression pattern and I thought that would involve clustering. I am sorry I didn't make it clear from my Question. But we also want to build up a gene regulatory network. I'll check the papers you suggested. I'm not sure what you mean by dynamic/static. We have expression profiles for 3 different time points.

[Manu Prestat]

Ok, so actually, you want to do both: [1] clustering your samples according to their expression profiles, and [2] infer the genetic regulatory network. For the second step, you have to think about dealing your time condition, if you want to modelize that explicitly (dynamic model, e.g. Dynamic Bayesian Network) or not (static model, e.g. Correlation/Mutual Information Based Networks, Bayesian Networks...etc). Unfortunately, I'm not familiar with dynamic ones...