Selecting "interesting genes" from a network based analysis

Question

[Vinay]

After performing gene co-expression network analysis in a case-control study, I am ending up with a bunch of genes whose expression, along with its partners, is dysregulated in two conditions. These highly connected genes are considered as putative "hub" biomarkers. Since, it is not possible for us to consider all of them for experimental validation (for inhibition), what other analysis/tests can be performed to choose one or few of them? I know there is no so caller "master regulator", but somehow if I need to pick only "interesting" genes, what other statistical analysis can be performed?

All suggestions and comments are appreciated.

[Sam]

For me, I will approach it in two direction:

1. Try to look into each genes and see which one has a function that is most relevant to my condition

2. Try to rank the genes by the number of other interesting genes they connect to. The higher the connection, the more likely for it to be in the center of the network

Answer 1

[michael.ante]

Hi vinay.plp,

Since you have a network, you might have a look at topological indexes, like the "betweenes centrality" or the "pairwise disconnectivity index" to rank your genes. As Sam mentioned in his comment, you can also apply the Google page rank algorithm to your network. Have than a look at the top-ranked ones.

You can use external databases to enrich your gene set (e.g. Omim or GO) and filter out classes of interests.

Cheers,

Michael