I'm not sure exactly what you are trying to do, can you explain more specifically what you are trying to figure out?

In terms of creating a null hypothesis, a standard way to generate a null hypothesis for networks is to scramble the network (while preserving the degree of every node, in this case, your miRNA) and calculate the statistic on the new scrambled network. If you do this many thousands of times, you can get p-values for the statistic calculated.

Joshua broyde can you tell me how to scramble a non weighted network. Is there any short method if no. of nodes are very high.

I also do not really understand your need, nor how your network was built (how do you infer a link between miR-miR or miR-Protein_gene?

An easy way to scramble the network is just to exchange the microRNA names and Protein_gene name, keeping nodes and links identical.

We've built a miR-miR network based on predicted targets (not enough validated targets, at least at that time). We further analyzed the network. Our work could inspire you.

Title: A statistically inferred microRNA network identifies breast cancer target miR-940 as an actin cytoskeleton regulator

Link: http://www.nature.com/articles/srep08336

http://www.nature.com/articles/srep08336

Article A statistically inferred microRNA network identifies breast ...

Dear Aporov,

What is your aim with the network? What kind of data do you have?

You mentioned that you have expression data to construct a condition-specific network - if I get it right. If your aim is to find a master regulator miRNA than try to find anticorrelation with miRNA and targets. Construct a metric such as in percentage how many of the  mRNAs which are targeted by a miRNA is down-regulated. You can prioritize your miRNAs based on this way.

If you want to do some more network analysis based method my first question would be what kind of gene-gene interaction network have you used? If it is a coexpression-based prediction than you will get very weak predictions from your data. This is not good for an educated guess of the cell function (This is the best case you can get with network analysis). You can use some protein-protein interaction data to construct interaction networks and see which part of the interactome is regulated mostly of your miRNAs. Then you can use pathway analysis to get an idea, which part of the cellular function is affected.

Some tools to use: 

Cytoscape (I used it with few thousand nodes) http://www.cytoscape.org/

Igraph (network analysis package in R and python)  http://igraph.org/

Pajek (network visualization tool- I am not familiar with it but I heard it is good also) http://mrvar.fdv.uni-lj.si/pajek/

Some databases: 

miRNA -targets:

TARBASE http://diana.imis.athena-innovation.gr/DianaTools/index.php?r=tarbase/index

miTARBASE: http://mirtarbase.mbc.nctu.edu.tw/

PPI:

BIOGRID http://thebiogrid.org/

INTACT http://www.ebi.ac.uk/intact/

Reactome http://www.reactome.org/

SignaLink:  http://www.signalink.org/

Pathways: KEGG-metabolic part, Reactome to analyze any pathway, and GO, because every reviewer would ask you for GO

Use Bioconductor in R (https://www.bioconductor.org/) and look for packages for pathway analysis.

Reactome: https://www.bioconductor.org/packages/release/bioc/vignettes/ReactomePA/inst/doc/ReactomePA.html

miRNA path check this resource also:

https://www.bioconductor.org/packages/devel/bioc/vignettes/miRNApath/inst/doc/miRNApath.pdf

Please describe your problem a bit more deeply to help you.

Good luck!

Dezső

Apoorv Gupta, the easiest way to scramble a network is i to you represent your graph as  two columns (in your case, the left column might be miRNA, and the right column, the gene it affects). To scramble the network, simple randomize the right column. This preserves the degree of each node (i.e. each miRNA still is predicted to regulate the same number of genes) but now it is random. You can do the same thing with Protein-Protein interactions. For further graphical analysis, I recommend either the igraph package in R, or using cytoscape if you are doing things more manually.

Dear Laurent Guyon Sir my network is miRNA - gene- gene interactive network. There are many R tools available which take validated targets from Xgmml file in no. databases together. Your work is so inspiring and try to use your method in scrambling a network.

Joshua Broyde Thank you I understand that you want me to do degree preserving randomization. Can I used cytoscape plugin Network Randomizer App for it.

Apoorv, yes: http://apps.cytoscape.org/apps/randomnetworks. If this doesn't suffice, google "Randomize networks cytoscape" to see other packages. If not, you could use R,python or any other programing language