I would like to know how to proceed from my gene expression data to building a network from that expression data .
So far what im doing is checking the differential expression of genes and see how it changes ..but that wont give me a global picture i suppose ..So it would be helpful if anyone can suggest me how do i proceed what kind of approach should i take .Suggestion would be really appreciable
so i wont be doing any ranking of deferentially expressed genes which i have from Deseq2 , i could just put the huge list to string DB? lets say i have 1000 diff expressed genes , would it be meaningful to pass the gene list to string DB?
Well the basic principle of plotting a network of related genes is that you must try to plot a network of RELATED genes. So rather than passing a RANDOM gene set we first make a set of genes with something in COMMON (e.g. a differentially expressed gene set, or co-expression gene set, or tissue specific gene set etc. ) in the hope that a biologically meaningful network can be constructed.
The key word here...is HOPE.
I would rank the significantly differentially expressed genes by padj and select then the top 1000. Hopefully you will get some nice networks in there. But expect to see lots of unconnected genes.
That would depend on your experimental design and what kind of biological questions you have in mind, I would say.
You can check BioLayout 3D package (http://www.biolayout.org/) and check their tutorials. You can build some network together with the visualization through that package and I also found their tutorials really helpful.
Biolayout 3D has been deprecated and is no longer supported..Miru (https://kajeka.com/) is its successor and is faster and more optimised than its predecessor..
If you have access to the enzyme annotations for your genes, you can use the KEGG API which can link enzymes to kegg pathways (http://www.kegg.jp/kegg/rest/keggapi.html). You can use R (or your favorite language) to access the API.
Perhaps a bit late, but given a single differential expression result, I typically go to a resource like GeneMANIA to generate a network based on more comprehensive co-expression analysis and then import that into Cytoscape. Then I overlay my diff exp data onto that network to assess the "functional" modules being regulated in my dataset.
First: Get the list of genes that are most significant deferentially expressed using a statistical package like DESEQ2
Second: Pass the gene list to StringDB
Third: Import the output of StringDB to Cytoscape
so i wont be doing any ranking of deferentially expressed genes which i have from Deseq2 , i could just put the huge list to string DB? lets say i have 1000 diff expressed genes , would it be meaningful to pass the gene list to string DB?
Well the basic principle of plotting a network of related genes is that you must try to plot a network of RELATED genes. So rather than passing a RANDOM gene set we first make a set of genes with something in COMMON (e.g. a differentially expressed gene set, or co-expression gene set, or tissue specific gene set etc. ) in the hope that a biologically meaningful network can be constructed.
The key word here...is HOPE.
I would rank the significantly differentially expressed genes by padj and select then the top 1000. Hopefully you will get some nice networks in there. But expect to see lots of unconnected genes.