Hello! May I ask what's the different principle between protein-protein interaction and mRNA-mRNA interaction? Can I use some R packages for prediction of protein-protein interaction, for mRNA-mRNA network? e.g.STRINGdb
I will be grateful short and clear answer is given, with detail links or publication provided.
By mRNA-mRNA interactions, I presume that you are referring to co-expression networks and not physical (i.e. molecular) mRNA-mRNA interactions? Co-expression networks are just based on correlation, a statistical parameter, and have no direct basis on molecular mRNA-mRNA interactions. Genes that are highly correlated to each other in a well-powered study are likely to be part of the same pathway, but not always.
For PPINs (protein-protein interaction networks), one can still just construct these based on correlation.
STRING is a bit different because it can build co-expression networks for either gene or protein expression data, and it uses known, curated protein interaction databases in order to do this.
From my own experience of working with other scientists in both gene and protein co-expression networks, there is a lot of confusion about how best to interpret the results. Network analysis is promising but its current implementation is limited until we have more curated databases that can be used as 'scaffolds' with which to build and guide the construction of these networks.
For a bit of further reading, take a look at this recent post: A: Which method do you use to build adjacency matrix of a genes co-expression netwo
Kevin
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
Yes. I only mean co-expression networks, if Im interested in RNA-sEQ. If I understand correctly, you mean STRING can also be used in mRNA network? A bioinoformatician in our group did is to input the pre-defined gene list, based on the module connection(in String for protein interaction), and delete that the genes are not highly interacted，then put into enrichment analysis. This raise me a question, if the mRNA interaction is not same as protein interaction(based on I have only RNA-Seq data), I guess this kind of enrichment for my case will be something wrong, right?
I don't believe it is incorrect to do it that way (as you and your colleague have done) - one must just understand what it means based on the methods that you are doing. Whilst mRNA transcription (RNA-seq) is not quite reflective of the proteome, it is a still an excellent measure of it without actually conducting a protein expression study.
As an example, 'everyone' performs pathway analysis using microarray and/or RNA-seq gene expression data using KEGG, Biocarta, Reactome, et cetera, but all of these databases are also based on protein interactions and protein pathways.
One could argue that, by using STRING, one is performing 'supervised' network analysis; whereas, if you do not use STRING and just build a co-expression network baed on just correlation between RNA-seq expression values, then that would be 'unsupervised'. It depends on your perspective.