Hi there,

I was just recently asked to review some normalization techniques for miRNA data but I haven't really worked with R much (don't really understand what is going on in the calcNormFactors function) and I am only a second year undergrad student so I am a little lost when it comes to reading journal articles. I was wondering if anyone had any idea how to implement the TMM normalization technique which was written about in a journal article by Mark Robinson and Alicia Oshlack? I need to implement this in MATLAB for a matrix of miRNA count data so if anyone could point me in the right direction that would be great.

Katherine

If you need to implement it in MATLAB, then you will have to either a.) read and understand the paper you cite, and then implement it, or b.) convert the R implementation in edgeR to MATLAB. As a neophyte, neither will be easy, but I assume that is the point?

If you are to choose b.), then the easiest thing to do is to get the source code. You can either go here, using 'readonly' as both username and password, browse to edgeR/R/calcNormFactors.R and saving as a text file, or downloading the tar.gz file, untarring and ungzipping it, or directly from subversion, again using readonly as username and password.

The source code is well commented, so you should be able to follow in general what the intent of the code is. The three main functions you should concern yourself with are calcNormFactors.default(), .calcFactorQuantile(), and .calcFactorWeighted(). The first is the main function, the second is used to figure out which column to use if a reference isn't specified, and the third is the workhorse function that calculates the TMM normalization factors.

The edgeR package, and R/Bioconductor in general, contain provide an enormous amount of code for the analysis of miRNA and mRNA data. I used to be a MATLAB user/programmer myself long ago, and I can vouch that sooner or later it will pay you to make the switch to using R rather than trying to reproduce things for yourself in MATLAB.

It might be easier to port the MATLAB calculations to R instead of porting the R libraries to MATLAB. I've looked into the TMM normalization before and unless you have a good feel for R + C, you can have a look at the source code for edgeR here: https://hedgehog.fhcrc.org/bioconductor/trunk/madman/Rpacks/edgeR/ the username/password is readonly/readonly There is also a very well done users guide for edgeR that can be found here: http://www.bioconductor.org/packages/release/bioc/html/edgeR.html