[BioC] limma

Sean Davis sdavis2 at mail.nih.gov
Mon Aug 20 23:25:56 CEST 2012

On Mon, Aug 20, 2012 at 5:19 PM, Novak Jaroslav
<jaroslav.novak at gmail.com> wrote:
> Fantastic and fascinating speed. I am hard pressed by deadline, so it
> means a lot!!! and have no experience with limma yet. Question: I
> assume expression matrix should be txt tab delimited, but what format?
> 1st col ID, 2nd etc expression log2, first row ID tab sample1 tab
> sample2 tab etc??

Hi, Novak.

It is best to keep the emails on the list so that you get the best
help you can from the many contributors to the list.

The format in the text file is not important, really.  Once the data
are in R, you can manipulate them as you see fit.  To answer your
implied question, a "matrix" in R is a rectangular data structure with
only one data type in it (in your case, it will be numeric), so you
will need to remove any columns (in R, probably) that do not contain
numeric values such as those with gene names, etc.  The same goes for
the first row.  You can use rownames() and colnames() to keep things
straight if you need to.  Ultimately, you will want to learn about
ExpressionSets to get around the shortcomings of data integrity with
simple matrices.

I hate to break it to you, but you will probably need to do a little
reading on limma and R before you can proceed with the task at hand.
The limma user guide is quite good reading, though.

Hope that gets you going.


> thanks again!
> Jaroslav
> On Mon, Aug 20, 2012 at 5:08 PM, Sean Davis <sdavis2 at mail.nih.gov> wrote:
>> On Mon, Aug 20, 2012 at 5:03 PM, Novak Jaroslav
>> <jaroslav.novak at gmail.com> wrote:
>>> I want to use limma (3x2 factorial experiment), I have a very large
>>> data set, which cannot be handled as CEL files by my computer.
>>> However, I have all expression values. How can I create eset file
>>> using my summary expression values? Thanks for help! /Jaroslav
>> Hi, Novak.
>> There really isn't a need to create an ExpressionSet for the purposes
>> of running limma.  Limma is quite happy with using a matrix as input.
>> If you do want to create an ExpressionSet, be sure to read the
>> documentation associated with the Biobase package.
>> Sean

More information about the Bioconductor mailing list