[BioC] Limma-Targets file for Imagene data

[Sean Davis]

On 8/16/06 4:15 PM, "Jailwala, Parthav" <pjailwal at mcw.edu> wrote:

> Hi,
> 
>  
> 
> I am new to this usergroup and also a new user of Limma. I have a
> question related to the targets.txt file which is input to Limma for
> Imagene generated intensity data.
> 
> For each slide, I have a Cy3 file and a Cy5 file. In the targets file, I
> specify the respective files under the columns 'FileNameCy3' and
> 'FileNameCy5'. As I understand from the limma users' guide, the
> FileNameCy3 files correspond to the 'Green ' channel and the FileNameCy5
> files correspond to the 'Red' channel.
> 
> My question is, while calculating the M in the MA algorithm, is the
> ratio always going to be log2(Red-FileNameCy5 / Green-FileNameCy3) ?
> 
> I particularly wish to know this, as we have a dye-flip design, where
> consequtive slides have the dye-assignments reversed for Stimulated and
> unstimulated samples. So in the resulting output file after
> normalization (if the ratios are always Cy5/Cy3), I should be actually
> taking the negative of the log ratios for those dye-flipped slides to
> arrive at the 'Stimulated/Unstimulated' ratios for comparison.

In limma, the effect of dye is part of the "analysis", not the
normalization.  Just leave the red/green columns as they are.  So, yes, some
of your ratios will be upside-down from what you want.  However, when you
set up your design matrix for analysis, the dye swap nature can be accounted
for.  There are a number of examples in the documentation that use two-color
arrays with dye-swaps; you may want to check those out to convince yourself
that this is an OK way to go.

Sean