[BioC] Limma: import files with genes missing

[Gordon K Smyth]

Dear Daniel,

> Date: Mon, 14 Apr 2008 10:56:13 +0100
> From: Daniel Brewer <daniel.brewer at icr.ac.uk>
> Subject: [BioC] Limma: import files with genes missing
> To: bioconductor at stat.math.ethz.ch
> Message-ID: <48032A3D.20300 at icr.ac.uk>
> Content-Type: text/plain; charset=ISO-8859-1
>
> I am trying to use limma to import a series of bluefuse post processed
> files.  This is the command I am using:
>
> CGHraw <- read.maimages(CGHFiles,source="bluefuse",
> wt.fun=f,annotation=c("ID","NAME","POSITION","CHROMOSOME"),other.columns=c("NORMFACTOR","COPY
> #"))
>
> The problem is that each of these post processed files only contain the
> genes that pass the QA and so different files have different probes.

That sounds to me just horrible.  Personally, I wouldn't use an image 
analysis program which prevented me from making my own decisions about the 
quality of the data.  As I've said many times on this list, I think that 
wholesale spot filtering is almost always a counter-productive practice.

> I believe that limma finds all the common probes between the files and 
> then only imports for them.

Actually limma assumes that all files have the same probes.  The help file 
for read.maimages() says

"Warning: All image analysis files being read are assumed to contain data 
for the same genelist in the same order. No checking is done to confirm 
that this is true. Probe annotation information is read from the first 
file only."

> I would like to do the reverse, importing all the probes that appear in 
> any file and setting the value where they do not appear to NA.

No, limma does not provide any automatic way to do this.

If you can get a full gene list, and figure out which probes have been 
removed from each file, then you could set up a full size matrix, read in 
each array separately, assign each to right rows and columns, and then 
you'd have what you want.

The best solution by far would be to re-run BlueFuse and tell it to output 
all the data.

> Any way of doing this?
>
> I have tried importing the files separately and then merging, but that
> just uses the the probe list from the first file.

Correct.

Best wishes
Gordon

> Thanks
>
> -- 
> **************************************************************
> Daniel Brewer, Ph.D.
> Institute of Cancer Research
> Email: daniel.brewer at icr.ac.uk
> **************************************************************