[BioC] Re-mapped Affy CDF files

James W. MacDonald jmacdon at med.umich.edu
Mon Jan 16 20:03:20 CET 2006

Hi Karl,

Interesting results. I see about the same thing when I analyze a set of 
data using both mappings; the probesets with big differences are usually 
consistent, whereas the probesets with smaller differences may vary.

The only time this would likely make much difference is when you have an 
experiment where there really are very limited differences between the 
groups (often the case with brain research, which I think is one of the 
reasons the MBNI folks started doing these things).

Anyway, it looks like we are going to be making these cdfenvs and probe 
packages available on BioC. Hopefully this will increase 



Dykema, Karl wrote:
> Jim,
> I've spent some time investigating the re-mapped CDF files you asked
> about last week on the BioC mailing list.
> We used a quick-and-dirty (but surprisingly effective) categorical
> approach to address this problem.
> Basically, if you compare a sample that contains an extra chromosome
> (i.e. 3 copies of chromosome 7) to a sample that contains a normal
> chromosome number (i.e. two copies) many of the genes that map to that
> chromosome will show relatively increased expression.
> So the process is:
> 1) create gene lists that map genes to chromosomes
> 2) preprocess gene expression data
> 3) simply subtract the gene expression profile of the normal tissue from
> the gene expression profile of a sample suspected of harboring a
> chromosomal abnormality
> 4) see if there is an enrichment of positive gene expression values in
> any of the chromosome derived gene lists.  This would indicate a
> chromosome gain has occurred for that chromosome.  Likewise, an
> enrichment of negative gene expression values would indicate a
> chromosome loss has occurred.
> 5) see (http://genomebiology.com/2002/3/12/RESEARCH/0075) for more
> details of this quick method.
> The advantage of this categorical approach is that certain tumors
> contain very reproducible sets of chromosomal abnormalities and can
> serve as positive controls.
> In this case, the tumor samples are all papillary renal cell (kidney)
> carcinomas and type of cancer has been shown to previously shown to
> produce gains of chromosomes 7, 16, and 17.
> The data was preprocessed using RMA using both the U of M version 6
> Entrez CDF and the standard CDF included in BioC. Each tumor sample was
> compared to a pooled normal kidney reference and enrichment scored using
> a simple t.test of gene that map to each chromosome. Plotted is a
> heatmap of resulting t.score (red high, blue low)
> It is easy to each the enrichment of positive values of genes that map
> to chromosomes 7,16,17 using either preprocessing method.
> While this anecdotal evidence and does not prove that the custom CDF
> files are any 'better' or 'worse' than the old Affy mappings, it
> suggests to me that they are "reasonable" (as opposed to "unreasonable")
> I'd be interested to get your comments. Thanks. This was not cc'ed to
> the BioC mailing list. If you like you can forward it on.
> -------------------------------
> Karl Dykema
> Bioinformatics Programmer/Analyst
> Laboratory of Computational Biology
> Van Andel Research Institute
> 333 Bostwick Ave. NE
> Grand Rapids, MI 49503
> (616) 234-5554

James W. MacDonald
Affymetrix and cDNA Microarray Core
University of Michigan Cancer Center
1500 E. Medical Center Drive
7410 CCGC
Ann Arbor MI 48109

More information about the Bioconductor mailing list