[BioC] Selecting probe pairs for analysis

[Laurent Gautier]

Hee Siew Wan wrote:
> Hi All,
> 
> I'm interested in calculating the expression measure using only 8
> pairs of probes from each probe set of Arabidopsis genechip (i.e
> instead of using the whole 11 pairs of a probe set). After searching
> through the archive, I found that I can create a new cdf environment
> that excludes the pairs that I'm not interested in. However, when I
> tried using makecdfenv to create the new CDF package, I get a
> Segmentation fault. I'm using R Version 1.9.0 (2004-04-12) on UNIX
> platform.

This is not good that makecdfenv segfaults (presumably caused by the C 
level parser for CDF files, I'll see if I can guess where it happens),
but noodling with cdf environments at the CDF file level is not good either.

> I understand that the error occurs due to the file that I have. I
> have a ATH1-121501.CDF of type Channel Definition File (which I
> didn't have problem reading)

..the parsers were made to parse genuine Affymetrix files...

>                             and I modified this file by deleting 1
> pair of probes from 266455_at. I saved it as another .CDF. I opened
> the file using EditPadLite Version 5.3.0 and did the modification
> from there as well. I'm not very sure where did I make mistake(s).
> I'd appreciate any comment on this.

...well, the CDF files contain a bit of redundancy (like the total 
number of probe pairs in the chip in the header), and blindly deleting
some lines in the file may hurt (and it segfaults rather than possibly 
returning silently non-sense)... once again this is not 
adviced/supported (at least by myself)/recommended to do surgery at the 
CDF file level.

> Is there another way of reading certain probe pairs instead of
> deleting them in the CDF? I'd appreciate any help. Thanks.


The package 'altcdfenvs' is your friend.



> Cheers siew wan
> 
> 
> 
> ------------------------------------------------------------------------
> 
> 
> _______________________________________________ Bioconductor mailing
> list Bioconductor at stat.math.ethz.ch 
> https://stat.ethz.ch/mailman/listinfo/bioconductor