[BioC] problems with affy in unix

[Rafael A. Irizarry]

do you mind sending the cel filenames to see what could be causing the
problem. also send version number of R and affy.

breaking up would be the easist thing. the larger the groups the better.

if you dont mind leaving R, you may try ben bolstad's RMAexpress:
http://stat-www.berkeley.edu/users/bolstad/RMAExpress/RMAExpress.html

finally, a strategy i recently heard terry speed suggest is to "train"
rma on say 50 chips (get probe effect estimates and an "average 
distribution" from bolstad's quantile normalization), then go chip by chip
quantile normalizing to the average distribution and
fitting a robust procedure to a linear model 
but with known probe-effects. this
requires some coding but it would solve your problem in a more elegant
way. btw, this strategy should work for other multi-chip measures such as
li  and wong's.


On Sat, 19 Jul 2003, fwan wrote:

> Thanks, Dr. James:
> 
> For the 2nd question, The ReadAffy did go wrong. I tried justRMA on the same
> cel files of class BCR and there are sample names instead of just "H".
> samples names are cel file names.
> 
> for the 3rd question, I still need some help. I have 335 cel files in total
> and I even tried justRMA and it always went wrong when it read in the 127th
> file, saying something like:
> could not open Hyperdip-5-C.cel (it is not problem with this cel file, i
> deleted it and justRMA went wrong again at another 127th file with same
> error message). My question is: if I break 335 into 3 groups ( approximately
> 110 cel files in a group and each group contains a few classes of lukemia
> cel files defined by saint jude childrens. there are 9 classes:
> BCR,NORMAL,HYPERDODIP,MLL..etc) and then run justRMA on each group and merge
> the final expression files later, is this process justified for the later
> data analysis ? I know this expression file will be different from the
> expression file i can get from justRMA on 335 cel files at one time.
> however, upgrading the mainframe might not be possible right now.
> 
> thanks again for your help. good weekend.
> 
> regards,
> 
> w.f
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor
>