We constantly experienced problems with rma function with soybean chip. Since the possible reason being the chip is too huge, i wonder whether there's a way to reduce the cel file size by taking only part of the raw intensity info for normalization. Any one can comment /addvise on that? Many thanks, Dianjing

On Thu, Nov 04, 2004 at 02:08:09PM -0500, Dianjing Guo wrote: > We constantly experienced problems with rma function with soybean chip. > Since the possible reason being the chip is too huge, i wonder whether > there's a way to reduce the cel file size by taking only part of the raw > intensity info for normalization. Any one can comment /addvise on that? That does not seem like a very good idea. I have not seen any postings that suggest that size is the issue; have you made them? None of this needs to be mysterious in any way. You should 1) make sure you have an up to date R, and an up to date version of the package. If you get errors, such as segmentation faults then you can use R -d gdb provided you have compiled R with the -g option (and if not then you will need to recompile it). From there you can track down the source of the bug and it can be fixed. For other bugs (such as problems in R code) there are options such as using debug etc. It is generally much better to figure out what is wrong, and why than to invent rather peculiar one-off solutions. Robert > > Many thanks, > Dianjing > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor -- +--------------------------------------------------------------------- ------+ | Robert Gentleman phone : (617) 632-5250 | | Associate Professor fax: (617) 632-2444 | | Department of Biostatistics office: M1B20 | | Harvard School of Public Health email: rgentlem@jimmy.harvard.edu | +--------------------------------------------------------------------- ------+

Hi Dianjing, I also do not think that the problem is the size of the chip since we have applied rma to a set of about 20 X3p chips which also contain more than 60,000 probe sets. And it perfectly worked on a computer with 2 GB RAM. Using just.rma it even worked on a 512 MB RAM machine. If you are however really interested in using just a subset of the probe sets, it will be possible to make your own cdf environment with altcdfenvs, then tell R that your AffyBatch object has this cdfName and then apply rma only to the probe sets in your alternative cdf environment. Please let me know if you are interested in a function that makes such an alternative cdf environment since I have written a couple of such functions for one of my collegues who was interested in using just some of the probe sets. Best, Holger > On Thu, Nov 04, 2004 at 02:08:09PM -0500, Dianjing Guo wrote: > > We constantly experienced problems with rma function with soybean chip. > > Since the possible reason being the chip is too huge, i wonder whether > > there's a way to reduce the cel file size by taking only part of the raw > > intensity info for normalization. Any one can comment /addvise on that? > > That does not seem like a very good idea. I have not seen any > postings that suggest that size is the issue; have you made them? > None of this needs to be mysterious in any way. > > You should 1) make sure you have an up to date R, and an up to date > version of the package. If you get errors, such as segmentation > faults then you can use > R -d gdb > provided you have compiled R with the -g option (and if not then you > will need to recompile it). From there you can track down the source > of the bug and it can be fixed. > > For other bugs (such as problems in R code) there are options such > as using debug etc. > > It is generally much better to figure out what is wrong, and why > than to invent rather peculiar one-off solutions. > > Robert > > > > > Many thanks, > > Dianjing > > > > _______________________________________________ > > Bioconductor mailing list > > Bioconductor@stat.math.ethz.ch > > https://stat.ethz.ch/mailman/listinfo/bioconductor > > -- > +--------------------------------------------------------------------- ------+ > | Robert Gentleman phone : (617) 632-5250 > | > | Associate Professor fax: (617) 632-2444 > | > | Department of Biostatistics office: M1B20 > | > | Harvard School of Public Health email: rgentlem@jimmy.harvard.edu > | > +--------------------------------------------------------------------- ------+ > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > -- Geschenkt: 3 Monate GMX ProMail + 3 Top-Spielfilme auf DVD ++ Jetzt kostenlos testen http://www.gmx.net/de/go/mail ++

Hi Laurent, Thanks so much for your great suggestion. I used the following expresso command and it worked! But i have one more question regarding the normalize.method choice: should i use "quantiles" or "quantiles.robust"? What's the difference between them? Following is my command: > eset<-expresso(data, normalize.method="quantiles", bgcorrect.method="pmonly", summary.method="medianpolish") Also, I'd like to thank Robert Gentleman, Vince Carey, and Holger Schwender for their kind help regarding this issue Laurent Gautier wrote: > Robert Gentleman wrote: > >> On Thu, Nov 04, 2004 at 02:08:09PM -0500, Dianjing Guo wrote: >> >>> We constantly experienced problems with rma function with soybean >>> chip. Since the possible reason being the chip is too huge, i wonder >>> whether there's a way to reduce the cel file size by taking only >>> part of the raw intensity info for normalization. Any one can >>> comment /addvise on that? >> >> >> >> That does not seem like a very good idea. > > > That's right. This is _really_ not a good idea, unless you really know > the guts of the 'affy' package (there is a rewrite of some of the > package on its way that will make that this kind of tricks more easy, > but we are not there yet). > >> I have not seen any >> postings that suggest that size is the issue; have you made them? > > > According to Dian-Jing's previous post, the segfault occurs when the > summary values are computed. I do not think either that the size is an > issue: the tough part for memory usage is usually the handling of > probe level data. > Robert is probably right: there is memory leak or an array > out-of-bound problem. At first sight I think that the problem comes > from somewhere in 'do_RMA' (file rma2.c), but it is hard to tell > (comment on line 410 is a hint of an out-of-bound thing, but it refers > to a value '200' that I cannot see anywhere). > > If Dian-Jing is not into all, the use of 'expresso' (see my previous > mail) is segfault safe (currently at the cost of a bit of memory > usage, but this will improve very soon). > >> None of this needs to be mysterious in any way. >> You should 1) make sure you have an up to date R, and an up to date >> version of the package. If you get errors, such as segmentation >> faults then you can use R -d gdb provided you have compiled R >> with the -g option (and if not then you >> will need to recompile it). From there you can track down the source >> of the bug and it can be fixed. >> >> For other bugs (such as problems in R code) there are options such >> as using debug etc. >> It is generally much better to figure out what is wrong, and why >> than to invent rather peculiar one-off solutions. >> >> Robert >> >> >>> Many thanks, >>> Dianjing >>> >>> _______________________________________________ >>> Bioconductor mailing list >>> Bioconductor@stat.math.ethz.ch >>> https://stat.ethz.ch/mailman/listinfo/bioconductor >> >> >>