This is in response to the various answers I received to my questions below. First, thank you all for your comments about duplicating work with the command I used. Secondly, I am running Windows XP with R 1.7.1. I should point out that according to the May 15th posting, Laurent Gautier's method does not run on Windows. I would also like to point out that I had already visited the FAQs and the past posing for this list before my current posting and tried ?memroy.size. I have sucessfully run this analysis using rma, so I would assume justRMA should work too. However, as my questions below will show, I am concerned about how to determine the amount of memory I will need in the future. I apparently did not make my question specific enough. So... I know how to check the memory and have increased the memory already. My two questions are: 1. Concerning the error below, 83200kb is less than the allocated 1024 Mb. Does this mean that it requires 83200Kb for each chip (ie I need 83200kb for each of 38 chips or a total of about 3000 Mb) or does this mean that there are other memory usages that are causing this memory issue? ie, is this due to the data processing for the analysis I have tried to perform? 2. Sort of the reverse of that question, is there an estimate for the amount of memory needed based on the number of chips being analyzed? I could go to another computer, buy more memory, or something else, but what do I need in order to be able to do the analysis that I want here or I might chose to do in the future? Is there some way to approximate these requirements? This is assuming I wanted to use rma (or an estimate for justRMA would be good too). If I decide later that I want to use one of the other analyses in Affy, is there some way, other than trial and error, to guess/approximate the amount of memory I will need? Thank you all, Michael Barnes On Wed, Jun 25, 2003 at 08:56:12AM -0400, Michael Barnes wrote: > What are the memory requirments when running Bioconductor's Affy > package? I have 38 U95a chips and when I try to process the data I > get: > > expresso(Data,widget=TRUE) > background correction: rma > normalization: quantiles > PM/MM correction : pmonly > expression values: medianpolish > background correcting...Error: cannot allocate vector of size 83200 Kb > In addition: Warning message: > Reached total allocation of 1024Mb: see help(memory.size) > > Is this saying that the vector size for each chip is 83Mb? Otherwise, > what is going on? > > In addition, is this a reasonable analysis or is this duplicating the > functions used in rma as may be suggested in "Textual description of > Affy", pg 14? I don't quite understand from the vignette. > > Michael Barnes > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://www.stat.math.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://www.stat.math.ethz.ch/mailman/listinfo/bioconductor

James W. MacDonald UMCCC Microarray Core Facility 1500 E. Medical Center Drive 7410 CCGC Ann Arbor MI 48109 734-647-5623 >>> "Michael Barnes" <michael.barnes@cchmc.org> 06/25/03 02:37PM >>> >This is in response to the various answers I received to my questions >below. First, thank you all for your comments about duplicating work >with the command I used. Secondly, I am running Windows XP with R >1.7.1. >I should point out that according to the May 15th posting, Laurent >Gautier's method does not run on Windows. You are incorrect; Laurent contributed source code so it must be compiled before use. This is not a trivial task on Windoze, so Laurent was simply letting people know that if they didn't have the R tools, they would be out of luck. However, with the correct tools it is doable. >1. Concerning the error below, 83200kb is less than the allocated 1024 >Mb. Does this mean that it requires 83200Kb for each chip (ie I need >83200kb for each of 38 chips or a total of about 3000 Mb) or does this >mean that there are other memory usages that are causing this memory >issue? ie, is this due to the data processing for the analysis I have >tried to perform? It does not mean that each chip requires 83 Mb. All it means is that R was trying to allocate 83 Mb and ran out of memory. The memory requirement is not linearly related to chip number, so you cannot simply divide to get memory required per chip. >2. Sort of the reverse of that question, is there an estimate for the >amount of memory needed based on the number of chips being analyzed? I >could go to another computer, buy more memory, or something else, but >what do I need in order to be able to do the analysis that I want here >or I might chose to do in the future? Is there some way to approximate >these requirements? This is assuming I wanted to use rma (or an >estimate for justRMA would be good too). If I decide later that I want >to use one of the other analyses in Affy, is there some way, other than >trial and error, to guess/approximate the amount of memory I will need? Rafael requested a table of memory requirements, which I am working on. However, ReadAffy gets really slow when the number of chips gets big, so I only have info for justRMA. R ver is 1.7.1, affy ver is 1.3.0 # Chips Max mem required Final memory used (to hold exprSet in memory) 59 U95A 310 Mb 212 Mb 76 U95A 529 Mb 338 Mb 96 U95A 582 Mb 370 Mb 46 U133A 345 Mb 301 Mb This was all done on a 1.13 GHz PIII laptop running WinXP with 512 Mb RAM (768 Mb page file). Needless to say, I was hitting the page file really hard. Free memory dipped to under 1 Kb with the 96 chip run ;-D I am testing ReadAffy/rma using a 2.4 GHz PIV Desktop running WinXP with 2 Gb RAM. So far all I have is info for the 59 U95A chips: #Chips Max mem required Final memory used (exprSet and AffyBatch) 59 U95A 1.8 Gb 807 Mb Jim

On Wed, Jun 25, 2003 at 02:37:44PM -0400, Michael Barnes wrote: > This is in response to the various answers I received to my questions > below. First, thank you all for your comments about duplicating work > with the command I used. Secondly, I am running Windows XP with R > 1.7.1. > > I should point out that according to the May 15th posting, Laurent > Gautier's method does not run on Windows. It does run on Windows, but you do have to build the package yourself. It was done this way because the add-on is very hackish (altought effective). This means that one is expected to have a fair knowledge of what is going on (otherwise it easy for one to shoot oneself in the foot). To minimize such incident, I only released the source for the package assuming that one able to build the package should avoid pitfalls. This was a temporary solution, better things are coming. Currently, the creation of the AffyBatch object is a very memory hungry step (way beyond what is required to have the object in memory). What you can do is to ask someone with a lot of memory to create the AffyBatch for you and save it as a R object. You just have to fetch this R data file and load it. Hopin' it helps, L.