>Date: Fri, 04 Jun 2004 11:59:39 -0400 >To: "Park, Richard" <richard.park@joslin.harvard.edu> >From: Naomi Altman <naomi@stat.psu.edu> >Subject: RE: [BioC] large amount of slides > >Feedback on this proposal would be appreciated: > >1) Start with quantile normalization of probes for a random subset of the >data. > >2) Use these slides to define the probe distribution F. > >3) Use F to do quantile normalization of probes on all of the arrays. > >4) Use a within array robust method (e.g. Tukey biweight) to combine the >probes into genes. > >There was previous discussion of the use of within slide probe combination >versus between-slide (median polish) under the topic "median polish vs >mas". The upshot was that within slide normalization cannot detect >probewise outliers adequately. This probably means that step 4 could be >improved upon in some clever way that uses several arrays but is faster >than median polish on all arrays. > >--Naomi > > >At 11:40 AM 6/4/2004 -0400, you wrote: >>Hi Vada, >>I would caution you on doing rma on that many datasets. I have noticed a >>trend in rma, that things get even more underestimated as the number and >>variance of the data increases. I have been doing an analysis on immune >>cell types for about 100 cel files. My computer (windows 2000, 2gb of >>ram, 2.6 pentium 4) gives out around 70 datasets, I am pretty sure that >>my problem is that windows 2000 has a maximum allocation of 1gb. >> >>But if most of your data is pretty related (i.e. same tissues, just a ko >>vs wt) you should be fine w/ rma. I would caution against using rma on >>data that is very different. >> >>hth, >>richard >> >>-----Original Message----- >>From: Vada Wilcox [mailto:v_wilcox@hotmail.com] >>Sent: Friday, June 04, 2004 11:06 AM >>To: bioconductor@stat.math.ethz.ch >>Subject: [BioC] large amount of slides >> >> >>Dear all, >> >>I have been using RMA succesfully for a while now, but in the past I have >>only used it on a small amount of slides. I would like to do my study on a >>larger scale now, with data (series of experiments) from other researchers >>as well. My questions is the following: if I want to study, let's say 200 >>slides, do I have to read them all into R at once (so together I mean, with >>read.affy() in package affy), or is it OK to read them series by series (so >>all wild types and controls of one researcher at a time)? >> >>If it is really necessary to read all of them in at one time how much RAM >>would I need (for let's say 200 CELfiles) and how can I raise the RAM? I now >>it's possible to raise it by using 'max vsize = ...' but I haven't been able >>to do it succesfully for 200 experiments though. Can somebody help me on >>this? >> >>Many thanks in advance, >> >>Vada >> >>_________________________________________________________________ >> >>http://toolbar.msn.click-url.com/go/onm00200415ave/direct/01/ >> >>_______________________________________________ >>Bioconductor mailing list >>Bioconductor@stat.math.ethz.ch >>https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor >> >>_______________________________________________ >>Bioconductor mailing list >>Bioconductor@stat.math.ethz.ch >>https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor > >Naomi S. Altman 814-865-3791 (voice) >Associate Professor >Bioinformatics Consulting Center >Dept. of Statistics 814-863-7114 (fax) >Penn State University 814-865-1348 (Statistics) >University Park, PA 16802-2111 Naomi S. Altman 814-865-3791 (voice) Associate Professor Bioinformatics Consulting Center Dept. of Statistics 814-863-7114 (fax) Penn State University 814-865-1348 (Statistics) University Park, PA 16802-2111