Dear Dr. Irizarry At the CHI meeting in Baltimore I enjoyed your interesting talk about gcrma, which currently seems to be the best algorithm for condesation/normalization of CEL files, as the affycomp results suggest. For this reason my colleagues and I were eager to test gcrma with our own datasets containing between 20 and 130 HGU133 chips. My colleague tested rma and gcrma with the following setting: HP xw8000 Dual Xeon 2.8 GHz with 2 GB RAM RedHat 8.0 with kernel 2.4.18-SMP R-1.8.1 and Bioconductor 1.3 compiled by us for above settings Using this setup, rma can process the data really fast: 21 HGU133A data: about 1 minute 40 HGU133A data: about 1.5 minutes 130 HGU133A data: about 2.5 minutes For gcrma we got the following results: 21 HGU133A data: about 45 minutes using 500 MB RAM 40 HGU133A data: about 90 minutes using 900 MB RAM 130 HGU133A data: the usual error: cannot allocate vector of size blabla Since we will soon switch to the new Affymetrix HG-U133_Plus_2 GeneChips, things will getting worse. My questions are the following: Do you intend to optimize the behavior of gcrma, e.g. by rewriting it in C? In the meantime, which setup would be sufficient for gcrma to handle 130 HGU133A data? Do you think that a 64bit processor machine would be helpful? Could the Dual G5 Mac be an option? Thank you in advance for your help. P.S. Please reply also to me since I am not subscribed to the mailing list. Best regards Christian Stratowa ============================================== Christian Stratowa, PhD Boehringer Ingelheim Austria Dept NCE Lead Discovery - Bioinformatics Dr. Boehringergasse 5-11 A-1121 Vienna, Austria Tel.: ++43-1-80105-2470 Fax: ++43-1-80105-2366 email: christian.stratowa@vie.boehringer-ingelheim.com

we have recently developed an approach to get around the slow numerical integration. by mid december it should be ready to use by others. -r On Wed, 26 Nov 2003 Christian.Stratowa@vie.boehringer-ingelheim.com wrote: > Dear Dr. Irizarry > > At the CHI meeting in Baltimore I enjoyed your interesting talk about gcrma, > > which currently seems to be the best algorithm for condesation/normalization > > of CEL files, as the affycomp results suggest. For this reason my colleagues > > and I were eager to test gcrma with our own datasets containing between 20 > and 130 HGU133 chips. > > My colleague tested rma and gcrma with the following setting: > HP xw8000 Dual Xeon 2.8 GHz with 2 GB RAM > RedHat 8.0 with kernel 2.4.18-SMP > R-1.8.1 and Bioconductor 1.3 compiled by us for above settings > > Using this setup, rma can process the data really fast: > 21 HGU133A data: about 1 minute > 40 HGU133A data: about 1.5 minutes > 130 HGU133A data: about 2.5 minutes > > For gcrma we got the following results: > 21 HGU133A data: about 45 minutes using 500 MB RAM > 40 HGU133A data: about 90 minutes using 900 MB RAM > 130 HGU133A data: the usual error: cannot allocate vector of size blabla > > Since we will soon switch to the new Affymetrix HG-U133_Plus_2 > GeneChips, things will getting worse. > > My questions are the following: > Do you intend to optimize the behavior of gcrma, e.g. by rewriting it in C? > In the meantime, which setup would be sufficient for gcrma to handle > 130 HGU133A data? Do you think that a 64bit processor machine would > be helpful? Could the Dual G5 Mac be an option? > > Thank you in advance for your help. > P.S. Please reply also to me since I am not subscribed to the mailing list. > > Best regards > Christian Stratowa > > ============================================== > Christian Stratowa, PhD > Boehringer Ingelheim Austria > Dept NCE Lead Discovery - Bioinformatics > Dr. Boehringergasse 5-11 > A-1121 Vienna, Austria > Tel.: ++43-1-80105-2470 > Fax: ++43-1-80105-2366 > email: christian.stratowa@vie.boehringer-ingelheim.com > > >