Dear all, I am trying to run the PLIER algorithm on 51 Affymetrix of HGU-133 plus 2.0 type using justPlier() function in plier package. After about 20 minutes, the process crashes with a core dump of 2 GB in size. I am running R-2.0.1 and plier-0.03 on a Linux cluster with emulated 64-bit processing and 10 GB of RAM. I do very much appreciate Crispin Miller and other's work and the fact that plier package is still in development, but this is a bit frustrating because I am able to do RMA and GCRMA on the same dataset on my laptop with only 512 MB of RAM. 1) Can anyone guess why this crashes ? 2) Has anyone been successful with dataset of such size in R ? 3) Can anyone suggest alternatives ? Do you think a GCOS system on a desktop with 4 GB RAM running Windows OS be capable of doing this ? Thank you. Regards, -- Adaikalavan Ramasamy ramasamy@cancer.org.uk Centre for Statistics in Medicine http://www.ihs.ox.ac.uk/csm/ Cancer Research UK Tel : 07906 308 566 Old Road Campus, Headington, Oxford Fax : 01865 226 962

I would just like to mention that I solved this problem. I believe it is most likely to be related to memory issues but I have not done a systematic investigation to justify this claim. I would also like to thank Crispin Miller whose suggestion (via private communication) helped me somewhat narrow down the problem. Just one more observation that I would like to make. HGU-133 plus 2 chips have more than double the number of probesets compared to its predecessor HGU-133A. If the preprocessing is of order(n^2), where n is number of probesets, then we will need 4 times the computing power to work on the same number of arrays. But if it is of order(n^3), we need 8 times more computing power. My recent experiences suggest the later or more extreme. Does anyone have a more informed opinion ? Thank you. Regards, Adai On Wed, 2005-03-23 at 17:01 +0000, Adaikalavan Ramasamy wrote: > Dear all, > > I am trying to run the PLIER algorithm on 51 Affymetrix of HGU-133 plus > 2.0 type using justPlier() function in plier package. After about 20 > minutes, the process crashes with a core dump of 2 GB in size. I am > running R-2.0.1 and plier-0.03 on a Linux cluster with emulated 64-bit > processing and 10 GB of RAM. > > I do very much appreciate Crispin Miller and other's work and the fact > that plier package is still in development, but this is a bit > frustrating because I am able to do RMA and GCRMA on the same dataset on > my laptop with only 512 MB of RAM. > > 1) Can anyone guess why this crashes ? > 2) Has anyone been successful with dataset of such size in R ? > 3) Can anyone suggest alternatives ? Do you think a GCOS system on a > desktop with 4 GB RAM running Windows OS be capable of doing this ? > > Thank you. > > Regards, >