I'm currently working on an experiment involving chips MG_U74Av2, MG_U74Bv2 and MG_U74Cv2. The RMA normalization treats every types of chips separately (only one .CDF file is required). Thus, I need to perform RMA three times and the result is certainly different from the one I would obtained performing RMA on all data simultaneously (especially for quantile normalization). Has someone already meet this problem? Are the results between the two methodolgies very different? More generally, do we observe differences between chips A, B and C (expression level, variability,...)? Thank you very much for your response! Regards Nicolas

Technically you are probably correct; the results for a particular gene may be different if you run rma on the three chip types separately or if you made a big AffyBatch with all three and ran rma once. However, this doesn't matter. The expression value for a given gene on a given chip only makes sense in comparison to the expression value of the same gene on a different chip (of the same type). In other words, an expression value of say, 14.5323 doesn't mean anything without another expression value to compare it to. Since the A, B, and C chips don't contain the same genes you will not be comparing the results from chip A to chip B. Therefore, it doesn't matter if you run rma thrice or only once. Jim James W. MacDonald UMCCC Microarray Core Facility 1500 E. Medical Center Drive 7410 CCGC Ann Arbor MI 48109 734-647-5623 >>> "Antille,Nicolas,LAUSANNE,NRC-BAS" <nicolas.antille@rdls.nestle.com> 06/16/03 08:39AM >>> I'm currently working on an experiment involving chips MG_U74Av2, MG_U74Bv2 and MG_U74Cv2. The RMA normalization treats every types of chips separately (only one .CDF file is required). Thus, I need to perform RMA three times and the result is certainly different from the one I would obtained performing RMA on all data simultaneously (especially for quantile normalization). Has someone already meet this problem? Are the results between the two methodolgies very different? More generally, do we observe differences between chips A, B and C (expression level, variability,...)? Thank you very much for your response! Regards Nicolas _______________________________________________ Bioconductor mailing list Bioconductor@stat.math.ethz.ch https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor

Hi Nicolas, Yes, I agree with Jim, since the 3 chips are testing different sets of genes, it is not necessarily that important to normalize everything together, since expression values in microarrray's are relative. But I can see the need of normalizing everything together if you are trying to link pathways and gene information from all 3 types of mouse chips together? Or trying to perform some sort of clustering or linkage analysis with all of the information? The simplest thing is probably performing rma on each chip separately, b/c of how the functions are dealt with in bioconductor and conduct analysis on each of set of chips. But if you really wanted to see everything together, you could probably get the expression values out of rma (you can play w/ the settings of rma) and then after you have the expression values for each probe you can then combine the sets of a,b,and c and then perform a normalization technique to make the different sets comparable. HTH, Richard Park -----Original Message----- From: Antille,Nicolas,LAUSANNE,NRC-BAS [mailto:nicolas.antille@rdls.nestle.com] Sent: Monday, June 16, 2003 8:39 AM To: 'bioconductor@stat.math.ethz.ch' Subject: [BioC] chips A,B and C I'm currently working on an experiment involving chips MG_U74Av2, MG_U74Bv2 and MG_U74Cv2. The RMA normalization treats every types of chips separately (only one .CDF file is required). Thus, I need to perform RMA three times and the result is certainly different from the one I would obtained performing RMA on all data simultaneously (especially for quantile normalization). Has someone already meet this problem? Are the results between the two methodolgies very different? More generally, do we observe differences between chips A, B and C (expression level, variability,...)? Thank you very much for your response! Regards Nicolas _______________________________________________ Bioconductor mailing list Bioconductor@stat.math.ethz.ch https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor