Dear list, I have one question about merging (Combining) microarray datasets. I have 50 samples, and each sample was hybridized to 2 different platforms--- GPL3694 (5352 probes) and GPL3696 (also 5352 probes) in duplicate. There were 168 probes identical between these two platforms. I want to merge (combine) two hybridization files of each sample before performing normalization for all 50 samples. What appropriate method can be used to do this? Can the common 168 probes provide some information? Thanks! --- LiGang

2009/6/23 Alberto Goldoni <alberto.goldoni1975@gmail.com> > Try to use rankProd package contains functions for the analysis of gene > expression microarray data, in > particular the identi cation of di erentially expressed genes. RankProd > utilizes the so called rank > product non-parametric method (Breitling et al., 2004, FEBS Letters 573:83) > to identify up-regulated > or down-regulated genes under one condition against another condition, e.g. > two di erent treatments, > two di erent tissue types, etc.It is indicated for "Multiple origins". > Follow the vignette. > > Best regards. > > 2009/6/23 Hui Yu <sheen2006@gmail.com> > > In my opinion, you should normalize, at least with regard to the two >> different platforms, before attempting to merge duplicated >> measurements of identical samples. Most probabably you'll end up >> averaging the measurements of the duplicatedly-measured probes, and >> collating the measurements for different genes from different >> platforms into one single data column. These operations require you >> ensure that the measurements from the two separate platforms are >> comparable by the time of merging. >> You may want to normalize only the platform-specific bias before the >> merging; a loess-like calibration based on the 168*50*2 data-points >> may suffice for this purpose. You have 168*50 pairs of data-points, >> where each pair of data-points, one from GPL3694 and the other from >> GPL3696, should be theoretically identical. A loess regression will >> suppress the artificial difference between GPL3696 and GPL3694. >> After the merging, which resulted in 50 data-columns each >> corresponding to a different sample, you can implement whatever >> normalization that came up to your mind in ordinary cases. Given the >> fact that you have only ~5000 probes on each chip, which account for >> only 25% of the total number of human genes, however, it may be >> riskful to apply any global normalization that is based on the >> 'invariable-total-expression' assumption. Maybe you need addtional >> information to normalize between samples. >> Any other opinions? >> >> On Fri, Jun 19, 2009 at 10:06 AM, LiGang<luzifer.li@gmail.com> wrote: >> > Dear list, >> > >> > I have one question about merging (Combining) microarray datasets. >> > >> > I have 50 samples, and each sample was hybridized to 2 different >> platforms--- >> > GPL3694 (5352 probes) and GPL3696 (also 5352 probes) in duplicate. >> > There were 168 probes identical between these two platforms. >> > >> > I want to merge (combine) two hybridization files of each sample before >> > performing normalization for all 50 samples. >> > What appropriate method can be used to do this? >> > Can the common 168 probes provide some information? >> > >> > Thanks! >> > --- >> > LiGang >> > >> > _______________________________________________ >> > Bioconductor mailing list >> > Bioconductor@stat.math.ethz.ch >> > https://stat.ethz.ch/mailman/listinfo/bioconductor >> > Search the archives: >> http://news.gmane.org/gmane.science.biology.informatics.conductor >> > >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor@stat.math.ethz.ch >> https://stat.ethz.ch/mailman/listinfo/bioconductor >> Search the archives: >> http://news.gmane.org/gmane.science.biology.informatics.conductor >> > > > > -- > ----------------------------------------------------- > Dr. Alberto Goldoni > Bologna, Italy > ----------------------------------------------------- > -- ----------------------------------------------------- Dr. Alberto Goldoni Bologna, Italy ----------------------------------------------------- [[alternative HTML version deleted]]

In my opinion, you should normalize, at least with regard to the two different platforms, before attempting to merge duplicated measurements of identical samples. Most probabably you'll end up averaging the measurements of the duplicatedly-measured probes, and collating the measurements for different genes from different platforms into one single data column. These operations require you ensure that the measurements from the two separate platforms are comparable by the time of merging. You may want to normalize only the platform-specific bias before the merging; a loess-like calibration based on the 168*50*2 data-points may suffice for this purpose. You have 168*50 pairs of data-points, where each pair of data-points, one from GPL3694 and the other from GPL3696, should be theoretically identical. A loess regression will suppress the artificial difference between GPL3696 and GPL3694. After the merging, which resulted in 50 data-columns each corresponding to a different sample, you can implement whatever normalization that came up to your mind in ordinary cases. Given the fact that you have only ~5000 probes on each chip, which account for only 25% of the total number of human genes, however, it may be riskful to apply any global normalization that is based on the 'invariable-total-expression' assumption. Maybe you need addtional information to normalize between samples. Any other opinions? On Fri, Jun 19, 2009 at 10:06 AM, LiGang<luzifer.li at="" gmail.com=""> wrote: > Dear list, > > I have one question about merging (Combining) microarray datasets. > > I have 50 samples, and each sample was hybridized to 2 different platforms--- > GPL3694 (5352 probes) and GPL3696 (also 5352 probes) in duplicate. > There were 168 probes identical between these two platforms. > > I want to merge (combine) two hybridization files of each sample before > performing normalization for all 50 samples. > What appropriate method can be used to do this? > Can the common 168 probes provide some information? > > Thanks! > --- > LiGang > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor >