To add to this; these data are almost surely MAS5 processed data, as I don't know of any other algorithm that gives the detection p-value. In addition, the range of 0 - 9000 indicates that these data are not logged (which is the next step for you). People normally use log base 2 so that a difference of 1 or -1 indicates two-fold up or down regulation. MAS5 data are normalized after the fact, so you should log transform and then look at plots of the densities to see if they look as if they have been normalized or not. The default is to do a scale normalization, so you are just looking for the densities to be in same general vicinity rather than overlaying each other. If you could get the original celfiles, you would be much better off. Best, Jim michael watson (IAH-C) wrote: > This is definitely processed data, and without access to the original data or a description of the analysis methodology, your options are limited. > > Personally, I'd do a test for normality on the "Signal" values, and if they turn out to be normal, I'd run a simple t-test (control vs treatment) on each gene and correct the p-values for multiple testing. > > Simple stuff, but it should work. > ________________________________________ > From: bioconductor-bounces at stat.math.ethz.ch [bioconductor- bounces at stat.math.ethz.ch] On Behalf Of avehna [avhena at gmail.com] > Sent: 16 February 2010 19:47 > To: bioconductor at stat.math.ethz.ch > Subject: [BioC] DNA micro-array normalization > > Hi There, > > I've got a DNA microarray dataset that looks like this: > > * Probe Signal Detection > Detection_p-value Descriptions* > AFFX-BioB-5_at 181 P > 0.00011 "E. coli GEN=bioB DB_XREF=gb:J04423.1" > AFFX-BioB-M_at 227.3 P 0.000044 > "E. coli GEN=bioB DB_XREF=gb:J04423.1" > AFFX-BioC-5_at 499.2 P > 0.000052 "E. coli GEN=bioC DB_XREF=gb:J04423.1" > > I have control and treatment with 3 replicas for each one of them. > > But I'm not sure whether these data have been already normalized, and on the > other hand, this is not the typical affymetrix format... > > Could you help me in this regard? What is the typical signal range for rough > affymetrix data? (these data range from 0 to 9000) > > If the data have been already normalized, Can I calculate the mean (for > treatment and control) followed by the differential expression of genes > without taking into account the "Detection" column? > > (I guess I will need to build my ExpressionSet from scratch) > > Thanks a lot (I'm a newbie in bioconductor and micro-array analysis). I will > appreciate you help! > > Avhena > > [[alternative HTML version deleted]] > > _______________________________________________ > 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 > > _______________________________________________ > 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 -- James W. MacDonald, M.S. Biostatistician Douglas Lab University of Michigan Department of Human Genetics 5912 Buhl 1241 E. Catherine St. Ann Arbor MI 48109-5618 734-615-7826 ********************************************************** Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues

Avehna: I'd try with lmFit / eBayes from limma, since the moderated t-test typically provides better power for such small sample sizes. Also, I'd look at the output of "meanSdPlot" (from the vsn package) and "multidensity" or "multiecdf" (from the geneplotter package) to see whether the data need (i) transformation and (ii) between-array normalisation. For both, "vsn2" from the vsn package is one possibility. Michael: one more :) -- I guess fortune(117) and fortune(234) apply. Less opaquely, - I don't know of a test that has power to reject Normality on a sample of size 3 or 6. - Normality of the data is a sufficient condition for some (important) theoretical properties of the t-test, but it is not necessary for it to provide good enough type I error control and power in applications. Best wishes Wolfgang michael watson (IAH-C) scripsit 02/16/2010 09:34 PM: > This is definitely processed data, and without access to the original data or a description of the analysis methodology, your options are limited. > > Personally, I'd do a test for normality on the "Signal" values, and if they turn out to be normal, I'd run a simple t-test (control vs treatment) on each gene and correct the p-values for multiple testing. > > Simple stuff, but it should work. > ________________________________________ > From: bioconductor-bounces at stat.math.ethz.ch [bioconductor- bounces at stat.math.ethz.ch] On Behalf Of avehna [avhena at gmail.com] > Sent: 16 February 2010 19:47 > To: bioconductor at stat.math.ethz.ch > Subject: [BioC] DNA micro-array normalization > > Hi There, > > I've got a DNA microarray dataset that looks like this: > > * Probe Signal Detection > Detection_p-value Descriptions* > AFFX-BioB-5_at 181 P > 0.00011 "E. coli GEN=bioB DB_XREF=gb:J04423.1" > AFFX-BioB-M_at 227.3 P 0.000044 > "E. coli GEN=bioB DB_XREF=gb:J04423.1" > AFFX-BioC-5_at 499.2 P > 0.000052 "E. coli GEN=bioC DB_XREF=gb:J04423.1" > > I have control and treatment with 3 replicas for each one of them. > > But I'm not sure whether these data have been already normalized, and on the > other hand, this is not the typical affymetrix format... > > Could you help me in this regard? What is the typical signal range for rough > affymetrix data? (these data range from 0 to 9000) > > If the data have been already normalized, Can I calculate the mean (for > treatment and control) followed by the differential expression of genes > without taking into account the "Detection" column? > > (I guess I will need to build my ExpressionSet from scratch) > > Thanks a lot (I'm a newbie in bioconductor and micro-array analysis). I will > appreciate you help! > > Avhena > > [[alternative HTML version deleted]] > > _______________________________________________ > 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 > > _______________________________________________ > 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 -- Best wishes Wolfgang -- Wolfgang Huber EMBL http://www.embl.de/research/units/genome_biology/huber/contact

Sorry! It wasn't my intention... Thank you for your help, Avhena On Wed, Feb 17, 2010 at 11:42 AM, James W. MacDonald <jmacdon@med.umich.edu>wrote: > Hi Avehna, > > Please don't take conversations off list. The list is considered to be a > resource that people in your situation can use in the future to answer > questions themselves. > > avehna wrote: > > On Tue, Feb 16, 2010 at 3:50 PM, James W. MacDonald < >> jmacdon@med.umich.edu <mailto:jmacdon@med.umich.edu>> wrote: >> >> To add to this; these data are almost surely MAS5 processed data, as >> I don't know of any other algorithm that gives the detection >> p-value. In addition, the range of 0 - 9000 indicates that these >> data are not logged (which is the next step for you). People >> normally use log base 2 so that a difference of 1 or -1 indicates >> two-fold up or down regulation. >> >> >> OK. But in this case what would be the reference point? Wouldn't be the up >> or down regulation respect to the control? Before writing to the list I have >> browsed several tutorials and I'm still missing this part. Should it be >> log2(treatment/control)? (It's not clear what I have read) >> > > Yes. Or since you have already taken logs, it will be log2(treatment) - > log2(control), which you will notice is the numerator of your t-statistic. > > > > >> >> MAS5 data are normalized after the fact, so you should log transform >> and then look at plots of the densities to see if they look as if >> they have been normalized or not. The default is to do a scale >> normalization, so you are just looking for the densities to be in >> same general vicinity rather than overlaying each other. >> >> >> OK. Could you send me some helpful references about that? >> > > > http://media.affymetrix.com/support/technical/whitepapers/sadd_white paper.pdf > > Best, > > Jim > > > >> >> If you could get the original celfiles, you would be much better off. >> >> >> I will try! >> >> Best and thank you so much for your help, >> >> Avhena. >> >> Best, >> >> Jim >> >> >> >> >> michael watson (IAH-C) wrote: >> >> This is definitely processed data, and without access to the >> original data or a description of the analysis methodology, your >> options are limited. >> >> Personally, I'd do a test for normality on the "Signal" values, >> and if they turn out to be normal, I'd run a simple t-test >> (control vs treatment) on each gene and correct the p-values for >> multiple testing. >> >> Simple stuff, but it should work. >> ________________________________________ >> From: bioconductor-bounces@stat.math.ethz.ch >> <mailto:bioconductor-bounces@stat.math.ethz.ch> >> >> [bioconductor-bounces@stat.math.ethz.ch >> <mailto:bioconductor-bounces@stat.math.ethz.ch>] On Behalf Of >> avehna [avhena@gmail.com <mailto:avhena@gmail.com>] >> >> Sent: 16 February 2010 19:47 >> To: bioconductor@stat.math.ethz.ch >> <mailto:bioconductor@stat.math.ethz.ch> >> >> Subject: [BioC] DNA micro-array normalization >> >> Hi There, >> >> I've got a DNA microarray dataset that looks like this: >> >> * Probe Signal Detection >> Detection_p-value Descriptions* >> AFFX-BioB-5_at 181 P >> 0.00011 "E. coli GEN=bioB DB_XREF=gb:J04423.1" >> AFFX-BioB-M_at 227.3 P 0.000044 >> "E. coli GEN=bioB DB_XREF=gb:J04423.1" >> AFFX-BioC-5_at 499.2 P >> 0.000052 "E. coli GEN=bioC DB_XREF=gb:J04423.1" >> >> I have control and treatment with 3 replicas for each one of them. >> >> But I'm not sure whether these data have been already >> normalized, and on the >> other hand, this is not the typical affymetrix format... >> >> Could you help me in this regard? What is the typical signal >> range for rough >> affymetrix data? (these data range from 0 to 9000) >> >> If the data have been already normalized, Can I calculate the >> mean (for >> treatment and control) followed by the differential expression >> of genes >> without taking into account the "Detection" column? >> >> (I guess I will need to build my ExpressionSet from scratch) >> >> Thanks a lot (I'm a newbie in bioconductor and micro-array >> analysis). I will >> appreciate you help! >> >> Avhena >> >> [[alternative HTML version deleted]] >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor@stat.math.ethz.ch >> <mailto: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 >> <mailto:bioconductor@stat.math.ethz.ch> >> >> https://stat.ethz.ch/mailman/listinfo/bioconductor >> Search the archives: >> http://news.gmane.org/gmane.science.biology.informatics.conductor >> >> >> -- James W. MacDonald, M.S. >> Biostatistician >> Douglas Lab >> University of Michigan >> Department of Human Genetics >> 5912 Buhl >> 1241 E. Catherine St. >> Ann Arbor MI 48109-5618 >> 734-615-7826 >> ********************************************************** >> Electronic Mail is not secure, may not be read every day, and should >> not be used for urgent or sensitive issues >> >> >> > -- > James W. MacDonald, M.S. > Biostatistician > Douglas Lab > University of Michigan > Department of Human Genetics > 5912 Buhl > 1241 E. Catherine St. > Ann Arbor MI 48109-5618 > 734-615-7826 > ********************************************************** > Electronic Mail is not secure, may not be read every day, and should not be > used for urgent or sensitive issues > [[alternative HTML version deleted]]