One other question--are these replicate spots (i.e., the same DNA) or different oligos/clones for the same gene? Sean On May 10, 2005, at 5:59 AM, Ankit Pal wrote: > Dear Mick, > Thanks a lot for the reply. > I am interested in the spots individually but for > further analysis of the spots I need a single > representative value for each gene. > I have looked up the manual, I did not find a way to > combine replicate spots into a single value. > Could you tell me what is the method or which section > of the manual is it present in. > t will be of great help to me. > Thank you, > -Ankit > > > --- "michael watson (IAH-C)" > <michael.watson@bbsrc.ac.uk> wrote: >> There are ways of combining replicate spots in >> limma, and it is all in the user guide :-) >> >> However, many people, myself included, prefer things >> reported on a spot-by-spot basis. If all replicate >> spots for a particular gene are reported as >> significant, I take that as further proof that i) >> the gene is differentially expressed, ii) my arrays >> are of good quality, iii) my experimental procedure >> was of good quality. Think about the case where >> only one out of two spots is reported - is that >> because one of the spots was of poor quality? Or >> because the values for each spot differ by a lot? >> You would lose this valuable information if you just >> took the average between replicates. >> >> If you *really* want an average value for each spot, >> simply take the average M value from the output of >> toTapble. >> >> Mick >> >> >> -----Original Message----- >> From: bioconductor-bounces@stat.math.ethz.ch on >> behalf of Ankit Pal >> Sent: Tue 10/05/2005 6:15 AM >> To: bioconductor@stat.math.ethz.ch >> Cc: >> Subject: [BioC] Limma final gene expression report >> >> Dear All, >> While looking at the Limma user guide, I came across >> the following example >> >>> targets <- readTargets("SwirlSample.txt") >>> RG <- read.maimages(targets$FileName, >> source="spot") >> >>> RG$genes <- readGAL() >>> RG$printer <- getLayout(RG$genes) >>> MA <- normalizeWithinArrays(RG) >>> MA <- normalizeBetweenArrays(MA) >>> fit <- lmFit(MA, design=c(-1,1,-1,1)) >>> fit <- eBayes(fit) >>> options(digits=3) >>> topTable(fit, n=30, adjust="fdr") >> ID Name M A t P.Value B >> control BMP2 -2.21 12.1 -21.1 0.000357 7.96 >> control BMP2 -2.30 13.1 -20.3 0.000357 7.78 >> control Dlx3 -2.18 13.3 -20.0 0.000357 7.71 >> control Dlx3 -2.18 13.5 -19.6 0.000357 7.62 >> fb94h06 20-L12 1.27 12.0 14.1 0.002067 5.78 >> fb40h07 7-D14 1.35 13.8 13.5 0.002067 5.54 >> >> I have omitted a few rows and columns. >> Here we see that after all the data transformations, >> we get an output where the ranking for the probes in >> an array is done on the basis of the B value. >> Notice that there are reapeating names for genes, >> therefore for a set of replicates, within and across >> arrays, each spot is reported separately as an >> individual entity. >> In the case of BMP2 from the above example, which >> result do I consider? >> Is there a way in which I can get a single result >> for >> a set of replicates. >> I am new to this package, so please do let me know >> if >> there is a problem in my understanding the concept. >> Thank you, >> -Ankit >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor@stat.math.ethz.ch >> https://stat.ethz.ch/mailman/listinfo/bioconductor >> >> >> >> > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor

Dear Ankit, In addition to Mick's arguments, you also need to be careful when averaging over spots if you use the empirical bayes methods in limma, and you have different number of replicates for different genes. If, for example, gene A has 1 spot but gene B has four replicated spots, then the variance of the mean of the replicates is very different for gene B and for gene A, and would violate some of the assumptions behind the Empirical Bayes procedure in limma. I think this is mentioned in the user guide (and/or Gordon Smyth's paper), and has also been mentioned in this list. If you really want an average, you can do as Mick suggests: > > If you *really* want an average value for each spot, > > simply take the average M value from the output of > > toTapble. you can use something like tapply(the.top.table.M.values, the.top.table.gene.identifiers, mean) Best, R. On Tuesday 10 May 2005 11:59, Ankit Pal wrote: > Dear Mick, > Thanks a lot for the reply. > I am interested in the spots individually but for > further analysis of the spots I need a single > representative value for each gene. > I have looked up the manual, I did not find a way to > combine replicate spots into a single value. > Could you tell me what is the method or which section > of the manual is it present in. > t will be of great help to me. > Thank you, > -Ankit > > > --- "michael watson (IAH-C)" > > <michael.watson@bbsrc.ac.uk> wrote: > > There are ways of combining replicate spots in > > limma, and it is all in the user guide :-) > > > > However, many people, myself included, prefer things > > reported on a spot-by-spot basis. If all replicate > > spots for a particular gene are reported as > > significant, I take that as further proof that i) > > the gene is differentially expressed, ii) my arrays > > are of good quality, iii) my experimental procedure > > was of good quality. Think about the case where > > only one out of two spots is reported - is that > > because one of the spots was of poor quality? Or > > because the values for each spot differ by a lot? > > You would lose this valuable information if you just > > took the average between replicates. > > > > If you *really* want an average value for each spot, > > simply take the average M value from the output of > > toTapble. > > > > Mick > > > > > > -----Original Message----- > > From: bioconductor-bounces@stat.math.ethz.ch on > > behalf of Ankit Pal > > Sent: Tue 10/05/2005 6:15 AM > > To: bioconductor@stat.math.ethz.ch > > Cc: > > Subject: [BioC] Limma final gene expression report > > > > Dear All, > > While looking at the Limma user guide, I came across > > the following example > > > > > targets <- readTargets("SwirlSample.txt") > > > RG <- read.maimages(targets$FileName, > > > > source="spot") > > > > > RG$genes <- readGAL() > > > RG$printer <- getLayout(RG$genes) > > > MA <- normalizeWithinArrays(RG) > > > MA <- normalizeBetweenArrays(MA) > > > fit <- lmFit(MA, design=c(-1,1,-1,1)) > > > fit <- eBayes(fit) > > > options(digits=3) > > > topTable(fit, n=30, adjust="fdr") > > > > ID Name M A t P.Value B > > control BMP2 -2.21 12.1 -21.1 0.000357 7.96 > > control BMP2 -2.30 13.1 -20.3 0.000357 7.78 > > control Dlx3 -2.18 13.3 -20.0 0.000357 7.71 > > control Dlx3 -2.18 13.5 -19.6 0.000357 7.62 > > fb94h06 20-L12 1.27 12.0 14.1 0.002067 5.78 > > fb40h07 7-D14 1.35 13.8 13.5 0.002067 5.54 > > > > I have omitted a few rows and columns. > > Here we see that after all the data transformations, > > we get an output where the ranking for the probes in > > an array is done on the basis of the B value. > > Notice that there are reapeating names for genes, > > therefore for a set of replicates, within and across > > arrays, each spot is reported separately as an > > individual entity. > > In the case of BMP2 from the above example, which > > result do I consider? > > Is there a way in which I can get a single result > > for > > a set of replicates. > > I am new to this package, so please do let me know > > if > > there is a problem in my understanding the concept. > > Thank you, > > -Ankit > > > > _______________________________________________ > > Bioconductor mailing list > > Bioconductor@stat.math.ethz.ch > > https://stat.ethz.ch/mailman/listinfo/bioconductor > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor -- Ram?n D?az-Uriarte Bioinformatics Unit Centro Nacional de Investigaciones Oncol?gicas (CNIO) (Spanish National Cancer Center) Melchor Fern?ndez Almagro, 3 28029 Madrid (Spain) Fax: +-34-91-224-6972 Phone: +-34-91-224-6900 http://ligarto.org/rdiaz PGP KeyID: 0xE89B3462 (http://ligarto.org/rdiaz/0xE89B3462.asc) **NOTA DE CONFIDENCIALIDAD** Este correo electr?nico, y en su caso los ficheros adjuntos, pueden contener informaci?n protegida para el uso exclusivo de su destinatario. Se proh?be la distribuci?n, reproducci?n o cualquier otro tipo de transmisi?n por parte de otra persona que no sea el destinatario. Si usted recibe por error este correo, se ruega comunicarlo al remitente y borrar el mensaje recibido. **CONFIDENTIALITY NOTICE** This email communication and any attachments may contain confidential and privileged information for the sole use of the designated recipient named above. Distribution, reproduction or any other use of this transmission by any party other than the intended recipient is prohibited. If you are not the intended recipient please contact the sender and delete all copies.

On May 10, 2005, at 7:45 AM, Ankit Pal wrote: > Dear Sean, > I agree averaging is not a good idea. > So how do I get a single value for a set of replicate > probes? > In case of different values for the same gene which > result do I consider to be representative of the > whole? That is the problem, isn't it. If you have duplicate spots (same sequence), you will need to look at the quality, etc., to see which you believe. If you have oligos that map to the same gene but behave differently, you will need to look at spot quality as well as other issues like the cross-hybridization potential (which often requires blasting), location in the gene (3' bias?), and splice variants that may be tissue specific. As I said, all of these require a bit of human intervention. In practice, though, you have to validate array results biologically--that is the real answer to your question. The "representative" spot is the one that validates; sometimes that will be the one that suggests differential expression, and sometimes not. Sean > It would definately help at the clustering level. > -Ankit