Dear Bioconductors users, I make a post doc in the oceanic research station of Roscoff (France). I use the limma bioconductor for analysing microarray results. There is on my microarrays three replicates for each spot (thus I use correlation or average to deal with these values). My problem is that we have several probes for each gene on the microarray: we designed 1 to 5 different probes for each gene, and I do not know how to use them: Should I average them ? It's difficult because the number of probes vary depending of each gene (1-5) and I cannot precise the number of "dups". I can also use correlation, but I have the same problem of variable number of probes. Has somebody already dealt with this type of problem of variable multi-probes by gene ? thank you in advance, Christophe Boutte

We found that on our arrays, if we avoided e.g. domains common to many genes in a family, probes from the same gene had the same expression pattern on the logarithmic scale, although different average detection. (Try graphing the probes by condition and see if the profiles are parallel.) Given that satisfactory result, for each gene we picked the probe that was about 2nd highest on most conditions. This was arbitrary - we felt that probes that had higher level of detections were likely to avoid being below detection in tissues where they actually expressed, and I worried that the highest probe might include some extreme values that were just errors. --Naomi At 04:19 AM 5/21/2007, Christophe Boutte wrote: >Dear Bioconductors users, > >I make a post doc in the oceanic research station of Roscoff (France). > >I use the limma bioconductor for analysing microarray results. >There is on my microarrays three replicates for each spot (thus I use >correlation or average to deal with these values). >My problem is that we have several probes for each gene on the >microarray: we designed 1 to 5 different probes for each gene, and I do >not know how to use them: >Should I average them ? It's difficult because the number of probes vary >depending of each gene (1-5) and I cannot precise the number of "dups". >I can also use correlation, but I have the same problem of variable >number of probes. >Has somebody already dealt with this type of problem of variable >multi-probes by gene ? > >thank you in advance, > >Christophe Boutte > >_______________________________________________ >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 Naomi S. Altman 814-865-3791 (voice) Associate Professor Dept. of Statistics 814-863-7114 (fax) Penn State University 814-865-1348 (Statistics) University Park, PA 16802-2111

Hi Christophe, This is one of the classic questions on the list and I suggest that you go through the archives (http://dir.gmane.org/gmane.science.biology.informatics.conductor) for more answers. It depends on why you would have differences if the probes don't follow the same profiles, as in Naomi's case, and which questions you are trying to answer. For differential expression, you can often deal with multiple probes per gene independently. If you are looking for gene set enrichment (such as GO analyses), then the statistics expect a measurement per gene. As I said, the archives contain a lot of information on that particular question. Francois On Mon, 2007-05-21 at 10:19 +0200, Christophe Boutte wrote: > Dear Bioconductors users, > > I make a post doc in the oceanic research station of Roscoff (France). > > I use the limma bioconductor for analysing microarray results. > There is on my microarrays three replicates for each spot (thus I use > correlation or average to deal with these values). > My problem is that we have several probes for each gene on the > microarray: we designed 1 to 5 different probes for each gene, and I do > not know how to use them: > Should I average them ? It's difficult because the number of probes vary > depending of each gene (1-5) and I cannot precise the number of "dups". > I can also use correlation, but I have the same problem of variable > number of probes. > Has somebody already dealt with this type of problem of variable > multi-probes by gene ? > > thank you in advance, > > Christophe Boutte > > _______________________________________________ > 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