Hi all, I have an unusual microarray design. The experiment is meant to compare two conditions. We have 4 slides of microarrays: all four considered as biological replicates. However, one slide contains 2 replicate arrays (we named top and bottom). These two come from the same culture. The design was made this way for frugality. So in total we have 8 samples. I read the limma user guide and there was no example of such design. There are duplicate spots in each array but in random fashion so there is no way to compute the duplicateCorrelation command. I wonder how would I build the design. Any help is appreciated. Yolande [[alternative HTML version deleted]]

I cannot speak for the experience of others, but some small experiments we did with Agilent arabidopsis arrays with 2 arrays per slide did not find a slide effect. We did find a batch effect - i.e. arrays run in a batch are more similar than arrays in different batches. But arrays within slide did not appear to be more similar than arrays within batch. Regarding the duplicate spots - just sort the genes by geneId before analysis and use spacing=1. The effect of spot duplication is usually much larger than the effect of position on the array. --naomi At 01:44 PM 1/23/2009, Yolande Tra wrote: >Hi all, > >I have an unusual microarray design. The experiment is meant to >compare two conditions. We have 4 slides of microarrays: all four >considered as biological replicates. However, one slide contains 2 >replicate arrays (we named top and bottom). These two come from the >same culture. The design was made this way for frugality. So in >total we have 8 samples. I read the limma user guide and there was >no example of such design. There are duplicate spots in each array >but in random fashion so there is no way to compute the >duplicateCorrelation command. I wonder how would I build the design. >Any help is appreciated. > >Yolande > > [[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 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 Naomi, Thank you for your reply. Would you say then we should test for batch effect and slide effect using the command lm and a formula with main effect and interaction or we can just use a design matrix where each array is considered as a biological replicate. Looking forward to your insight. Yolande -----Original Message----- From: Naomi Altman [mailto:naomi@stat.psu.edu] Sent: Fri 1/23/2009 2:14 PM To: Yolande Tra; bioconductor@stat.math.ethz.ch Subject: Re: [BioC] Limma: design matrix I cannot speak for the experience of others, but some small experiments we did with Agilent arabidopsis arrays with 2 arrays per slide did not find a slide effect. We did find a batch effect - i.e. arrays run in a batch are more similar than arrays in different batches. But arrays within slide did not appear to be more similar than arrays within batch. Regarding the duplicate spots - just sort the genes by geneId before analysis and use spacing=1. The effect of spot duplication is usually much larger than the effect of position on the array. --naomi At 01:44 PM 1/23/2009, Yolande Tra wrote: >Hi all, > >I have an unusual microarray design. The experiment is meant to >compare two conditions. We have 4 slides of microarrays: all four >considered as biological replicates. However, one slide contains 2 >replicate arrays (we named top and bottom). These two come from the >same culture. The design was made this way for frugality. So in >total we have 8 samples. I read the limma user guide and there was >no example of such design. There are duplicate spots in each array >but in random fashion so there is no way to compute the >duplicateCorrelation command. I wonder how would I build the design. >Any help is appreciated. > >Yolande > > [[alternative HTML version deleted]] > >_______________________________________________ >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 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 [[alternative HTML version deleted]]

I guess the question is how stringently you want to do this. We did a pretest using arrays not in the study, designed the study to minimize batch effects, and then did not include terms for slide or batch effects. If your arrays were run in batches, then I would definitely allow for a batch effect, but maybe not a slide effect. --Naomi At 05:36 PM 1/25/2009, Yolande Tra wrote: >Hi Naomi, > >Thank you for your reply. >Would you say then we should test for batch effect and slide effect >using the command lm and a formula with main effect and interaction >or we can just use a design matrix where each array is considered as >a biological replicate. >Looking forward to your insight. > >Yolande > > > >-----Original Message----- >From: Naomi Altman [mailto:naomi at stat.psu.edu] >Sent: Fri 1/23/2009 2:14 PM >To: Yolande Tra; bioconductor at stat.math.ethz.ch >Subject: Re: [BioC] Limma: design matrix > >I cannot speak for the experience of others, but some small >experiments we did with Agilent arabidopsis arrays with 2 arrays per >slide did not find a slide effect. We did find a batch effect - i.e. >arrays run in a batch are more similar than arrays in different >batches. But arrays within slide did not appear to be more similar >than arrays within batch. > >Regarding the duplicate spots - just sort the genes by geneId before >analysis and use spacing=1. The effect of spot duplication is >usually much larger than the effect of position on the array. > >--naomi > > > >At 01:44 PM 1/23/2009, Yolande Tra wrote: > > >Hi all, > > > >I have an unusual microarray design. The experiment is meant to > >compare two conditions. We have 4 slides of microarrays: all four > >considered as biological replicates. However, one slide contains 2 > >replicate arrays (we named top and bottom). These two come from the > >same culture. The design was made this way for frugality. So in > >total we have 8 samples. I read the limma user guide and there was > >no example of such design. There are duplicate spots in each array > >but in random fashion so there is no way to compute the > >duplicateCorrelation command. I wonder how would I build the design. > >Any help is appreciated. > > > >Yolande > > > > [[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 > >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 > > > > [[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 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

Thank you for the insight. Yolande -----Original Message----- From: Naomi Altman [mailto:naomi@stat.psu.edu] Sent: Mon 1/26/2009 9:53 PM To: Yolande Tra; Naomi Altman; bioconductor@stat.math.ethz.ch Subject: Re: [BioC] Limma: design matrix I guess the question is how stringently you want to do this. We did a pretest using arrays not in the study, designed the study to minimize batch effects, and then did not include terms for slide or batch effects. If your arrays were run in batches, then I would definitely allow for a batch effect, but maybe not a slide effect. --Naomi At 05:36 PM 1/25/2009, Yolande Tra wrote: >Hi Naomi, > >Thank you for your reply. >Would you say then we should test for batch effect and slide effect >using the command lm and a formula with main effect and interaction >or we can just use a design matrix where each array is considered as >a biological replicate. >Looking forward to your insight. > >Yolande > > > >-----Original Message----- >From: Naomi Altman [mailto:naomi@stat.psu.edu] >Sent: Fri 1/23/2009 2:14 PM >To: Yolande Tra; bioconductor@stat.math.ethz.ch >Subject: Re: [BioC] Limma: design matrix > >I cannot speak for the experience of others, but some small >experiments we did with Agilent arabidopsis arrays with 2 arrays per >slide did not find a slide effect. We did find a batch effect - i.e. >arrays run in a batch are more similar than arrays in different >batches. But arrays within slide did not appear to be more similar >than arrays within batch. > >Regarding the duplicate spots - just sort the genes by geneId before >analysis and use spacing=1. The effect of spot duplication is >usually much larger than the effect of position on the array. > >--naomi > > > >At 01:44 PM 1/23/2009, Yolande Tra wrote: > > >Hi all, > > > >I have an unusual microarray design. The experiment is meant to > >compare two conditions. We have 4 slides of microarrays: all four > >considered as biological replicates. However, one slide contains 2 > >replicate arrays (we named top and bottom). These two come from the > >same culture. The design was made this way for frugality. So in > >total we have 8 samples. I read the limma user guide and there was > >no example of such design. There are duplicate spots in each array > >but in random fashion so there is no way to compute the > >duplicateCorrelation command. I wonder how would I build the design. > >Any help is appreciated. > > > >Yolande > > > > [[alternative HTML version deleted]] > > > >_______________________________________________ > >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 > >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 > > > > [[alternative HTML version deleted]] > >_______________________________________________ >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 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 [[alternative HTML version deleted]]