how to handle technical replicates and biological replicates using limma
1
0
Entering edit mode
Ren Na ▴ 250
@ren-na-870
Last seen 9.6 years ago
Dr. Gordon Smyth, Thank you for answering my question. I still have some questions about how to fit mouse effect for my data, My purpose is to find significantly expressed genes between mutant and wild type mice. I tried to figure it out by reading "Limma user's guide" and came up with the following solution, is it right? or doesn't make any sense? My targets file is: SlideNumber FileName Cy3 Cy5 Target1 Target2 1 1391.spot wt1 mu1 wt1 mu1 2 1392.spot mu1 wt1 mu1 wt1 3 1340.spot wt2 mu1 wt2 mu1 4 1341.spot mu1 wt2 mu1 wt2 5 1395.spot wt3 mu1 wt3 mu1 6 1396.spot mu1 wt3 mu1 wt3 7 1393.spot wt1 mu2 wt1 mu2 8 1394.spot mu2 wt1 mu2 wt1 9 1371.spot wt2 mu2 wt2 mu2 10 1372.spot mu2 wt2 mu2 wt2 11 1338.spot wt3 mu2 wt3 mu2 12 1339.spot mu2 wt3 mu2 wt3 13 1387.spot wt1 mu3 wt1 mu3 14 1388.spot mu3 wt1 mu3 wt1 15 1399.spot wt2 mu3 wt2 mu3 16 1390.spot mu3 wt2 mu3 wt2 17 1397.spot wt3 mu3 wt3 mu3 18 1398.spot mu3 wt3 mu3 wt3 mu1, mu2, and mu3 are different mice which are biological replicates, and wt1, wt2 and wt3 are different mice which are biological replicates. > targets <- readTargets() # I try to include mouse effect in the following way: > t1 <- model.matrix(~0+factor(targets$Target1)) > t2 <- model.matrix(~0+factor(targets$Target2)) > t3 <- t1+t2 > design <- cbind(Dye=1, MuvsWt=c(1,-1,1,-1,1,-1,1,-1,1,-1,1,-1,1,-1,1,-1,1,-1), t3 ) > pair <- c(1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9) > corfit <- duplicateCorrelation(MA,design,block=pair) > fit <- lmFit(MA,design,block=pair,correlation=corfit$consensus) > fit <- eBayes(fit) Thanks again for your time. Ren
• 971 views
ADD COMMENT
0
Entering edit mode
@gordon-smyth
Last seen 1 hour ago
WEHI, Melbourne, Australia
Dear Ren, No I would not use the blocking feature of limma in this case. Your "blocks" are not independent because the same mice appear in more than one dye-swap pair. This experiment could be analysed using log-ratios or log-expression values. I would personally do a log-ratio analysis fitting effects for each mouse: design <- modelMatrix(targets, ref="wt1") fit <- lmFit(MA, design) cont.matrix <- makeContrasts(muvswt=(mu1+mu2+mu3-wt2-wt3)/3, levels=design) fit2 <- contrasts.fit(fit, cont.matrix) fit2 <- eBayes(fit2) The linear model 'fit2' estimates the average difference between the mutant and wt mice. If you want to add a dye-effect, you would insert design <- cbind(Dye=1,design) after the first line above. For this sort of analysis it would be helpful to consult a local statistician. Gordon At 08:37 AM 11/08/2004, Ren Na wrote: >Dr. Gordon Smyth, > >Thank you for answering my question. >I still have some questions about how to fit mouse effect for my data, My >purpose is to find significantly expressed genes between mutant and wild >type mice. I tried to figure it out by reading "Limma user's guide" and >came up with the following solution, is it right? or doesn't make any sense? >My targets file is: >SlideNumber FileName Cy3 Cy5 Target1 Target2 >1 1391.spot wt1 mu1 wt1 > mu1 >2 1392.spot mu1 wt1 mu1 > wt1 >3 1340.spot wt2 mu1 wt2 > mu1 >4 1341.spot mu1 wt2 mu1 > wt2 >5 1395.spot wt3 mu1 wt3 > mu1 >6 1396.spot mu1 wt3 mu1 > wt3 >7 1393.spot wt1 mu2 wt1 > mu2 >8 1394.spot mu2 wt1 mu2 > wt1 >9 1371.spot wt2 mu2 wt2 > mu2 >10 1372.spot mu2 wt2 mu2 > wt2 >11 1338.spot wt3 mu2 wt3 > mu2 >12 1339.spot mu2 wt3 mu2 > wt3 >13 1387.spot wt1 mu3 wt1 > mu3 >14 1388.spot mu3 wt1 mu3 > wt1 >15 1399.spot wt2 mu3 wt2 > mu3 >16 1390.spot mu3 wt2 mu3 > wt2 >17 1397.spot wt3 mu3 wt3 > mu3 >18 1398.spot mu3 wt3 mu3 > wt3 >mu1, mu2, and mu3 are different mice which are biological replicates, and >wt1, wt2 and wt3 are different mice which are biological replicates. > > targets <- readTargets() ># I try to include mouse effect in the following way: > > t1 <- model.matrix(~0+factor(targets$Target1)) > > t2 <- model.matrix(~0+factor(targets$Target2)) > > t3 <- t1+t2 You can't add design matrices in this way. > > design <- cbind(Dye=1, > MuvsWt=c(1,-1,1,-1,1,-1,1,-1,1,-1,1,-1,1,-1,1,-1,1,-1), t3 ) > > pair <- c(1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9) > > corfit <- duplicateCorrelation(MA,design,block=pair) > > fit <- lmFit(MA,design,block=pair,correlation=corfit$consensus) > > fit <- eBayes(fit) >Thanks again for your time. > >Ren
ADD COMMENT

Login before adding your answer.

Traffic: 554 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6