Agilent spike-in probes
1
0
Entering edit mode
@sean-davis-490
Last seen 12 weeks ago
United States
On Sat, Mar 29, 2008 at 11:39 PM, Srinivas Iyyer <srini_iyyer_bio at="" yahoo.com=""> wrote: > dear sean, > i apologize for sending this email and attached > figures to you. I am not sure if I can send figures as > attachment to mailing list. I wanted to see expert > opinion on this particular topic because this is first > time i am analyzing agilent chip data. > Would you please look into my design, code and figures > and let me know if this method okay. > > Spike-in probes are for QC purposes, if so why I am > getting spike-in probes as top candidates. Is there a > way to suppress them. > Thank you and I appreciate your help. > > > > dear group, > > I have agilent 4x44 (G4112F) chips. the hybs are done > as a paired design. sample obtained from patient > before and after treatment. 40 patient are in the > study. chip was hybridized with before treated(cy3) > and after treated (cy5) rna. > > I used LIMMA for normalizing and to calcuate > differentially expressed. > > in the first step, I did not go for background > subtraction and observed a blown-out ma plot. I'm not sure what "blown-out" means, but Agilent typically does background subtraction automatically (you'll need to look at the specific image extraction protocol to check). If you use the gProcessedSignal and rProcessedSignal (these are not the defaults in limma), you will probably get the benefit of their spatially-detrended loess background subtraction. > when i did background subtraction, i observed a more > compact ma. For q-q plot points at intersection are > not many suggesting that many genes are differentially > expressed. (figures are attach > > my main concern is, of top100 (from toptable > number=100), most of the probesets are spikein > probesets. (+)E1A_r60_a22 , DCP_22_6,DCP_22_7 and so > on. This could be dye bias, but I'm not sure. You didn't do dye swaps, so you cannot separate signal from dye bias. In any case, you will need to do some QC. Agilent provides a huge amount of QC and plots on the scanner machine. You can always look there to see what they do. Also, their technical manuals are pretty good at giving direction about the technology and the array data processing. > These spike-in probes are highly differentlly > expressed. > > > my targets file > > filename cy3 cy5 > patient1 before after > patient2 before after > ...... > patient40 before after > > my design matrix: > desin <- modelMatrix(targets,ref='before') > > desin > after > [1,] 1 > [2,] 1 > [3,] 1 > [4,] 1 > [5,] 1 > [6,] 1 > [7,] 1 > [8,] 1 > > RG2 <- backgroundCorrect(RG,method='subtract') > MA2 <- normalizeWithinArrays(RG2,method='loess') > plotDensities(MA2) > boxplot(MA2$M~col(MA2$M),names=colnames(MA2$M)) > MA2a <- normalizeBetweenArrays(MA2,method='scale') These are two-color arrays. Do you really need to do the between-array normalization? You might, but I think you might spend some time proving to yourself that is the case. > fit.b <- lmFit(MA2a,design) > fit.b <- eBayes(fit.b) > topTable(fit.b,number=50,adjust.method='BH')[,c(5,9,10,11,12,13)] > > my questions are: > > 1. for this paired sample (cy3,cy5) design, is my > limma model matrix okay. > 2. how to avoid getting spike-in . I never saw > spike-in getting into top-table. is there some mistake > going on at some place. is it normal for spike-in > probes to come as top differentially expressed probes. It happens, yes. I would definitely do some QC, though. It doesn't look like you have done any in your code here. > 3. are the attached figures (MA plot and q-q plot) > reflect a good normalized data. The qq plot does not really tell you about normalization. The single MA plot looks OK. You will want to look at all of the MA plots and some more extensive QC. > 4. my chip is hgug4112F. I do not see annotation file > on bioconductor. I think the hgug4112a annotation package is what you want. You'll want to double-check that with a few lookups to be sure. Sean
Annotation Normalization GO hgug4112a limma Annotation Normalization GO hgug4112a limma • 883 views
ADD COMMENT
0
Entering edit mode
Naomi Altman ★ 6.0k
@naomi-altman-380
Last seen 2.9 years ago
United States
In my experience with Agilent arabdopsis arrays, some of the Agilent spike-ins bind only to one of the dyes (or bind much more strongly to one). I always remove the controls before doing differential expression analysis. Naomi At 08:29 AM 3/30/2008, Sean Davis wrote: >On Sat, Mar 29, 2008 at 11:39 PM, Srinivas Iyyer ><srini_iyyer_bio at="" yahoo.com=""> wrote: > > dear sean, > > i apologize for sending this email and attached > > figures to you. I am not sure if I can send figures as > > attachment to mailing list. I wanted to see expert > > opinion on this particular topic because this is first > > time i am analyzing agilent chip data. > > Would you please look into my design, code and figures > > and let me know if this method okay. > > > > Spike-in probes are for QC purposes, if so why I am > > getting spike-in probes as top candidates. Is there a > > way to suppress them. > > Thank you and I appreciate your help. > > > > > > > > dear group, > > > > I have agilent 4x44 (G4112F) chips. the hybs are done > > as a paired design. sample obtained from patient > > before and after treatment. 40 patient are in the > > study. chip was hybridized with before treated(cy3) > > and after treated (cy5) rna. > > > > I used LIMMA for normalizing and to calcuate > > differentially expressed. > > > > in the first step, I did not go for background > > subtraction and observed a blown-out ma plot. > >I'm not sure what "blown-out" means, but Agilent typically does >background subtraction automatically (you'll need to look at the >specific image extraction protocol to check). If you use the >gProcessedSignal and rProcessedSignal (these are not the defaults in >limma), you will probably get the benefit of their spatially- detrended >loess background subtraction. > > > when i did background subtraction, i observed a more > > compact ma. For q-q plot points at intersection are > > not many suggesting that many genes are differentially > > expressed. (figures are attach > > > > my main concern is, of top100 (from toptable > > number=100), most of the probesets are spikein > > probesets. (+)E1A_r60_a22 , DCP_22_6,DCP_22_7 and so > > on. > >This could be dye bias, but I'm not sure. You didn't do dye swaps, so >you cannot separate signal from dye bias. In any case, you will need >to do some QC. Agilent provides a huge amount of QC and plots on the >scanner machine. You can always look there to see what they do. >Also, their technical manuals are pretty good at giving direction >about the technology and the array data processing. > > > These spike-in probes are highly differentlly > > expressed. > > > > > > my targets file > > > > filename cy3 cy5 > > patient1 before after > > patient2 before after > > ...... > > patient40 before after > > > > my design matrix: > > desin <- modelMatrix(targets,ref='before') > > > desin > > after > > [1,] 1 > > [2,] 1 > > [3,] 1 > > [4,] 1 > > [5,] 1 > > [6,] 1 > > [7,] 1 > > [8,] 1 > > > > RG2 <- backgroundCorrect(RG,method='subtract') > > MA2 <- normalizeWithinArrays(RG2,method='loess') > > plotDensities(MA2) > > boxplot(MA2$M~col(MA2$M),names=colnames(MA2$M)) > > MA2a <- normalizeBetweenArrays(MA2,method='scale') > >These are two-color arrays. Do you really need to do the >between-array normalization? You might, but I think you might spend >some time proving to yourself that is the case. > > > fit.b <- lmFit(MA2a,design) > > fit.b <- eBayes(fit.b) > > topTable(fit.b,number=50,adjust.method='BH')[,c(5,9,10,11,12,13)] > > > > my questions are: > > > > 1. for this paired sample (cy3,cy5) design, is my > > limma model matrix okay. > > 2. how to avoid getting spike-in . I never saw > > spike-in getting into top-table. is there some mistake > > going on at some place. is it normal for spike-in > > probes to come as top differentially expressed probes. > >It happens, yes. I would definitely do some QC, though. It doesn't >look like you have done any in your code here. > > > 3. are the attached figures (MA plot and q-q plot) > > reflect a good normalized data. > >The qq plot does not really tell you about normalization. The single >MA plot looks OK. You will want to look at all of the MA plots and >some more extensive QC. > > > 4. my chip is hgug4112F. I do not see annotation file > > on bioconductor. > >I think the hgug4112a annotation package is what you want. You'll >want to double-check that with a few lookups to be sure. > >Sean > >_______________________________________________ >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
ADD COMMENT

Login before adding your answer.

Traffic: 728 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