[BioC] (no subject)
Naomi Altman
naomi at stat.psu.edu
Sat Mar 13 18:34:10 CET 2010
Dear Ana,
There are several very good books on microarray analysis - several of
them written by contributers to this email list and using
Bioconductor. You might also want to read up on ANOVA for unbalanced
data. I cannot recommend any particular text, as they mostly focus
on balanced designs. There is a text by Searle focusing on
unbalanced data, but I have only dipped into it.
Regards,
Naomi
At 03:00 AM 3/13/2010, Ana Staninska wrote:
>Dear Naomi,
>Thank you very very much for your very helpful answers.
>Could you maybe tell me what did you mean that if the dye swap
>correlation between two treatments is -0.2 I am in trouble. What is
>considered to be a good dye swap correlation (I calculate it using
>duplicateCorrelation function in limma). Also what is considered as
>a good correlation between duplicate spots (after normalization) ?
>I know that the easiest way out is to ask a statistician to do the
>analysis, but I would like to learn it myself to do it (I am a
>mathematician, so I think I should be able to learn it). Could you
>maybe point out a literature that I could read and learn a proper
>way of dealing with any kind of microarrays.
>
>
>Thank you very much one more time, Best, Ana
>
> > Date: Fri, 12 Mar 2010 16:29:03 -0500
> > To: staninska at hotmail.com; naomi at stat.psu.edu;
> bioconductor at stat.math.ethz.ch
> > From: naomi at stat.psu.edu
> > Subject: RE: [BioC] (no subject)
> >
> > Dear Ana,
> > I actually meant that you should average dye swaps, not spots,
> > although either is OK as long as you use corfit for the other.
> >
> > If there are no technical replicates for some biological reps, the
> > analysis is much more complicated. This really requires a
> > statistical consultant and someone who will do some detailed
> > preliminary analyses.
> >
> > Naomi
> >
> > p.s. I hope that the correlation of -0.2 for the dye swaps is for
> > R-G. If it is for treatment A - treatment B, you have a problem.
> >
> > At 03:08 PM 3/12/2010, Ana Staninska wrote:
> > >Dear Naomi,
> > >
> > >Thank you very much for your answer. I just have few follow up question.
> > >
> > >How big should be the correlation on my duplicate spots in order to
> > >"safetly" average them?
> > >Before the normalization, the correlation on my duplicate spots is
> > >around 0.7-0.8, but after normalization
> > >it is only around 0.4-0.6. Which I think it is not the best.
> > >Probably I should mention that the correlation of dye swapped arrays
> > >is around -0.2.
> > >
> > >Also, for some of the experiments, we had to remove certain arrays,
> > >and therefore not all of my biological replicates are dye swapped.
> > >In that case I think I should use the contrast matrix to average of
> > >the treated vs non-treated comparisons.
> > >Isn't then better to use the corfit$consensus on my duplicate spots?
> > >
> > >Thank you very much in advance,
> > >
> > >All the best,
> > >Ana
> > >
> > >
> > >
> > >
> > >
> > > > Date: Fri, 12 Mar 2010 12:28:06 -0500
> > > > To: staninska at hotmail.com; bioconductor at stat.math.ethz.ch
> > > > From: naomi at stat.psu.edu
> > > > Subject: Re: [BioC] (no subject)
> > > >
> > > > The estimated error variance used for the test denominator will be an
> > > > average of technical and biological replication, and therefore not
> > > > really appropriate for your analysis. However, you could average the
> > > > 2 technical replicates prior to running limma which would give you
> > > > the right error structure.
> > > >
> > > > --Naomi
> > > >
> > > > At 12:04 PM 3/12/2010, Ana Staninska wrote:
> > > >
> > > > >Dear Bioconductor,
> > > > >I have a simple experiment that I have to analyze in order to find
> > > > >differentially expressed genes. I have 10 biological replicates, and
> > > > >each biological replicate has two technical replicates which appear
> > > > >as dye swapped. So in total I have 20 arrays. Each of the probes are
> > > > >spotted twice on the array (on the left and on the right hand side).
> > > > >I use limma to do my analysis. I know at the moment it is not
> > > > >possible to treat duplicate spots, technical replicates and
> > > > >biological replicates, but I though if I use the
> > > > >duplicateCorrelation function on my duplicate spots, and then to use
> > > > >a contrast matrix to average of all of the Treated vs Non-treated
> > > > >biological samples, I could address all 3 replications. Am I correct?
> > > > >
> > > > >
> > > > >I am sending a copy of my code, if someone could look at it at tell
> > > > >me whether I made somewhere a mistake.
> > > > >Thank you very much in advance,
> > > > >Sincerely Ana Staninska
> > > > >
> > > > >
> > > > > library(limma)> library(statmod)> library(marray)>
> > > > > library(convert)> library(hexbin)> library(gridBase)>
> > > > > library(RColorBrewer)> > targets <-
> > > > > readTargets("Lysi_270705.txt")> > ### Only manually removed ot
> > > > > absent spots are given 0 weight ###> RGa <- read.maimages(targets,
> > > > > source="genepix", wt.fun=wtflags(weight=0,
> > > > > cutoff=-75), other.columns=c("F635 SD","B635 SD","F532 SD","B532
> > > > > SD","B532 Mean","B635 Mean","F Pixels","B Pixels"))Read
> > > > > LYSI270705_1_200905.gpr Read LYSI270705_1dw_200905.gpr Read
> > > > > LYSI270705_2_200905.gpr Read LYSI270705_2dw_200905.gpr Read
> > > > > LYSI270705_3_121005.gpr Read LYSI270705_3dw_121005.gpr Read
> > > > > LYSI270705_4_121005.gpr Read LYSI270705_4dw_121005.gpr Read
> > > > > LYSI270705_5_121005.gpr Read LYSI270705_5dw__121005.gpr Read
> > > > > LYSI270705_6_121005.gpr Read LYSI270705_6dw__121005.gpr Read
> > > > > LYSI270705_7_151001.gpr Read LYSI270705_7dw_151005.gpr Read
> > > > > LYSI270705_8_151005.gpr Read LYSI270705_8dw_151005.gpr Read
> > > > > LYSI270705_9_151005.gpr Read LYSI270705_9dw_151005.gpr Read
> LYSI270705!
> > > > > _10_151005.gpr Read LYSI270705_10dw_151005.gpr > for(i in
> > > > > 1:nrow(RGa)){+ for(j in
> > > > > 1:ncol(RGa)){+ if(RGa$Rb[i,j]+RGa$R[i,j]+ RGa$G[i,j]+
> > > > > RGa$Gb[i,j] ==0)+ RGa$weights[i,j]<-0+ }+ }> >
> > > > > ####################################################> ###
> > > > > Background Correction = Normexp + offset 25 ####>
> > > > > ####################################################> > RG
> > > > > <-backgroundCorrect(RGa, method="normexp", , normexp.method="mle",
> > > > > offset=25)Green channelCorrected array 1 Corrected array 2
> > > > > Corrected array 3 Corrected array 4 Corrected array 5 Corrected
> > > > > array 6 Corrected array 7 Corrected array 8 Corrected array 9
> > > > > Corrected array 10 Corrected array 11 Corrected array 12 Corrected
> > > > > array 13 Corrected array 14 Corrected array 15 Corrected array 16
> > > > > Corrected array 17 Corrected array 18 Corrected array 19 Corrected
> > > > > array 20 Red channelCorrected array 1 Corrected array 2 Corrected
> > > > > array 3 Corrected array 4 Corrected array 5 Corrected array 6
> > > > > Corrected array 7 Corrected array 8 Corrected array !
> > > > > 9 Corrected array 10 Corrected array 11 Corrected array 12
> Corrected a
> > > > >rray 13 Corrected array 14 Corrected array 15 Corrected array 16
> > > > >Corrected array 17 Corrected array 18 Corrected array 19 Corrected
> > > > >array 20 > ####################################################>
> > > > >##### normalize Within arrays #########>
> > > > >####################################################> > MA
> > > > ><-normalizeWithinArrays(RG, method="loess")> >
> > > > >####################################################> ######
> > > > >Contrast Matrix ############>
> > > > >####################################################> >
> > > > >design<-cbind( + MU1vsWT1=c(
> > > > >1,-1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0),+ MU2vsWT2=c(0,0,
> > > > >1,-1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0),+ MU3vsWT3=c(0,0,0,0,
> > > > >1,-1,0,0,0,0,0,0,0,0,0,0,0,0,0,0),+ MU4vsWT4=c(0,0,0,0,0,0,
> > > > >1,-1,0,0,0,0,0,0,0,0,0,0,0,0),+ MU5vsWT5=c(0,0,0,0,0,0,0,0,
> > > > >1,-1,0,0,0,0,0,0,0,0,0,0),+ MU6vsWT6=c(0,0,0,0,0,0,0,0,0,0,
> > > > >1,-1,0,0,0,0,0,0,0,0),
> > > > >+ MU7vsWT7=c(0,0,0,0,0,0,0,0,0,0!
> > > > > ,0,0,
> > > > > 1,-1,0,0,0,0,0,0),+ MU8vsWT8=c(0,0,0,0,0,0,0,0,0,0,0,0,0,0,
> > > > > 1,-1,0,0,0,0),+ MU9vsWT9=c(0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
> > > > > 1,-1,0,0),+ MU10vsWT10=c(0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
> > > > > 1,-1))> > cont.matrix <-
> > > > >
> > > > > ####################################################>
> > > > > ### Duplicate Correlations on duplicate spots ####>
> > > > > ####################################################> >
> > > > > corfit<-duplicateCorrelation(MA, ndups=2, spacing=192)> >
> > > > > ####################################################> ##### Linear
> > > > > Fit Model and Contrasts fit #######>
> > > > > ####################################################> >
> > > > > fit<-lmFit(MA, design, ndups=2, spacing=192,
> > > > > cor=corfit$consensus)> > fit<-contrasts.fit(fit, cont.matrix)> >
> > > > > ####################################################>
> > > > > ######### eBayes Statistics ###############> #################!
> > > > > ###################################> > fit<-eBayes(fit)> >
> ###########
> > > > >###################################################> ### Writing
> > > > >the Results ######>
> > > > >##############################################################>
> > > > >TTnew<-topTable(fit,coef=1, number=100, adjust="BH")
> > > > >
> > > > >
> > > > >Ana StaninskaHelmholtz-Zentrum MuenchenDepartment of Scientific
> > > > >ComputingNeuherberg, Deutschland+49 (0) 89 3187 2656
> > > > >
> > > > > [[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
> > > >
> >
> > 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
More information about the Bioconductor
mailing list