[BioC] (no subject)
James W. MacDonald
jmacdon at med.umich.edu
Sat Mar 13 14:54:21 CET 2010
Hi Ana,
You need to re-read what Naomi said. A correlation between dye swaps
would be expected. What she was warning about was a correlation between
treatments.
As for a good book, have you read the BioC monograph?
http://bioconductor.org/pub/docs/mogr/
The case studies might be of interest as well
http://bioconductor.org/pub/biocases/
Best,
Jim
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
**********************************************************
Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues
More information about the Bioconductor
mailing list