[BioC] limma - batch dates incorrporated into t-test

[James W. MacDonald]

Hi Helen,

On Thursday, September 12, 2013 9:02:25 AM, Helen Smith wrote:
> Hi All,
>
> I am new to limma and R programming.
>
> I have the CEL.files uploaded, normalised as an expression set, and have put together the target.txt file containing the following layout:
> Name    FileName             Target   Date
> 5              0207_DaB7_H_ea_Blast1.CEL     blastocyst            13/02/2007
> 6              0207_DaB8_H_ea_Blast2.CEL     blastocyst            13/02/2007
> 4              0207_DaB6_H_ea_4cell_r3.CEL  4cell       16/02/2007
> 16           0307_DaB1_H_ea_oocyte_r1.CEL             oocyte  23/03/2007
> 17           0307_DaB3_H_ea_oocyte_r3.CEL             oocyte  23/03/2007
> 18           0307_DaB5_H_ea_4cell_r2.CEL  4cell       23/03/2007
> 21           1007_DaB4_H_ea_4cell_r1.CEL  4cell       24/10/2007
> 22           1007_DaB9_H_ea_Blast3.CEL     blastocyst            24/10/2007
> 23           1007_DaB21_H_ea_oocyte9856.CEL        oocyte  24/10/2007
> 24           1007_DaB22_H_ea_oocyte9858.CEL        oocyte  24/10/2007
> 1              0108_DaB18_H_ea_4cell_0.9_r4.CEL       4cell       10/01/2008
> 2              0108_DaB20_H_ea_4cell_0.9_r6.CEL       4cell       10/01/2008
> 3              0108_DaB23_H_ea_Blast_2.5_r4.CEL      blastocyst            10/01/2008
> 10           0213_SueK(2)4_H_ea_005_B4.CEL           blastomere         01/03/2013
> 14           0213_SueK(2)8_H_ea_005_B8.CEL           blastomere         01/03/2013
> 15           0213_SueK(2)10_H_ea_002_8CE.CEL      8cell       01/03/2013
> 7              0213_SueK(2)1_H_ea_005_B1.CEL           blastomere         06/03/2013
> 8              0213_SueK(2)2_H_ea_005_B2.CEL           blastomere         06/03/2013
> 9              0213_SueK(2)3_H_ea_005_B3.CEL           blastomere         06/03/2013
> 11           0213_SueK(2)5_H_ea_005_B5.CEL           blastomere         06/03/2013
> 12           0213_SueK(2)6_H_ea_005_B6.CEL           blastomere         06/03/2013
> 13           0213_SueK(2)7_H_ea_005_B7.CEL           blastomere         06/03/2013
> 19           0713_Suek(3)2_Ep_H_8C_006_11.CEL    8cell       02/07/2013
> 20           0713_Suek(3)3_Ep_H_8C_004_11.CEL    8cell       02/07/2013
>
>
> But am struggling with setting up the design matrix and contrasts. My limited programming ability is probably playing the largest role, and I feel like smashing the computer against the wall (probably a sign of slight frustration )

It's really quite simple ;-D. You just make a design matrix that 
includes the dates, and then make a contrasts matrix that doesn't use 
the dates to make comparisons.

The simplest way to do this is to use model.matrix() and 
makeContrasts().

Assume you read in your targets.txt file like so:

targets <- read.table("targets.txt", header=TRUE, stringsAsFactors = 
FALSE)

we have to do the stringsAsFactors business because your 4cell and 
8cell sample designations won't work with makeContrasts() (they cannot 
start with a number).

targets$Targets <- factor(gsub("4", "four_", gsub("8", "eight_", 
targets$Targets)))

then

design <- model.matrix(~0+Target+Date, targets)

Now you probably have column names with an extra 'targets' in the name, 
which nobody likes, so we remove

colnames(design) <- gsub("targets", "", colnames(design))

contrast <- makeContrasts(blastomere - oocyte, blastomere - eight_cell, 
blastomere - four_cell, <add all the contrasts you want here>, levels = 
design)

and now you proceed on to the lmFit/contrasts.fit/eBayes steps.

Best,

Jim


>
> The batch dates are included in the Target file, as I would like to add them as a factor when computing the T-test. But this isn’t a contrast I would like to make on its own (does that make sense?)
>
> I would like to make all possible combinations of contrasts, listed as targets.
>
> Any help would be fantastic!!!!!
> Thank you in advance,
> Helen
>
>
>
>
> 	[[alternative HTML version deleted]]
>
>
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor

--
James W. MacDonald, M.S.
Biostatistician
University of Washington
Environmental and Occupational Health Sciences
4225 Roosevelt Way NE, # 100
Seattle WA 98105-6099