[BioC] unbalanced block design using limma

[Gordon K Smyth]

Dear Juan Lin,

Only one of your samples has a technical replicate, and this is not enough 
information to estimate the duplicate (within sample) correlation, so you 
can't use the duplicateCorrelation() function.  You need to either average 
the two technical replicates (there is a function avearrays in the 
developmental version of limma for this purpose), or just choose the 
technical replicate that looks of highest quality and drop the other one.

Best wishes
Gordon


> Date: Fri, 24 Sep 2010 15:17:28 -0400
> From: "Juan Lin" <julin at aecom.yu.edu>
> To: <bioconductor at stat.math.ethz.ch>
> Subject: [BioC] unbalanced block design using limma
> Message-ID: <002801cb5c1d$20f9cd40$62ed67c0$@yu.edu>
> Content-Type: text/plain
>
> Dear List,
>
>
>
> I am using limma to analyze a affy ST1.0 dataset.
>
>
>
> I have 6 patients from which I have two tissue types: tumor and non-tumor.
> One of the sample has a technical repeat, resulting 12 arrays.
>
> I would like to compare tumor vs. non-tumor and incorporate the pairing and
> technical repeats information into the analysis. However, I got error when I
> tried to calculate the correlation using duplicatedCorrelation
> function(please see code below). Therefore, I cannot do limma fit as usual.
>
>
>
>> library(limma)
>
>> target<-read.table("target.txt",header=T,sep="\t")
>
>
>
>> target
>
>          FileName  Type Tumor Patient Repeat
>
> 1   8671-596NT.CEL ADENO     N       1      1
>
> 2    8672-543T.CEL ADENO     Y       2      2
>
> 3    8674-558T.CEL ADENO     Y       3      3
>
> 4   8675-558NT.CEL ADENO     N       3      4
>
> 5    8676-560T.CEL ADENO     Y       4      5
>
> 6   8677-560NT.CEL ADENO     N       4      6
>
> 7  8679-596NT2.CEL ADENO     N       1      1
>
> 8    8686-507T.CEL ADENO     Y       5      7
>
> 9   8687-507NT.CEL ADENO     N       5      8
>
> 10 8688-507NTb.CEL ADENO     N       5      9
>
> 11   8689-538T.CEL ADENO     Y       6     10
>
> 12  8690-538NT.CEL ADENO     N       6     11
>
>
>
>> F<-factor(target$Tumor,levels=c("N","Y"))
>
>> Patient<-factor(target$Patient)
>
>> design<-model.matrix(~Patient+F)
>
>> design
>
>   (Intercept) Patient2 Patient3 Patient4 Patient5 Patient6 FY
>
> 1            1        0        0        0        0        0  0
>
> 2            1        1        0        0        0        0  1
>
> 3            1        0        1        0        0        0  1
>
> 4            1        0        1        0        0        0  0
>
> 5            1        0        0        1        0        0  1
>
> 6            1        0        0        1        0        0  0
>
> 7            1        0        0        0        0        0  0
>
> 8            1        0        0        0        1        0  1
>
> 9            1        0        0        0        1        0  0
>
> 10           1        0        0        0        1        0  0
>
> 11           1        0        0        0        0        1  1
>
> 12           1        0        0        0        0        1  0
>
>
>
>> corfit<-duplicateCorrelation(intensity,design,ndups=1,block=target$Repeat)
>
> Loading required package: statmod
>
>
>
>> fit<-lmFit(intensity,design,block=target$Repeat,cor=corfit$consensus)
>
> Error in backsolve(cholV, t(M), transpose = TRUE) :
>
>  NA/NaN/Inf in foreign function call (arg 1)fit<-eBayes(fit)
>
>
>
> Hope someone can help me to figure out the problem . Thanks!
>
>
>
> Juan Lin
>
> AECOM

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}