Question: Test for differential variability with missMethyl (DiffVar) using block design
0
gravatar for Simone
22 months ago by
Simone170
Simone170 wrote:

Hello,

I would like to test for differential methylation variability using missMethyl, but with a block design to take sample pairs/quartets into account.
If I wanted to test for differential (mean) methylation with my experimental design, I would use a code like this for example:

design <- model.matrix(~0+A1+A2+B1+B2)

blocks <- id
dupcor <- duplicateCorrelation(object=mval, design=design, block=blocks)
fit <- lmFit(mval, design, blocks, correlation=dupcor$consensus)

contrast.matrix <- makeContrasts(
  contrastA=A2-A1,
  contrastB=B2-B1,
  contrast1=B1-A1,
  contrast2=B2-A2,
  levels=design)

fit2 <- contrasts.fit(fit, contrast.matrix)
fit2 <- eBayes(fit2)

topTable(fit2, coef="contrastA")
topTable(fit2, coef="contrastB")
topTable(fit2, coef="contrast1")
topTable(fit2, coef="contrast2")

This works fine.
However, when I try to "translate" this into code for missMethyl, the following happens:

> fitvar <- varFit(mval, design, blocks, correlation=dupcor$consensus)
Error in varFit(mval, design, blocks, correlation = dupcor$consensus) : 
  unused argument (correlation = dupcor$consensus)

I used blocks because I have samples from the same (or a subset of the same) individuals across all conditions and would like to take this into account. As the block design didn't work, I also tried to simply add the ids (which are factors) as covariates to the design, but then I got the following warning and no results (all NA):

> design <- model.matrix(~0+A1+A2+B1+B2+id)
> [...]
> fitvar <- varFit(mval, design=design)
Warning messages:
1: Partial NA coefficients for 466566 probe(s) 

Does anybody know how I could properly test for differential variability taking the information about the individuals into account?

Best,
Simone

> sessionInfo()
R version 3.2.1 (2015-06-18)
Platform: x86_64-unknown-linux-gnu (64-bit)
Running under: Red Hat Enterprise Linux Server 7.3 (Maipo)

locale:
 [1] LC_CTYPE=en_GB.UTF-8       LC_NUMERIC=C               LC_TIME=en_GB.UTF-8        LC_COLLATE=en_GB.UTF-8     LC_MONETARY=en_GB.UTF-8   
 [6] LC_MESSAGES=en_GB.UTF-8    LC_PAPER=en_GB.UTF-8       LC_NAME=C                  LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_GB.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] limma_3.26.9     missMethyl_1.4.0 RSQLite_1.0.0    DBI_0.5-1   
ADD COMMENTlink modified 11 months ago by lhuang730 • written 22 months ago by Simone170
Answer: Test for differential variability with missMethyl (DiffVar) using block design
0
gravatar for lhuang7
11 months ago by
lhuang730
United States
lhuang730 wrote:

I got the similar warning message when calling varFit under a contrast design matrix setting (multiple comparison groups).

> fitvar <- varFit(mval, design = design)
Warning messages:
1: Partial NA coefficients for 466566 probe(s)

The issue turns out to be a typo within the function varFit.default() in the source code DiffVar.R:

if(is.null(coef)) 
   coef <- c(1, ncol(design))
z <- getLeveneResiduals(data, design = design[,coef], type = type)

As you can see, the coefficients only take the first and the last comparison group. It should be coef <- 1:ncol(design).

I hope the authors can fix it in the next release but meanwhile you could just modify this line of code by yourself and re-install the package from the fixed source file. HTH.

 

 

 

ADD COMMENTlink modified 11 months ago • written 11 months ago by lhuang730
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 126 users visited in the last hour