Correct model specification for paired design

[David Fisher]

Hi folks,

I have conducted an experiment with a paired design, where we looked at WGBS data of 5 individuals before and after a treatment.

Previously I've loaded the data like this:

all_IDs = methRead(file.list,
                sample.id = list("T52_1","T53_1","T54_1","T55_1","T56_1",
                                 "T52_2","T53_2","T54_2","T55_2","T56_2"),
                assembly="new_ass",
                pipeline = "bismarkCoverage",
                treatment = c(rep(c(0,1), each = 5)), 
                dbtype="tabix",  
                mincov = 5) 

[some intermediate steps]

And analysed like this:

all_dmb = calculateDiffMeth(all_IDs_fu,  #"_fu" suffix indicates filtered and unified
                            covariates = NULL,
                            overdispersion = "MN",
                            #test = "F",
                            mc.cores = 32,
                            suffix = "fuMNtestC")

However, I am not certain this is properly accounting for the paired nature of the design. Can anyone comment on whether it is or not? If not, how does one specify a paired design, would I use the individual ID as a covariate like this:

covariates =c( "2", "3", "4", "5", "6", "2", "3", "4", "5", "6")

Its done in DSS, see section 3.4.4 of https://www.bioconductor.org/packages/devel/bioc/vignettes/DSS/inst/doc/DSS.html

Thanks for your help!

David

[Alexander Blume]

Hi David, 

As recently answered by Altuna (here and here), methylKit does not account for paired designs per se. But as you suggested, the pair information could be included as a covariate. 
Just make sure to provide the covariate as data.frame (see vignette ),  probably you should also convert the IDs to factors as mentioned in the DSS manual :

covariates = data.frame( ind_id = factor(c( "2", "3", "4", "5", "6", "2", "3", "4", "5", "6")))

We are not sure this works, maybe you could also double-check with DSS and report the results here?

Best,
Alex


https://bioconductor.org/packages/release/bioc/vignettes/methylKit/inst/doc/methylKit.html#38_Accounting_for_covariates

[David Fisher]

OK thank you, and apologies for not searching properly before asking.

David