Hey,

Is it possible to correct for both a location and batch effect using Limma?

I have 4 surfaces: PET, PE, Glass and Water,

5 locations: A, B, C, D ,E,

and 2 batches: 1, 2

enter OTU<-read.csv("filterotu.csv",header=T,row.names=1) 
targets<-read.csv("sample.csv",header=T,row.names=1)
tax<-read.csv("filter_tax.csv", header=T, row.names=1)
tax <- as.matrix(tax)
group<-factor(targets$surface)
location<-factor(targets$location)
batch<-factor(targets$Batch)
dge<- DGEList(counts=OTU,samples=targets, genes=tax, group=group)
keep<- filterByExpr(dge, min.count = 1)
dge <- dge[keep,,keep.lib.sizes=FALSE]
dge <- calcNormFactors(dge)
design <- model.matrix(~0+group)

I tried using batch=batch and batch2=location but that didn't work.

toplot<-(removeBatchEffect(cpm(dge,log=TRUE,prior.count=5),batch=batch,batch2=location,design=design)) error: "Partial NA coefficients for 7319 probe(s)"

Then I tried the duplicatecorrelation function

corfit <- duplicateCorrelation(v,design,block=batch) toplot <- removeBatchEffect(cpm(dge,log=TRUE,prior.count=5), batch=location, design=design, block=batch, correlation=corfit$consensus)

but when I plotted the MDS it was exactly the same as just removing batch=location alone, as if the block wasn't having any effect.

Anyone got any suggestions?