Hi, 

I am trying to use RUVs to normalise my data. I have 12 samples which are replicated in pairs. 

i.e. I can specify my differences like this:

differences = matrix(data= seq(1,12,by=1), byrow=TRUE, nrow=6)

where each row is a biological replicate

I create a "counts" matrix which is n X 12 matrix of the read counts of my genes. I create a vector "geneNames" which lists all the genes in the count matrix. 

I call the RUVs function as such:

RUVs(counts, geneNames, k=1, differences)

However, I am getting the following error:

Error in Y[scIdx[ii, jj], , drop = FALSE] : 
  no 'dimnames' attribute for array

Please note my "counts" matrix is not a data.frame but simply a numeric matrix. 

I then tried using the newSeqExpressionSet function:

x = as.factor(rep(c(1,2,3,4,5,6), each=2))

set = newSeqExpressionSet(as.matrix(counts), phenoData = data.frame(x,row.names=geneNames))

and run 

RUVs(set, genes, k=1, differences)

However, I get the following error:

Error in Y[scIdx[ii, jj], , drop = FALSE] : subscript out of bounds

Could someone suggest to me how I could modify my commands to make this work ? 

Thank you ! 

Hi Asif,

my guess is that your count matrix doesn't have the "rownames" attribute. You should modify your matrix to have row names the same gene IDs that you use in the "genes" vector. Alternatively, you can use a numeric index instead of a character string to identify genes.  

I hope this helps.

davide

I plotted the PCA for my normalized counts but it doesn't give anything sensible. I did normalization using DESeq's size factor estimation algorithm and it gives me sensible results. The function call I made is below:

ruv_de = RUVs(data.matrix(de), genes, k=1, differences)

x=rep(c(1,2,3,4,5,6),2)

EDASeq::plotPCA(data.matrix(ruv_de$normalizedCounts), col=colors[x], xlim=c(-0.5,0.5), cex=0.90)

Is there something I am doing wrong ?