I am using the variancePartition package's dream() pipeline, and have encountered something strange. I have 300 samples with 2009 genes. I have uploaded a TMM normalized cpm matrix from edgeR, written a formula, and run

vobjDream = voomWithDreamWeights( geneExpr, form, metadata )

It gave me the warning message:

"Warning message:
In regularize.values(x, y, ties, missing(ties)) :
  collapsing to unique 'x' values"

I inspected the vobjDream object and noticed that the last 45 samples have NA encoded for all genes.

Has anyone encountered this problem, does anyone know what may be happening here? The formula itself was

form <- ~ (1|factDx) + (1|Library.Work.) + RIN + (1|Sex) + PMI + (1|Smoking) + (1|Ethanol)

and worked fine in the standard application's fitExtractVarModel() pipeline.