My question is similar to Minfi: estimateCellCounts rgSet, which was not answered after 6 months.

I have beta values from a level 3 TCGA dataset. They are stored in a matrix, where each row corresponds to a locus and each column is a sample. I tried to use the estimateCellCounts() to infer the proportion of cell types in my data but this functions needs a an object of class RGChannelSet as its rgSet argument. 

Is there a way to infer the cell proportions using only a matrix of beta values?

One solution might be to use the projectCellType() function which is called by estimateCellCounts function. But projectCellType is not exported by minfi package and it is not documents.

I could not find a "safe" solution using minfi. An "unsafe" solution is using the minfi:::projectCellType (not exported) function as illustrated here.