I’m having trouble with the estimateCellCounts function in minfi.. I am running it on EPIC data, and while the estimates of cell composition seem great, and correlate well with both Houseman’s original code and EpiDish prediction, the normalized dataset in the output (when returnAll = TRUE) is always reduced to 452567 probes. I’ve tried the following variations which all gave the same outcome:
Minfi_cellcounts_corr <- estimateCellCounts(RGset_282, compositeCellType = "Blood",
processMethod = "auto", probeSelect = "auto",
cellTypes = c("Bcell", "CD4T", "CD8T", "Gran", "Mono", "NK"),
referencePlatform = c("IlluminaHumanMethylation450k"),
returnAll = TRUE, meanPlot = FALSE, verbose = TRUE)
Minfi_cellcounts_corr2 <- estimateCellCounts(RGset_282, compositeCellType = "Blood",
processMethod = "auto", probeSelect = "auto",
cellTypes = c("Bcell", "CD4T", "CD8T", "Gran", "Mono", "NK"),
returnAll = TRUE, meanPlot = FALSE, verbose = TRUE)
Minfi_cellcounts_corr3 <- estimateCellCounts(RGset_282, compositeCellType = "Blood",
processMethod = "auto", probeSelect = "auto",
cellTypes = c("Bcell", "CD4T", "CD8T", "Gran", "Mono", "NK"),
convertArray=TRUE,
returnAll = TRUE, meanPlot = FALSE, verbose = TRUE)
When I attempted to switch the reference platform to EPIC using the code below, I got the following error:
Minfi_cellcounts_corr2 <- estimateCellCounts(RGset_282, compositeCellType = "Blood",
processMethod = "auto", probeSelect = "auto",
cellTypes = c("Bcell", "CD4T", "CD8T", "Gran", "Mono", "NK"),
referencePlatform = c("IlluminaHumanMethylationEPIC"
),
returnAll = TRUE, meanPlot = FALSE, verbose = TRUE)
Failed with error: 'Package 'genefilter' version 1.62.0 cannot be unloaded:
Error in unloadNamespace(package) : namespace 'genefilter' is imported by 'sva', 'minfi' so cannot be unloaded
'
Error in estimateCellCounts(RGset_282, compositeCellType = "Blood", processMethod = "auto", :
Could not find reference data package for compositeCellType 'Blood' and referencePlatform 'EPIC' (inferred package name is 'FlowSorted.Blood.EPIC')
Does anyone have any idea how I can fix this? I saw in a forum from a year ago Kasper said that when you use FlowSorted.Blood.EPIC the command looks for a dataset that doesn’t exist – is there a way around this to get the full EPIC dataset adjusted for cell composition? I am using minfi version 1.27.6 in R version 3.5.0.
Thanks in advance