Issue with knownDoublets

[bjstewart1]

the knownDoublets option in scDblFinder is is throwing an error when presented with knownDoublets and also samples

Reproducible example:

library(scDblFinder)
sce <- mockDoubletSCE()
sce$type <- sce$type %in% "doublet"
sce$channel <- c(rep("sample1", floor(ncol(sce)/2)), rep("sample2", ceiling(ncol(sce)/2)))[1:ncol(sce)]
scldbl <- scDblFinder(sce = sce, 
                      samples = "channel",
                      knownDoublets = "type")

yields
Error in value[[3L]](cond) : An error occured while processing sample 'cluster1': Error in .checkColArg(sce, knownDoublets):knownDoubletsshould have a length equal to the number of columns insce.

but

scldbl <- scDblFinder(sce = sce, 
                      #samples = "channel",
                      knownDoublets = "type")

succeeds
Clustering cells... 4 clusters Creating ~5000 artifical doublets... Dimensional reduction Finding KNN... Evaluating cell neighborhoods... Training model... Finding threshold... Threshold found:0.425 19 (3.7%) doublets called

[plger]

thanks, will look into it in the next couple of days. I'm assuming you're using the latest github version?
Note that the current usage of knownDoublets actually decreases performance if (as is often the case) a sizeable fraction of the known doublets are homotypic. More flexibility on their usage will be added soon.

[bjstewart1]

yes this is after
BiocManager::install("plger/scDblFinder")
on 19-7-21
thanks

[plger]

should be fixed in the latest push