extracting the number of cells that are considered doublets

[tanasa]

Dear all,

Would you please advise how shall I extract the number of potential dublets in a sample (ie. given a list of cells in the sample to assign to each cell the labels "doublet / no dublet"). Thanks a lot,

Bogdan

[rcorces]

Thats a bit of an odd question. Normally, you run addDoubletScores() and then filterDoublets() which removes all of the doublets from your project based on the provided params. So I guess you would be asking how to identify which cells are doublets before running filterDoublets() but after running addDoubletScores()? You could relatively easily step through the code in filterDoublets() to get this answer.

ArchR/R/FilterCells.R

Line 45 in 968e442

filterDoublets <- function(ArchRProj = NULL, cutEnrich = 1, cutScore = -Inf, filterRatio = 1){

[tanasa]

Thank you Ryan.

Just for my own information please, as I have been using ArchR only for a short while.

Assuming that we do not use ArchR or any other dublet scoring algorithm, what is the typical range of median number of fragments per cell that we shall see in a successful experiment ?

For example, shall we have an experiment where the median_number_of_fragments per cell is ~ 90 000, would it indicate dublets or rather triplets ? Thanks a lot !

[rcorces]

Number of fragments is a very poor predicter of doublets. We showed this in Fig. 1h of the ArchR manuscript: