Not enough training samples for any cell types at root of cell type hierarchy!

[Dan Schnell]

What would be the most likely cause(s) of the "Not enough training samples for any cell types at root of cell type hierarchy!

Dataset I'm working with has 46 clusters, ranging from 11 to 600+ barcodes. All barcodes have a cluster assignment--no "Unknown"s.

Markers not in db or "Ok - marker_score N/A" have been removed.

Does the column in pdata that contains the clusterID that matches those used in the marker file need to have a specific name? Be a factor?

code:
library(org.Mm.eg.db)
set.seed(260)
marker_file_path <- paste(in1DataPath,"SMART-Markers2.txt",sep="")
SMART_classifier <- train_cell_classifier(cds = SMART_cds,
marker_file = marker_file_path,
db=org.Mm.eg.db,
cds_gene_id_type = "SYMBOL",
num_unknown = 500,
marker_file_gene_id_type = "SYMBOL")

Have tried values of num_unknown from 5 to 1000--all produce the error message (+ some extras w/ 5).

Thanks very much in advance for any guidance!

[Hannah Pliner]

The most likely culprit here would be some sort of mismatch on your gene names. I recently updated the monocle3 branch with a fix for the issue of varying version numbers on ENSEMBL IDs (i.e. if your data has the piece after the dot in ENS#####.#) Try pulling the latest and see if that solves the problem?

Best,

Hannah

[Schnell, Daniel (Dan)]

Thanks very much for your reply Hannah.  I will update the installation of monocle3.

 

Also, I saw some recent posts about using db=”none”in the checkmarkers function.  Doing so results in significantly fewer markers being flagged as being problematic, so maybe that will also help.

 

Best,

Dan

--
You received this message because you are subscribed to the Google Groups "garnett-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to garnett-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/garnett-users/8c2693d5-98de-4383-9336-b8d830dd7868%40googlegroups.com.