Clarification on Multinomial_Label

[domenico-somma]

Thanks for this interesting tool. I am testing it on 1 sample (formed from CD45+ cells of tissue), I used either the web page and the terminal.
If I check the Multinomial_Label in the starcat_results_scores file these are the all the label I can find:
CD4_CM CD4_EM CD4_Naive CD8_CM CD8_EM CD8_Naive CD8_TEMRA gdT MAIT Treg
Which surprised me, as I was expecting some doublets, artifact, etc.

If I check starcat_results_usage file, for example for this cell

TCACACCTCCTTCACG-1 the highest score is: Doublet-Myeloid with 0.32282695,
but the same cell in the starcat_results_scores file as been assigned as CD4_CM under the Multinomial_Label column

Is this expected? Did I misunderstand anything?

Thanks

[dylkot]

Hi @domenico-somma, yes this is expected behavior. The multinomial classifier was only trained with labeled T cell classes based on surface markers and so only outputs the labels you mentioned. The starcat_results_usage file can be used to identify doublets based on the various Doublet programs. But we currently don't have a discrete label classifier impelemented.