search

dviraran / SingleR

SingleR: Single-cell RNA-seq cell types Recognition (legacy version)
GNU General Public License v3.0
263 stars 98 forks source link

Confliciting single.label and main.label #108

Open danshu opened 5 years ago

danshu commented 5 years ago

HI,

I noticed that for my mouse data, some cells have conflicting single.label and main.label annotations. For example, one cell got single.label annotation as "T cells (T.ETP)" while its main.label is "Stem cells". Is this normal and what are the possible reasons?

Best, Quan

dviraran commented 5 years ago
  1. This is not really conflicting, T.ETP are early T-cell precursors, and 'stem cells' in the ImmGen reference refer to hematopoietic stem cells and progenitors. So this actually looks like a good catch by SingleR.
  2. Yes, of course there can be conflicts. SingleR does not work with hierarchy (if you are interested why - because in reality, there is no real cell types hierarchy), and the difference between labels and main.labels is just that the reference samples for main.labels are all cells that correspond to this cell type. The interpretation from your result is that the expression profile of your cells is more progenitor-like than T-cell-like (which is of course true).
danshu commented 5 years ago

Thanks for your explanations. Here are a few other cells that have conflicting annotations. "T_cell:CD8+" "NK_cell" "B_cell:immature" "Epithelial_cells" "T_cell:Treg:Naive" "Neutrophils" I was confused because I expect single.labels are annotated after main.labels are determined. If single.labels and main.labels are actually independent of each other, then those cells are just some low-quality cells or doublets.