new UMAP and mistake label after Reference Mapping #4040
-
Hello Satija lab, |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
Hi, In transferring the labels from the reference to query dataset, we will always assign a cell the label with the highest prediction score. For cell types that are unique to the query dataset, this often is the "next-closest" reference cell type but is not as ideal as knowing their actual identity (e.g. neutrophils). If you know ahead of time which cells are neutrophils, you could remove them before mapping to avoid misannotation. An alternative is to manually relabel though cells after mapping (e.g. using functions like |
Beta Was this translation helpful? Give feedback.
Hi,
In transferring the labels from the reference to query dataset, we will always assign a cell the label with the highest prediction score. For cell types that are unique to the query dataset, this often is the "next-closest" reference cell type but is not as ideal as knowing their actual identity (e.g. neutrophils). If you know ahead of time which cells are neutrophils, you could remove them before mapping to avoid misannotation. An alternative is to manually relabel though cells after mapping (e.g. using functions like
SetIdent
to map specific cells to new ident labels). This would allow you to keep the umap plot with all the data but enable relabeling of neutrophils specifically.