Are the confidence scores of celltype predicted by different reference models comparable？

[hxpGit512]

Dear Celltypist Team, Are the confidence scores of celltype predicted by different reference models comparable? whether I could choose the most suitable model to annotate new datasets by comparing the average confidence scores of each cluster annotated by different models? in addition, If a cell is annotated as “unassigned”or “heterogeneous”, what does its confidence score mean？

[ChuanXu1]

@hxpGit512, you can only compare scores under one model but not across models. In cases of “unassigned”or “heterogeneous”, the score of a cell represents the probability when predicting this cell into its most likely cell type in the model.

[hxpGit512]

已收件

[ChuanXu1]

I will close the issue as this should be solved. Please let us know if you still have problems.

[hxpGit512]

已收件