Closed DarioS closed 5 years ago
I strongly recommend going over supplementary information 2. There is a long discussion there about low correlations and unknowns. We use different measures for confidence and it is possible for the user to use them, however, at least as a first pass it is not recommended.
SingleR is not meant to be viewed solely by the colors in the tSNE plots, this is only a first step in understanding the complexity of the data. We provide different tools and visualizations for further analyses which can assist in this endeavor. For example, one of the strengths of SingleR is to help detect previously uncharacterized cells. You can see in the paper our journey to identify a new macrophage state which SingleR identified as associated with two previously characterized states.
I found the supplementary information in the folder named "vignettes" and it answers my question.
Looking at the the web application's incorporated data sets, it seems that every cell is assigned to a cell type. SingleR should be able to report Unknown for cells which have little correlation to any of the cell types in the reference database.