How to plot the heat map of a subset of cell types using plotHeatmap from scater package

Question

[FantasticAI]

Hi

I'm using plotHeatmap (library(scater)) to visualize the accuracy of cell type annotation output from SingleR. In total I got 28 labels, but as you can see on the top of the plot, the color different is hard to distinguish due to too many labels. I wonder whether there is a simple way to show only 5 to 7 labels in this plot, this is because I only care some specific cell types only, not all of them. For example "AT1 cells, AT2 cells and B cells" are the three cells I'm interested. How to only plot for only three labels?

 Library(scater)  
 plotHeatmap(sce.data.combined.sct, order_columns_by = "labels", features=feature)

[FantasticAI]

One of the potential solution is to subset from the sce object based on the ColData = "AT1 Cells, AT2 Cells, B Cells", then use plotHeatmap function. But how to subset the columns that are associated with those three cell types?

Answer 1

[jared.andrews07]

Gonna cheat and just recommend dittoSeq. To do what you want will be something like:

dittoHeatmap(sce.data.combined.sct, genes = feature, order.by = "labels", annot.by = "labels", 
  cells.use = sce.data.combined.sct$labels %in% c("AT1 cells", "AT2 cells", "B cells"), scale = "none")

You may also be interested in setting scaled.to.max = TRUE, which is useful for single-cell data given their enrichment for 0s. Setting scale = "row" will plots z-scores instead. I also usually pass complex = TRUE and use_raster = TRUE with single cell data, as it makes the resulting image much easier to edit in Illustrator or such.

[FantasticAI]

Thank you for the great answer! It works really well and what a powerful package it is, thank you.