search

powellgenomicslab / Nebulosa

R package to visualize gene expression data based on weighted kernel density estimation
89 stars 8 forks source link

Can Nebulosa be used from seurat object normalized via NormalizeData() instead of SCTransform()? #18

Open levinhein opened 2 years ago

levinhein commented 2 years ago

Hello! From this tutorial, the steps start with creating the seurat object, where the normalization step is via SCTransform.

Can I use a seurat object that is already normalized, clustered, cell-identified, and fully processed? When I skip the pre-processing and go directly to use plot_density(), the plot shows a PCA plot, whereas if I use the FeaturePlot, it shows a UMAP plot.

How can I make it so that the already-processed seurat object will show the density plot with already-clustered UMAP visualization?

https://bioconductor.org/packages/devel/bioc/vignettes/Nebulosa/inst/doc/nebulosa_seurat.html#

joseah commented 2 years ago

Hi @levinhein, you can calculate the density from NormalizeData() as you mention. By default, Nebulosa takes the latest reduction stored in the Seurat object. I believe Seurat uses a different strategy to obtain the "default" reduction. For consistency, Nebulosa should adopt this same approach to avoid the issue you're experiencing. In the meantime, you will need to explicitly provide the reduction name you want to plot via the reduction parameter in the plot_density() function. I believe what you're looking for at the moment is plot_density(..., reduction = "umap"). Cheers.