Tool:KNetL versus tSNE and UMAP for dimensionality reduction in single cell RNA-seq

Question

[alireza.khodadadi.j]

KNetL map (pronounced like "nettle") is a new dimensionality reduction developed for scRNA-Seq. It's capable of zooming and shows more detail than tSNE or UMAP.

To see more details: https://www.biorxiv.org/content/10.1101/2020.05.05.078550v1

KNetL map is available through iCellR package:

• https://github.com/rezakj/iCellR (GitHub)
• https://CRAN.R-project.org/package=iCellR (CRAN)

How to run KNetL, tSNE and UMAP using iCellR:

my.obj <- run.knetl(my.obj, dims = 1:20, k = 100)
my.obj <- run.pc.tsne(my.obj, dims = 1:10)
my.obj <- run.umap(my.obj, dims = 1:10)

Here we use a k value of 100, but this might not be ideal for your data.

Important Note:

• Because KNetL has a very high resolution it's best to set the dims value to 20 (this usually works best for most data) and 10 for UMPA and tSNE.
• In order to zoom, use the k value. For data with more than 5,000 cells use a k value of about 400. A k value of 400 usually works for large datasets, but you can adjust it for your intended resolution. Just like a microscope, you need to zoom in and out to see the intended amount of details.