Using scRNA-seq to draw concrete evidence about transitional cluster

Question

[odi]

Hi all! In my research, i suspect that there is a transitional cell type in the organ that i am studying. Now, i have gone through the process of single cell analysis and my dimensionality reduction plot (UMAP) display a cluster that could potentially be this cell type... right now i have it as unknown. This transitional cell type clusters between cell type A and cell type B. Considering we are saying that this transitional cell type exists as a result of travel from cell type A to B; the transitional cell type is in the middle. Our clustering seems to show this. Our gene expression profile also seems to show the transitional cluster expressing both cell type A and B genes. However, i know this is not concrete enough to define this as a transitional cluster. I am new to single cell so i would love some suggestions. Right now, i am stuck on whether the gene profile expression should be 50% from Cell type A and 50% from cell type B for it to be transitional? But that doesn't sound right... will trajectory analysis help or even i am thinking RNA velocity analysis?

Please all suggestions would be helpful!

Answer 1

[yura.grabovska]

Yes you could look at velocity using something like velocyto + scvelo, or you could try PHATE analysis which will allow you to construct expression trajectories in pseudotime. One caveat is that if you're dealing with specific cell niches, velocity methods can sometimes be confounded by quiescent or senescent cell populations

It's important to note that single-cell expression analysis won't definitively prove the existence of a transitory cell type. Because a single-cell library is a snapshot of cells in time you have vague information on transcriptional plasticity and spatial organisation of cells - eg. cycling cells clustering together as a single cluster/having a single metaprogram when in reality a random proportion of dividing cells is undergoing cell cycle transition across the sample. Proving a discrete state in this way is difficult with scRNAseq alone.

To prove the existence of transitional clusters you probably would need to use other modalities like spatial, or barcoding the cells to show that an individual cell can transit from A to B. If you're studying cancer, you could maybe look at copy-number to see if there's a clonal copy number pattern.

[odi]

Thank you for your response. I definitely plan to use trajectory/pseudotime analysis and maybe make some diffusion map plots. But the spatial would be great. We actually recently got a machine for that so hopefully in 3 months i can get that up and running.