Page 1 of 1

Differential Expression between cell population

PostPosted: Fri Jan 08, 2021 8:17 am
by gilgifri
Hello,
I am analyzing a CyTOF experiment with two biological conditions, each has 3 replicates. After clustering the cells, there are ~10 cell populations.
Does anyone has recommendations for a tool or for a statistical test for differential expression between cell populations (each cell population includes 2 conditions, and 3 replicates for each condition)?
Thank you

Re: Differential Expression between cell population

PostPosted: Sun Jan 10, 2021 9:40 pm
by sgranjeaud
Hi,
IMHO the most rigorous approach is the diffcyt package as featured in the workflow of Nowicka. Look at the Differential analysis paragraph in the middle of the article. I would advice you to read the article carefully in order to understand the whole workflow.
As this experiment ended up with 10 pop, I feel like it does not exploit the discovery power of CyTOF.
As the design consists in only 3 replicates per condition, I feel like this experiment is under powered. How much confidence do you have in an estimation based on 3 measures?
Best,
Samuel

Re: Differential Expression between cell population

PostPosted: Mon Jan 11, 2021 8:06 am
by MCOlivier
Hi ^^
You could also try MEM, which highlights the differentially expressed markers beetween clusters.
https://github.com/cytolab/mem

Best,
Olivier

Re: Differential Expression between cell population

PostPosted: Mon Jan 11, 2021 3:04 pm
by gilgifri
Thank you for the replies!!
Actually I used the CyTOF workflow of Nowicka et al. (https://pubmed.ncbi.nlm.nih.gov/28663787/). However, it supports only differential expression between conditions, inside a cell population. I am interested also in comparing expression between different cell populations.
I will look also into MEM.
Thank you

Re: Differential Expression between cell population

PostPosted: Mon Jan 11, 2021 4:24 pm
by markrobinsonca
@gilgifri: Just to say that while the Nowicka workflow doesn't do DE between subpopulations/clusters, it does put the data in a SingleCellExperiment object, which allows you to directly use, for example, the findMarkers() function from the scran package. For example, since you likely have multiple samples, you can add sample as a blocking factor.

More details here:
https://rdrr.io/bioc/scran/man/findMarkers.html

Best, Mark