Quantify transcriptomic shift between two conditions for each cluster

Question

[Wocher33]

Hi all,

I have scRNAseq samples from 2 conditions (control and drug-treated). If I cluster them and plot them on a UMAP, I clearly see that some cell clusters from the drug-treated sample show a strong shift and cluster seperatly, whereas others are not affected by the drug-treatment and nicely cluster with the control cells. Besides the visualization I would like to quantify this transcriptomic shift and assess a "score" of how strongly each cluster is affected by the drug.The cells in my sample are heterogenous in cell number and cell size(amount of genes dected per cell size), so just taking the amount of DEGs wont work since the amount of DEGs is heavily influenced by these factors. Is there a way to do this?

Cheers and thanks for thinking about my question.

[Wocher33]

For people looking at this question 1 year later. Differential abundance testing is a good way to go. The miloR package does a wonderful job.

Answer 1

[bigomics.team]

Yes. Some clusters are smaller/bigger so you cannot really compare their p-values. I would resample the same number of genes (say 100) for each group and do the test on these equal groups.

[Wocher33]

Sorry, for the late reply. The problem I face is, that the cells have different amount of expressed genes. Eg. cell cluster 1 has cells that have tipically 8,000 expressed genes, whereas cluster 2 has 5,000 expressed genes. If I just resample the genes, I will miss out on what is happening in the rest of the 3,000 genes or genes that are specifically expressed in one of the clusters.

[bigomics.team]

Could you please share your UMAP plot? How many cells do you have in your samples? Can you explain better what do you mean by cell size?