How to find differences between single cell RNA seq samples beyond cluster information and differential expression?

Question

[tushard1987]

Consider a single cell RNA seq data with a control and a test sample. These are integrated to minimize the batch effect. Reads have been normalized to the sequencing depth. One can compare the samples for their cluster composition. Further one can perform cluster-wise differential expression followed by enrichment analysis to find out genes and pathways up or down regulated in the samples. One may also perform trajectory analysis to study transition between cell types.

Apart from this, are there any ways to compare and find differences between samples in a single cell RNA seq data? I am looking for methods/tools/parameters that can tell me differences if any between the samples particularly in the absence of any hypothesis (even if I have a hypothesis in mind, I would still be interested in unbiased way of finding differences between the samples)

[ATpoint]

I would take the findings that you have and try to build a hypothesis for downstream validation or functional testing. These single-cell applications are on the one hand very powerful because they provide a detailed look over a population of cells, but on the other hand there is a lot of noise that will smuggle into the results, and this accumulates the more analysis you stack on top of each other. Unless you want to publish a highly descriptive paper you will have to narrow down at some point. I would not believe that clustering, DE composition analysis does not suggest at least some interesting aspects that encourage you to dig deeper.

Answer 1

[andrew.j.skelton73]

Providing you have replication, one could test for proportional expansion / contraction of a given cell type. If you're confident in the biology and understanding of sample clusters (cell types), then you could explore differential ligand - receptor interactions with approaches like cellchat (paper, tool).

Ultimately, there are a number of approaches to analyse single cell data but at some point, you'll have to tie that to a hypothesis to progress. Single cell data gives really great resolution compared to historical bulk RNA experiments, but it shouldn't be treated as the end point, rather a steppingstone towards expanding an area of investigation which can be solidified in other independent experiments.