Seurat integrated analysis - methods for cell-type specific responses

Question

[jomo018]

I am following Seurat vignette: immune_alignment with my own treated vs control data sets.

The vignette offers two methods for identifying differential expressed genes across conditions. One is based on comparing averaged expression between conditions and the other is based on FindMarkers which eventually computes fold-change between conditions. The vignette shows many genes common to both methods (per one specific cluster).

In my data sets, the resulting two sets of DE genes are quite different. What could be the reason for this? More generally, what is the core difference between methods?

In my case, for the cluster under examination, the number of cells mapped to control is larger by a factor of 7 compared with those mapped to treatment. Could this be a related factor?

[Haci]

Technically, they are not comparing DE results obtained with the averaged expression approach to those obtained with theFindMarkers approach. The former is performed for CD4 Naive T cells and CD14 Monocytes for stim vs ctrl comparison and the latter is used for B cells, again for stim vs ctrl. Nevertheless, all three DE lists, at a first glance, kind of overlap and is speculated to result from "a conserved interferon response pathway". Actually, the averaged expression approach is not a DE analysis attempt per se, as the authors of the vignette state, but is just "a way to look broadly at these changes" and does not provide a list of DE genes but just a scatter plot, which can be used to infer DE-related information. The FindMarkers() can be run using one of the 9 statistical tests, some of which take metrics like variance/dispersion and not just average expression, and therefore should be the way to go in terms of deciding the DE genes between cell clusters/types/states of interest.