How to perform scRNA seq analysis?

Question

[bioinfo]

Hello,

I am trying to analyze some single cell RNA seq data. I have two samples. One is from a wild type spleen and the other spleen has a gene knockout. I have some questions regarding the analysis.

1. From going through Seurat tutorials it seems like I would have to integrate the 2 samples. Should I first load the 2 datasets in seurat, do the QC analysis, normalize, FindVariableFeatures and then select the features for integration and integrate? Would it be appropriate to merge them first for the QC analysis? On the seurat tutorial for merging it says to normalize and identify variable features for each dataset independently so would merging the datasets cause an issue?

2. Would a VlnPlot or a FeaturePlot for the knockout gene be more appropriate to show that the knock out was successful?

3. I would like to do differential gene expression analysis to compare one cell type between wild type and knockout. Would it be better to use the FindMarkers functions in Seurat using the MAST test or would it be better to use pseudobulk?

Thank you

Answer 1

[rpolicastro]

1) With scRNA-seq integration you usually perform the cell QC for each sample separately, and then integrate afterwards.

2) I would use a violin+jitter plot at first, since it makes it easier to see if there are any outlying cells for the expression of your KO gene.

3) pseudobulk is generally better for differential expression in scRNA-seq because cells from the same sample are more appropriately thought of as technical as opposed to biological replicates. However, you need more than one sample in each group for pseudobulk to make sense. In your case it's probably better to just do a simple wilcoxon rank-sum test and validating important DEGs further computationally or experimentally.

[bioinfo]

Thank you for your help!