I am analyzing a single cell dataset of 38 samples of 11 individuals (6 cases and 5 controls). I did differential gene expression analysis using DESeq2. When I did a heat map plot before DE analysis I noticed that my pca plot (for pseudo bulk data) and Heatmap plot show that the cases and controls are not well separated; Heatmap plot shows that some of the controls are more correlated to cases rather than other controls.
I am not sure if I should remove those samples. In terms of the analysis I would like to know if there is any way to improve this situation? I did apply more strict threshold but it didn't work.
And I also would like to know if I should remove those samples? If I do so I should at least remove 4 to 5 samples which decrease the power of my study.
I really appreciate any comment on this!