I am working with two drop-seq datasets that I would like to integrate using MNN_correct and then perform DE, reproducing the method presented in the MNN paper. I would appreciate your opinion on my rationale.
I have processed both datasets and then corrected using mnn_correct. I found the clusters on the corrected data and then I identified the cell type of my interest based on the corrected values for its gene markers. Then, I gave a cell-type label for my cell type of interest.
I then subsetted my data for the cell type of interest and created a factor merging the original study variable (GSE1, GSE2) X the feeding condition variable (Fed, FD). Reproducing the logic of the paper. My aim is to get the effect of food deprivation for this cell type using all the cells, then I would make two contrasts always comparing the cells inside the same dataset: GSE1.FedvsGSE1.FD, GSE2.FedvsGSE2.FD.
I intend to use Deseq2, which is more suitable for dropseq data. I am not sure if will be able to get the means of the two contrasts in Deseq2 the same way it was done in Limma by the authors, I'm still trying to figure that part out. If it's not possible, do you guys think it would be valid to get the average of the two DE lists?
What do you guys think so far? Would be valid to get the mean of two independent GSEs since the cells of interest were identified with the MNNcorrected data? I would like to hear your thoughts on how I could proceed to perform DE for the feeding condition for this cell type in two GSEs.
Thanks a lot, All the best.