Pseudo-bulk DEG analysis with Multi-tissue in scRNA-seq

Question

[joonhong kwon]

Hi all,

I have 4 datasets (Brain region A (Control & Disease), Brain region B (Control & Disease)) and performed scRNA-seq by integrating all data.

And I want to compare Disease and Control in each brain region using pseudo-bulk DEG analysis. (ex. Region A Disease vs Region A Control)

Even if area A and area B are integrated, is it ok to perform DEG analysis for each brain region with the code below?

dds <- DESeqDataSetFromMatrix(cluster_counts, 
                            colData = cluster_metadata, 
                            design = ~ group_id)
contrast <- c("group_id", "Region A Disease", "Region A Control")
### or contrast <- c("group_id", "Region B Disease", "Region B Control")
res <- results(dds, 
             contrast = contrast,
             alpha = 0.05)

group_id contains 'Region A Disease', 'Region A Control', 'Region B Disease', and 'Region B Control'.

Or should I do it another way?

Thanks in advance,

Joonhong

[ATpoint]

Are these regions from the same experiment or is this independent studies?

[joonhong kwon]

Regions A and B are from the same patient (individual).

[ATpoint]

So then it should be possible to quantitatively compare them. But you mention pseudobulk...if it is the same patient, isn't it then 1 vs 1 so no replication as required for pseudobulks?

[joonhong kwon]

I have a total 24 samples from 12 individual (6 Region A Disease, 6 Region A Control, 6 Region B Disease and 6 Region B Control).

[ATpoint]

Then it should be fine. I usually do pseudobulk in these situations to avoid overly much (but meaningless) statistical power due to the large number of cells, and rather focus on the meaningful biological replication which is the individuals.

[joonhong kwon]

Thank you for your help.