Single cell - integration after demultiplexing samples?

Question

[txema.heredia]

Hi,

I have a single-cell RNA-seq (10x) project. It is an analysis of WT vs mutant samples. The project's design is:

• 2 10x runs / batch. Each contining:
  • 1x WT + 1x mutant donors

I have successfully run donor demultiplexing on each of the separate runs. Using Seurat, I have also run integration of both runs/batches. However, there are some notable differences by genotype.

Should I run the integration step, not at the run/batch level but at the donor level?

Thanks

[bk11]

Your choice of batch covariate depends upon which level of variation you want to retain and which level be removed. You typically have samples from different subjects or various locations in the tissues. These effects can be very meaningful to study. Therefore, the choice of batch covariates will depend upon your data integration goal. You need to decide what variation you want to see: 1) differences between individuals or 2) Variations on common cell type?

[txema.heredia]

Thanks.

As far as I've seen in my data, there are wide differences in cell type composition between the two genotypes. And they are consistent in both batches/runs. I will try to re-do the integration using donor instead of batch to see how it fares.