Number of reads per gene in pseudobulk with more or fewer nuclei

Question

[WouterDeCoster]

Hi,

We are planning an experiment with single nuclei sequencing of a large cohort, but are wondering about the best choice of number of nuclei to sequence. Specifically, the number of reads we will sequence is fixed per sample (budget reasons) - so loading more nuclei will reduce the number of reads per nucleus.

If we were to use pseudobulk (or do you recommend anything else?) for the analysis, would it matter to have deeper coverage for 2500 nuclei of type A or more shallow coverage for 5000 nuclei of type A before summing the read counts? Unless I oversee something, it would still be sampling molecules from a library and the final gene count distribution would average out towards the same distribution.

Does that make any sense?