Hi everyone,

I am working on a small pilot study using single-nucleus RNA-seq in an Activity-Based Anorexia (ABA) mouse model. We have a limited number of samples (5 hypothalamic punches pooled into an ABA sample and 5 pooled into a control sample).

We initially started from the raw Cell Ranger output and applied cell calling, QC filtering, and doublet removal. After these steps, the number of nuclei decreased substantially (approximately 22,000 nuclei per sample down to about 8,000).

I have a few questions:

1. In a pilot study with limited biological replication, how would you evaluate whether the QC pipeline is removing mostly low-quality nuclei versus potentially removing biologically relevant signal?

2. How much cell loss after QC and doublet removal would you consider typical or acceptable for snRNA-seq data from hypothalamic tissue?

3. Would it be reasonable to perform a sensitivity analysis by running the same downstream workflow on both:

   • the raw matrices
   • the filtered matrices,

   and then compare cell type composition, clustering, and major biological findings? or would you pick one over the other

4. If the major conclusions differ substantially between the two approaches, how would you determine which preprocessing strategy is more appropriate?

Any thoughts or recommendations would be greatly appreciated.