How to Analyze Subtypes Within a Rare Cell Population After Integration?

[baiquanCi]

Hi everyone,

I’ve performed integration across multiple scRNA-seq samples (from different individuals or experimental conditions) using Seurat’s standard integration workflow (FindIntegrationAnchors() → IntegrateData()). After clustering and annotating major cell types in the integrated dataset, I’m now interested in analyzing subtypes within a rare cell population (e.g., monocytes or cDCs).

Here’s the issue: this rare population contains very few cells in each sample, and when I subset the integrated object to just this population and then attempt to split by sample and re-integrate, the integration often fails — either due to insufficient anchors or poor-quality UMAP/clustering results.

My question is:
In this case, is it acceptable to just use the subset of cells from the original integrated object, set DefaultAssay to "integrated", and continue with PCA, clustering, and UMAP directly? Or is there a better strategy to handle rare populations when re-integration isn't feasible?
Any suggestions, best practices, or references would be very much appreciated!

Thanks in advance