HD Visium: How to compare cancer cells between samples?

[pikapika505]

I’m working with spatial transcriptomics data generated using the HD Visium platform. I have four samples: A, A1, D, and D1. A and D were processed together on one slide, and A1 and D1 were processed together on another slide. In each sample, I analyzed the data separately using the Seurat V5 pipeline and manually annotated cancer cell clusters (confirmed via H&E staining), so I’m confident about the cancer cell labels.

Now, I need to identify differentially expressed genes between cancer cells from A/A1 (primary cancer) and D/D1 (metastatic cancer).

My question: How should I approach differential expression analysis when I already know which cells are cancer cells in each sample?

I feel like I shouldn’t need to integrate the datasets or apply batch correction, since I’m comparing only cancer cells and their labels are confidently assigned. But at the same time, it feels wrong to directly compare cancer cells from different samples without any integration or correction step, especially considering that the samples were processed on different slides and days.

Has anyone faced a similar situation? Is pseudobulk the recommended way here? I’d really appreciate any advice or shared experience.