Single-cell–level analysis of bin2cell in high-resolution Visium HD spatial transcriptomics and its compatibility with RCTD

[yaolei-6]

I am currently using bin2cell to analyze Visium HD data and then applying RCTD for deconvolution. However, I have observed that nearly 70% of the spatial locations are dominated by a single cell type in the deconvolution results.

After reviewing the comment you shared, I am wondering whether this suggests that, for high-resolution spatial data such as Visium HD, bin2cell is more suitable for reconstructing single-cell–level objects followed by direct cell-type annotation, rather than performing deconvolution.

I would appreciate your guidance or recommendations regarding best practices for applying bin2cell and downstream analysis strategies in Visium HD data, as I am currently encountering this issue.

[nadavyayon]

hello, yes indeed it also highly depends on the tissue type / section thickness / FFPE vs OCT and other factors
But in all the above, a 2um pixel resolution in xy and 5-10um in z will lead to mixed profiles. Even in Xenium which is around 200nm resolution there are mixtures of transcripts between cells. In our hands, deconvolution is definitely beneficial in specific cases. I did not try RCTD but we did use TACCO https://www.nature.com/articles/s41587-023-01657-3