Hi,
I’ve been exploring punkst for visualizing MERFISH/Xenium-style spatial transcriptomics data, and I noticed something I want to clarify.
In the FICTURE paper and documentation, there is a map command that allows users to integrate a reference scRNA-seq dataset (e.g., Allen Brain Atlas h5ad) with a query dataset (MERFISH/Xenium counts):
ficture map
--reference /path/to/reference_cell_types.h5ad
--query /path/to/MERFISH_counts.csv
--output /path/to/output_directory
--n-processors 8
This is useful for label transfer / reference-guided mapping, where query pixels/cells are assigned to known cell types.
My question is:
Does punkst include or plan to include a similar map function?
Or is punkst meant to focus purely on visualization, with the assumption that mapping/classification should be done with external tools (e.g., FICTURE, Seurat, scANVI, cell2location), and then the results imported into punkst (via cells.csv and transcripts.csv)?
If punkst does not currently handle reference-based mapping, are there any recommended workflows for combining punkst with FICTURE or other tools?
Thanks for clarifying — I want to make sure I’m not missing a hidden feature, and it would help other users understand the division between analysis and visualization.
This is part of an effort to do supervised learning through punkst.
Thanks
Hi,
I’ve been exploring punkst for visualizing MERFISH/Xenium-style spatial transcriptomics data, and I noticed something I want to clarify.
In the FICTURE paper and documentation, there is a map command that allows users to integrate a reference scRNA-seq dataset (e.g., Allen Brain Atlas h5ad) with a query dataset (MERFISH/Xenium counts):
ficture map
--reference /path/to/reference_cell_types.h5ad
--query /path/to/MERFISH_counts.csv
--output /path/to/output_directory
--n-processors 8
This is useful for label transfer / reference-guided mapping, where query pixels/cells are assigned to known cell types.
My question is:
Does punkst include or plan to include a similar map function?
Or is punkst meant to focus purely on visualization, with the assumption that mapping/classification should be done with external tools (e.g., FICTURE, Seurat, scANVI, cell2location), and then the results imported into punkst (via cells.csv and transcripts.csv)?
If punkst does not currently handle reference-based mapping, are there any recommended workflows for combining punkst with FICTURE or other tools?
Thanks for clarifying — I want to make sure I’m not missing a hidden feature, and it would help other users understand the division between analysis and visualization.
This is part of an effort to do supervised learning through punkst.
Thanks