Proseg missing cells along the border

[james-bole-pan]

Hi there,

I'm running Proseg on my Xenium data using the default settings:

proseg transcripts.csv.gz --xenium

The tool has been working great overall—it's significantly improved my UMAP clustering results (incredible tool, thank you!).

However, I noticed that a subset of cells along the upper border of the view (shown in grey in the image below) are missing from the cell-metadata.csv.gz output, even though their cell_ids are present in the original transcripts.csv.gz file. I'm wondering if this might be due to a boundary-handling setting or filtering step within Proseg.

Is there a parameter I should adjust to ensure that all cells, including those at the edges, are included in the output?

Thanks so much for your time and for developing such a powerful tool!

[dcjones]

Thanks for tying proseg. It's not obvious to me what's going wrong with your data, I'm afraid. Proseg will by default filter out transcripts that have a qv score lower than 20, and transcripts that are very far from any cells.

I suppose the qv score filtering could explain this if fovs at the top of the slide had some kind of issue. You can try --min-qv 0 to avoid that filtering.

It will also exclude cells from the input that have zero overlapping transcripts. That might also explain this if no transcripts were generated for the cells in these upper regions.