Do you have an example of running Xenium DAPI with transcript information through this model

[abs51295]

Hey,

I really liked reading your manuscript that was published today. Can you give me an example of how to use DAPI staining data, cell membrane staining data and transcript measurement data from the xenium outputs to get cell segmentation working. I want to compare it with baysor for my dataset. Also, does it help to include membrane staining data along with DAPI and transcripts for cell segmentation?

Thanks

[maxpmx]

Hi,

Thank you for your interest in CelloType and for reading our manuscript! For Xenium data that includes membrane staining, we actually recommend using our DAPI+Membrane workflow. Here's why and how:

1. For Xenium data with membrane staining:

• Use the DAPI+Membrane workflow (https://cellotype.readthedocs.io/en/latest/inference_large.html)
• This approach is optimized for large images and provides single-cell level segmentation
• The membrane signal provides strong cell boundary information, making it more reliable than transcript-based segmentation

2. Regarding DAPI+membrane vs DAPI+transcripts:

• When membrane staining is available, we recommend using it
• Membrane staining provides direct visualization of cell boundaries
• DAPI+transcripts segmentation is currently only supported at the patch level images, rather than whole slide images

[abs51295]

Hey,

Thanks for the response. Unfortunately we don't have membrane staining data for Xenium. Can you provide an example of running the segmentation based on DAPI+ transcripts? We have TMA cores that we need to run segmentation on.

[Thapeachydude]

Hello,

do you perhaps have a comparison of CelloType vs the onboard Xenium DAPI + membrane staining?
Also since Xenium membrane staining comes with a cocktail of antibodies (saved as 3 separate images), which of them would you suggest to use for CelloType? The onboard Xenium segmentation seems to rely a lot on RNA 18S info, less the CD45 etc.

Cheers

[maxpmx]

Hello,

Yes, we have internally compared CelloType results with the Xenium onboard segmentation. From our analysis, CelloType’s segmentation results show better consistency with the actual cell morphology visible in the images.

For CelloType, we recommend using:

• morphology_focus_0000.ome.tif (DAPI image) as the nuclear channel
• morphology_focus_0001.ome.tif (ATP1A1/E-Cadherin/CD45 combined) as the membrane channel

[Thapeachydude]

Thanks a lot!