search

mahmoodlab / HEST

HEST: Bringing Spatial Transcriptomics and Histopathology together - NeurIPS 2024
Other
149 stars 12 forks source link

Xenium datasets coordinates #46

Closed flyfyyfyy closed 1 week ago

flyfyyfyy commented 1 month ago

Hi,

The Xenium are imaging-based ST technology so that the Xenium dataset's spatial coordinates are not supposed to be grid. Why does the TENX99 dataset have the grid coordinates.

I check the raw data from 10x genomics replicate 1, https://www.10xgenomics.com/datasets/ffpe-human-breast-using-the-entire-sample-area-1-standard. The cells matrix in raw data.

image

And I only get 280 genes for each spot and 892965 spots in raw data. But 23060 spots in HEST TENX99. So, how do you guys pre-process the Xenium dataset here?

guillaumejaume commented 1 month ago

Hi @flyfyyfyy,

HEST-1k was initially designed to match H&E crops and ST spots. Xenium transcripts have been pooled into pseudo-Visium (squared spots). As it seems that many people are interested in Xenium, we will be releasing aligned H&E with Xenium/DAPI at cell-level + the list of all transcripts.

Having perfect alignment takes time as Xenium explorer tools don't scale well. We are using a sota whole-slide registration instead. More information here: https://github.com/mahmoodlab/HEST/issues/38.

The release is expected for next Friday.

kennethahah commented 1 month ago

Hi @guillaumejaume

Which sota whole-slide registration tool did you use to align the HE image with Xenium?

guillaumejaume commented 1 month ago

A slightly modified version of Valis, https://github.com/MathOnco/valis.

pauldoucet commented 1 month ago

Hi @flyfyyfyy, @kennethahah,

The aligned transcripts are now available on HF along with 121 new samples.

We apologize for coming back to you this late, registration took more time than expected.

guillaumejaume commented 1 month ago

Tutorial has been updated: https://github.com/mahmoodlab/HEST/blob/main/tutorials/2-Interacting-with-HEST-1k.ipynb