Smaller | Cell types → Cell modules - cell neighborhoods → Niches/tissue domain | Larger
Other ressource → https://github.com/crazyhottommy/awesome_spatial_omics
- This paper recommends cell volume normalization for imaging-based techniques, especially when the list of probes is small
- Best practices Bioconductor
- scverse / squidpy
- Giotto suite
- Vitessce
- Voyager from Pachter lab
- Multiple sample analysis with BASS
- SpaVAE to do everything: reduction, visualization, clustering, batch integration, denoising, differential expression, spatial interpolation, and resolution enhancement.
- Baysor
- Cellpose to tune to highly specific system with few training examples
- Able to do supersampling / restoration with cellpose 3
- DeepCell
- Best than SOTA segmentation - Nature Methods 2024 - Bo Wang
- Bin2Cell to segment VisiumHD data
- Proseg
- ComSeg uses transcripts to segment
- FICTURE
- https://bioimage.io/#/ segmentation repo
- Detection of SVGs with PROST
- BANKSY unifies cell typing and tissue domain segmentation for scalable spatial omics data analysis
- staryfish
- TISSUE
- CellCharter. Looking at the most stable solutions. There is a hierarchy of niches at different layers/hierarches.
- SpatialGLUE for cell niches with multi-omics
- smoothclust from lukas weber
- SpaTopic
- PASTE/PASTE2
- Super resolution 10X Visium with TESLA
- Super resolution Visium with istar
- Subspot resolution with BayesSPACE
- Doing gene imputation is not recommended
- Sprod
- https://www.biorxiv.org/content/10.1101/2024.04.03.586404v1.full
- Benchmark claims that RCTD and Cell2location are the best
- Benchmark on spatial clustering
- Other review here, Cell2location again
- Cell2location is the best in the open problems
- https://huggingface.co/owkin/phikon-v2
- Bioptimus: https://huggingface.co/bioptimus/H-optimus-0
- Interesting datasets embedding CNN + human features from DeepCell https://exploredata.deepcell.com/cell-visualizations/9/versions/1