ENACT for Visium HD analysis

I explore the ENACT (https://doi.org/10.1101/2024.10.17.618905) preprint, which discusses segmentation and cell-type-calling methods for Visium HD data.

Transcript

1. ENACT: End-to-End Analysis of Visium High Definition (HD) Data Presented by Nick Eagles 11/22/2024 Kamel et al., bioRxiv, 2024 DOI: 10.1101/2024.10.17.618905
3. Bin-to-cell assignment methods How are bins that cover multiple cells handled? - Naive: drop them completely - Weight by area: assign part of expression to each cell based on the area of the bin they occupy - Weight by transcript: distribute transcripts differently based on expression profiles of each cell - Weight by cluster: distribute transcripts differently based on expression profiles of each cell’s cluster
4. Synthetic Visium HD Datasets 1. Xenium-based human colorectal cancer (below) Epithelial Immune Stromal 2. seqFISH+-based mouse small intestine (not shown) Pathologist-annotated cell types
5. Bin-to-cell method performance: Xenium Naive WB area WB tx WB cluster Naive WB area WB tx WB cluster Naive WB area WB tx WB cluster Precision Recall F1 score Nuclei Whole cell
6. Bin-to-cell method performance: seqFISH+ Precision Recall F1 score Naive WB area WB tx WB cluster Naive WB area WB tx WB cluster Naive WB area WB tx WB cluster
7. Bin-to-cell method runtime Naive WB area WB tx WB cluster Time (seconds per patch)
8. Which cell annotation method is best?
9. A B
10. A B
11. Comparison to bin2cell Polański et al., DOI: 10.1093/bioinformatics/btae546