BANKSY unifies cell typing and tissue domain segmentation for scalable spatial omics data analysis
Date:
Talk Abstract I will introduce a spatial clustering algorithm named BANKSY, which applies to any spatial (RNA/protein/…) assay, scales to 2M cells, and accurately detects both cell types and spatial domains. BANKSY operates on the philosophy that ‘a cell is known by the company it keeps’ and works by embedding cells in a product space of neighborhood transcriptome features (expression mean, texture) and the cell’s own transcriptome. We will discuss various applications of this approach, ranging from spatial cell typing to domain segmentation, and even spatial batch correction.
Older talks on the same topics can be found here:
- Human Cell Atlas Asia 2021:
- Singapore Single Cell Network
You can find the talk slides here.
- Shyam Prabhakar’s talk at the Simons Institute:
You can also watch Shyam Prabhakar talk about this work at the Simons Institute seminar, or a shorter version at the STOC workshop here.