BANKSY unifies cell typing and tissue domain segmentation for scalable spatial omics data analysis

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Spatial omics technologies generate high-resolution data on cellular composition and spatial organization within tissues.
• Current methods often analyze cell types and tissue domains separately, limiting integrated insights.
• Accurate analysis of spatial omics data is crucial for understanding complex biological systems.

Purpose of the Study:

• To develop a unified computational framework for analyzing spatial omics data.
• To integrate cell state and microenvironment information for improved clustering and domain segmentation.
• To provide a scalable and efficient tool for processing large-scale spatial omics datasets.

Main Methods:

• Developed BANKSY (Building Aggregates with a Neighborhood Kernel and Spatial Yardstick), an algorithm that embeds cells in a product space.
• The product space represents individual cell transcriptomes and their local neighborhood transcriptomes.
• Applied spatial feature augmentation to RNA (imaging, sequencing) and protein (imaging) datasets.

Main Results:

• BANKSY improved performance in both cell typing and domain segmentation tasks across diverse datasets.
• Identified novel niche-dependent cell states in the mouse brain.
• Demonstrated superior performance compared to existing methods on benchmark datasets.
• Showcased utility in quality control and batch effect correction for spatial transcriptomics.

Conclusions:

• BANKSY offers an accurate, biologically motivated, and versatile framework for spatial omics analysis.
• The algorithm is scalable, enabling analysis of millions of cells.
• BANKSY advances the integrated analysis of cell state and tissue microenvironment.