SEraster

SEraster aggregates cellular information into spatial pixels to reduce computational resource requirements for spatial omics data analysis while maintaining analytical performance.

Key Features:

• Rasterization preprocessing: Reduces the number of spatial points in spatial omics datasets by aggregating single cells' gene expression or cell-type labels into equally sized pixels based on a user-defined resolution.
• SpatialExperiment integration: Built on the R/Bioconductor S4 class SpatialExperiment for compatibility with existing spatial omics analysis workflows.
• Downstream analysis compatibility: Can be incorporated with other packages for spatial omics analyses such as detecting spatially variable genes and characterizing cell-type spatial co-enrichment across length scales.
• Scalability to large datasets: Enables analysis of spatial omics datasets with over a million cells, demonstrated on a mouse pup dataset to identify tissue-level and cell-type-specific spatially variable genes and spatially co-enriched cell types.

Scientific Applications:

• Spatially variable gene detection: Preprocesses spatial omics data prior to spatial variable gene expression analysis to reduce computational requirements while maintaining high performance compared to other down-sampling approaches.
• Cell-type spatial co-enrichment analysis: Integrates with existing analysis tools to characterize cell-type spatial co-enrichment patterns across length scales.
• Tissue structure characterization: Identifies spatially co-enriched cell types that recapitulate expected organ structures in complex tissues.

Methodology:

SEraster implements a rasterization preprocessing approach where single cells' gene expression or cell-type labels are aggregated into equally sized pixels based on a user-defined resolution.