SEraster: a rasterization preprocessing framework for scalable spatial omics data analysis

Abstract

MotivationSpatial omics data demand computational analysis but many analysis tools have computational resource requirements that increase with the number of cells analyzed. This presents scalability challenges as researchers use spatial omics technologies to profile millions of cells.ResultsTo enhance the scalability of spatial omics data analysis, we developed a rasterization preprocessing framework called SEraster that aggregates cellular information into spatial pixels. We apply SEraster to both real and simulated spatial omics data prior to spatial variable gene expression analysis to demonstrate that such preprocessing can reduce resource requirements while maintaining high performance. We further integrate SEraster with existing analysis tools to characterize cell-type spatial cooccurrence. Finally, we apply SEraster to enable analysis of a mouse pup spatial omics dataset with over a million cells to identify tissue-level and cell-type-specific spatially variable genes as well as cooccurring cell-types that recapitulate expected organ structures.Availability and implementationSource code is available on GitHub (https://github.com/JEFworks-Lab/SEraster) with additional tutorials athttps://JEF.works/SEraster.

Gene count normalization in single-cell imaging-based spatially resolved transcriptomics

“…We used nnSVG with SEraster preprocessing to identify SVGs after gene count normalization with the ventricle-skewed gene panel and compared the results to those obtained with the full gene panel (Fig. 5 A) [ 27 , 28 ]. Comparing p -values with the ventricle-skewed gene panel to those with the full gene panel, we find that while there is some discordance in p -values without normalization and with cell volume normalization, likely reflective of the stochasticity of the nnSVG algorithm, this discordance is more pronounced with the evaluated count-based normalization methods.…”

“…To identify spatially variable genes after gene expression normalization, we first rasterized the normalized gene expression data using SEraster for computational tractability, using a bin size of 50 μm, and averaging counts within each bin [ 28 ]. Rasterized gene expression was then log 10 transformed with a pseudocount of 1.…”

Gene count normalization in single-cell imaging-based spatially resolved transcriptomics

“…To investigate the impact of different normalization methods with skewed gene panels on downstream analyses specific to SRT data, we sought to identify spatially variable genes (SVGs), genes with highly spatially correlated expression patterns. We used nnSVG with SEraster preprocessing to identify SVGs after gene count normalization with the ventricle-skewed gene panel and compared the results to those obtained with the full gene panel (Figure 5A) [27], [28]. Comparing p-values with the ventricle-skewed gene panel to those with the full gene panel, we find that while there is some discordance in p-values without normalization and with cell volume normalization, likely reflective of the stochasticity of the nnSVG algorithm, this discordance is more pronounced with the evaluated count-based normalization methods.…”

“…To investigate the impact of different normalization methods with skewed gene panels on downstream analyses specific to SRT data, we sought to identify spatially variable genes (SVGs), genes with highly spatially correlated expression patterns. We used nnSVG with SEraster preprocessing to identify SVGs after gene count normalization with the ventricle-skewed gene panel and compared the results to those obtained with the full gene panel (Figure 5A) [27], [28].…”

SpotSweeper: spatially-aware quality control for spatial transcriptomics

“…While SpotSweeper currently uses multiscale variance of mitochondrial ratio to detect these artifacts, it is possible that a similar approach utilizing the negative control genes normally included in image-based methods may be useful for detecting damaged tissue sections. In addition, rasterization techniques that aggregate mRNA counts into spatial pixels [ 32 ] will increase compatibility with the current SpotSweeper workflow, while ensuring the scalability of our approaches for imaging-based platforms. Moreover, the current implementation of SpotSweeper should be amenable to future spot-based technological advancements, such as the VisiumHD platform from 10x Genomics [ 33 ], that have substantially increase spatial resolution with complete tissue coverage.…”