Integrating gene expression and imaging data across Visium capture areas with visiumStitched

Abstract

AbstractBackgroundVisium is a widely-used spatially-resolved transcriptomics assay available from 10x Genomics. Standard Visium capture areas (6.5mm by 6.5mm) limit the survey of larger tissue structures, but combining overlapping images and associated gene expression data allow for more complex study designs. Current software can handle nested or partial image overlaps, but is designed for merging up to two capture areas, and cannot account for some technical scenarios related to capture area alignment.ResultsWe generated Visium data from a postmortem human tissue sample such that two capture areas were partially overlapping and a third one was adjacent. We developed the R/Bioconductor packagevisiumStitched, which facilitates stitching the images together withFiji(ImageJ), and constructingSpatialExperimentR objects with the stitched images and gene expression data.visiumStitchedconstructs an artificial hexagonal array grid which allows seamless downstream analyses such as spatially-aware clustering without discarding data from overlapping spots. Data stitched withvisiumStitchedcan then be interactively visualized withspatialLIBD.ConclusionsvisiumStitchedprovides a simple, but flexible framework to handle various multi-capture area study design scenarios. Specifically, it resolves a data processing step without disrupting analysis workflows and without discarding data from overlapping spots.visiumStichedrelies on affine transformations byFiji, which have limitations and are less accurate when aligning against an atlas or other situations.visiumStichedprovides an easy-to-use solution which expands possibilities for designing multi-capture area study designs.