Spatial omics is a technique that studies molecular characteristics within the spatial context of intact tissues. It helps people understand how cells are organized and interact, which is crucial for understanding how the body works and how diseases develop. Typically, spatial omics is done on 2D tissue slices. However, these slices do not fully capture the complex molecular dynamics within the tissue, which is important for studying developmental biology and tumor microenvironment.
To get a complete picture, scientists align data from multiple 2D slices to create a 3D view of the tissue. This 3D view gives a better understanding of the tissue’s ecosystem. For example, one study used this technique to create 3D models of embryos and larvae, which helped identify different functional areas and analyze cell changes during development [1]. However, 2D data alone cannot show the molecular details above and below the slice, limiting our understanding of 3D cell interactions and organ development.
Aligning slices accurately is challenging because adjacent slices often do not cover the exact same area. Even small misalignments can distort the 3D molecular characteristics. Furthermore, researchers often work with large tissue samples, but current technology can only capture small areas (up to 15 cm²), making it challenging to study larger slices [2]. For instance, according to reports on breast cancer staging, tumors can be larger than 5 cm at stage T3, which surpasses the current capture limits of available technology [3].
To solve these problems, scientists use a method called tissue stitching, similar to how images are stitched together [4]. This involves combining partially overlapping tissue slices to create a larger, continuous slice [5]. From a technical standpoint, the stitching task can be treated as an extension of the alignment task, but it deals more effectively with technical displacements and enriches mutual information between slices. This method also allows combining data from different platforms, providing complementary spatial resolutions and genomic coverages, which is particularly useful in creating comprehensive tissue profiles.
The method we proposed is SANTO, a novel method for aligning and stitching spatial omics data. SANTO works in two steps: a coarse step to quickly find the initial positions of two slices and their overlapping regions, and a fine step to refine the alignment using detailed spatial and omics patterns. SANTO outperforms existing methods in various challenging tasks, including stitching slices from different platforms with different resolutions. It enables advanced analyses such as identifying new cell types, predicting gene expressions, and studying cell interactions in tumors. For example, by using two slices of breast cancer samples from different platforms (10x Xenium and Visium), SANTO successfully stitches these slices, leveraging their supercellular and subcellular resolutions to conduct downstream analyses. This includes identifying novel cell types, predicting undetected genes’ expressions, and understanding cell-cell communication within the tumor microenvironment.
SANTO also has potential applications beyond just stitching and aligning spatial omics data. It can be used for 3D-to-3D spatiotemporal alignment, as demonstrated in a study of mouse embryonic development. By aligning samples from different developmental stages, researchers can gain insights into tissue development over time. Additionally, SANTO can handle cross-modality alignment, such as aligning spatial transcriptomic and epigenomic data from mouse brain samples, to study interactions between different types of omics data.
Overall, SANTO offers robust solutions for alignment and stitching challenges of spatial omics. Its ability to integrate data from different platforms and resolutions, combined with its potential for future enhancements, positions it as a powerful tool for advancing our understanding of tissue ecosystems and improving clinical outcomes. SANTO is available at: https://github.com/leihouyeung/SANTO.
Paper link: https://www.nature.com/articles/s41467-024-50308-x
Figure 1. A. Overview of SANTO. B. SANTO can be applied to many scenarios, e.g., cross-platform stitching, 3D-to-3D spatiotemporal alignment, and cross-modality alignment.
Reference:
[1] M. Wang et al., “A single-cell 3D spatiotemporal multi-omics atlas from Drosophila embryogenesis to metamorphosis,” bioRxiv, p. 2024.02.06.577903, Jan. 2024, doi: 10.1101/2024.02.06.577903.
[2] A. Chen et al., “Spatiotemporal transcriptomic atlas of mouse organogenesis using DNA nanoball-patterned arrays,” Cell, vol. 185, no. 10, pp. 1777-1792.e21, 2022, doi: 10.1016/j.cell.2022.04.003.
[3] S. Paik, “Development and clinical utility of a 21-gene recurrence score prognostic assay in patients with early breast cancer treated with tamoxifen,” Oncologist, vol. 12, no. 6, pp. 631–635, 2007.
[4] X. Liu, R. Zeira, and B. J. Raphael, “Partial alignment of multislice spatially resolved transcriptomics data,” Genome Res., vol. 33, no. 7, pp. 1124–1132, 2023, doi: 10.1101/gr.277670.123.
[5] M. Brown and D. G. Lowe, “Automatic panoramic image stitching using invariant features,” Int. J. Comput. Vis., vol. 74, pp. 59–73, 2007.