Construction of a 3D whole organism spatial atlas by joint modeling of multiple slices

Abstract

Spatial transcriptomics (ST) technologies are revolutionizing the way that researchers explore the spatial architecture of tissues. Currently, ST data analysis is often restricted to 2D space within a single tissue slice, limiting our capacity to understand biological processes that take place in 3D space. Here, we present STitch3D, a unified computational framework that integrates multiple 2D tissue slices to reconstruct 3D cellular structures from the tissue level to the whole organism level. By jointly modeling multiple 2D tissue slices and integrating them with cell-type-specific expression profiles derived from single-cell RNA-sequencing data, STitch3D simultaneously identifies 3D spatial regions with coherent gene expression levels and reveals 3D distributions of cell types. STitch3D distinguishes biological variation among slices from batch effects, and effectively borrows shared information across slices to assemble powerful 3D models of tissues. Through comprehensive experiments using diverse datasets, we demonstrate the performance of STitch3D in building comprehensive 3D tissue architectures of the mouse brain, the human heart, and the Drosophila embryo, which allow 3D analysis in the entire tissue region or even the whole organism. To gain deeper biological insights, the outputs of STitch3D can be used for downstream tasks, such as inference of spatial trajectories, identification of spatially variable genes enriched in tissue regions or subregions, denoising or imputation of spatial gene expressions, as well as generation of virtual tissue slices.

A single-cell 3D spatiotemporal multi-omics atlas fromDrosophilaembryogenesis to metamorphosis

“…Moreover, the unique organization of our datasets can serve as a source of inspiration for the development of multiple types of bioinformatic algorithms and methods and can serve as benchmarking resources for such algorithms. The organism-wide 3D high-resolution features of our previous Stereo-seq datasets have already facilitated the development of several approaches for various purposes, including quantitative spatiotemporal modeling of single-cell spatial transcriptomic datasets 14 , visualization and analysis of spatial omics data 114 , construction of databases and optimization of their accessibility 115 , alignment of 2D spatial transcriptomic sections for 3D modeling 116,117 , and more. The new dataset features in this study can further assist in the development of bioinformatic approaches in many other aspects, such as cell segmentation of spatial transcriptomic data, integration of multi-omics data, spatial mapping of cell types, machine learning-based cell type and age prediction, and cell lineage tracing, among others.…”

Navigating the landscapes of spatial transcriptomics: How computational methods guide the way

“…This reduces batch effects and improves the performance of spatial deconvolution and clustering tasks. Similarly, STitch3D (Wang, Zhao, et al, 2023) and SpatialPrompt (Asish Kumar et al, 2023) also integrate consecutive slices and perform spatial domain recognition and cell‐type deconvolution tasks. However, they use different architectures: graph attention networks for STitch3D and non‐negative ridge regression for SpatialPrompt.…”

Spatially Resolved Transcriptomics Mining in 3D and Virtual Reality Environments with VR-Omics

“…3d). While tools for working with z-stacked SRT datasets in 3D such as STich3D 30 are available, it is more limited in its functionality than VR-Omics. It has currently only demonstrated compatibility with sequencingbased SRT platforms and does not offer a GUI 20 .…”

“…Users will require knowledge of Python to run the bioinformatics analysis of datasets prior to visualisation within STich3D, a task handled in its entirety by the AW in VR-Omics. VR-Omics offers additional levels of analysis such as identification of SVGs within the AW while STitch3D requires the user to output the results and load them into a separate package for analysis 30 .…”