Orchestrating spot and cellular resolution Spatial Transcriptomics workflow with Bioconductor – SpatialExperimentIO, STexampleData 2.0, and ggspavis 2.0
Author(s): Yixing Dong,Lukas M Weber
Affiliation(s): Biomedical Data Science Center - Lausanne University Hospital
The recent advancements in spatial transcriptomics (SRT) have been primarily catalyzed by innovative technologies. Among these, Visium CytAssist (10X Genomics) and imaging-based SRTs like Xenium (also by 10X Genomics) offer varying resolutions, catering to both spot and cellular levels of analysis, respectively. Despite these breakthroughs, a noticeable gap exists within the Bioconductor community on essential infrastructure for reading, accessing, and visualizing spatial datasets. During this presentation, we showcase newly devised data-wrangling methodologies aimed at streamlining SRT workflows across spot and cellular resolutions. Leveraging datasets from 10X Genomics as published by Janesick et al. (2023), which include computationally annotated Chromium, Visium, and two Xenium replicates as adjacent sections, we illustrate the visualization of biological insights using established Bioconductor packages such as BayesSpace, RCTD (work in progress), and SingleR. Our initial step involves the development of a reader package called SpatialExperimentIO, designed to facilitate the retrieval of new imaging-based spatial transcriptomics technology data in the form of SpatialExperiment objects. These example datasets are conveniently housed in the updated STexampleData package available on ExperimentHub. For visualization purposes, we have refined existing ggspavis functions, incorporating zoom-in capabilities to explore regions of interest, as well as on-plot text annotation functionalities. Additionally, enhancements have been made to Visium visualization techniques, including the overlay of empty spots or scatter pie representations of cell-type deconvolution on H&E stain images. We have also expanded the range of built-in color palettes to ensure better contrasts. Furthermore, new visualization functions for quality control and dimensionality reduction have been introduced to improve the completeness of the overall analytical workflow. Slides available at: https://f1000research.com/slides/13-393