Fig. 1: Overview of Sopa.
From: Sopa: a technology-invariant pipeline for analyses of image-based spatial omics
a The pipeline input consists of experimental files of any image-based spatial omics. It is transformed into a SpatialData object, on which we can optionally select a region of interest (ROI) interactively. b Afterwards, the data is split into overlapping patches, and segmentation is run on each patch (for instance, Cellpose, Baysor, or a custom segmentation tool). Since patches are overlapping, some cells can be segmented multiple times on different patches. Therefore, these conflicts have to be resolved: two boundaries with a significant overlap are merged into one cell, while two cells barely touching are kept separate. The next step is aggregation, i.e., counting transcripts and averaging each channel intensity inside each cell. This allows annotation, either based on transcripts (using Tangram) or on channel intensities. c Afterwards, Sopa outputs a user-friendly report and files to be opened in the Xenium Explorer (whatever the input technology). d All data files are kept for further analysis in Sopa, such as spatial statistics, or integration with community tools.
