5 tools to visualize spatial scRNA-seq data

5 tools to visualize spatial scRNA-seq data


Adding to the existing complexity of single-cell resolution datasets, the spatial dimension also presents another challenge for the visualization and manipulation of the data, combining physiological or histological data with the clustering of gene expression profiles. Here we present 5 open-source tools developed by academics to help make sense of the data visually.


Giotto Viewer

Giotto Viewer is the second and independent module of a comprehensive package for spatial transcriptomics analysis. With the whole toolbox designed from the start for spatial transcriptomics data rather than a spatial add-on for an existing scRNA-seq analysis pipeline, Giotto viewer can handle large datasets to display data simultaneously in the physical and gene expression space, while sharing annotations and IDs for easier navigation. While the main purpose of Giotto Viewer is to accommodate data from a large array of spatial transcriptomics technologies, the Giotto Analyzer package also has the ability to use external scRNA-seq data to enrich spatial data that does not initially hqve single-cell resolution.

You can find both Giotto packages on the Harvard website.


ST Viewer

Developed by the SciLife Lab in Sweden, ST Viewer aims to be a user-friendly software for visualization and real-time interaction of spatial transcriptomics data without pre-required coding knowledge. It combines the count matrix from scRNA-seq data with histological hematoxylin and eosin-stained tissue imaging to display complex data in a intuitive way with highly customizable options. A separate package is available for deeper analyses of selected region or tissues, including differential expression analysis and spot classification.

The source code for ST viewer is available under a MIT license on GitHub, as well as the ST analysis package.


Vitessce Image Viewer

Vitessce is a highly modular tool developed at Harvard to combine and visualize multiple sources of single-cell data simultaneously. It is particularly useful for multi-omics studies, with spatial images captured with different technologies at different resolutions displayed as independent layers in the same space. Vitessce is a rapidly evolving software aiming to be integrated in the Human BioMolecular Atlas Program (HuBMAP.

The is the original package of Vitessce with demo datasets, but you can find more information on the newer Beta version.

Check out this article if you want to read more about the progress of spatially resolved single-cell data across different methods!


If you are looking to fine-tune the resolution of your spatial data down to individual cells, have a look at BayesSpace. Current spatial transcriptomics technologies do not yet technically reach single-cell resolution: whether using barcoded spots or beads, their diameters usually cover several, or even dozens, of cells. To improve the resolution of the resulting dataset, researchers at the Fred Hutchinson Cancer Research Center developed BayesSpace, a statistical method focusing on better clustering to identify accurate tissue structures. In their own words, the method has been made accessible by minimizing user input (such as the preselection of marker genes) and parameter tuning.

BayesSpace can be found on the open-source software database Bioconductor and the source code on GitHub.



This visualization tool is rather strikingly different from the ones above: rather than gathering spatial data from spatial transcriptomics technologies, CSOmap (Cellular Spatial Organization mapper) suggests a de novo reconstruction using scRNA-seq and known receptor-ligand complexes. Ranking cells by their relative abundances of ligands and receptors with their respective affinities, the mathematical model assumes cell with high potentials for interactivity for each other are grouped spatially, but also compete for the lack of space in the normal physiological context. Using scRNA-seq data, CSOmap then suggest a de novo 3D model of a tissue or organ, where individual cell-cell interactions and specific ligand-receptor interactions can be ranked by importance. With such a model available in silico, CSOmap offers the interesting feature of modeling a disturbance in the ligand-receptor network, by overexpressing or knocking down a particular molecule and observing the effect on spatial organization at tissue levels.

The R package for CSOmap is on GitHub.

Laisser un commentaire

Votre adresse de messagerie ne sera pas publiée.