visium_analysis

Informing biologically relevant signal from spatial transcriptomic data

visium_analysis is a Python tool for downstream processing of Visium spatial transcriptomics data. It enables neighborhood distance analysis, cell-type deconvolution evaluation, and pathway activity scoring across spatial spots.

---

Reference and Links

• Author: Emir Radkevich
• Preprint: bioRxiv (2024.09.09.610361) | DOI: 10.1101/2024.09.09.610361
• Publication: Used in a 2026 Science publication.
• Claude skill: Link to the skill.
• Tutorials & Extensions: Example notebooks are available below and pending pull request to squidpy_notebooks.

---

Usage and Workflow

The visium_analysis/ directory contains the core scripts to perform downstream analysis on Visium spatial transcriptomics slides:

1. Spatial Neighborhood Analysis (distance.py)

Analyzes cell distribution and gene expression dynamics as a function of radial distance from a region of interest (in 100 um increments, moving both inward and outward):

• Cell Composition: Maps changes in cell-type composition from the core to surrounding neighborhoods.
• Expression Dynamics: Runs linear regression to track how gene expression changes across spatial distance gradients.

2. Deconvolution and Structural Comparison (deconvolution.py)

Uses deconvolution results to compute PCA and UMAP embeddings, comparing structural compartments (e.g., distinct immune aggregates) by:

• Calculating cell-type proportion differences.
• Computing Mann-Whitney U-tests per cell type across compartments.
• Assessing expression of key receptor-ligand pairs within aggregates.

3. Pathway Activity Scoring (run_decoupler.py)

Leverages decoupler to infer footprint-based activity scores across Visium spots using established reference databases:

• PROGENy (Pathway activities)
• DoRothEA (Transcription factor activities)
• CytoSig (Cytokine signaling)
• MSigDB (Molecular signatures)

Installation

We suggest using a separate conda environment for the package:

conda create --name va python=3.12
conda activate va
pip install git+https://github.com/ismms-himc/visium_analysis.git

Or it can be installed in a python virtual environment:

mkdir venvs && cd venvs
python3.12 -m venv va
source va/bin/activate
pip install git+https://github.com/ismms-himc/visium_analysis.git