spatialcore 0.5.4

Standardized spatial statistics tools for computational biology

SpatialCore

Standardized spatial statistics for computational biology.

---

🎯 The Mission

The Problem

Tools for spatial biology analysis are fragmented. Implementations and complex functions often differ between languages (R vs Python) and across packages, making reproducibility difficult and benchmarking impossible. We believe simple statistical tools solve many of the core problems of spatial biology, and we have desinged them to be intuative, easy to use, and scaleable for millions of cells.

The Solution

SpatialCore serves as a package for computational biologists, by computational biologists. It provides robust, standardized implementations of core spatial statistics that ensure identical results across platforms, wrapping high-performance libraries where available.

The Goal

To make spatial analysis engineering boring, so you can focus on the exciting biology. Standardized. Scalable. Reproducible.

📖 See the full documentation: for modules, examples, and benchmarks mcap91.github.io/SpatialCore

⚖️ Philosophy

• Reproducibility: Same inputs = Same outputs. Period.
• Scalability: Built for the era of millions of cells (Xenium/CosMx).
• Transparency: Thin wrappers, not black boxes. We verify, we don't obfuscate.
• Documentation: Clear docstrings with academic references.

What we are NOT:

• Inventing new methods, alogithims, or complex solutions.
• Replacing Scanpy, Seurat, or other methods.

---

---

📦 Installation

Recommended: Conda/Mamba Environment

For the best experience, we recommend using a conda or mamba environment:

# Create environment with Python 3.11
mamba create -n spatialcore python=3.11
mamba activate spatialcore

# Install SpatialCore
pip install spatialcore

This installs all core Python dependencies including CellTypist for custom model training for cell type annotation.

Run upgrade to get the latest modules, features, and fixes

# Activate environment
mamba activate spatialcore

# Upgrade
pip install --upgrade spatialcore

R Requirements

SpatialCore uses R for certain operations that are statistically optimized or perform better in R. The r_bridge module handles R integration via subprocess (no rpy2 required).

Install R packages in your environment:

# If using conda/mamba (recommended)
mamba install -c conda-forge r-base r-sf r-concaveman r-dplyr r-purrr r-jsonlite

# If using system R (Linux/macOS)
sudo apt-get install r-base  # Ubuntu/Debian
R -e "install.packages(c('sf', 'concaveman', 'dplyr', 'purrr', 'jsonlite'), repos='https://cloud.r-project.org/')"

Verify R is configured correctly:

from spatialcore.r_bridge import check_r_available, get_r_version
print(check_r_available())  # True
print(get_r_version())      # R version 4.x.x

How r_bridge Works

The r_bridge automatically detects your environment:

Environment	R Execution Method
Conda/Mamba	mamba run -n env_name Rscript ...
System R	Rscript directly

No manual configuration needed - it just works.

---

🚀 Quick Start

import scanpy as sc
import spatialcore

# Check what's available in your installation
spatialcore.print_info()
# SpatialCore v0.4.5
# Available modules: core, annotation, nmf, r_bridge, spatial, ...

# Load your spatial data
adata = sc.read_h5ad("spatial_data.h5ad")

# Spatial domain detection on B cells
from spatialcore.spatial import make_spatial_domains
from spatialcore.plotting import plot_domains

adata = make_spatial_domains(
    adata,
    filter_expression="hieratype_ontology_name == 'B cell'",
    output_column="bcell_domain",
    domain_prefix="Bcell",
    platform="cosmx",
)

# Visualize spatial domains
plot_domains(adata, domain_col="bcell_domain", title="B Cell Domains")

---

🧩 Modules & Features

Module	Status	Features
spatialcore.annotation	✅ Available	CellTypist wrappers, custom model training, benchmarking
spatialcore.nmf	✅ Available	Spatial non-negative matrix factorization (spaNMF)
spatialcore.spatial	✅ Available	Moran's I, Lee's L, spatial autocorrelation
spatialcore.domains	✅ Available	Neighborhood profiling, niche identification, domain detection
spatialcore.thresholding	✅ Available	Cell classification, oncogene thresholding
spatialcore.r_bridge	✅ Available	Seurat integration, R interoperability via subprocess
spatialcore.diffusion	🔜 Coming soon	Diffusion maps, pseudotime analysis

---

📚 Terminology

We strictly define our spatial units to ensure clarity:

Term	Definition
Neighborhood	The immediate spatial vicinity of a cell (e.g., k-Nearest Neighbors or fixed radius).
Niche	A functional microenvironment defined by a specific composition of cell types (e.g., "Tumor-immune border").
Domain	A macroscopic, continuous tissue region with shared structural characteristics (e.g., "Cortex", "Medulla").

---

🤝 Ecosystem Integration

SpatialCore is designed to play nice with others. It fits seamlessly into the existing Python spatial biology stack:

• Scanpy: The backbone for single-cell analysis.
• Squidpy: Advanced spatial omics analysis.
• Seurat: Direct R interoperability for teams working across languages.

---

📝 Citation

If SpatialCore aids your research, please cite:

@software{spatialcore,
  title = {SpatialCore: Standardized spatial statistics for computational biology},
  url = {https://github.com/mcap91/SpatialCore},
  license = {Apache-2.0}
}

License

Apache License 2.0

The SpatialCore name and trademarks are reserved to ensure the community can rely on the "Standardized" quality of the core library. You are free to use, modify, and distribute the code, including for commercial use.