Installation

PyPI

SPLISOSM can be installed via pip:

# from PyPI (stable version)
$ pip install splisosm

# or from GitHub (latest version)
$ pip install git+https://github.com/JiayuSuPKU/SPLISOSM.git#egg=splisosm

Minimal software dependencies:

torch
scipy
scikit-learn
pandas
tqdm
matplotlib
anndata

Additional dependencies

SplisosmFFT requires the spatialdata package to be installed, and SplisosmNP’s differential usage test optionally supports the gpytorch backend for Gaussian process regression. To install these additional dependencies, you can use the following command:

$ pip install "splisosm[sdata, gp]"

For spatially variable gene expression test, we also provide a Python wrapper, splisosm.utils.run_sparkx(), for SPARK-X. To use this functionality, ensure that the conda environment where SPLISOSM is installed has R (>=4.0) available in PATH, and that the R package SPARK-X is properly installed.

# install R and the R package SPARK-X in the conda environment
$ conda install -c conda-forge r-base r-devtools
$ Rscript -e "devtools::install_github('xzhoulab/SPARK')"

# install the rpy2 package to interface R from python
$ pip install rpy2
$ pip install splisosm # if not already installed

# test whether SPARK-X is correctly configured
$ python -c "import numpy as np; from splisosm.utils import run_sparkx; run_sparkx(np.random.randn(10,5), np.random.rand(10,2))"

Note

Our new gene-level spatial variability test, HSIC-GC, is available as a standalone function (splisosm.utils.run_hsic_gc()). It can be used as a drop-in replacement for SPARK-X, where we optimize the spatial kernel to achieve higher statistical power while maintaining computational efficiency.

from splisosm.utils import run_hsic_gc
import numpy as np

# gene expression matrix: (n_spot, n_gene)
gene_counts = np.random.randn(100, 50)

# spatial coordinates: (n_spot, 2)
coordinates = np.random.rand(100, 2)

# run HSIC-GC test (default: Liu's eigenvalue method for the null)
test_results = run_hsic_gc(gene_counts, coordinates)
print(test_results['statistic'])  # test statistics, (n_gene,)
print(test_results['pvalue'])     # p-values, (n_gene,)

# for large datasets, you can also choose alternative null approximations via
# (1) moment-matching normal approximation (no eigen-decomposition)
test_results = run_hsic_gc(gene_counts, coordinates, null_method="trace")

# (2) or restrict eigenvalue null to the top-k eigenvalues
test_results = run_hsic_gc(
    gene_counts, coordinates,
    null_method="eig",
    null_configs={"approx_rank": 100}
)