stimwrap - a python interface for STIM

stimwrap is a Python wrapper for the spatial transcriptomics library STIM. It provides an interface for running STIM commands from python, as well as for extracting datasets and their attributes from n5 containers (backed by n5 or AnnData) that are created by STIM.

stimwrap is created and maintained by Nikos Karaiskos and Daniel León-Periñán.

Quick start

We provide an example notebook under notebooks/openst_example.ipynb, and also a Google Colab notebook

These will walk you through downloading Open-ST data, running STIM in a Python notebook (via stimwrap bindings), and seamlessly running some downstream analyses.

We provide another notebook under notebooks/visium_example.ipynb, and also a 2nd Google Colab notebook showcasing the 10x Visium adult mouse brain dataset showcased in all our tutorials.

Installation

We recommend installing stimwrwap in the same conda environment with STIM. To install stimwrap try:

pip install stimwrap

or if you do not have sudo rights:

pip install --user stimwrap

Check if the library is successfully installed:

python -c import stimwrap as sw

Usage

The following assumes that the file container.n5 contains the datasets and their attributes as created by STIM:

import stimwrap as sw
pucks = sw.Container('/path/to/container.n5')

Print the names of the datasets:

print(pucks.get_dataset_names())

Focus on a specific puck and extract the relevant information:

puck_name = pucks.get_dataset_names()[0]
puck = pucks.get_dataset(puck_name)

Get the puck locations either directly from the puck:

locations = puck['locations']

or fetch them from the container:

locations = pucks.get_dataset(puck_name)['locations']

The examples above assume that the dataset is N5-backed. For AnnData-backed datasets, the key for the puck locations might be:

locations = pucks.get_dataset(puck_name)['spatial']

which will try to access the obsm/spatial variable from the dataset. Alternatively, we recommend using the official AnnData package for handling these files.

Fetch gene expression

It is possible to get the expression vector of a single gene:

hpca_vec = pucks.get_dataset(puck_name).get_gene_expression(gene='Hpca')

or the whole gene expression matrix:

dge = pucks.get_dataset(puck_name).get_gene_expression()

Fetch dataset attributes

STIM stores the dataset attributes in the n5 container. These can be directly accessed with stimwrap:

puck.get_attribute(attribute='geneList')

In N5-backed STIM, available options might also include: barcodeList and metadataList.

Fetch aligned locations

In the case where multiple consecutive sections are obtained and aligned with STIM, the aligned locations can be obtained with:

aligned_locations = puck.get_aligned_locations(transformation='model_sift')

Store aligned locations

The aligned locations can be stored in the N5 or AnnData-backed object, for seamless downstream analysis:

aligned_locations = puck.apply_save_transform(transformation='model_sift')