scSpatial 0.1.2

Package used to analyse spatial transcriptomics data with build in segmentation and information transfer

scSpatial

This software aims to allow analysis of 2D spatial transcriptomics data by integrating segmentation methods and information transfer methods in a graphical UI. Interactions with your images are allowed by the multi dimentional image viewer Napari. This package provides segmentation of the cells and mapping of gene expression data. This is controlled by a widget in the Napari viewer.

1. Segmentation

For high resolution gene expression analysis, where the exact position of a gene is known (such as in HybISS), segmentation of the cells are critical to allow cell level analysis of gene expression. We include several options for cell segmentation:

• Segmentation using Cellpose (nuclei)
• Segmentation using Cellpose (cytoplasm)
• Loading of 3rd party segmentation (ImageJ, Ilastik, CellProfiler etc.)

Cellpose can be used with data including only nuclei staining or images with both nuclei and cytoplasm markers. To achive high gene mapping ratios, segmentation using cytoplasm markers are prefered! When loading 3rd party segmentation images, these should have unique pixel values for each object in the segmentation, with the value 0 reserved for background.

2. Information transfer

With a segmented image, it is possible to map genes to objects (cells) creating a table with gene expression for each object. This information is similar to the data in scRNA-seq datasets, and information transfer between a reference dataset and the spatial mapping can be performed using the Tangram method. Here we have implemented the BoneFight modification from Linnarssons lab which allows for cell cluster based computation.

3. Visualization

Currently, we support visualization of gene expression and cell type prediction. This section will be expanded as new tools are implemented.

Getting started

Installation

Install Miniconda or Anaconda and create a virtual environment with python 3.7 and activate it.

conda create -n scSpatial python=3.7
conda activate scSpatial

Install scSpatial using pip:

pip install scSpatial

Next we install some dependencies that are not available via pip:

pip install git+https://github.com/linnarsson-lab/BoneFight.git@8c1ec1f
pip install git+https://github.com/linnarsson-lab/loompy.git@e0963fb

Running napari with the scSpatial widget

To open Napari with the scSpatial widget, run the scSpatial command in the terminal:

scSpatial

If you want to start napari from within your own pipeline/script, import the scSpatial.app.App class and instantiate it.

from scSpatial.app import App
app = App()

Change Log

v1.0.2

• Bumped version of cellpose to > 2.0.4
• Removed resample during cell pose model evaluation to speed up segmentation.

v1.0.1

• Constrained Cellpose to version 1.0.2 as later versions cased errors.

V1.0.0

Initial release