SpatialRNA: a Python package for easy application of Graph Neural Network models on single-molecule spatial...

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Image-based spatial transcriptomics (iST) provides high-resolution gene expression data with preserved spatial context.
• Graph Neural Networks (GNNs) show promise for analyzing complex molecular and cellular phenotypes in tissues.

Purpose of the Study:

• To present SpatialRNA, a Python package for generating (sub)graphs from tissue samples for GNN analysis.
• To facilitate the application of GNN models for detecting spatial domains in iST data.

Main Methods:

• Development of the SpatialRNA Python package for subgraph generation from iST data.
• Integration with the PyG framework for efficient GNN model application.
• Provision of comprehensive tutorials and workflows for user guidance.

Main Results:

• SpatialRNA enables scalable segmentation of tissues into spatial domains.
• The tool aids in the biological interpretation of iST data and molecular microenvironments.

Conclusions:

• SpatialRNA simplifies the application of GNNs to large iST datasets.
• This tool enhances the understanding of tissue complexity and cellular interactions through spatial domain identification.