Integrating spatial transcriptomics data across different conditions, technologies and developmental stages.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Spatial transcriptomics (ST) data generation is rapidly increasing.
• Integrating diverse ST datasets is crucial for comprehensive biological understanding.

Purpose of the Study:

• To present STAligner, a novel graph attention neural network for ST data integration and alignment.
• To enable spatially aware data integration, domain identification, and comparative analysis.

Main Methods:

• Developed STAligner, a graph attention neural network.
• Applied STAligner to human cortex, mouse olfactory bulb, mouse hippocampus, and mouse organogenesis datasets.
• Utilized STAligner for capturing shared tissue structures, disease-related substructures, and developmental changes.

Main Results:

• STAligner successfully integrated and aligned multiple ST datasets from varied sources.
• The method identified shared tissue structures, disease-specific alterations in Alzheimer's models, and dynamic developmental changes.
• Identified shared spatial domains and nearest-neighbor pairs facilitated 3D reconstruction.

Conclusions:

• STAligner provides a powerful tool for spatially aware integration and analysis of diverse spatial transcriptomics data.
• The approach enhances understanding of tissue organization, disease pathology, and developmental processes.
• STAligner improves 3D reconstruction accuracy through structure-guided registration.