Accurate and efficient integrative reference-informed spatial domain detection for spatial transcriptomics.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Spatially resolved transcriptomics (SRT) enables detailed mapping of tissue structure and function.
• Increasing SRT dataset size and complexity necessitate advanced computational methods for analysis.

Purpose of the Study:

• To introduce Integrative and Reference-informed Tissue Segmentation (IRIS), a novel computational method.
• To accurately and efficiently detect spatial domains in SRT data.
• To integrate multiple SRT slices and account for cross-slice correlations.

Main Methods:

• IRIS leverages single-cell RNA sequencing data for reference-informed domain detection.
• The method integrates multiple SRT slices, considering intra- and inter-slice correlations.
• Validation performed on six diverse SRT datasets across different technologies, tissues, species, and resolutions.

Main Results:

• IRIS demonstrated significant accuracy gains (39-1,083%) and speed improvements (4.6-666.0) on moderate datasets.
• IRIS is the only method applicable to large-scale datasets like Stereo-seq and 10x Xenium.
• The method successfully revealed intricate brain structures, tumor microenvironment heterogeneity, and testicular changes in diabetes.

Conclusions:

• IRIS provides a powerful and efficient computational approach for spatial domain detection in SRT studies.
• The method enhances biological interpretability and scalability for analyzing complex spatial transcriptomics data.
• IRIS facilitates discoveries in diverse biological contexts, from neuroanatomy to disease pathology.