Efficient integration of spatial omics data for joint domain detection, matching, and alignment with stMSA

Area of Science:

• Genomics
• Proteomics
• Systems Biology
• Computational Biology

Background:

• Spatial omics (SO) offers insights into molecular features within tissue architecture.
• Increasing SO datasets necessitate advanced analytical methods for comprehensive biological understanding.
• Current methods often overlook complex spatial patterns and hinder multi-omics integration.

Purpose of the Study:

• To introduce stMSA (SpaTial Multi-Slice/omics Analysis), a novel deep graph contrastive learning model.
• To develop a method for batch-corrected spatial omics data integration that preserves spatial patterns.
• To enable robust analysis across diverse experimental conditions and omics layers.

Main Methods:

• Utilized a deep graph contrastive learning framework with graph auto-encoder techniques.
• Incorporated intra- and interbatch relationships for enhanced data integration.
• Developed stMSA for spatial pattern retention and batch effect correction.

Main Results:

• stMSA outperformed existing methods in distinguishing tissue structures across diverse slices.
• The model demonstrated effectiveness in integrating spatial proteomics and transcriptomics data.
• Achieved superior cross-slice matching and alignment for 3D tissue reconstruction.

Conclusions:

• stMSA provides accurate batch-corrected representations while preserving critical spatial information.
• The method facilitates deeper biological insights from complex spatial omics datasets.
• stMSA is a powerful tool for multi-omics integration and 3D spatial tissue analysis.