Integrating multi-modal information to detect spatial domains of spatial transcriptomics by graph attention network.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Spatially resolved transcriptomic technologies offer insights into tissue architecture and function.
• Existing methods struggle to fully utilize multimodal spatial and histological data.

Purpose of the Study:

• To develop a novel method, SPAtially embedded Deep Attentional graph Clustering (SpaDAC), for enhanced spatial transcriptomics analysis.
• To leverage multimodal information for accurate spatial domain identification and gene expression profile reconstruction.

Main Methods:

• SpaDAC constructs multi-view graph modules to capture spatial location and morphological connectives.
• The method learns low-dimensional embeddings for spatial transcriptomics data.
• SpaDAC integrates gene expression, spatial location, and histology image data.

Main Results:

• SpaDAC outperforms existing algorithms on benchmark spatial transcriptomics datasets.
• The method successfully identifies spatial domains and reconstructs denoised gene expression profiles.
• SpaDAC demonstrates efficient utilization of spatial and histological information.

Conclusions:

• SpaDAC is a valuable tool for spatial domain detection in tissues.
• The method enhances the comprehension of tissue architecture and cellular microenvironments.
• SpaDAC provides a robust approach for analyzing complex spatial transcriptomics data.