m2ST

m2ST performs dual multi-scale graph clustering for spatially resolved transcriptomics data to improve spatial domain annotation accuracy.

Key Features:

• Multi-scale masked graph autoencoder: Extracts representations across different scales from spatial transcriptomic data using a random masking mechanism for node features and a scaled cosine error as the loss function.
• Tailored multi-scale clustering framework: Integrates scale-common and scale-specific information exploration into the clustering process for more robust annotation performance.
• Dynamic scale weighting: Utilizes Shannon entropy to dynamically adjust the importance of different scales during clustering.

Scientific Applications:

• Spatial domain annotation: Identifies spatial domains in spatially resolved transcriptomics data with improved accuracy compared to existing methods.
• Multi-scale transcriptomic analysis: Enables analysis of organism functions and complex mechanisms by integrating information across multiple scales rather than relying on single-scale representations.

Methodology:

m2ST first uses a multi-scale masked graph autoencoder to extract representations across different scales from spatial transcriptomic data, introducing a random masking mechanism for node features and using a scaled cosine error as the loss function. A tailored multi-scale clustering framework integrates scale-common and scale-specific information exploration into the clustering process. Shannon entropy is utilized to dynamically adjust the importance of different scales.