Revealing Tissue Heterogeneity and Spatial Dark Genes from Spatially Resolved Transcriptomics by Multiview Graph...

Area of Science:

• Computational biology
• Genomics
• Cancer research

Background:

• Spatially resolved transcriptomics (SRT) offers comprehensive gene expression profiling and spatial composition insights.
• Understanding tumor organizational complexity and immune escape mechanisms requires advanced analytical tools.

Purpose of the Study:

• To develop a novel multiview graph attention fusion model, stMGATF, for integrating diverse data types in SRT.
• To enhance the characterization of spatial domains and elucidate tumor immune escape mechanisms.

Main Methods:

• stMGATF integrates gene expression, histological images, spatial location, and gene association data.
• Utilizes SimCLRv2 for visual feature extraction and edge feature enhanced graph attention networks.
• Employs a global attention mechanism for adaptive integration of multiple data views into a low-dimensional representation.

Main Results:

• stMGATF demonstrates robustness across diverse SRT datasets, outperforming existing methods in spatial domain detection and data denoising.
• Successfully elucidates tissue heterogeneity and extracts 3-dimensional gene expression domains.
• Identifies 'spatial dark genes' and predicts tumor-driving transcription factors, offering insights into immune escape.

Conclusions:

• stMGATF provides a powerful framework for analyzing complex spatial transcriptomic data.
• The model aids in uncovering tumor heterogeneity and immune escape mechanisms, paving the way for new immunotherapeutic strategies.