SpaGCN

SpaGCN integrates gene expression, spatial coordinates, and histological information using graph convolutional networks to identify spatial domains and detect spatially variable genes in spatially resolved transcriptomics (SRT) data.

Key Features:

• Integration of Multimodal Data: Combines gene expression, spatial coordinates, and histological information for joint analysis of SRT datasets.
• Graph Convolutional Network Approach: Uses graph convolutional networks (GCNs) to aggregate gene expression profiles from each spot with information from neighboring spots.
• Domain-Guided Differential Expression Analysis: Performs domain-guided differential expression (DE) analysis to identify genes enriched in identified spatial domains.
• Comparative Performance: Demonstrated improved detection of spatially enriched genes in evaluations across seven SRT datasets compared with existing methods.
• Transferability of Results: Identified spatially enriched genes transfer across datasets, enabling cross-study application of findings.
• Computational Efficiency and Platform Independence: Provides computational efficiency for rapid analyses and platform-independent applicability across SRT studies.

Scientific Applications:

• Tissue Microenvironment Characterization: Elucidates spatial gene expression variations to characterize tissue microenvironments.
• Disease Mechanism Exploration: Detects spatially variable genes relevant to diseases with spatial heterogeneity, such as cancer.
• Biomarker Discovery: Identifies domain-specific genes that can serve as candidate biomarkers for diagnostic or therapeutic research.

Methodology:

Integrates gene expression, spatial coordinates, and histological information into a spot graph analyzed via graph convolutional networks that aggregate neighboring spot profiles, followed by domain-guided differential expression (DE) analysis and benchmarking across seven SRT datasets.