3D-spaGNN-E

3D-spaGNN-E develops spatially resolved gene neighborhood network embedding to identify subcellular gene proximity relationships and key subcellular motifs in cell-cell communication by combining 3D imaging-based spatial transcriptomics and graph-based deep learning.

Key Features:

• 3D Transcript Detection: Detects single transcripts in 3D cell culture samples.
• Gene Proximity Identification: Identifies subcellular gene proximity relationships from 3D spatial data.
• Graph Autoencoder Embedding: Projects gene proximity relationships into a latent space using a graph autoencoder.
• Explainability Analysis: Applies explainability analysis to identify subcellular cell-cell communication motifs.
• Cell Type Classification: Distinguishes CD4+ and CD8+ T cells in coculture systems.
• Immune Suppression Profiling: Distinguishes strong and weak suppression of different immune cells through local gene proximity and autoencoder embedding changes.
• Regional Comparison: Compares astrocyte-neuron cell-cell communication across different brain regions.

Scientific Applications:

• Mesenchymal Stem Cell Analysis: Identifies changes in local gene proximity near borders between homotypic and heterotypic cells in hydrogel-cultured mesenchymal stem cells.
• Immune Cell Coculture Studies: Analyzes mesenchymal stem cell-peripheral blood mononuclear cell coculture systems to identify T cell subtypes and immune suppression states.
• Brain Tissue Analysis: Compares astrocyte-neuron cell-cell communication in mouse hypothalamus and cortex using 3D multiplexed-error-robust fluorescence in situ hybridization (MERFISH) data.

Methodology:

3D-spaGNN-E detects single transcripts in 3D cell culture samples and identifies subcellular gene proximity relationships. A graph autoencoder projects the gene proximity relationships into a latent space. Explainability analysis is then applied to identify subcellular cell-cell communication motifs.