scEGG

scEGG leverages exogenous gene information, such as gene networks, to generate discriminative cell representations for deep clustering of single-cell transcriptomic data.

Key Features:

• Attention-Enhanced Graph Autoencoder: Captures topological signal patterns among cells efficiently.
• Random Walk on Protein-Protein Interaction Network: Acquires gene embeddings from exogenous network information.
• Gene-Cell Cooperative Embedding Integration: Reconstructs integrated embeddings to yield discriminative representations for clustering.

Scientific Applications:

• Single-Cell Transcriptomic Clustering: Groups cells based on integrated transcriptomic and exogenous gene network information.
• Disease Research: Provides enhanced insights into cell characteristics to lay the foundation for early diagnosis and treatment of diseases.

Methodology:

scEGG uses an attention-enhanced graph autoencoder to capture topological signal patterns among cells, performs random walk on a protein-protein interaction network to acquire gene embeddings, and integrates and reconstructs gene-cell cooperative embeddings for discriminative clustering.