scGCN

scGCN applies a graph convolutional network (GCN) to transfer cell-type labels and other annotations across heterogeneous single-cell omics datasets.

Key Features:

• Graph-Based Model: scGCN employs a graph convolutional network (GCN) framework to model complex relationships within single-cell data.
• Robust Knowledge Transfer: Provides improved accuracy for transferring labels from labeled datasets to newly generated datasets.
• Versatility Across Datasets: Benchmarked across 30 diverse single-cell omics datasets and demonstrated performance across cells from various tissues, platforms, species, and molecular layers.
• Implementation: Implemented as an integrated workflow in Python.

Scientific Applications:

• Cross-dataset integration: Explore and integrate data from different biological contexts, such as tissue types or species.
• Cellular heterogeneity analysis: Enhance understanding of cellular heterogeneity by leveraging multi-layered molecular profiles.
• Translational and personalized medicine: Facilitate cross-dataset analyses to support discoveries relevant to personalized medicine.

Methodology:

Constructs a graph representation of single-cell data with nodes as individual cells and edges encoding similarity or connectivity, then applies a GCN to learn and transfer labels across datasets with varying characteristics.