BBKNN

BBKNN removes batch effects and integrates single-cell RNA sequencing (scRNA-Seq) datasets to enable combined analysis across experimental batches.

Key Features:

• Graph-Based Methodology: Employs a graph-based Batch Balanced K-Nearest Neighbors approach to integrate datasets while accounting for batch labels.
• Integration with scanpy: Interfaces with the scanpy workflow to operate on scRNA-Seq preprocessed data and downstream analyses.
• Dimensionality Reduction and Clustering Support: Produces neighborhood graphs that are used for dimensionality-reduced visualization and clustering of single cells.
• Large-Scale Cross-Atlas Integration: Has been demonstrated on large-scale integrative studies including multiple murine atlases.
• Facilitates Regulatory Inference: Enables integration across experimentally diverse datasets to support inference of transcription factor networks.

Scientific Applications:

• Cross-batch dataset integration: Combines scRNA-Seq datasets from different experimental batches, conditions, or platforms for joint analysis.
• Cellular heterogeneity analysis: Supports identification of global markers of cell type and organ specificity across integrated datasets.
• Comparative atlas studies: Enables comparison and integration of data across murine atlases and other large-scale collections.
• Gene regulatory exploration: Provides integrated data contexts suitable for downstream inference of transcription factor networks and regulatory programs.

Methodology:

Implements a graph-based Batch Balanced K-Nearest Neighbors algorithm to construct batch-balanced kNN graphs that supply neighborhoods for dimensionality-reduced visualization and clustering within the scanpy workflow.