coupleCoC

coupleCoC applies a coupled co-clustering-based unsupervised transfer learning algorithm to integrate multimodal single-cell genomic datasets and jointly co-cluster cells and genomic features to reduce noise and enable knowledge transfer across datasets.

Key Features:

• Multimodal Data Integration: Integrates multiple single-cell genomic data types, including scATAC-seq with scRNA-seq, sc-methylation with scRNA-seq, and supports cross-species datasets such as mouse and human.
• Simultaneous Co-Clustering: Performs concurrent co-clustering of cells and genomic features to group similar features and cells together.
• Unsupervised Transfer Learning: Leverages unsupervised transfer learning to transfer information across datasets without requiring labeled data.
• Improved Clustering Performance: Enhances matching of cell subpopulations across multimodal single-cell genomic datasets and improves clustering performance.
• Computational Efficiency: Maintains computational efficiency to enable scaling to large single-cell genomic datasets.

Scientific Applications:

• Integrative single-cell analysis: Provides a unified view of cellular states by combining scRNA-seq, scATAC-seq, and sc-methylation data.
• Cell subpopulation identification: Facilitates identification of novel or refined cell subpopulations through joint clustering.
• Cross-species comparison: Enables comparison of genomic patterns and cell subpopulations between mouse and human datasets.

Methodology:

Implements an information theoretic co-clustering framework that simultaneously co-clusters cells and genomic features using a coupled co-clustering-based unsupervised transfer learning approach, aligning similar genomic features across datasets to reduce noise and facilitate knowledge transfer.