dropClust

dropClust performs scalable clustering of droplet-based single-cell RNA sequencing (scRNA-seq) data to identify cell types and rare subpopulations.

Key Features:

• Scalability: Leverages Locality Sensitive Hashing (LSH) for approximate nearest neighbor search to handle large-scale single-cell datasets efficiently.
• Accuracy and Speed: Enhances execution time while maintaining high clustering accuracy compared to existing methods.
• Detectability of Minor Cell Subtypes: Sensitive detection of rare or less abundant cell subtypes within complex datasets.

Scientific Applications:

• Cell Type Identification: Clusters cells by gene expression profiles to identify and characterize distinct cell types.
• Subtype Discovery: Enables discovery of rare cellular subpopulations and minor subtypes within heterogeneous samples.
• Comparative Studies: Facilitates comparison of cellular compositions across conditions, time points, or treatments.

Methodology:

dropClust employs Locality Sensitive Hashing (LSH), an approximate nearest neighbor search technique that approximates distances to preserve locality and efficiently group cells into clusters based on transcriptomic profiles.