SHARP

SHARP applies ensemble random projection and multi-layer meta-clustering to perform fast, scalable dimensionality reduction and clustering of single-cell RNA sequencing (scRNA-seq) data while preserving cell-to-cell distances.

Key Features:

• Ensemble Random Projection (RP): Reduces dimensionality of large-scale scRNA-seq data while preserving critical cell-to-cell distances in the reduced-dimensional space.
• Multi-layer Meta-clustering: Integrates multiple clustering layers to achieve high clustering accuracy on large datasets.
• Scalability: Capable of processing datasets containing up to 10 million cells.
• R-based Implementation: Provided as an R-based solution for clustering very large single-cell RNA-seq datasets.
• Benchmarking: Demonstrated superior speed and accuracy in comprehensive benchmarking on 17 public scRNA-seq datasets, notably for datasets larger than 40,000 cells.

Scientific Applications:

• Large-scale scRNA-seq clustering: Clustering of millions of single cells to identify cell populations and heterogeneity.
• Dimensionality reduction for single-cell transcriptomics: Generating reduced-dimensional representations that preserve intercellular distances for downstream analysis.
• Method benchmarking: Comparative evaluation of speed and accuracy across public scRNA-seq datasets.

Methodology:

Ensemble random projection (RP) for dimensionality reduction, preservation of cell-to-cell distances in the reduced space, and multi-layer meta-clustering; performance assessed on 17 public scRNA-seq datasets.