FastPG

FastPG performs unsupervised clustering of large-scale single-cell datasets (including millions of cells) to identify distinct cell populations while retaining high cell assignment accuracy.

Key Features:

• Unsupervised clustering: Performs unsupervised clustering of single-cell datasets at million-cell scale.
• Graph-based algorithm: Builds upon the graph-based algorithm PhenoGraph to perform graph-based clustering.
• High speed: Achieves clustering speeds up to 27 times faster than PhenoGraph.
• Accuracy: Maintains high cell assignment accuracy and demonstrated superior accuracy compared with FlowSOM and PARC in comparative tests.
• No subsampling: Operates without requiring dataset subsampling, enabling analysis of complete datasets.

Scientific Applications:

• Genomics: Clustering of single-cell genomics data to resolve cellular heterogeneity.
• Transcriptomics: Identification of distinct cell populations in single-cell transcriptomics datasets.
• Proteomics: Analysis of single-cell proteomics datasets to detect proteomic heterogeneity.
• Developmental biology: Resolving cell population dynamics during development.
• Cancer research: Characterizing tumor cellular heterogeneity.
• Immunology: Studying immune cell population structure and heterogeneity.

Methodology:

Extends the graph-based algorithm PhenoGraph to perform unsupervised graph-based clustering of large-scale single-cell datasets without subsampling, reporting up to 27-fold speed improvements over PhenoGraph.