Fig. 1.

(a) Summary of the clustering performance of PARC and other competitive clustering methods on various single-cell datasets, including flow cytometry, mass cytometry (CyTOF), imaging cytometry, and scRNA-seq data. (b) Overview of PARC workflow for large-scale single-cell analysis on multiple types of high-dimensional single-cell data. The enabling features include fast graph construction by HNSW, 2-step data-driven graph refinement and pruning, and accelerated community detection by Leiden algorithm.