PARC: ultrafast and accurate clustering of phenotypic data of millions of single cells

Abstract

Motivation: New single-cell technologies continue to fuel the explosive growth in the scale of heterogeneous single-cell data. However, existing computational methods are inadequately scalable to large datasets and therefore cannot uncover the complex cellular heterogeneity. Results:We introduce a highly scalable graph-based clustering algorithm PARCphenotyping by accelerated refined community-partitioning -for ultralarge-scale, high-dimensional single-cell data (> 1 million cells). Using large single cell mass cytometry, RNA-seq and imaging-based biophysical data, we demonstrate that PARC consistently outperforms state-of-the-art clustering algorithms without sub-sampling of cells, including Phenograph, FlowSOM, and Flock, in terms of both speed and ability to robustly detect rare cell populations. For example, PARC can cluster a single cell data set of 1.1M cells within 13 minutes, compared to >2 hours to the next fastest graph-clustering algorithm, Phenograph. Our work presents a scalable algorithm to cope with increasingly large-scale single-cell analysis.Availability and Implementation: https://github.com/ShobiStassen/PARC Contact: tsia@hku.hk

Comprehensive investigations revealed consistent pathophysiological alterations after vaccination with COVID-19 vaccines

“…Then principal component analysis (PCA) was performed utilizing the HVGs and Harmony algorithm was used to remove batch effects 25 . We used the PARC approach to identify clusters 37 and selected features by "FeatureSelec-tionByEnrichment" function from cytograph2 algorithm 38 , followed by another round of PCA, Harmony, and PARC. Subsequently, we calculated K nearest neighbors in a KNN graph, performed uniform manifold approximation and projection (UMAP) by Pegasuspy 39 , and identified clusters by PARC.…”

New interpretable machine-learning method for single-cell data reveals correlates of clinical response to cancer immunotherapy

“…26 We varied the number of components from 5 to 115. We applied FAUST, DEPECHE, 27 flowMeans, 10,28 FlowSOM, 6,29 k-means, PARC, 30 and Pheno-Graph 31,32 to the simulated datasets (supplemental experimental procedures A.7 shows an example of the simulated data). We provided k-means with the true number of clusters in each simulation iteration, a required parameter setting.…”

Generalized and scalable trajectory inference in single-cell omics data with VIA

“…Algorithm. VIA first represents the single-cell data as a cluster graph (i.e., each node is a cluster of single cells), computed by our recently developed data-driven community-detection algorithm, PARC, which allows scalable clustering whilst preserving global properties of the topology needed for accurate TI 14 (Step 1 in Fig. 1).…”

“…VIA first represents the singlecell data in a k-nearest-neighbor (KNN) graph where each node is a cluster of single cells. The clusters are computed by our recently developed clustering algorithm, PARC 14 . In brief, PARC is built on hierarchical navigable small world 46 accelerated KNN graph construction and a fast community-detection algorithm (Leiden method 47 ), which is further refined by data-driven pruning.…”