clustcells: PhenoGraph clustering

In dibbelab/gficf: Gene Frequency - Inverse Cell Frequency

PhenoGraph clustering

Description

R implementation of the PhenoGraph algorithm

Usage

clustcells(
  data,
  from.embedded = F,
  k = 15,
  dist.method = "manhattan",
  nt = 2,
  community.algo = "louvian",
  store.graph = T,
  seed = 180582,
  verbose = TRUE,
  resolution = 0.8,
  n.start = 10,
  n.iter = 10
)

Arguments

data	list; Input data (gficf object)
from.embedded	logical; Use embeddedd (UMAP or tSNA) space for clustering cells. Best results are usually obtained not using the embedded space.
k	integer; number of nearest neighbours (default:15)
dist.method	character; Dist to use for K-nn. Type of distance metric to use to find nearest neighbors. One of: "euclidean" (the default) "cosine" "manhattan" "hamming" (very slow)
nt	integer; Number of cpus to use for k-nn search
community.algo	characthers; Community algorithm to use for clustering. Supported are: "louvian" (the default, the original Louvian method) "louvian 2" (Louvian with modularity optimization from Seurat) "louvian 3" (Louvain algorithm with multilevel refinement from Seurat) "leiden" (Leiden algorithm see Traag et al. 2019) "walktrap" "fastgreedy"
store.graph	logical; Store produced phenograph in the gficf object
seed	integer; Seed to use for replication.
verbose	logical; Increase verbosity.
resolution	Value of the resolution parameter, use a value above (below) 1.0 if you want to obtain a larger (smaller) number of communities (used only for leiden and louvian 2 or 3 methods).
n.start	Number of random starts (used only for louvian 2 or 3 methods).
n.iter	Maximal number of iterations per random start (used only for louvian 2 or 3 methods).

Details

A custom R implementation of the PhenoGraph (http://www.cell.com/cell/abstract/S0092-8674(15)00637-6) algorithm, which is a clustering method designed for high-dimensional single-cell data analysis. It works by creating a graph ("network") representing phenotypic similarities between cells by calclating the Jaccard coefficient between nearest-neighbor sets, and then identifying communities using the well known Louvain method (https://sites.google.com/site/findcommunities/) in this graph.

That version used PCA or LSA reduced meta-cells and multithreading annoy version for K-nn search (from uwot package).

Value

the updated gficf object

dibbelab/gficf documentation built on Nov. 2, 2022, 2:28 a.m.