cellClustering: Do clustering of cells

In maplesword/RefSimSpec:

Description

This function does clustering of cells based on their input feature matrix (e.g. RSS matrix). Possible clustering methods include k-means and k-medoids (PAM, or Partitioning Around Medoid). It is also allowed to only select a subset of samples (cells) for clustering, and then train an SVM-based classifier to assign clusters for the remaining cells.

Usage

cellClustering(input, clustMethod = c("pam", "kmeans", "pca-kmeans"),
  numClust = 20, pcNum = 50, maxCells = 40000, seedIdx = NULL,
  threads = 1, verbose = TRUE)

Arguments

input	The input expression matrix, with rows representing SAMPLES (cells) and columns representing FEATURES.
clustMethod	The method used to do cell clustering.
numClust	The expected number of clusters.
pcNum	When PCA is used for dimension reduction, it defines the number of dimensions.
maxCells	The maximum number of cells selected for clustering. NULL or NA for unlimitation.
seedIdx	The indices of samples (cells) used for clustering. The remaining samples will be assigned to clusters based on the SVM-classification model.

Value

A list with two elements: 'clust' is the vector of class labels; 'seedset' is the vector of indices of samples (cells) used in the initial clustering

maplesword/RefSimSpec documentation built on May 23, 2019, 1:47 p.m.