SOUP: SOUP

In lingxuez/SOUP: Semi-soft Clustering of Single Cell Data

Description

A semi-soft clustering algorithm for single cells.

Usage

SOUP(expr, Ks = 3, type = "log", i.pure = NULL, ext.prop = NULL,
  pure.prop = 0.5, nPC = NULL, nstart = 50, verbose = FALSE)

Arguments

expr	a cell-by-gene expression matrix, either the raw counts or log-transformed expressions.
Ks	number of clusters, can be a single integer or a list of integers.
type	"log" if expr has been normalized and log-transformed (default), or "count" if expr contains the raw counts. It is recommended to use the log scale, which usually gives better results in practice.
i.pure	(optional) the indices of the pure cells. By default is NULL, and SOUP will infer the pure list. If the list is already known (for example, from previous runs), then providing it will reduce the computation time.
ext.prop	(optional) the proportion of extreme neighbors for each cell, such that ext.prop*n.cells is roughly the number of pure cells per cluster. By default, ext.prop=0.1 for less than 1,000 cells, and ext.prop=0.05 for larger datasets.
pure.prop	(optional) the expected proportion of pure cells in the data. By default pure.prop=0.5.
nPC	(optional) the number of principal components to be used, which is at least K (default). Sometimes it can be helpful to use a slightly larger nPC.

Value

A list containing

Ks

the list of K, number of clusters, used by SOUP.

memberships

a list of cell-by-K membership matrices, one per K in Ks.

centers

a list of K-by-gene cluster center matrices, one per K in Ks.

purity

a vector containing the purity scores of all cells.

i.pure

the indices of pure cells with the highest purity scores.

lingxuez/SOUP documentation built on May 28, 2019, 3:38 p.m.