runCORE: runCORE

In IMB-Computational-Genomics-Lab/ascend: ascend - Analysis of Single Cell Expression, Normalisation, and Differential expression

Description

This function determines the optimal number of clusters for a dataset. This function first generates a distance matrix and a hclust object, and then cuts the tree at different heights.

Usage

runCORE(object, ..., conservative, nres, remove.outliers)

## S4 method for signature 'EMSet'
runCORE(object, conservative = TRUE, dims = 20,
  nres = 40, remove.outliers = FALSE)

Arguments

object	An EMSet object that has undergone PCA reduction.
...	...
conservative	Use conservative (more stable) clustering result (TRUE or FALSE). Default: TRUE.
nres	Number of resolutions to test, ranging from 20 to 100. Default: 40.
remove.outliers	Remove cells that weren't assigned a cluster with dynamicTreeCut. This is indicative of outlier cells within the sample. Default: FALSE.
dims	Number of PC components to use in distance matrix generation. Default: 20

Details

This will return an EMSet with the following objects:

DistanceMatrix

A distance matrix.

Hclust

A hclust object.

PutativeClusters

Cluster identities generated by dynamicTreeCut.

ClusteringMatrix

A matrix containing a cluster identities from cutting at 40 different heights.

Clusters

Optimum cluster identities for each cell.

NumberOfClusters

Number of clusters.

OptimalTreeHeight

Optimal tree height used to generate cluster identities.

KeyStats

A dataframe containing information on each generated clustering result, that is used to determine the optimal cluster number.

Value

An EMSet with cluster information loaded into the clusterAnalysis and colInfo slots

Examples

# Load example EMSet
em_set <- ascend::analyzed_set

# Run CORE with default parameters
em_set <- runCORE(em_set, conservative = TRUE,
dims = 20, nres = 40, remove.outliers = TRUE)

IMB-Computational-Genomics-Lab/ascend documentation built on Aug. 29, 2019, 4:10 a.m.