cluster_consensus: Cluster the consensus matrix.

Description Usage Arguments Details Value Author(s) See Also Examples

Description

This function will perform hierarchical clustering on the noise consensus matrix allowing the user to investigate the appropriate number of clusters, k, considering the noise within the experiment.

Usage

1
cluster_consensus(consensus_matrix, cluster_num, method = "complete")

Arguments

consensus_matrix

A noise consensus output by compute_consensus().

cluster_num

The number of clusters expected from the hierarchical clustering of the noise consensus matrix.

method

The hierarchical clustering method to be used on the consensus.

Details

We have found it useful to identify the optimal number of clusters in terms of resiliance to noise by examining these metrics by cutting hierarchical clustering dendograms of the noise consensus and comparing the results to the original clustering labels. To do this create a vector containing each number of clusters one wishes to examine (the function automatically determines the results for the dataset as a single cluster) and then cluster the consensus with this function.

Frequently one will want to assess multiple possible cluster number situations at once. In this case it is recommended that one use a lapply in conjunction with a vector of all biologically reasonable cluster numbers to fulfill the task of attempting to identify the optimal cluster number.

Value

The output is a vector of cluster labels based on hierarchical clustering of the noise consensus. In the event that a vector is supplied for number of clusters in conjunction with lapply, then the output is a data.frame of the cluster labels for each of the various number of clusters deemed biologically reasonable by the user.

Author(s)

David T. Severson <david_severson@hms.harvard.edu>

Maintainer: Benjamin Schuster-Boeckler <benjamin.schuster-boeckler@ludwig.ox.ac.uk>

See Also

compute_consensus report_cluster_metrics report_cell_metrics

Examples

1
2
3
4
5

seversond12/BEARscc documentation built on June 8, 2020, 8:04 a.m.