EnsembleClustering: Ensemble clustering via graph partitioning (CSPA / HGPA /...

In MosaiClusteR: An Umbrella Framework for Multi-Source and Multi-Omics Clustering

Ensemble clustering via graph partitioning (CSPA / HGPA / MCLA)

Description

Cluster-based ensemble clustering that combines several single-source partitions into a single consensus partition through the graph-partitioning algorithms of Strehl and Ghosh: the Cluster-based Similarity Partitioning Algorithm (CSPA), the HyperGraph Partitioning Algorithm (HGPA) and the Meta-CLustering Algorithm (MCLA), as well as a "Best" selection strategy.

External executable required. This method does not run in pure R. It writes the stacked label matrix to disk and delegates the actual graph partitioning to an external graph-partitioning executable (e.g. a compiled CSPA/HGPA/MCLA binary backed by METIS / gpmetis / shmetis) or to a MATLAB session. These binaries are not bundled with the package; the user must supply and configure them. The path to the executable is selected through the executable argument. When executable=FALSE (the default), or when the configured executable cannot be located, the function stops with an informative error rather than attempting to launch an unavailable program.

Usage

EnsembleClustering(
  List,
  type = c("data", "dist", "clust"),
  distmeasure = c("tanimoto", "tanimoto"),
  normalize = c(FALSE, FALSE),
  method = c(NULL, NULL),
  clust = "agnes",
  linkage = c("flexible", "flexible"),
  alpha = 0.625,
  nrclusters = c(7, 7),
  gap = FALSE,
  maxK = 15,
  ensembleMethod = c("CSPA", "HGPA", "MCLA", "Best"),
  waitingtime = 300,
  file_number = 0,
  executable = FALSE
)

Arguments

List	A list of data matrices. It is assumed the rows are corresponding with the objects.
type	indicates whether the provided matrices in "List" are either data matrices, distance matrices or clustering results obtained from the data. If type="dist" the calculation of the distance matrices is skipped and if type="clust" the single source clustering is skipped. Type should be one of "data", "dist" or "clust".
distmeasure	A vector of the distance measures to be used on each data matrix. Should be one of "tanimoto", "euclidean", "jaccard", "hamming". Defaults to c("tanimoto","tanimoto").
normalize	Logical. Indicates whether to normalize the distance matrices or not, defaults to c(FALSE, FALSE) for two data sets. This is recommended if different distance types are used. More details on normalization in Normalization.
method	A method of normalization. Should be one of "Quantile","Fisher-Yates", "standardize","Range" or any of the first letters of these names. Default is c(NULL,NULL) for two data sets.
clust	Choice of clustering function (character). Defaults to "agnes".
linkage	Choice of inter group dissimilarity (character) for each data set. Defaults to c("flexible", "flexible") for two data sets.
alpha	The parameter alpha to be used in the "flexible" linkage of the agnes function. Defaults to 0.625 and is only used if the linkage is set to "flexible".
nrclusters	The number of clusters to divide each individual dendrogram in. Default is c(7,7) for two data sets.
gap	Logical. Whether the optimal number of clusters should be determined with the gap statistic. Default is FALSE.
maxK	The maximal number of clusters to investigate in the gap statistic. Default is 15.
ensembleMethod	A character string indicating the consensus function to use. One of "CSPA", "HGPA", "MCLA" or "Best". Defaults to "CSPA".
waitingtime	The maximum number of seconds to wait for the external program to write its result before stopping. Defaults to 300.
file_number	An identifier appended to the temporary files exchanged with the external program. Defaults to 00.
executable	Either FALSE (the default) or the path/flag identifying the external graph-partitioning executable (or MATLAB) to use. When FALSE the function stops, since no consensus can be computed in pure R. Precompiled binaries are not bundled with the package.

Value

The returned value is a list of two elements:

DistM	A NULL object
Clust	The resulting consensus clustering

The value has class 'Ensemble'.

References

Strehl A. and Ghosh J. (2002). Cluster ensembles - a knowledge reuse framework for combining multiple partitions. Journal of Machine Learning Research, 3, 583-617.

Examples

## Not run: 
data(mosaic_toy)
L <- mosaic_toy$List

# Requires an external graph-partitioning executable / MATLAB to be
# configured through the 'executable' argument.
res <- EnsembleClustering(List = L, type = "data",
  distmeasure = c("euclidean", "euclidean"), normalize = c(FALSE, FALSE),
  method = c(NULL, NULL), clust = "agnes",
  linkage = c("flexible", "flexible"), nrclusters = c(7, 7),
  gap = FALSE, maxK = 15, ensembleMethod = "CSPA",
  executable = "./EnsembleClusteringC")

## End(Not run)

MosaiClusteR documentation built on Oct. 25, 2026, 1:08 a.m.