Merges hierarchical consensus modules that are too close as measured by the correlation of their eigengenes.
hierarchicalMergeCloseModules(
# input data
multiExpr,
multiExpr.imputed = NULL,
labels, # Optional starting eigengenes
MEs = NULL,
unassdColor = if (is.numeric(labels)) 0 else "grey",
# If missing data are present, impute them?
impute = TRUE,
# Options for eigengene network construction
networkOptions,
# Options for constructing the consensus
consensusTree,
calibrateMESimilarities = FALSE,
# Merging options
cutHeight = 0.2,
iterate = TRUE,
# Output options
relabel = FALSE,
colorSeq = NULL,
getNewMEs = TRUE,
getNewUnassdME = TRUE,
# Options controlling behaviour of the function
trapErrors = FALSE,
verbose = 1, indent = 0)
Expression data in the multi-set format (see multiData). A vector of
lists, one per set. Each set must contain a component data that contains the expression data, with
rows corresponding to samples and columns to genes or probes.
If multiExpr contain missing data, this argument can be used to supply the
expression data with missing data imputed. If not given, the impute.knn function will
be used to impute the missing data within each module (see imputeByModule.
A vector (numeric, character or a factor) giving module labels for genes (variables) in multiExpr.
If module eigengenes have been calculated before, the user can save some computational time
by inputting them. MEs should have the same format as multiExpr.
If they are not given, they will be calculated.
The label (value in labels)
that represents unassigned genes. Module of this label will
not enter the module eigengene clustering and will not be merged with other modules.
Should missing values be imputed in eigengene calculation? If imputation is disabled, the
presence of NA entries will cause the eigengene calculation to fail and eigengenes will be
replaced by their hubgene approximation. See moduleEigengenes for more details.
A single list of class NetworkOptions giving options for network calculation for all of the
networks, or a multiData structure containing one such list for each input data set.
A list specifying the consensus calculation. See newConsensusTree for details.
Logical: should module eigengene similarities be calibrated? This setting overrides the calibration options
in consensusTree.
Maximum dissimilarity (i.e., 1-correlation) that qualifies modules for merging.
Controls whether the merging procedure should be repeated until there is no change. If FALSE, only one iteration will be executed.
Controls whether, after merging, color labels should be ordered by module size.
Color labels to be used for relabeling. Defaults to the standard color order used
in this package if colors are not numeric, and to integers starting from 1 if
colors is numeric.
Controls whether module eigengenes of merged modules should be calculated and returned.
When doing module eigengene manipulations, the function does not normally
calculate the eigengene of the 'module' of unassigned ('grey') genes. Setting this option to
TRUE will force the calculation of the unassigned eigengene in the returned newMEs, but not
in the returned oldMEs.
Controls whether computational errors in calculating module eigengenes, their
dissimilarity, and merging trees should be trapped. If TRUE, errors will be trapped and the
function will return the input colors. If FALSE, errors will cause the function to stop.
Controls verbosity of printed progress messages. 0 means silent, up to (about) 5 the verbosity gradually increases.
A single non-negative integer controlling indentation of printed messages. 0 means no indentation, each unit above that adds two spaces.
If no errors occurred, a list with components
Labels for the genes corresponding to merged modules. The function attempts to
mimic the mode of the input labels: if the input labels is numeric, character and
factor, respectively, so is the output. Note, however, that if the function performs relabeling, a
standard sequence of labels will be used: integers starting at 1 if the input labels is
numeric, and a sequence of color labels otherwise (see colorSeq above).
Hierarchical clustering dendrogram (average linkage) of the eigengenes of the most
recently computed tree. If iterate was set TRUE, this will be the dendrogram of the merged
modules, otherwise it will be the dendrogram of the original modules.
Hierarchical clustering dendrogram (average linkage) of the eigengenes of the original modules.
The input cutHeight.
Module eigengenes of the original modules in the sets given by useSets.
Module eigengenes of the merged modules in the sets given by useSets.
A logical set to TRUE.
If an error occurred and trapErrors==TRUE, the list only contains these components:
A copy of the input colors.
a logical set to FALSE.
This function merges input modules
that are closely related. The similarities are quantified by correlations of module eigengenes; a
``consensus'' similarity is calculated using hierarchicalConsensusMEDissimilarity
according to the recipe in consensusTree. Once the
(dis-)similarities are calculated, average linkage hierarchical clustering of the module eigengenes is
performed, the dendrogram is cut at the height cutHeight and modules on each branch are merged.
The process is (optionally) repeated until no more modules are merged.
If, for a particular module, the module eigengene calculation fails, a hubgene approximation will be used.
The user should be aware that if a computational error occurs and trapErrors==TRUE,
the returned list (see below) will not contain all of the components returned upon normal execution.
multiSetMEs for calculation of (consensus) module eigengenes across multiple data sets;
newConsensusTree for information about consensus trees;
hierarchicalConsensusMEDissimilarity for calculation of hierarchical consensus eigengene
dissimilarity.