This function repeatedly resamples the samples (rows) in supplied data and identifies modules on the resampled data.
sampledBlockwiseModules(
datExpr,
nRuns,
startRunIndex = 1,
endRunIndex = startRunIndex + nRuns - skipUnsampledCalculation,
replace = FALSE,
fraction = if (replace) 1.0 else 0.63,
randomSeed = 12345,
checkSoftPower = TRUE,
nPowerCheckSamples = 2000,
skipUnsampledCalculation = FALSE,
corType = "pearson",
power = 6,
networkType = "unsigned",
saveTOMs = FALSE,
saveTOMFileBase = "TOM",
...,
verbose = 2, indent = 0)
A list with one component per run. Each component is a list with the following components:
The output of the function blockwiseModules
applied to a resampled data set.
Indices of the samples selected for the resampled data step for this run.
Actual soft-thresholding powers used in this run.
Expression data. A matrix (preferred) or data frame in which columns are genes and rows ar samples.
Number of sampled network construction and module identification runs. If skipUnsampledCalculation
is FALSE
,
one extra calculation (the first) will contain the unsampled calculation.
Number to be assigned to the start run. The run number or index is used to make saved files unique. It is also used in setting the seed for each run to allow the runs to be replicated in smaller or larger batches.
Number (index) of the last run. If given, nRuns
is ignored.
Logical: should samples (observations or rows in entries in multiExpr
) be sampled with replacement?
Fraction of samples to sample for each run.
Integer specifying the random seed. If non-NULL, the random number generator state is saved before the seed is set
and restored at the end of the function. If NULL
, the random number generator state is not saved nor
changed at the start, and not restored at the end.
Logical: should the soft-tresholding power be adjusted to approximately match the connectivity distribution of the sampled data set and the full data set?
Number of genes to be sampled from the full data set to calculate connectivity and match soft-tresholding powers.
Logical: should a calculation on original (not resampled) data be skipped?
Character string specifying the correlation to be used. Allowed values are (unique
abbreviations of) "pearson"
and "bicor"
, corresponding to Pearson and bidweight
midcorrelation, respectively. Missing values are handled using the pairwise.complete.obs
option.
Soft-thresholding power for network construction.
network type. Allowed values are (unique abbreviations of) "unsigned"
,
"signed"
, "signed hybrid"
. See adjacency
.
Logical: should the networks (topological overlaps) be saved for each run? Note that for large data sets (tens of thousands of nodes) the TOM files are rather large.
Character string giving the base of the file names for TOMs. The actual file names will consist of a
concatenation of saveTOMFileBase
and "-run-<run
number>-Block-<block number>.RData"
.
Other arguments to blockwiseModules
.
integer level of verbosity. Zero means silent, higher values make the output progressively more and more verbose.
indentation for diagnostic messages. Zero means no indentation, each unit adds two spaces.
Peter Langfelder
For each run, samples (but not genes) are randomly sampled to obtain a perturbed data set; a full network analysis and module identification is carried out, and the results are returned in a list with one component per run.
For each run, the soft-thresholding power can optionally be adjusted such that the mean adjacency in the re-sampled data set equals the mean adjacency in the original data.
An application of this function is described in the motivational example section of
Langfelder P, Horvath S (2012) Fast R Functions for Robust Correlations and Hierarchical Clustering. Journal of Statistical Software 46(11) 1-17; PMID: 23050260 PMCID: PMC3465711
blockwiseModules
for the underlying network analysis and module identification;
sampledHierarchicalConsensusModules
for a similar resampling analysis of consensus networks.