| Title: | Parallelized Minimum Redundancy, Maximum Relevance (mRMR) |
|---|---|
| Description: | Computes mutual information matrices from continuous, categorical and survival variables, as well as feature selection with minimum redundancy, maximum relevance (mRMR) and a new ensemble mRMR technique. Published in De Jay et al. (2013) <doi:10.1093/bioinformatics/btt383>. |
| Authors: | Nicolas De Jay [aut], Simon Papillon-Cavanagh [aut], Catharina Olsen [aut], Gianluca Bontempi [aut], Bo Li [aut], Christopher Eeles [ctb], Benjamin Haibe-Kains [aut, cre] |
| Maintainer: | Benjamin Haibe-Kains <[email protected]> |
| License: | Artistic-2.0 |
| Version: | 2.1.2 |
| Built: | 2024-09-05 03:21:38 UTC |
| Source: | https://github.com/bhklab/mrmre |
The adjency matrix is a directed matrix of 0's and 1's indicating if there is a link between features.
## S4 method for signature 'mRMRe.Network' adjacencyMatrix(object)## S4 method for signature 'mRMRe.Network' adjacencyMatrix(object)
object |
a |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Build an mRMR-based network and display adjacency matrix (topology) network <- new("mRMRe.Network", data = feature_data, target_indices = c(1, 2), levels = c(2, 1), layers = 1) adjacencyMatrix(network)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Build an mRMR-based network and display adjacency matrix (topology) network <- new("mRMRe.Network", data = feature_data, target_indices = c(1, 2), levels = c(2, 1), layers = 1) adjacencyMatrix(network)
The causality data is compute using the co-information lattice algorithm on each V-structure (feature, target, feature). Given that this procedure is computed for each pair of features, the minimum result is kept. A negative score indicates putative causality of the feature to the target.
## S4 method for signature 'mRMRe.Filter' causality(object) ## S4 method for signature 'mRMRe.Network' causality(object)## S4 method for signature 'mRMRe.Filter' causality(object) ## S4 method for signature 'mRMRe.Network' causality(object)
object |
a |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) causality(filter)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) causality(filter)
This dataset contains gene expression of 200 cancer cell lines for which sensitivity (IC50) to Camptothecin was measured (release 2).
data(cgps)data(cgps)
The cgps dataset is composed of three objects
Dataframe containing gene annotations
Matrix containing expressions of 1000 genes; cell lines in rows, genes in columns
Drug sensitivity measurements (IC50) for Camptothecin
Camptothecin is a drug mainly used in colorectal cancer.
http://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-783
http://www.nature.com/nature/journal/v483/n7391/extref/nature11005-s2.zip
Garnett MJ et al. "Systematic identification of genomic markers of drug sensitivity in cancer cells", Nature, 483:570-575, 2012.
set.thread.count(2) data(cgps) message("Gene expression data:") print(cgps.ge[1:3, 1:3]) message("Gene annotations:") print(head(cgps.annot)) message("Drug sensitivity (IC50) values:") print(head(cgps.ic50))set.thread.count(2) data(cgps) message("Gene expression data:") print(cgps.ge[1:3, 1:3]) message("Gene annotations:") print(head(cgps.annot)) message("Drug sensitivity (IC50) values:") print(head(cgps.ic50))
Correlate is a function that cestimates correlation between two variables, which can be either continuous, categorical (ordered factor) or censored (survival data).
correlate(X, Y, method = c("pearson", "spearman", "kendall", "frequency", "cramersv", "cindex"),strata, weights, outX = TRUE, bootstrap_count = 0, alpha = 0.05, alternative = c("two.sided", "less", "greater"))correlate(X, Y, method = c("pearson", "spearman", "kendall", "frequency", "cramersv", "cindex"),strata, weights, outX = TRUE, bootstrap_count = 0, alpha = 0.05, alternative = c("two.sided", "less", "greater"))
X |
Vector of type numeric, ordered factor, or Surv. |
Y |
Vector of type numeric, ordered factor, or Surv of same length as |
method |
One of the following values: pearson, spearman, kendall, frequency, cramersv, or cindex. |
strata |
Vector of type factor corresponding to the sample strata. |
weights |
Vector of type numeric corresponding to the sample weights. |
outX |
For cindex, if set to |
bootstrap_count |
If set to |
alpha |
The probability of Type I error that is, rejecting a null hypothesis when it is in fact true |
alternative |
a character string specifying the alternative hypothesis,must be one of two.sided (default), greater or less. You can specify just the initial letter. |
The correlate function could be used to measure correlation between any types of variables:
Pearson, Spearman, Kendall or concordance index
concordance index (Somers' Dxy)
concordance index (Somers' Dxy)
Carmer's V
concordance index (Somers' Dxy)
concordance index (Somers' Dxy)
Part of the code underlying correlate is also used in mim method of the mRMRe.Data object because correlations are used to build the mutual information matrix in order for feature selection to take place. This is why these two functions have many argiuments in common.
estimate |
point estimate |
se |
standard error |
lower |
lower confidence bound |
upper |
upper confidence bound |
p |
p-value |
n |
sample size |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) ## load data data(cgps) ## spearman correlation coefficent between the first gene and Camptothecin IC50 correlate(X=cgps.ge[ ,1], Y=cgps.ic50, method="spearman") ## concordance index between the first gene and Camptothecin IC50 correlate(X=cgps.ge[ ,1], Y=cgps.ic50, method="cindex")set.thread.count(2) ## load data data(cgps) ## spearman correlation coefficent between the first gene and Camptothecin IC50 correlate(X=cgps.ge[ ,1], Y=cgps.ic50, method="spearman") ## concordance index between the first gene and Camptothecin IC50 correlate(X=cgps.ge[ ,1], Y=cgps.ic50, method="cindex")
Export the concordance index
samplesA |
... |
samplesB |
... |
samplesC |
... |
samplesD |
... |
sampleStrata |
... |
sampleWeights |
... |
sampleStratumCount |
... |
outX |
... |
ratio |
... |
concordantWeights |
... |
discordantWeights |
... |
uninformativeWeights |
... |
relevantWeights |
... |
No return. Modifies the ratio argument by reference.
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
Export the filters
childrenCountPerLevel |
... |
dataMatrix |
... |
priorsMatrix |
... |
priorsWeight |
... |
sampleStrata |
... |
sampleWeights |
... |
featureTypes |
... |
sampleCount |
... |
featureCount |
... |
sampleStratumCount |
... |
targetFeatureIndices |
... |
continuousEstimator |
... |
outX |
... |
bootstrapCount |
... |
miMatrix |
... |
Exhaustively computes the minimum redudancy maximim relevance features from the mutual information matrix, returing a list of solutions where each item is a numeric index of selected features.
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
Export the filters
solutionCount |
... |
solutionLength |
... |
dataMatrix |
... |
priorsMatrix |
... |
priorsWeight |
... |
sampleStrata |
... |
sampleWeights |
... |
featureTypes |
... |
sampleCount |
... |
featureCount |
... |
sampleStratumCount |
... |
targetFeatureIndices |
... |
continuousEstimator |
... |
outX |
... |
bootstrapCount |
... |
miMatrix |
... |
Bootstraps and estimate of the minimum redundancy maximim relevance features from the mutual information, returning a list where each item is a numeric vector of selected feature indices.
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
Export mim
dataMatrix |
... |
priorsMatrix |
... |
priorsWeight |
... |
sampleStrata |
... |
sampleWeights |
... |
featureTypes |
... |
sampleCount |
... |
featureCount |
... |
sampleStratumCount |
... |
continuousEstimator |
... |
outX |
... |
bootstrapCount |
... |
miMatrix |
... |
Returns the mutual information matrix.
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
The feature count is simply the total number of feature considered in the mRMRe procedure.
## S4 method for signature 'mRMRe.Data' featureCount(object) ## S4 method for signature 'mRMRe.Filter' featureCount(object) ## S4 method for signature 'mRMRe.Network' featureCount(object)## S4 method for signature 'mRMRe.Data' featureCount(object) ## S4 method for signature 'mRMRe.Filter' featureCount(object) ## S4 method for signature 'mRMRe.Network' featureCount(object)
object |
a |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) featureCount(feature_data) filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) featureCount(filter)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) featureCount(feature_data) filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) featureCount(filter)
the featureData consists of the numerical value of each feature for each sample considered
## S4 method for signature 'mRMRe.Data' featureData(object)## S4 method for signature 'mRMRe.Data' featureData(object)
object |
a |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) featureData(feature_data)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) featureData(feature_data)
featureNames are the names of the features given as input to the mRMRe procedure.
## S4 method for signature 'mRMRe.Data' featureNames(object) ## S4 method for signature 'mRMRe.Filter' featureNames(object) ## S4 method for signature 'mRMRe.Network' featureNames(object)## S4 method for signature 'mRMRe.Data' featureNames(object) ## S4 method for signature 'mRMRe.Filter' featureNames(object) ## S4 method for signature 'mRMRe.Network' featureNames(object)
object |
a |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) featureNames(feature_data) filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) featureNames(filter)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) featureNames(feature_data) filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) featureNames(filter)
This methods allows you to get the number of cores currently accessible to openMP
thread_count |
number of OPENMP threads to be used |
Return the current number of cores accessible to openMP for C level parallelization
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
This methods allows you to retrieve the number of cores currently accessible to openMP
get.thread.count()get.thread.count()
Return the number of cores accessible to openMP for C level parallelization.
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
get.thread.count()get.thread.count()
In both mRMRe.Filter and mRMRe.Network objects, a sparse mutual information matrix is computed for the mRMRe procedure and this lazy-evaluated matrix is returned. In the context of a a mRMRe.Data 'mim', the full pairwise mutual information matrix is computed and returned.
## S4 method for signature 'mRMRe.Data' mim(object, prior_weight, continuous_estimator, outX, bootstrap_count) ## S4 method for signature 'mRMRe.Filter' mim(object, method) ## S4 method for signature 'mRMRe.Network' mim(object)## S4 method for signature 'mRMRe.Data' mim(object, prior_weight, continuous_estimator, outX, bootstrap_count) ## S4 method for signature 'mRMRe.Filter' mim(object, method) ## S4 method for signature 'mRMRe.Network' mim(object)
object |
a |
prior_weight |
a numeric value [0,1] of indicating the impact of priors (mRMRe.Data only). |
continuous_estimator |
an estimator of the mutual information between features: either "pearson", "spearman", "kendall", "frequency" (mRMRe.Data only). |
outX |
a boolean used in the concordance index estimator to keep or throw out ties (mRMRe.Data only). |
bootstrap_count |
an integer indicating the number of bootstrap resampling used in estimation (mRMRe.Data only). |
method |
either "mi" or "cor"; the latter will return the correlation coefficients (rho) while the former will return the mutual information (-0.5 * log(1 - (rho^2))). |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Calculate the pairwise mutual information matrix mim(feature_data) filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) # Obtain the sparse (lazy-evaluated) mutual information matrix. mim(filter)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Calculate the pairwise mutual information matrix mim(feature_data) filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) # Obtain the sparse (lazy-evaluated) mutual information matrix. mim(filter)
"mRMRe.Data"
mRMRe.Data is the class containing datasets. Most if not all of the routines in the mRMRe package use mRMRe.Data objects as primary input.
Such an object is instantiated with a data frame containing the sample sets and optionally, stratum, weight vectors and a prior matrix. In addition to basic accession functions, we describe several methods which serve to manipulate the contents of the dataset.
Note that mRMR.data function is a wrapper to easily create mRMRe.Data objects.
Objects are created via calls of the form new("mRMRe.Data", data, strata, weights, priors).
data: is expected to be a data frame with samples and features respectively organized as rows and columns. The columns
have to be of type :numeric, ordered factor, Surv and respectively interpreted as :continuous, discrete and survival variables.
strata: is expected to be a vector of type :ordered factor with the strata associated to the samples provided
in data.
weights: is expected to be a vector of type :numeric with the weights associated to the samples provided
in data.
priors: is expected to be a matrix of type :numeric where priors[i, j]: denotes an forced association between
features i and j in data. The latter takes into consideration the directionality of the relationship and must be a value
between 0 and 1.
The mim method computes and returns a mutual information matrix. A correlation between continuous features is estimated
using an estimator specified in continuous_estimator; currently, :pearson, spearman, kendall, frequency are supported.
The estimator for discrete features is Cramer's V and for all other combinations, concordance index.
When outX is set to TRUE, ties are ignored when computing the concordance index and otherwise, these are considered.
The correlations are first computed per strata and these are then combined by the inverse variance weight mean of the estimates
using a bootstrap_count number of bootstraps if the former parameter is greater than 0, and by the relative weights of each
strata otherwise. The resulting correlation is then summated with the corresponding value in the priors matrix with the
latter being weighed for a proportion prior_weight of a final, biased correlation.
sample_names:Object of class "character" containing the sample names.
feature_names:Object of class "character" containing the feature names.
feature_types:Object of class "numeric" containing the internal representation of features/variables: 1 for numeric, 2 for ordered factor, and 3 for survival data
data:Object of class "matrix" containing the internal representation of the data set.
strata:Object of class "numeric" containing the feature strata.
weights:Object of class "numeric" containing sample weights.
priors:Object of class "matrix" containing the priors.
signature(object = "mRMRe.Data"): Returns the number of features.
signature(object = "mRMRe.Data"): Returns a data frame corresponding to the data set.
signature(object = "mRMRe.Data"): Returns a vector containing the feature names.
signature(object = "mRMRe.Data", prior_weight = 0,
continuous_estimator = c("pearson", "spearman", "kendall", "frequency"),
outX = TRUE, bootstrap_count = 0): Computes and returns the mutual information matrix.
signature(object = "mRMRe.Data"): Returns a matrix containing the priors.
signature(object = "mRMRe.Data", value): Sets the prior matrix.
signature(object = "mRMRe.Data"): Returns the number of samples.
signature(object = "mRMRe.Data"): Returns a vector containing sample names.
signature(object = "mRMRe.Data"): Returns a vector containing sample strata.
signature(object = "mRMRe.Data", value): Sets the sample strata.
signature(object = "mRMRe.Data"): Returns a vector containing sample weights.
signature(object = "mRMRe.Data"): Sets the sample weights.
signature(object = "mRMRe.Data", row_indices, column_indices): Returns another data object containing only the specified samples and features (rows and columns, respectively.)
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
mRMRe.Filter-class, mRMRe.Network-class
showClass("mRMRe.Data") set.thread.count(2) ## load data data(cgps) ## equivalent ways of building an mRMRe.Data object ge <- mRMR.data(data = data.frame(cgps.ge[ , 1:10, drop=FALSE])) ge <- new("mRMRe.Data", data = data.frame(cgps.ge[ , 1:10, drop=FALSE])) ## print data print(featureData(ge)[1:3, 1:3]) ## print feature names print(featureNames(ge)) ## print the first sample names print(head(sampleNames(ge))) ## print the first sample weights print(head(sampleWeights(ge)))showClass("mRMRe.Data") set.thread.count(2) ## load data data(cgps) ## equivalent ways of building an mRMRe.Data object ge <- mRMR.data(data = data.frame(cgps.ge[ , 1:10, drop=FALSE])) ge <- new("mRMRe.Data", data = data.frame(cgps.ge[ , 1:10, drop=FALSE])) ## print data print(featureData(ge)[1:3, 1:3]) ## print feature names print(featureNames(ge)) ## print the first sample names print(head(sampleNames(ge))) ## print the first sample weights print(head(sampleWeights(ge)))
"mRMRe.Filter"
mRMRe.Filter is a wrapper for various variants of the maximum relevance minimum redundancy (mRMR) feature selection/filter.
Note that mRMR.classic and mRMR.ensemble functions are wrappers to easily perform classical (single) and ensemble mRMR feature selection.
Objects are created via calls of the form new("mRMRe.Filter", data, prior_weight,
target_indices, levels, method, continuous_estimator, outX, bootstrap_count).
data: is expected to be a mRMRe.Data object.
target_indices: is expected to be a vector of type integer containing the indices of the features
that will serve as targets for the feature selections.
levels: is expected to be a vector of type integer containing the number of children of each element
at each level of the resulting filter tree.
method: is expected to be either exhaustive or bootstrap. The former uses the whole dataset to pick siblings in the tree according
to the mRMR metric, while the latter perform the classical mRMR feature selection on several bootrstap selections of the dataset.
continuous_estimator: it specifies the estimators for correlation between two continuous variables; value is either pearson, spearman, kendall, frequency,
outX: set to TRUE (default value) to not count pairs of
observations tied on x as a relevant pair. This results in a
Goodman-Kruskal gamma type rank correlation.
bootstrap_count: Number of bootstraps to statistically compare
the mRMR scores of each solution.
Since a mutual information matrix must be computed in order for feature selection to take place, the remaining arguments
are identical to those required by the mim method of the mRMRe.Data object.
filters:Object of class "list" containing for each target a solutions matrix.
mi_matrix:Object of class "matrix" containing the combined mutual information matrix of the relevant targets.
causality_list:Object of class "list" containing for each target a vector of causality coefficients between the target and its predictors.
sample_names:Object of class "character" containing the sample names.
feature_names:Object of class "character" containing the feature names.
target_indices:Object of class "integer" containing the target indices.
fixed_feature_count:Object of class "integer" containing the number of fixed features.
levels:Object of class "integer" containing the desired topology of the tree.
scores:Object of class "list" containing the mRMR score of selected features, respective to filters.
signature(object = "mRMRe.Filter"): ...
signature(object = "mRMRe.Filter"): Returns the number of features.
signature(object = "mRMRe.Filter"): Returns a vector containing the feature names.
signature(object = "mRMRe.Filter"): Returns the potentially partial mutual information matrix used for feature selection.
signature(object = "mRMRe.Filter"): Returns the number of samples.
signature(object = "mRMRe.Filter"): Returns a vector containing sample names.
signature(object = "mRMRe.Filter", mi_threshold = -Inf,
causality_threshold = Inf):
Returns a matrix in which each column represents a different solution (path from root of the tree to a leaf.)
signature(object = "mRMRe.Filter"): Returns a vector containing the target indices.
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
Ding, C. and Peng, H. (2005). "Minimum redundancy feature selection from microarray gene expression data". Journal of bioinformatics and computational biology, 3(2):185–205.
showClass("mRMRe.Filter") set.thread.count(2) ## load data data(cgps) ## build an mRMRe.Data object ge <- mRMR.data(data = data.frame(cgps.ge[ , 1:100, drop=FALSE])) ## perform a classic (single) mRMR to select the 10 genes the most correlated with ## the first gene but the less correlated between each other exect <- system.time(fs <- new("mRMRe.Filter", data = ge, target_indices = 1, levels = c(8, 1, 1, 1, 1))) print(exect) ## print the index of the selected features for each distinct mRMR solutions print(solutions(fs)[[1]]) ## print the names of the selected features for each distinct mRMR solutions print(apply(solutions(fs)[[1]], 2, function(x, y) { return(y[x]) }, y=featureNames(ge)))showClass("mRMRe.Filter") set.thread.count(2) ## load data data(cgps) ## build an mRMRe.Data object ge <- mRMR.data(data = data.frame(cgps.ge[ , 1:100, drop=FALSE])) ## perform a classic (single) mRMR to select the 10 genes the most correlated with ## the first gene but the less correlated between each other exect <- system.time(fs <- new("mRMRe.Filter", data = ge, target_indices = 1, levels = c(8, 1, 1, 1, 1))) print(exect) ## print the index of the selected features for each distinct mRMR solutions print(solutions(fs)[[1]]) ## print the names of the selected features for each distinct mRMR solutions print(apply(solutions(fs)[[1]], 2, function(x, y) { return(y[x]) }, y=featureNames(ge)))
"mRMRe.Network"
mRMRe.Network is a wrapper for inferring a network of features based on mRMR feature selection.
Objects are created via calls of the form new("mRMRe.Network", data, prior_weight,
target_indices, levels, layers, ..., mi_threshold, causality_threshold).
layers: is expected to be an integer specifying the number of layers of network inference desired. When multiple
layers are desired, the elements of the solutions found in the last step of feature selection are used as the targets of the next
step.
Since networking involves filter processing, the remaining arguments are identical to those required by solutions method of
the mRMRe.Filter object and mim method of the mRMRe.Data object.
topologies:Object of class "list" ~~
mi_matrix:Object of class "matrix" containing the combined mutual information matrix of the network elements.
causality_list:Object of class "list" containing for each target a vector of causality coefficients between the target and its predictors.
sample_names:Object of class "character" containing the sample names.
feature_names:Object of class "character" containing the feature names.
target_indices:Object of class "integer" containing the target indices.
signature(object = "mRMRe.Network"): Returns a matrix describing the topology of the network.
signature(object = "mRMRe.Network"): ...
signature(object = "mRMRe.Network"): Returns a list containing vectors containing causality coefficients between targets and predictors.
signature(object = "mRMRe.Network"): Returns a vector containing the feature names.
signature(object = "mRMRe.Network"): ...
signature(object = "mRMRe.Network"): Returns a vector containing sample names.
signature(object = "mRMRe.Network"): ...
signature(object = "mRMRe.Network"): ...
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
mRMRe.Filter-class, mRMRe.Data-class
showClass("mRMRe.Network") set.thread.count(2) ## load data data(cgps) ## build an mRMRe.Data object ge <- mRMR.data(data = data.frame(cgps.ge[ , 1:100, drop=FALSE])) ## build a network object with the 10 first genes and their children, ## 8 distinct mRMR feature selections of 5 genes for each gene exect <- system.time(netw <- new("mRMRe.Network", data = ge, target_indices = 1:10, levels = c(8, 1, 1, 1, 1), layers = 2)) print(exect) ## plot network using igraph ## Not run: visualize(netw)showClass("mRMRe.Network") set.thread.count(2) ## load data data(cgps) ## build an mRMRe.Data object ge <- mRMR.data(data = data.frame(cgps.ge[ , 1:100, drop=FALSE])) ## build a network object with the 10 first genes and their children, ## 8 distinct mRMR feature selections of 5 genes for each gene exect <- system.time(netw <- new("mRMRe.Network", data = ge, target_indices = 1:10, levels = c(8, 1, 1, 1, 1), layers = 2)) print(exect) ## plot network using igraph ## Not run: visualize(netw)
The priors matrix consists of a prior bias to be used in computation to mutual information between features.
## S4 method for signature 'mRMRe.Data' priors(object) ## S4 replacement method for signature 'mRMRe.Data' priors(object) <- value## S4 method for signature 'mRMRe.Data' priors(object) ## S4 replacement method for signature 'mRMRe.Data' priors(object) <- value
object |
a |
value |
a numeric matrix containing values from 0 to 1 (or NA), one per pairwise feature bias. |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) priors(feature_data)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) priors(feature_data)
The feature count is simply the total number of samples considered in the mRMRe procedure.
## S4 method for signature 'mRMRe.Data' sampleCount(object) ## S4 method for signature 'mRMRe.Filter' sampleCount(object) ## S4 method for signature 'mRMRe.Network' sampleCount(object)## S4 method for signature 'mRMRe.Data' sampleCount(object) ## S4 method for signature 'mRMRe.Filter' sampleCount(object) ## S4 method for signature 'mRMRe.Network' sampleCount(object)
object |
a |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) sampleCount(feature_data) filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) sampleCount(filter)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) sampleCount(feature_data) filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) sampleCount(filter)
sampleNames are the names of the samples given as input to the mRMRe procedure.
## S4 method for signature 'mRMRe.Data' sampleNames(object) ## S4 method for signature 'mRMRe.Filter' sampleNames(object) ## S4 method for signature 'mRMRe.Network' sampleNames(object)## S4 method for signature 'mRMRe.Data' sampleNames(object) ## S4 method for signature 'mRMRe.Filter' sampleNames(object) ## S4 method for signature 'mRMRe.Network' sampleNames(object)
object |
a |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) sampleNames(feature_data) filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) sampleNames(filter)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) sampleNames(feature_data) filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) sampleNames(filter)
The sampleStrata vector consists of a sampling stratification that will be used in computing mutual information between features. If known batch effects or sample stratification is present between samples, identify such subsets using this.
## S4 method for signature 'mRMRe.Data' sampleStrata(object) ## S4 replacement method for signature 'mRMRe.Data' sampleStrata(object) <- value## S4 method for signature 'mRMRe.Data' sampleStrata(object) ## S4 replacement method for signature 'mRMRe.Data' sampleStrata(object) <- value
object |
a |
value |
a factor vector identifying the stratification of samples. |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # No stratification (default) sampleStrata(feature_data) # Random stratification sampleStrata(feature_data) <- as.factor(sample(c(0,1), sampleCount(feature_data), replace=TRUE)) # Show result sampleStrata(feature_data)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # No stratification (default) sampleStrata(feature_data) # Random stratification sampleStrata(feature_data) <- as.factor(sample(c(0,1), sampleCount(feature_data), replace=TRUE)) # Show result sampleStrata(feature_data)
TODO
## S4 method for signature 'mRMRe.Data' sampleWeights(object) ## S4 replacement method for signature 'mRMRe.Data' sampleWeights(object) <- value## S4 method for signature 'mRMRe.Data' sampleWeights(object) ## S4 replacement method for signature 'mRMRe.Data' sampleWeights(object) <- value
object |
a |
value |
a numeric vector containing the biases of each sample in the mutual information computation. |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Uniform weight (default) sampleWeights(feature_data) # Random weighting sampleWeights(feature_data) <- runif(sampleCount(feature_data)) # Show result sampleWeights(feature_data)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Uniform weight (default) sampleWeights(feature_data) # Random weighting sampleWeights(feature_data) <- runif(sampleCount(feature_data)) # Show result sampleWeights(feature_data)
The scores method returns the scores of individual features in respect to previously selected features as per standard
mRMR procedure. For each target, the score of a feature is defined as the mutual information between the target and this feature
minus the average mutual information of previously selected features and this feature.
## S4 method for signature 'mRMRe.Data' scores(object, solutions) ## S4 method for signature 'mRMRe.Filter' scores(object) ## S4 method for signature 'mRMRe.Network' scores(object)## S4 method for signature 'mRMRe.Data' scores(object, solutions) ## S4 method for signature 'mRMRe.Filter' scores(object) ## S4 method for signature 'mRMRe.Network' scores(object)
object |
a |
solutions |
a set of solutions from mRMRe.Filter or mRMRe.Network to be used in computing the scores from a mRMRe.Data set. |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Create an mRMR filter and obtain the indices of selected features filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) scores(filter)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Create an mRMR filter and obtain the indices of selected features filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) scores(filter)
This methods allows you to set the number of cores currently accessible to openMP
thread_count |
number of OPENMP threads to be used |
No return, sets the number of cores available to openMP for C level parallelization.
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
This methods allows you to set the number of cores currently accessible to openMP
set.thread.count(thread_count)set.thread.count(thread_count)
thread_count |
number of OPENMP threads to be used |
No return, sets the number of cores available to openMP for C level parallelization.
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
# Access to number of available threads threads <- get.thread.count() # Force a single threaded openMP job set.thread.count(1) # Revert back to all accessible threads set.thread.count(threads)# Access to number of available threads threads <- get.thread.count() # Force a single threaded openMP job set.thread.count(1) # Revert back to all accessible threads set.thread.count(threads)
The 'solutions' method allows one to access the set of selected features resulting of the mRMR algorithm. More generally, the set of feature are identified by their indices in the inputed feature set (1 being the first feature (column)). At the network level, 'solutions' consists of the topology of the network, identifying which features is connected to others.
## S4 method for signature 'mRMRe.Filter' solutions(object, mi_threshold, causality_threshold, with_fixed_features) ## S4 method for signature 'mRMRe.Network' solutions(object)## S4 method for signature 'mRMRe.Filter' solutions(object, mi_threshold, causality_threshold, with_fixed_features) ## S4 method for signature 'mRMRe.Network' solutions(object)
object |
a |
mi_threshold |
a numeric value used in filtering the features based on their mRMR scores, features that do not pass the threshold will be set at NA. |
causality_threshold |
a numeric value used in filtering the features based on their causality scores, features that do not pass the threshold will be set at NA |
with_fixed_features |
a boolean indicating if fixed features are used in the computation, default TRUE |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Create an mRMR filter and obtain the indices of selected features filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) solutions(filter) # Build an mRMR-based network and obtain feature connections (topology) network <- new("mRMRe.Network", data = feature_data, target_indices = c(1, 2), levels = c(2, 1), layers = 1) solutions(network)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Create an mRMR filter and obtain the indices of selected features filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) solutions(filter) # Build an mRMR-based network and obtain feature connections (topology) network <- new("mRMRe.Network", data = feature_data, target_indices = c(1, 2), levels = c(2, 1), layers = 1) solutions(network)
This method is used to extract a subset of the current mRMRe.Data object.
## S4 method for signature 'mRMRe.Data' subsetData(object, row_indices, column_indices)## S4 method for signature 'mRMRe.Data' subsetData(object, row_indices, column_indices)
object |
a |
row_indices |
An integer vector of the rows to be included in the subset. |
column_indices |
An integer vector of the columns to be included in the subset. |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Subset the same dimensions, equivalent to making a copy feature_data_copy <- subsetData(feature_data, row_indices=sampleCount(feature_data), column_indices=featureCount(feature_data)) # Use only half of the samples feature_data_samples <- subsetData(feature_data, row_indices=sampleCount(feature_data)/2) # Use only half of the features feature_data_features <- subsetData(feature_data, column_indices=featureCount(feature_data))set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Subset the same dimensions, equivalent to making a copy feature_data_copy <- subsetData(feature_data, row_indices=sampleCount(feature_data), column_indices=featureCount(feature_data)) # Use only half of the samples feature_data_samples <- subsetData(feature_data, row_indices=sampleCount(feature_data)/2) # Use only half of the features feature_data_features <- subsetData(feature_data, column_indices=featureCount(feature_data))
The 'target' method allows you to access the target of a mRMR procedure. In a mRMRe.Network setting, the target consists of the seed or the starting set of features given in the network building.
## S4 method for signature 'mRMRe.Filter' target(object) ## S4 method for signature 'mRMRe.Network' target(object)## S4 method for signature 'mRMRe.Filter' target(object) ## S4 method for signature 'mRMRe.Network' target(object)
object |
a |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Create an mRMR filter and obtain the targets of that filter filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) target(filter) # Build an mRMR-based network and obtain targets (seeds) of the network network <- new("mRMRe.Network", data = feature_data, target_indices = c(1, 2), levels = c(2, 1), layers = 1) target(network)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Create an mRMR filter and obtain the targets of that filter filter <- mRMR.classic("mRMRe.Filter", data = feature_data, target_indices = 3:5, feature_count = 2) target(filter) # Build an mRMR-based network and obtain targets (seeds) of the network network <- new("mRMRe.Network", data = feature_data, target_indices = c(1, 2), levels = c(2, 1), layers = 1) target(network)
The 'visualize' methods allows the visual display of an inferred mRMRe.Network topology.
## S4 method for signature 'mRMRe.Network' visualize(object)## S4 method for signature 'mRMRe.Network' visualize(object)
object |
a |
Nicolas De Jay, Simon Papillon-Cavanagh, Benjamin Haibe-Kains
set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Build an mRMR-based network and display it network <- new("mRMRe.Network", data = feature_data, target_indices = c(1), levels = c(3, 1), layers = 2) visualize(network)set.thread.count(2) data(cgps) feature_data <- mRMR.data(data = data.frame(cgps.ge)) # Build an mRMR-based network and display it network <- new("mRMRe.Network", data = feature_data, target_indices = c(1), levels = c(3, 1), layers = 2) visualize(network)