| Title: | Utilities to train and validate classifiers based on pair switching using the K-Top-Scoring-Pair (KTSP) algorithm |
|---|---|
| Description: | The package offer different classifiers based on comparisons of pair of features (TSP), using various decision rules (e.g., majority wins principle). |
| Authors: | Bahman Afsari <[email protected]>, Luigi Marchionni <[email protected]>, Wikum Dinalankara <[email protected]> |
| Maintainer: | Bahman Afsari <[email protected]>, Luigi Marchionni <[email protected]>, Wikum Dinalankara <[email protected]> |
| License: | GPL-2 |
| Version: | 1.41.0 |
| Built: | 2024-06-30 04:56:29 UTC |
| Source: | https://github.com/bioc/switchBox |
The switchBox package allows to train and apply a K-Top-Scoring-Pair (KTSP) classifier with learning mechanism proposed in Afsari et al (AOAS, 2014) and as used by Marchionni et al (BMC Genomics, 2013). KTSP is an extension of the TSP classifier described by Geman and colleagues (Bioinformatics, 2005). The TSP algorithm is a simple binary classifier based on the reversal ordering across phenotypes of two measurements (e.g. gene expression reversals from normal to cancer.
The switchBox package contains several utilities enabling to:
A) Filter the features to be used to develop the classifier (i.e., differentially expressed genes);
B) Compute the scores for all available feature pairs to identify the top performing TSP;
C) Compute the scores for selected feature pairs to identify the top performing TSP;
D) Identify the number of $K$ TSP to be used in the final classifier using the analysis of variance;
E) Compute individual TSP votes for one class or the other and combine the votes based on user defined methods;
F) Classify new samples based on the top KTSP based on various methods.
Bahman Afsari [email protected], Luigi Marchionni [email protected]
Afsari et al., "Rank Discriminants for Predicting Phenotypes from RNA Expression.", Annals of Applied Statistics, 2014, to appear.
Marchionni et al., "A simple and reproducible breast cancer prognostic test.", BMC Genomics, 2013, 14(1):336-342 http://www.ncbi.nlm.nih.gov/pubmed/23682826
Tan et al., "Simple decision rules for classifying human cancers from gene expression profiles.", Bioinformatics (2005) 21(20), 3896-3904. http://www.ncbi.nlm.nih.gov/pubmed/16105897
Xu et al., "Robust prostate cancer marker genes emerge from direct integration of inter-study microarray data" Bioinformatics (2005) 21(20), 3905-3911. http://www.ncbi.nlm.nih.gov/pubmed/16131522
Geman et al. "Classifying gene expression profiles from pairwise mRNA comparisons" Statistical applications in genetics and molecular biology (2004) 3.1 : 1071. http://www.ncbi.nlm.nih.gov/pubmed/16646797
KTSP.Classify classifies new test samples
using KTSP coming out of the function KTSP.Train.
This function was used in Marchionni et al, 2013, BMC Genomics,
and it is maintained only for backward compatibility.
It has been replaced by SWAP.KTSP.Classify.
KTSP.Classify(data, classifier, combineFunc)KTSP.Classify(data, classifier, combineFunc)
data |
the test data: a matrix in which the rows represent the genes and the columns the samples. |
classifier |
The output of |
combineFunc |
A user defined function to combine the predictions of the individual K TSPs. If missing the consensus classification among the majority of the TSPs will be used. |
Bahman Afsari [email protected], Luigi Marchionni [email protected]
See switchBox for the references.
KTSP.Train,
SWAP.KTSP.Classify,
################################################## ### Load gene expression data for the training set data(trainingData) ### Turn into a numeric vector with values equal to 0 and 1 trainingGroupNum <- as.numeric(trainingGroup) - 1 ### Show group variable for the TRAINING set table(trainingGroupNum) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- KTSP.Train(matTraining, trainingGroupNum, n=8) ### Show the classifier classifier ################################################## ### Testing on new data ### Load the example data for the TEST set data(testingData) ### Turn into a numeric vector with values equal to 0 and 1 testingGroupNum <- as.numeric(testingGroup) - 1 ### Show group variable for the TEST set table(testingGroupNum) ### Apply the classifier to one sample of the TEST set using ### sum of votes grearter than 2 testPrediction <- KTSP.Classify(matTesting, classifier, combineFunc = function(x) sum(x) < 2.5) ### Show prediction table(testPrediction, testingGroupNum)################################################## ### Load gene expression data for the training set data(trainingData) ### Turn into a numeric vector with values equal to 0 and 1 trainingGroupNum <- as.numeric(trainingGroup) - 1 ### Show group variable for the TRAINING set table(trainingGroupNum) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- KTSP.Train(matTraining, trainingGroupNum, n=8) ### Show the classifier classifier ################################################## ### Testing on new data ### Load the example data for the TEST set data(testingData) ### Turn into a numeric vector with values equal to 0 and 1 testingGroupNum <- as.numeric(testingGroup) - 1 ### Show group variable for the TEST set table(testingGroupNum) ### Apply the classifier to one sample of the TEST set using ### sum of votes grearter than 2 testPrediction <- KTSP.Classify(matTesting, classifier, combineFunc = function(x) sum(x) < 2.5) ### Show prediction table(testPrediction, testingGroupNum)
KTSP.Train trains a K-TSP classifier
for the specific phenotype of interest.
The classifiers resulting from using this function can be
passed to KTSP.Classify
for samples classification.
This function was used in Marchionni et al, 2013, BMC Genomics,
and it is maintained only for backward compatibility.
It has been replaced by SWAP.KTSP.Train.
KTSP.Train(data, situation, n)KTSP.Train(data, situation, n)
data |
the matrix of the values (usually gene expression) to be used to train the classifier. The columns represents samples and the rows represents the genes. |
situation |
an integer vector containing the training labels. Its elements should be one or zero. |
n |
The number of disjoint TSP used for classification. If before n pairs, the score drops to zero, the TSP with zero score are ignored. |
The KTSP classifier, a list containing the following elements:
TSPs |
a matrix containing TSPs indexes. |
score |
a vector containing TSPs scores. |
geneNames |
a matrix containing TSPs feature names. |
It should be passed to KTSP.Classify for classification of test samples.
Bahman Afsari [email protected], Luigi Marchionni [email protected]
See switchBox for the references.
KTSP.Classify,
SWAP.KTSP.Train,
################################################## ### Load gene expression data for the training set data(trainingData) ### Turn into a numeric vector with values equal to 0 and 1 trainingGroupNum <- as.numeric(trainingGroup) - 1 ### Show group variable for the TRAINING set table(trainingGroupNum) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- KTSP.Train(matTraining, trainingGroupNum, n=8) ### Show the classifier classifier################################################## ### Load gene expression data for the training set data(trainingData) ### Turn into a numeric vector with values equal to 0 and 1 trainingGroupNum <- as.numeric(trainingGroup) - 1 ### Show group variable for the TRAINING set table(trainingGroupNum) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- KTSP.Train(matTraining, trainingGroupNum, n=8) ### Show the classifier classifier
A numerical matrix containing gene expression matrix
for 70 genes and 307 breast cancer patients (test set data)
from the Buyse et al cohort
(see the mammaPrintData package).
data(testingData)data(testingData)
The matTesting matrix contains normalized
expression values for the 70 gene signature (rows)
across 307 samples (columns).
Group information (emph“bad” versus “good” prognosis)
is shown in colnames(matTesting).
This dataset corresponds to the breast cancer patients' cohort
published by Buyse and colleagues in JNCI (2006).
The gene expression matrix was obtained from
the mammaPrintData package as described
by Marchionni and colleagues in BMC Genomics (2013).
Bahman Afsari [email protected], Luigi Marchionni [email protected]
See switchBox for the references.
### Load gene expression data for the test set data(testingData) ### Show the class of the ``matTesting'' object class(matTesting) ### Show the dimentions of the ``matTesting'' matrix dim(matTesting) ### Show the first 10 sample names of the ``matTest'' matrix head(colnames(matTesting), n=10) testingGroup[1:10]### Load gene expression data for the test set data(testingData) ### Show the class of the ``matTesting'' object class(matTesting) ### Show the dimentions of the ``matTesting'' matrix dim(matTesting) ### Show the first 10 sample names of the ``matTest'' matrix head(colnames(matTesting), n=10) testingGroup[1:10]
A numerical matrix containing gene expression matrix
for 70 genes and 78 breast cancer patients (training set data)
from the Glas et al cohort
(see the mammaPrintData package).
data(trainingData)data(trainingData)
The matTraining matrix contains normalized
expression values for the 70 gene signature (rows)
across 78 samples (columns).
Group information (emph“bad” versus “good” prognosis)
is shown in colnames(matTraining).
This dataset corresponds to the breast cancer patients' cohort
published by Glas and colleagues in BMC Genomics (2006).
The gene expression matrix was obtained from
the mammaPrintData package as described
by Marchionni and colleagues in BMC Genomics (2013).
Bahman Afsari [email protected], Luigi Marchionni [email protected]
See switchBox for the references.
### Load gene expression data for the training set data(trainingData) ### Show the class of the ``matTraining'' object class(matTraining) ### Show the dimentions of the ``matTraining'' matrix dim(matTraining) ### Show the first 10 sample names of the ``matTraining'' matrix head(colnames(matTraining), n=10)### Load gene expression data for the training set data(trainingData) ### Show the class of the ``matTraining'' object class(matTraining) ### Show the dimentions of the ``matTraining'' matrix dim(matTraining) ### Show the first 10 sample names of the ``matTraining'' matrix head(colnames(matTraining), n=10)
SWAP.Calculate.BasicTSPScores calculates basic TSP
scores.
SWAP.Calculate.BasicTSPScores(phenoGroup, inputMat1, inputMat2 = NULL, classes = NULL, RestrictedPairs = NULL, handleTies = FALSE, verbose = FALSE, score_opts=list())SWAP.Calculate.BasicTSPScores(phenoGroup, inputMat1, inputMat2 = NULL, classes = NULL, RestrictedPairs = NULL, handleTies = FALSE, verbose = FALSE, score_opts=list())
phenoGroup |
is a factor containing the training phenotypes with two levels. |
inputMat1 |
is a numerical matrix containing the measurements (e.g., gene expression data) for choosing the first item of a top scoring pair. |
inputMat2 |
is a numerical matrix containing the
measurements for choosing the second item of a
top scoring pair. If |
classes |
is a character vector of length 2 providing the phenotype class labels (case followed by control). If NULL, the levels of phenoGroup will be taken as the labels. |
RestrictedPairs |
is a character matrix with two columns containing the feature pairs to be considered for score calculations. |
handleTies |
is a logical value indicating whether tie handling should be enabled or not. FALSE by default. |
verbose |
is a logical value indicating whether status messages will be printed or not throughout the function. FALSE by default. |
score_opts |
is a list of additional variables that will be passed on to the scoring function. |
The output is a list containing the following items:
labels |
the levels (phenotypes) in |
score |
is a vector containing the pair-wise scores. |
tieVote |
is a vector indicating the class the pair would vote for in the case of a tie. |
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
See SWAP.Calculate.SignedTSPScores
### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ### Compute the scores scores = SWAP.Calculate.BasicTSPScores(trainingGroup, matTraining[1:3, ]) # View the scores scores$score### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ### Compute the scores scores = SWAP.Calculate.BasicTSPScores(trainingGroup, matTraining[1:3, ]) # View the scores scores$score
SWAP.Calculate.SignedTSPScores calculates signed TSP
scores. The input provided to this function should be
already sanitized; to filter features and calculate
pairwise scores, use SWAP.CalculateScores instead.
SWAP.Calculate.SignedTSPScores(phenoGroup, inputMat1, inputMat2 = NULL, classes = NULL, RestrictedPairs = NULL, handleTies = FALSE, verbose = FALSE, score_opts=list())SWAP.Calculate.SignedTSPScores(phenoGroup, inputMat1, inputMat2 = NULL, classes = NULL, RestrictedPairs = NULL, handleTies = FALSE, verbose = FALSE, score_opts=list())
phenoGroup |
is a factor containing the training phenotypes with two levels. |
inputMat1 |
is a numerical matrix containing the measurements (e.g., gene expression data) for choosing the first item of a top scoring pair. |
inputMat2 |
is a numerical matrix containing the
measurements for choosing the second item of a
top scoring pair. If |
classes |
is a character vector of length 2 providing the phenotype class labels (case followed by control). If NULL, the levels of phenoGroup will be taken as the labels. |
RestrictedPairs |
is a character matrix with two columns containing the feature pairs to be considered for score calculations. |
handleTies |
is a logical value indicating whether tie handling should be enabled or not. FALSE by default. |
verbose |
is a logical value indicating whether status messages will be printed or not throughout the function. FALSE by default. |
score_opts |
is a list of additional variables that will be passed on to the scoring function. |
The output is a list containing the following items:
labels |
the levels (phenotypes) in |
score |
is a vector containing the pair-wise scores. |
tieVote |
is a vector indicating the class the pair would vote for in the case of a tie. |
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
See SWAP.Calculate.BasicTSPScores
### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ### Compute the scores scores = SWAP.Calculate.SignedTSPScores(trainingGroup, matTraining[1:3, ]) # View the scores scores$score### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ### Compute the scores scores = SWAP.Calculate.SignedTSPScores(trainingGroup, matTraining[1:3, ]) # View the scores scores$score
SWAP.CalculateScores calculates the pair-wise scores
between features pairs. The user may pass a filtering function
to reduce the number of starting features, or provide a restricted
set of pairs to limit the reported scores to this list. The user
can also pass a score-calculating function by either passing one of the
scoring functions available in the package(i.e.
SWAP.Calculate.SignedTSPScores and SWAP.Calculate.BasicTSPScores)
or a custom function.
SWAP.CalculateScores(inputMat, phenoGroup, classes = NULL, FilterFunc = SWAP.Filter.Wilcoxon, RestrictedPairs = NULL, handleTies = FALSE, verbose = FALSE, score_fn = signedTSPScores, score_opts = list(), ...)SWAP.CalculateScores(inputMat, phenoGroup, classes = NULL, FilterFunc = SWAP.Filter.Wilcoxon, RestrictedPairs = NULL, handleTies = FALSE, verbose = FALSE, score_fn = signedTSPScores, score_opts = list(), ...)
inputMat |
is a numerical matrix containing the
measurements (e.g., gene expression data)
to be used to build the K-TSP classifier.
The columns represent samples and the
rows represent the features (e.g., genes).
The number of columns must agree
with the length of |
phenoGroup |
is a factor containing the training phenotypes with two levels. |
classes |
is a character vector of length 2 providing the phenotype class labels (case followed by control). If NULL, the levels of phenoGroup will be taken as the labels. |
FilterFunc |
is a filtering function to reduce the
starting number of features to be used to identify the
Top Scoring Pairs (TSPs). The default filter is based on
the Wilcoxon rank-sum test
and alternative filtering functions can be passed too
(see |
RestrictedPairs |
is a character matrix with two columns
containing the feature pairs to be considered for score calculations.
Each row should contain a pair of feature names matching the
|
handleTies |
is a logical value indicating whether tie handling should be enabled or not. FALSE by default. |
verbose |
is a logical value indicating whether status messages will be printed or not throughout the function. FALSE by default. |
score_fn |
is a function for calculating TSP scores.
By default, the signed TSP scores as calculated by
|
score_opts |
is a list of additional variables that
will be passed on to the scoring function as the |
... |
Additional argument passed to the filtering
function |
The output is a list containing the following items:
labels |
the levels (phenotypes) in |
score |
is a vector containing the pair-wise scores. |
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
See SWAP.KTSP.Train,
SWAP.Filter.Wilcoxon,
SWAP.Calculate.BasicTSPScores,
SWAP.Calculate.SignedTSPScores,
and SWAP.KTSP.Statistics.
### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ### Compute the scores using all features (a matrix will be returned) scores <- SWAP.CalculateScores(matTraining, trainingGroup, FilterFunc=NULL)### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ### Compute the scores using all features (a matrix will be returned) scores <- SWAP.CalculateScores(matTraining, trainingGroup, FilterFunc=NULL)
SWAP.CalculateSignedScore calculates the pair-wise scores
between features pairs. The user may pass a filtering function
to reduce the number of starting features, or provide a restricted
set of pairs to limit the reported scores to this list.
SWAP.CalculateSignedScore(inputMat, phenoGroup, FilterFunc = SWAP.Filter.Wilcoxon, RestrictedPairs, handleTies = FALSE, verbose = FALSE, ...)SWAP.CalculateSignedScore(inputMat, phenoGroup, FilterFunc = SWAP.Filter.Wilcoxon, RestrictedPairs, handleTies = FALSE, verbose = FALSE, ...)
inputMat |
is a numerical matrix containing the
measurements (e.g., gene expression data)
to be used to build the K-TSP classifier.
The columns represent samples and the
rows represent the features (e.g., genes).
The number of columns must agree
with the length of |
phenoGroup |
is a factor containing the training phenotypes with two levels. |
FilterFunc |
is a filtering function to reduce the
starting number of features to be used to identify the
Top Scoring Pairs (TSPs). The default filter is based on
the Wilcoxon rank-sum test
and alternative filtering functions can be passed too
(see |
RestrictedPairs |
is a character matrix with two columns
containing the feature pairs to be considered for score calculations.
Each row should contain a pair of feature names matching the
|
handleTies |
is a logical value indicating whether tie handling should be enabled or not. FALSE by default. |
verbose |
is a logical value indicating whether status messages will be printed or not throughout the function. FALSE by default. |
... |
Additional argument passed to the filtering
function |
The output is a list containing the following items:
labels |
the levels (phenotypes) in |
score |
a matrix or a vector containing the pair-wise scores.
Basically, |
Note that the P, Q, and score
list elements are matrices when scores are computed
for all possible feature pairs, while they are vectors
when scores are computed for restricted pairs
defined by RestrictedPairs.
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
See SWAP.KTSP.Train,
SWAP.Filter.Wilcoxon,
and SWAP.KTSP.Statistics.
### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ### Compute the scores using all features (a matrix will be returned) scores <- SWAP.CalculateSignedScore(matTraining, trainingGroup, FilterFunc=NULL, ) ### Show scores class(scores) dim(scores$score) ### Get the scores for a couple of features diag(scores$score[ 1:3 , 5:7 ]) ### Compute the scores using the default filtering function for 20 features scores <- SWAP.CalculateSignedScore(matTraining, trainingGroup, featureNo=20) ### Show scores dim(scores$score) ### Creating some random pairs set.seed(123) somePairs <- matrix(sample(rownames(matTraining), 25, replace=FALSE), ncol=2) ### Compute the scores for restricted pairs (a vector will be returned) scores <- SWAP.CalculateSignedScore(matTraining, trainingGroup, FilterFunc = NULL, RestrictedPairs = somePairs ) ### Show scores class(scores$score) length(scores$score)### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ### Compute the scores using all features (a matrix will be returned) scores <- SWAP.CalculateSignedScore(matTraining, trainingGroup, FilterFunc=NULL, ) ### Show scores class(scores) dim(scores$score) ### Get the scores for a couple of features diag(scores$score[ 1:3 , 5:7 ]) ### Compute the scores using the default filtering function for 20 features scores <- SWAP.CalculateSignedScore(matTraining, trainingGroup, featureNo=20) ### Show scores dim(scores$score) ### Creating some random pairs set.seed(123) somePairs <- matrix(sample(rownames(matTraining), 25, replace=FALSE), ncol=2) ### Compute the scores for restricted pairs (a vector will be returned) scores <- SWAP.CalculateSignedScore(matTraining, trainingGroup, FilterFunc = NULL, RestrictedPairs = somePairs ) ### Show scores class(scores$score) length(scores$score)
SWAP.Filter.Wilcoxon filters the features to top
differential expressed
to be used for KTSP classifier implementation.
SWAP.Filter.Wilcoxon(phenoGroup, inputMat, featureNo = 100, UpDown = TRUE)SWAP.Filter.Wilcoxon(phenoGroup, inputMat, featureNo = 100, UpDown = TRUE)
phenoGroup |
a factor with levels containing training labels for the phenotype of interest. |
inputMat |
a numerical matrix containing feature
measurements to be used to implement the classifier
(e.g., the set of gene expression values).
The columns of this matrix correspond to samples and
must correspond to |
featureNo |
an integer specifying the number of different features to be returned. |
UpDown |
logical value specifying whether an equal proportion of features displaying opposite change across the two phenotypes should be returned (e.g.an equal number of up- and down-regulated genes). |
The names of the features that survived the statistical filtering, i.e. differential expressed features.
Bahman Afsari [email protected], Luigi Marchionni [email protected]
See switchBox for the references.
SWAP.KTSP.Classify,
SWAP.Filter.Wilcoxon,
SWAP.CalculateSignedScore
### Load gene expression data for the training set data(trainingData) ### Return equal numbers of up- and down- regulated features (default) SWAP.Filter.Wilcoxon(trainingGroup, matTraining, featureNo=10) ### Return the top 10 differentially expressed features irrispective to ### the direction of change. ### By setting the argument 'UpDown' equal to FALSE the number of ### up- and down- regulated features can be different SWAP.Filter.Wilcoxon(trainingGroup, matTraining, featureNo=10, UpDown=FALSE)### Load gene expression data for the training set data(trainingData) ### Return equal numbers of up- and down- regulated features (default) SWAP.Filter.Wilcoxon(trainingGroup, matTraining, featureNo=10) ### Return the top 10 differentially expressed features irrispective to ### the direction of change. ### By setting the argument 'UpDown' equal to FALSE the number of ### up- and down- regulated features can be different SWAP.Filter.Wilcoxon(trainingGroup, matTraining, featureNo=10, UpDown=FALSE)
Given a list of predicted labels and true labels, provides accuracy, sensitivity, specificity, balanced accuracy (i.e. (sensitivity+specificity)/2 ), and AUC if decision values are given.
SWAP.GetKTSP.PredictionStats(predictions, truth, classes=NULL, decision_values=NULL)SWAP.GetKTSP.PredictionStats(predictions, truth, classes=NULL, decision_values=NULL)
predictions |
is a vector or factor of predicted classes. |
truth |
is a vector or factor of the true class labels. |
classes |
is a character vector of length 2 providing the phenotype class labels (case followed by control). If NULL, the levels of phenoGroup will be taken as the labels. |
decision_values |
is a vector providing the decision values (such as sum of votes from a k-TSP classifier). Will be used to compute AUC if provided. |
A vector providing accuracy, sensitivity, specificity, and balanced accuracy
, and if decision_values is prodvided, area under the ROC curve (AUC).
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
### Load gene expression data data(trainingData) data(testingData) ### train 1-TSP classifier = SWAP.Train.1TSP(matTraining, trainingGroup) predictions = SWAP.KTSP.Classify(matTesting, classifier) ### get performance results SWAP.GetKTSP.PredictionStats(predictions, testingGroup)### Load gene expression data data(trainingData) data(testingData) ### train 1-TSP classifier = SWAP.Train.1TSP(matTraining, trainingGroup) predictions = SWAP.KTSP.Classify(matTesting, classifier) ### get performance results SWAP.GetKTSP.PredictionStats(predictions, testingGroup)
Given a kTSP classifier and data matrix and class labels, calculates the predictions and
vote sums and then applies SWAP.GetKTSP.PredictionStats.
SWAP.GetKTSP.Result(classifier, inputMat, Groups, classes=NULL, predictions=FALSE, decision_values=FALSE)SWAP.GetKTSP.Result(classifier, inputMat, Groups, classes=NULL, predictions=FALSE, decision_values=FALSE)
classifier |
a k-TSP classifier computed using
|
inputMat |
is a matrix of data with rows being the features (such as gene names, if the matrix if gene expression data) and columns being the samples. |
Groups |
is a factor or a vector providing the phenotype class
each sample belongs to. It should correspond to the order of samples
given by the columns of |
classes |
is a vetor of length 2 providing the two phenotype or
class labels of |
predictions |
is a logical indicating whether to return the predictions or not. |
decision_values |
is a logical indicating whether to return the decision values or not. |
A list with items:
stats |
A vector providing accuracy, sensitivity, specificity, balanced accuracy, and AUC. |
roc |
An ROC curve object produced by the |
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
### Load gene expression data data(trainingData) data(testingData) require(pROC) ### train 1-TSP classifier = SWAP.Train.1TSP(matTraining, trainingGroup) ### get performance results SWAP.GetKTSP.Result(classifier, matTesting, testingGroup)$stats### Load gene expression data data(trainingData) data(testingData) require(pROC) ### train 1-TSP classifier = SWAP.Train.1TSP(matTraining, trainingGroup) ### get performance results SWAP.GetKTSP.Result(classifier, matTesting, testingGroup)$stats
Trains a kTSP on given training data and provides
getkTSPResult output for both training and
testing data.
SWAP.GetKTSP.TrainTestResults(trainMat, trainGroup, testMat, testGroup, classes=NULL, predictions=FALSE, decision_values=FALSE, ...)SWAP.GetKTSP.TrainTestResults(trainMat, trainGroup, testMat, testGroup, classes=NULL, predictions=FALSE, decision_values=FALSE, ...)
trainMat |
is a matrix of data for training with rows being the features (such as gene names, if the matrix if gene expression data) and columns being the samples. |
trainGroup |
is a factor or a vector providing the phenotype class
each training sample belongs to. It should correspond to the order of samples
given by the columns of |
testMat |
is a matrix of data for testing with rows being the features (such as gene names, if the matrix if gene expression data) and columns being the samples. |
testGroup |
is a factor or a vector providing the phenotype class
each testing sample belongs to. It should correspond to the order of samples
given by the columns of |
classes |
is a vetor of length 2 providing the two phenotype or class labels. |
predictions |
is a logical indicating whether to return the predictions or not. |
decision_values |
is a logical indicating whether to return the decision values or not. |
... |
any further arguments to be passed on for k-TSP training. |
A list with items:
classifier |
The trained k-TSP classifier. |
train |
Training performance. |
train |
Testing performance. |
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
### Load gene expression data data(trainingData) data(testingData) require(pROC) ### perform training and testing result = SWAP.GetKTSP.TrainTestResults(matTraining, trainingGroup, matTesting, testingGroup, featureNo=100) ### view results result$train result$test### Load gene expression data data(trainingData) data(testingData) require(pROC) ### perform training and testing result = SWAP.GetKTSP.TrainTestResults(matTraining, trainingGroup, matTesting, testingGroup, featureNo=100) ### view results result$train result$test
SWAP.Kby.Measurement can be supplied to a
kTSP classifier training function to select
an optimal k by adding top-scoring pairs to
maximize a given measurement such as accuracy or
sensitivitiy over the training data.
SWAP.Kby.Measurement(inputMat, phenoGroup, scoreTable, classes, krange, k_opts=list(disjoint=TRUE, measurement="auc")SWAP.Kby.Measurement(inputMat, phenoGroup, scoreTable, classes, krange, k_opts=list(disjoint=TRUE, measurement="auc")
inputMat |
is a numerical matrix containing the measurements (e.g., gene expression data) to be used to build the K-TSP classifier. |
phenoGroup |
is a factor with two levels containing the phenotype information used to train the K-TSP classifier. |
scoreTable |
a data frame output of |
classes |
is a character vector of length 2 providing the phenotype class labels (case followed by control). If NULL, the levels of phenoGroup will be taken as the labels. |
krange |
an integer (or a vector of integers) defining the candidate number of Top Scoring Pairs (TSPs) from which the algorithm chooses to build the final classifier. |
k_opts |
is a list of additional variables:
|
A vector of indices of length k indicating
which pairs from scoreTable should be
selected.
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
SWAP.Kby.Ttest,
SWAP.MakeTSPTable
SWAP.Kby.Ttest can be supplied to a
kTSP classifier training function to select
an optimal k via performing t-tests.
SWAP.Kby.Ttest(inputMat, phenoGroup, scoreTable, classes, krange, k_opts=list())SWAP.Kby.Ttest(inputMat, phenoGroup, scoreTable, classes, krange, k_opts=list())
inputMat |
is a numerical matrix containing the measurements (e.g., gene expression data) to be used to build the K-TSP classifier. |
phenoGroup |
is a factor with two levels containing the phenotype information used to train the K-TSP classifier. |
scoreTable |
a data frame output of |
classes |
is a character vector of length 2 providing the phenotype class labels (case followed by control). If NULL, the levels of phenoGroup will be taken as the labels. |
krange |
an integer (or a vector of integers) defining the candidate number of Top Scoring Pairs (TSPs) from which the algorithm chooses to build the final classifier. |
k_opts |
is not used and is left for
conforming to the arguments of |
A vector of indices of length k indicating
which pairs from scoreTable should be
selected.
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
SWAP.Kby.Measurement,
SWAP.MakeTSPTable
SWAP.KTSP.Classify classifies new test samples
using KTSP coming out of the function SWAP.KTSP.Train.
SWAP.KTSP.Classify(inputMat, classifier, DecisionFunc)SWAP.KTSP.Classify(inputMat, classifier, DecisionFunc)
inputMat |
is a numerical matrix containing the
measurements (e.g., gene expression data)
to be used with a K-TSP classifier to classify the samples
in a specific class or the other.
In this numerical matrix the columns represent the samples
and the rows represent the features (e.g., genes)
used by the classification rule.
Note that |
classifier |
the classifier obtained by invoking
|
DecisionFunc |
is the function used to generate the final
classification prediction by combining the comparisons of the TSPs
in the classifier. By default each sample is classified
according to the class voted by the majority of the TSPs
(“majority wins” principle).
Different decision rules can be also specified using
alternative functions passed |
The SWAP.KTSP.Classify classifies new test samples
based on a specific decision rule. By default, each sample
is classified based on the the majority voting rule of
the comparisons of TSPs in the classifier.
Alternative rules can be defined by the user and passed
to SWAP.KTSP.Classify using the argument
DecisionFunc. A decision function takes
as its input a logical vector x corresponding
to the individual decision of each TSP
(TRUE if the first feature in the pair is
larger then the second, FALSE
in the opposite case). The output of the
DecisionFunction is a single logical
value summarizing all votes of the individual TSPs
(see examples below).
This function returns the predicted class for each sample in the form of a factor.
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
SWAP.KTSP.Train,
SWAP.Filter.Wilcoxon,
SWAP.CalculateSignedScore
################################################## ### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, krange=c(3, 5, 8:15)) ### Show the classifier classifier ### Apply the classifier to the TRAINING set using default decision rule trainingPrediction <- SWAP.KTSP.Classify(matTraining, classifier) ### Resubstitution performance in the TRAINING set ### Define a "positive" test result if needed table(trainingPrediction, trainingGroup) ### Use an alternative DecideFunction to classify each patient ### Here for instance at least two TSPs must agree trainingPrediction <- SWAP.KTSP.Classify(matTraining, classifier, DecisionFunc = function(x) sum(x) > 5.5 ) ### Contingency table for the TRAINING set table(trainingPrediction, trainingGroup) ################################################## ### Testing on new data ### Load the example data for the TEST set data(testingData) ### Show group variable for the TEST set table(testingGroup) ### Apply the classifier to one sample of the TEST set using default decision rule testPrediction <- SWAP.KTSP.Classify(matTesting[ , 1, drop=FALSE], classifier) ### Show prediction testPrediction ### Apply the classifier to the complete the TEST set ### using decision rule defined above (agreement of two TSPs) testPrediction <- SWAP.KTSP.Classify(matTesting, classifier, DecisionFunc = function(x) sum(x) > 5.5) ### Show prediction head(testPrediction, n=10) ### Contingency table for the TEST set table(testPrediction, testingGroup)################################################## ### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, krange=c(3, 5, 8:15)) ### Show the classifier classifier ### Apply the classifier to the TRAINING set using default decision rule trainingPrediction <- SWAP.KTSP.Classify(matTraining, classifier) ### Resubstitution performance in the TRAINING set ### Define a "positive" test result if needed table(trainingPrediction, trainingGroup) ### Use an alternative DecideFunction to classify each patient ### Here for instance at least two TSPs must agree trainingPrediction <- SWAP.KTSP.Classify(matTraining, classifier, DecisionFunc = function(x) sum(x) > 5.5 ) ### Contingency table for the TRAINING set table(trainingPrediction, trainingGroup) ################################################## ### Testing on new data ### Load the example data for the TEST set data(testingData) ### Show group variable for the TEST set table(testingGroup) ### Apply the classifier to one sample of the TEST set using default decision rule testPrediction <- SWAP.KTSP.Classify(matTesting[ , 1, drop=FALSE], classifier) ### Show prediction testPrediction ### Apply the classifier to the complete the TEST set ### using decision rule defined above (agreement of two TSPs) testPrediction <- SWAP.KTSP.Classify(matTesting, classifier, DecisionFunc = function(x) sum(x) > 5.5) ### Show prediction head(testPrediction, n=10) ### Contingency table for the TEST set table(testPrediction, testingGroup)
Partitions the data into k folds and applies
SWAP.GetKTSP.TrainTestResults for each fold.
Then it combines prediction votes by dividing the vote
sums by the number of TSPs in each fold to produce an
overall cross-validation result.
SWAP.KTSP.CV(inputMat, Groups, classes = NULL, k = 4, folds = NULL, randomize = TRUE, ...)SWAP.KTSP.CV(inputMat, Groups, classes = NULL, k = 4, folds = NULL, randomize = TRUE, ...)
inputMat |
is a matrix of data with rows being the features (such as gene names, if the matrix if gene expression data) and columns being the samples. |
Groups |
is a factor or a vector providing the phenotype class
each sample belongs to. It should correspond to the order of samples
given by the columns of |
classes |
is a vetor of length 2 providing the two phenotype or class labels. |
k |
an integer giving the number of folds to use. |
folds |
a list containing the samples to be used in each fold;
if |
randomize |
is a logical indicating whether to randomize the sample
order before diving into |
... |
any further arguments to be passed on for k-TSP training. |
A list with items:
folds |
A list containing the sample indices used in each fold. |
cv |
A list containing the classifier, training performance and testing performance for each fold. |
stats |
Overall cross-validation performance. |
roc |
ROC curve object for overall cross-validation performance. |
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
### Load gene expression data data(trainingData) data(testingData) require(pROC) ### perform leave one out cross-validation result = SWAP.KTSP.CV(matTraining, trainingGroup, featureNo=100) ### print results result$stats### Load gene expression data data(trainingData) data(testingData) require(pROC) ### perform leave one out cross-validation result = SWAP.KTSP.CV(matTraining, trainingGroup, featureNo=100) ### print results result$stats
Performs leave one out cross validation; then it combines prediction votes by dividing the vote sums by the number of TSPs in each fold to produce an overall cross-validation result.
SWAP.KTSP.LOO(inputMat, Groups, classes = NULL, ...)SWAP.KTSP.LOO(inputMat, Groups, classes = NULL, ...)
inputMat |
is a matrix of data with rows being the features (such as gene names, if the matrix if gene expression data) and columns being the samples. |
Groups |
is a factor or a vector providing the phenotype class
each sample belongs to. It should correspond to the order of samples
given by the columns of |
classes |
is a vetor of length 2 providing the two phenotype or class labels. |
... |
any further arguments to be passed on for k-TSP training. |
A list with items:
loo |
A list containing the classifier, training performance and testing performance for each fold. |
decision_values |
Decision values obtained for each left-out sample. |
predictions |
Predicted classes for each left-out sample. |
stats |
Overall peformance results. |
roc |
ROC curve object for overall performance. |
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
### Load gene expression data data(trainingData) data(testingData) require(pROC) ### perform leave one out cross-validation result = SWAP.KTSP.LOO(matTraining, trainingGroup, featureNo=100) ### print results result$stats### Load gene expression data data(trainingData) data(testingData) require(pROC) ### perform leave one out cross-validation result = SWAP.KTSP.LOO(matTraining, trainingGroup, featureNo=100) ### print results result$stats
SWAP.KTSP.Statistics computes the votes in favor
of one of the classes or the other for each TSP. This function
also computes the final, combined, consensus of all
TSP votes based on a specific decision rules. The default is the kTSP statistics, sum of the votes.
SWAP.KTSP.Statistics(inputMat, classifier, CombineFunc)SWAP.KTSP.Statistics(inputMat, classifier, CombineFunc)
inputMat |
is a numerical matrix containing the
measurements (e.g., gene expression data)
to be used to compute the individual TSP votes
and their consensus.
like the matrix used for training classifier
(in |
classifier |
the classifier obtained by invoking
|
CombineFunc |
is the function used to combine
the votes (i.e., comparisons) of individual TSPs contained in the classifier.
By default, the consensus is the count of the votes taking into
account the order of the features in each TSP.
Using this argument alternative aggregating functions
can be also passed to |
For each TSP in the KTSP classifier,
SWAP.KTSP.Statistics computes the vote in favor of
one of classes or the other.
This function also aggregates the individual TSP votes
and computes a final consensus of all TSP votes
based on specific combination rules.
By default, this combination is achieved by counting the
comparisons (votes) of TSPs as follows:
If the first feature is larger than the second one,
the TSP vote is positive, else the TSP vote is negative.
Different combination rules can also be specified
by defining an alternative combination function
and by passing it to SWAP.KTSP.Statistics
using the CombineFunc argument.
A combination function takes
as its input a logical vector x corresponding
to the sample TSP comparisons
(TRUE if the first feature in the pair is
larger then the second, FALSE
in the opposite case). The output of the
CombineFunction is a single value
summarizing the votes of all individual TSPs
(see examples below).
Note that CombineFunction function must operate on a logical
vector as input and the outcome must be real value number.
A list containing the following two components:
statistics |
a named vector containing the
aggregated summary statistics computed by
|
comparisons |
a logical matrix containing the individual
TSP votes ( |
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
SWAP.KTSP.Classify,
SWAP.Filter.Wilcoxon,
SWAP.CalculateSignedScore
################################################## ### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, FilterFunc = NULL, krange=8) ### Show the TSP in the classifier classifier$TSPs ################################################## ### Compute the TSP votes and combine them using various methods ### Here we will use the count of the signed TSP votes ktspStatDefault <- SWAP.KTSP.Statistics(inputMat = matTraining, classifier = classifier) ### Here we will use the sum of the TSP votes ktspStatSum <- SWAP.KTSP.Statistics(inputMat = matTraining, classifier = classifier, CombineFunc=sum) ### Here, for instance, we will apply a hard treshold equal to 2 ktspStatThreshold <- SWAP.KTSP.Statistics(inputMat = matTraining, classifier = classifier, CombineFunc = function(x) sum(x) > 2 ) ### Show components names(ktspStatDefault) ### Show some of the votes head(ktspStatDefault$comparisons[ , 1:2]) ### Show default statistics head(ktspStatDefault$statistics) ### Show statistics obtained using the sum head(ktspStatSum$statistics) ### Show statistics obtained using the hard threshold head(ktspStatThreshold) ### Make a heatmap showing the individual TSPs votes colorForRows <- as.character(1+as.numeric(trainingGroup)) heatmap(1*ktspStatDefault$comparisons, scale="none", margins = c(10, 5), cexCol=0.5, cexRow=0.5, labRow=trainingGroup, RowSideColors=colorForRows)################################################## ### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, FilterFunc = NULL, krange=8) ### Show the TSP in the classifier classifier$TSPs ################################################## ### Compute the TSP votes and combine them using various methods ### Here we will use the count of the signed TSP votes ktspStatDefault <- SWAP.KTSP.Statistics(inputMat = matTraining, classifier = classifier) ### Here we will use the sum of the TSP votes ktspStatSum <- SWAP.KTSP.Statistics(inputMat = matTraining, classifier = classifier, CombineFunc=sum) ### Here, for instance, we will apply a hard treshold equal to 2 ktspStatThreshold <- SWAP.KTSP.Statistics(inputMat = matTraining, classifier = classifier, CombineFunc = function(x) sum(x) > 2 ) ### Show components names(ktspStatDefault) ### Show some of the votes head(ktspStatDefault$comparisons[ , 1:2]) ### Show default statistics head(ktspStatDefault$statistics) ### Show statistics obtained using the sum head(ktspStatSum$statistics) ### Show statistics obtained using the hard threshold head(ktspStatThreshold) ### Make a heatmap showing the individual TSPs votes colorForRows <- as.character(1+as.numeric(trainingGroup)) heatmap(1*ktspStatDefault$comparisons, scale="none", margins = c(10, 5), cexCol=0.5, cexRow=0.5, labRow=trainingGroup, RowSideColors=colorForRows)
SWAP.KTSP.Train trains a binary K-TSP classifier.
The classifiers resulting from using this function can be
passed to SWAP.KTSP.Classify
for samples classification. Note that this function is
deprecated and we recommend SWAP.Train.KTSP for
training k-TSP classifiers.
SWAP.KTSP.Train(inputMat, phenoGroup, krange = 2:10, FilterFunc = SWAP.Filter.Wilcoxon, RestrictedPairs, handleTies = FALSE, verbose = FALSE, ...)SWAP.KTSP.Train(inputMat, phenoGroup, krange = 2:10, FilterFunc = SWAP.Filter.Wilcoxon, RestrictedPairs, handleTies = FALSE, verbose = FALSE, ...)
inputMat |
is a numerical matrix containing the
measurements (e.g., gene expression data)
to be used to build the K-TSP classifier.
The columns represent samples and the
rows represent the features (e.g., genes).
The number of columns must agree
with the length of |
phenoGroup |
is a factor with two levels containing
the phenotype information used to train the K-TSP classifier.
In order to identify the best TSP to be included in the classifier,
the features contained in |
krange |
an integer (or a vector of integers) defining the candidate number of Top Scoring Pairs (TSPs) from which the algorithm chooses to build the final classifier. The algorithm uses the mechanism in Afsari et al (AOAS, 2014) to select the number of pairs and pair of features. Default is the range from 2 to 10. |
FilterFunc |
is a filtering function to reduce the
starting number of features to be used to identify the
Top Scoring Pairs (TSP). The default filter is differential
expression test based on the Wilcoxon rank-sum test
and alternative filtering functions can be passed too
(see |
RestrictedPairs |
is a character matrix with two columns
containing the feature pairs to be considered for score calculations.
Each row should contain a pair of feature names matching the
rownames of |
handleTies |
is a logical value indicating whether tie handling should be enabled or not. FALSE by default. |
verbose |
is a logical value indicating whether status messages will be printed or not throughout the function. FALSE by default. |
... |
Additional argument passed to the filtering
function |
The KTSP classifier, in the form of a list, which contains the following components:
name |
The classifier name. |
TSPs |
A |
$score |
scores TSP for the top |
$label |
The class labels. These labels correspond to
the |
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
SWAP.KTSP.Classify,
SWAP.Filter.Wilcoxon,
SWAP.CalculateSignedScore
################################################## ### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, krange=c(3, 5, 8:15)) ### Show the classifier classifier ################################################## ### Train another classifier from the top 4 best features ### according to the deafault filtering function classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, FilterFunc=SWAP.Filter.Wilcoxon, featureNo=4) ### Show the classifier classifier ################################################## ### To use all features "FilterFunc" must be set to NULL classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, FilterFunc=NULL) ### Show the classifier classifier ################################################## ### Train a classifier using and alternative filtering function. ### For instance we can use the a "t.test" to selec the features ### with an absolute t-statistics larger than a specified quantile topRttest <- function(situation, data, quant = 0.75) { out <- apply(data, 1, function(x, ...) t.test(x ~ situation)$statistic ) names(out[ abs(out) > quantile(abs(out), quant) ]) } ### Show the top features selected topRttest(trainingGroup, matTraining, quant=0.95) ### Train a classifier using the alternative filtering function ### and also define the maximum number of TSP using "krange" classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, FilterFunc = topRttest, quant = 0.75, krange=c(15:30) ) ### Show the classifier classifier ################################################## ### Training with restricted pairs ### Define a set of specific pairs to be used for classifier development ### For this example we will a random set of features ### In a real example these pairs should be provided by the user. set.seed(123) somePairs <- matrix(sample(rownames(matTraining), 6^2, replace=FALSE), ncol=2) head(somePairs, n=3) dim(somePairs) ### Train a classifier using the restricted feature pairs and the default filtering classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, RestrictedPairs = somePairs, krange=3:16) ### Show the classifier classifier################################################## ### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, krange=c(3, 5, 8:15)) ### Show the classifier classifier ################################################## ### Train another classifier from the top 4 best features ### according to the deafault filtering function classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, FilterFunc=SWAP.Filter.Wilcoxon, featureNo=4) ### Show the classifier classifier ################################################## ### To use all features "FilterFunc" must be set to NULL classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, FilterFunc=NULL) ### Show the classifier classifier ################################################## ### Train a classifier using and alternative filtering function. ### For instance we can use the a "t.test" to selec the features ### with an absolute t-statistics larger than a specified quantile topRttest <- function(situation, data, quant = 0.75) { out <- apply(data, 1, function(x, ...) t.test(x ~ situation)$statistic ) names(out[ abs(out) > quantile(abs(out), quant) ]) } ### Show the top features selected topRttest(trainingGroup, matTraining, quant=0.95) ### Train a classifier using the alternative filtering function ### and also define the maximum number of TSP using "krange" classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, FilterFunc = topRttest, quant = 0.75, krange=c(15:30) ) ### Show the classifier classifier ################################################## ### Training with restricted pairs ### Define a set of specific pairs to be used for classifier development ### For this example we will a random set of features ### In a real example these pairs should be provided by the user. set.seed(123) somePairs <- matrix(sample(rownames(matTraining), 6^2, replace=FALSE), ncol=2) head(somePairs, n=3) dim(somePairs) ### Train a classifier using the restricted feature pairs and the default filtering classifier <- SWAP.KTSP.Train(matTraining, trainingGroup, RestrictedPairs = somePairs, krange=3:16) ### Show the classifier classifier
Given the output from SWAP.CalculateScores and a number maxk,
makes a table of the top maxk pairs. The output of this function
can be provided to a k-selection function such as SWAP.Kby.Ttest
or SWAP.Kby.Measurement to test out different k-selection methods.
SWAP.MakeTSPTable(Scores, maxk, disjoint = TRUE)SWAP.MakeTSPTable(Scores, maxk, disjoint = TRUE)
Scores |
is the output of a scoring function
such as |
maxk |
is an integer: the number of pairs to select. |
disjoint |
a logical indicating whether only disjoint pairs should be selected or not. |
A data frame of maxk pairs, their score and tieVote.
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
SWAP.Kby.Ttest,
SWAP.Kby.Measurement
### load gene expression data data(trainingData) ### calculate scores scores = SWAP.CalculateScores(matTraining, trainingGroup, featureNo=5) ### make top 5 pair table SWAP.MakeTSPTable(scores, 5, FALSE)### load gene expression data data(trainingData) ### calculate scores scores = SWAP.CalculateScores(matTraining, trainingGroup, featureNo=5) ### make top 5 pair table SWAP.MakeTSPTable(scores, 5, FALSE)
Plots two genes or features as a pair of boxplots; optionally, individual samples can be plotted on top of the boxplots as points; for this points can be colored by either gene, or class, or whether first gene < second gene is TRUE or FALSE for each sample.
SWAP.PlotKTSP.GenePairBoxplot(genes, inputMat, Groups=NULL, classes=NULL, points=FALSE, point_coloring="byGene", colors=c(), point_colors=c(), ...)SWAP.PlotKTSP.GenePairBoxplot(genes, inputMat, Groups=NULL, classes=NULL, points=FALSE, point_coloring="byGene", colors=c(), point_colors=c(), ...)
genes |
is a vector of length two providing the pair
(from the rownames of |
inputMat |
is a matrix of data with rows being the features (such as gene names, if the matrix if gene expression data) and columns being the samples. |
Groups |
is a factor or a vector providing the phenotype class
each sample belongs to. It should correspond to the order of samples
given by the columns of |
classes |
is a vetor of length 2 providing the two phenotype or
class labels of |
points |
is a logical value indicating whether to overlay the boxplot with points for individual samples or not. |
point_coloring |
can be either 'byGene' or 'byClass' indicating whether to color the points by gene/feature or by phenotype. A third option is 'byDirection' indicating to color the points by whether the first gene is less than the second gene. |
colors |
is a character vector indicating the color to be used for each boxplot. |
point_colors |
is a character vector indicating the color to be used for the points. |
... |
any further arguments are supplied to the |
Produces a pair of boxplots indicating the distribution of the measured values for the pair of features/genes.
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
SWAP.PlotKTSP.GenePairClassesBoxplot
### Load gene expression data data(trainingData) ### train 1-TSP classifier = SWAP.Train.1TSP(matTraining, trainingGroup) ### plot top pair SWAP.PlotKTSP.GenePairBoxplot(classifier$TSPs, matTraining, points=TRUE, point_coloring="byGene")### Load gene expression data data(trainingData) ### train 1-TSP classifier = SWAP.Train.1TSP(matTraining, trainingGroup) ### plot top pair SWAP.PlotKTSP.GenePairBoxplot(classifier$TSPs, matTraining, points=TRUE, point_coloring="byGene")
Plots two genes or features, each as a pair of boxplots seperated to two classes or phenotypes.
SWAP.PlotKTSP.GenePairClassesBoxplot(genes, inputMat, Groups, classes=NULL, points=FALSE, ordering="byGene", colors=c(), point_colors=c(), point_directions=FALSE, ...)SWAP.PlotKTSP.GenePairClassesBoxplot(genes, inputMat, Groups, classes=NULL, points=FALSE, ordering="byGene", colors=c(), point_colors=c(), point_directions=FALSE, ...)
genes |
is a vector of length two providing the pair
(from the rownames of |
inputMat |
is a matrix of data with rows being the features (such as gene names, if the matrix if gene expression data) and columns being the samples. |
Groups |
is a factor or a vector providing the phenotype class
each sample belongs to. It should correspond to the order of samples
given by the columns of |
classes |
is a vetor of length 2 providing the two phenotype or
class labels of |
points |
is a logical value indicating whether to overlay the boxplot with points for individual samples or not. |
ordering |
can be either 'byGene' or 'byClass' respectively indicating whether to plot two adjacent boxplots for each class/phenotype or two adjacent boxplots for each gene/features. |
colors |
is a character vector indicating the color to be used for each class or gene boxplots. |
point_colors |
is a character vector indicating the color to be used for the points. |
point_directions |
is a logical indicating whether to color the points by whether the first gene is less than the second gene. |
... |
any further arguments are supplied to the |
Produces a pair of boxplots indicating the distribution of the measured values for the pair of features/genes.
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
### Load gene expression data data(trainingData) ### train 1-TSP classifier = SWAP.Train.1TSP(matTraining, trainingGroup) ### plot top pair SWAP.PlotKTSP.GenePairClassesBoxplot(classifier$TSPs, matTraining, trainingGroup, levels(trainingGroup), points=TRUE, ordering="byGene")### Load gene expression data data(trainingData) ### train 1-TSP classifier = SWAP.Train.1TSP(matTraining, trainingGroup) ### plot top pair SWAP.PlotKTSP.GenePairClassesBoxplot(classifier$TSPs, matTraining, trainingGroup, levels(trainingGroup), points=TRUE, ordering="byGene")
Makes a scatter plot of a pair of features/genes.
SWAP.PlotKTSP.GenePairScatter(inputMat, Groups, classes, genes, colors=c(), legends=c(), ...)SWAP.PlotKTSP.GenePairScatter(inputMat, Groups, classes, genes, colors=c(), legends=c(), ...)
inputMat |
is a matrix of data with rows being the features (such as gene names, if the matrix if gene expression data) and columns being the samples. |
Groups |
is a factor or a vector providing the phenotype class
each sample belongs to. It should correspond to the order of samples
given by the columns of |
classes |
is a vetor of length 2 providing the two phenotype or
class labels of |
genes |
is a vector of length one or more providing the names
(from the rownames of |
colors |
is a character vector indicating the color to be used for each phenotype. |
legends |
is a character vector providing any additional information to be appended to the phenotype label in the legend. |
Produces a scatter plot containing points for each sample colored by the phenotype, with two axes being the measurements for the given two features.
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
### Load gene expression data data(trainingData) ### train 1-TSP classifier = SWAP.Train.1TSP(matTraining, trainingGroup) ### plot top pair SWAP.PlotKTSP.GenePairScatter(matTraining, trainingGroup, levels(trainingGroup), classifier$TSPs)### Load gene expression data data(trainingData) ### train 1-TSP classifier = SWAP.Train.1TSP(matTraining, trainingGroup) ### plot top pair SWAP.PlotKTSP.GenePairScatter(matTraining, trainingGroup, levels(trainingGroup), classifier$TSPs)
Makes line plots of one or more features seperated by phenotype.
SWAP.PlotKTSP.Genes(inputMat, Groups, classes, genes, colors=c(), legends=c(), ...)SWAP.PlotKTSP.Genes(inputMat, Groups, classes, genes, colors=c(), legends=c(), ...)
inputMat |
is a matrix of data with rows being the features (such as gene names, if the matrix if gene expression data) and columns being the samples. |
Groups |
is a factor or a vector providing the phenotype class
each sample belongs to. It should correspond to the order of samples
given by the columns of |
classes |
is a vetor of length 2 providing the two phenotype or
class labels of |
genes |
is a vector of length one or more providing the names
(from the rownames of |
Produces a plot containing a line for each feature plotted, the x-axis being the ordering of samples and the y-axis being the measured value (such as gene expression).
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
### Load gene expression data data(trainingData) ### train 1-TSP classifier = SWAP.Train.1TSP(matTraining, trainingGroup) ### plot top pair SWAP.PlotKTSP.Genes(matTraining, trainingGroup, levels(trainingGroup), classifier$TSPs)### Load gene expression data data(trainingData) ### train 1-TSP classifier = SWAP.Train.1TSP(matTraining, trainingGroup) ### plot top pair SWAP.PlotKTSP.Genes(matTraining, trainingGroup, levels(trainingGroup), classifier$TSPs)
Given the output from SWAP.GetKTSP.TrainTestResults(), plots the training and testing ROC curves.
SWAP.PlotKTSP.TrainTestROC(result, colors=c(), legends=c(), ...)SWAP.PlotKTSP.TrainTestROC(result, colors=c(), legends=c(), ...)
result |
is either the output from |
colors |
is a character vector indicating the color to be used for each curve. |
legends |
is a character vector providing any additional information to be appended to each curve label in the legend. |
Produces a plot with two ROC curves corresponding to training results and testing/validation results.
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
SWAP.GetkTSPTrainTestResults
### Load gene expression data data(trainingData) data(testingData) require(pROC) ### perform training and testing result = SWAP.GetKTSP.TrainTestResults(matTraining, trainingGroup, matTesting, testingGroup, featureNo=100) ### plot ROC curves SWAP.PlotKTSP.TrainTestROC(result)### Load gene expression data data(trainingData) data(testingData) require(pROC) ### perform training and testing result = SWAP.GetKTSP.TrainTestResults(matTraining, trainingGroup, matTesting, testingGroup, featureNo=100) ### plot ROC curves SWAP.PlotKTSP.TrainTestROC(result)
Given a k-TSP classifer and a matrix of data, plots a heatmap of the votes of the pairs computed on the given data.
SWAP.PlotKTSP.Votes(classifier, inputMat, Groups=NULL, CombineFunc, ...)SWAP.PlotKTSP.Votes(classifier, inputMat, Groups=NULL, CombineFunc, ...)
classifier |
is a k-TSP classifier produced by
|
inputMat |
is a matrix of data with rows being the features (such as gene names, if the matrix if gene expression data) and columns being the samples. |
Groups |
is a factor or a vector providing the phenotype class
each sample belongs to. It should correspond to the order of samples
given by the columns of |
CombineFunc |
is a function corresponding to the
|
Produces a heatmap where the color indicates a vote of 1 or 0 for a given sample by a top scoring pair.
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
SWAP.Train.1TSP trains a binary TSP classifier
with a single top scoring pair. The classifiers resulting
from using this function can be
passed to SWAP.KTSP.Classify
for samples classification.
SWAP.Train.1TSP(inputMat, phenoGroup, classes = NULL, FilterFunc = SWAP.Filter.Wilcoxon, RestrictedPairs = NULL, handleTies = FALSE, disjoint = TRUE, score_fn = signedTSPScores, score_opts = NULL, verbose = FALSE, ...)SWAP.Train.1TSP(inputMat, phenoGroup, classes = NULL, FilterFunc = SWAP.Filter.Wilcoxon, RestrictedPairs = NULL, handleTies = FALSE, disjoint = TRUE, score_fn = signedTSPScores, score_opts = NULL, verbose = FALSE, ...)
inputMat |
is a numerical matrix containing the
measurements (e.g., gene expression data)
to be used to build the K-TSP classifier.
The columns represent samples and the
rows represent the features (e.g., genes).
The number of columns must agree
with the length of |
phenoGroup |
is a factor with two levels containing
the phenotype information used to train the K-TSP classifier.
In order to identify the best TSP to be included in the classifier,
the features contained in |
classes |
is a character vector of length 2 providing the phenotype class labels (case followed by control). If NULL, the levels of phenoGroup will be taken as the labels. |
FilterFunc |
is a filtering function to reduce the
starting number of features to be used to identify the
Top Scoring Pairs (TSP). The default filter is differential
expression test based on the Wilcoxon rank-sum test
and alternative filtering functions can be passed too
(see |
RestrictedPairs |
is a character matrix with two columns
containing the feature pairs to be considered for score calculations.
Each row should contain a pair of feature names matching the
rownames of |
handleTies |
is a logical value indicating whether tie handling should be enabled or not. FALSE by default. |
disjoint |
is a logical value indicating whether only disjoint pairs should be considered in the final set of selected pairs; i.e. all features occur only once among the set of TSPs. |
score_fn |
is a function for calculating TSP scores.
By default, the signed TSP scores as calculated by
|
score_opts |
is a list of additional variables that
will be passed on to the scoring function as the |
verbose |
is a logical value indicating whether status messages will be printed or not throughout the function. FALSE by default. |
... |
Additional argument passed to the filtering
function |
The TSP classifier, in the form of a list, which contains the following components:
name |
The classifier name. |
TSPs |
A 1 by 2 matrix, containing
the feature names for the selected TSP. These names
correspond to the |
score |
scores TSP for the top TSP. |
label |
the class labels. These labels correspond to
the |
tieVote |
indicates which class the pair would vote for in case of a tie. |
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
SWAP.KTSP.Classify,
SWAP.Filter.Wilcoxon,
SWAP.CalculateSignedScore
################################################## ### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- SWAP.Train.1TSP(matTraining, trainingGroup) ### Show the classifier classifier################################################## ### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- SWAP.Train.1TSP(matTraining, trainingGroup) ### Show the classifier classifier
SWAP.Train.KTSP trains a binary K-TSP classifier.
The classifiers resulting from using this function can be
passed to SWAP.KTSP.Classify
for samples classification.
SWAP.Train.KTSP(inputMat, phenoGroup, classes = NULL, krange = 2:10, FilterFunc = SWAP.Filter.Wilcoxon, RestrictedPairs = NULL, handleTies = FALSE, disjoint = TRUE, k_selection_fn = KbyTtest, k_opts = list(), score_fn = signedTSPScores, score_opts = NULL, verbose = FALSE, ...)SWAP.Train.KTSP(inputMat, phenoGroup, classes = NULL, krange = 2:10, FilterFunc = SWAP.Filter.Wilcoxon, RestrictedPairs = NULL, handleTies = FALSE, disjoint = TRUE, k_selection_fn = KbyTtest, k_opts = list(), score_fn = signedTSPScores, score_opts = NULL, verbose = FALSE, ...)
inputMat |
is a numerical matrix containing the
measurements (e.g., gene expression data)
to be used to build the K-TSP classifier.
The columns represent samples and the
rows represent the features (e.g., genes).
The number of columns must agree
with the length of |
phenoGroup |
is a factor with two levels containing
the phenotype information used to train the K-TSP classifier.
In order to identify the best TSP to be included in the classifier,
the features contained in |
classes |
is a character vector of length 2 providing the phenotype class labels (case followed by control). If NULL, the levels of phenoGroup will be taken as the labels. |
krange |
an integer (or a vector of integers) defining the candidate number of Top Scoring Pairs (TSPs) from which the algorithm chooses to build the final classifier. The algorithm uses the mechanism in Afsari et al (AOAS, 2014) to select the number of pairs and pair of features. Default is the range from 2 to 10. |
FilterFunc |
is a filtering function to reduce the
starting number of features to be used to identify the
Top Scoring Pairs (TSP). The default filter is differential
expression test based on the Wilcoxon rank-sum test
and alternative filtering functions can be passed too
(see |
RestrictedPairs |
is a character matrix with two columns
containing the feature pairs to be considered for score calculations.
Each row should contain a pair of feature names matching the
rownames of |
handleTies |
is a logical value indicating whether tie handling should be enabled or not. FALSE by default. |
disjoint |
is a logical value indicating whether only disjoint pairs should be considered in the final set of selected pairs; i.e. all features occur only once among the set of TSPs. |
k_selection_fn |
is a function for selecting the optimal k
once the TSP scores have been calculated for all the candidate pairs.
This can be either |
k_opts |
a list of additional arguments to be passed on to a custom k selection function. |
score_fn |
is a function for calculating TSP scores.
By default, the signed TSP scores as calculated by
|
score_opts |
is a list of additional variables that
will be passed on to the scoring function as the |
verbose |
is a logical value indicating whether status messages will be printed or not throughout the function. FALSE by default. |
... |
Additional argument passed to the filtering
function |
The KTSP classifier, in the form of a list, which contains the following components:
name |
The classifier name. |
TSPs |
A |
score |
scores TSP for the top |
label |
the class labels. These labels correspond to
the |
tieVote |
indicates which class the pair would vote for in case of a tie. |
Bahman Afsari [email protected], Luigi Marchionni [email protected], Wikum Dinalankara [email protected]
See switchBox for the references.
SWAP.KTSP.Classify,
SWAP.Filter.Wilcoxon,
SWAP.CalculateSignedScore
################################################## ### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- SWAP.Train.KTSP(matTraining, trainingGroup) ### Show the classifier classifier ################################################## ### Train another classifier from the top 4 best features ### according to the deafault filtering function classifier <- SWAP.Train.KTSP(matTraining, trainingGroup, FilterFunc=SWAP.Filter.Wilcoxon, featureNo=4) ### Show the classifier classifier################################################## ### Load gene expression data for the training set data(trainingData) ### Show group variable for the TRAINING set table(trainingGroup) ################################################## ### Train a classifier using default filtering function based on the Wilcoxon test classifier <- SWAP.Train.KTSP(matTraining, trainingGroup) ### Show the classifier classifier ################################################## ### Train another classifier from the top 4 best features ### according to the deafault filtering function classifier <- SWAP.Train.KTSP(matTraining, trainingGroup, FilterFunc=SWAP.Filter.Wilcoxon, featureNo=4) ### Show the classifier classifier
A factor with two levels
describing the phenotypes for the testing data
(Buyse et al cohort,
(see the mammaPrintData package).
data(testingData)data(testingData)
The matTesting factor contains phenotypic information
for the 307 samples of the testing dataset.
This phenotype factor corresponds to the breast cancer patients' cohort
published by Buyse and colleagues in JNCI (2006).
The gene expression matrix was obtained from
the mammaPrintData package as described
by Marchionni and colleagues in BMC Genomics (2013).
Bahman Afsari [email protected], Luigi Marchionni [email protected]
See switchBox for the references.
### Load gene expression data for the test set data(testingData) ### Show the class of the ``testingGroup'' object class(testingGroup) ### Show group variable table(testingGroup)### Load gene expression data for the test set data(testingData) ### Show the class of the ``testingGroup'' object class(testingGroup) ### Show group variable table(testingGroup)
A factor with two levels
describing the phenotypes for the training data
(Glas et al cohort,
see the mammaPrintData package).
data(trainingData)data(trainingData)
The trainingGroup factor contains phenotypic information
for the 78 samples of the training dataset.
This phenotype factor corresponds to the breast cancer patients' cohort
published by Glas and colleagues in BMC Genomics (2006).
The information was obtained from
the mammaPrintData package as described
by Marchionni and colleagues in BMC Genomics (2013).
Bahman Afsari [email protected], Luigi Marchionni [email protected]
See switchBox for the references.
### Load gene expression data for the training set data(trainingData) ### Show the class of the ``trainingGroup'' object class(trainingGroup) ### Show group variable table(trainingGroup)### Load gene expression data for the training set data(trainingData) ### Show the class of the ``trainingGroup'' object class(trainingGroup) ### Show group variable table(trainingGroup)