fit.DRWPClassGM: Train a classifier using DRW-GM

Description Usage Arguments Details Value Author(s) References See Also Examples

View source: R/fit.DRWPClassGM.R

Description

DRW-GM is a disease classification method which performs pathway-based classifier construction and precise disease status prediction by joint analysis of genomic and metabolomic data and pathway topology.

Usage

1
2
3
4
5
fit.DRWPClassGM(xG, yG.class1, yG.class2, xM = NULL, yM.class1 = NULL, 
                yM.class2 = NULL, DEBUG = FALSE, pathSet, globalGraph,
                testStatistic = c("t-test", "SAM"), classifier = "Logistic", 
                normalize = TRUE, nFolds = 5, numTops = 50, iter = 1, 
                Gamma = 0.7, Alpha = 0.5, fdr.output = 0.2)

Arguments

xG

a p x n matrix of gene expression measurements with p genes and n samples.

yG.class1

a integer vector comprising the indexes of class 1 samples in xG.

yG.class2

a integer vector comprising the indexes of class 2 samples in xG.

xM

a m x n matrix of metabolite expression measurements with m metabolites and n samples.

yM.class1

a integer vector comprising the indexes of class 1 samples in xM.

yM.class2

a integer vector comprising the indexes of class 2 samples in xM.

DEBUG

Logical. Should debugging information be plotted.

pathSet

A list of pathways. Each pathway is represented as a vector of pathway member genes and metabolites.

globalGraph

An igraph R object contains the global directed gene-metabolite pathway graph.

testStatistic

The test method used to identify differential genes. For testStatistic="t-test", function caltScore is used. For testStatistic="SAM", function calSAMScore is used.

classifier

The method to train classifiers. The default is "Logistic". To use other methods, such as "libsvm", one should install the corresponding package in Weka.

normalize

Logical flag for xG and xM standardization, prior to fitting the model. Default is normalize=TRUE.

nFolds

The number of folds to split xG. Default is 5.

numTops

The number of pathway features used for feature selection. Default is 50.

iter

The number of runs to split xG for optimal classifier selection. Default is 1.

Gamma

A numeric value. The restart probability in directed random walk. Default is 0.7.

Alpha

A proportional coefficient to balance the initial weights of genes and metabolites, which are used to construct the initial weights W0 for directed random walk.

fdr.output

(Approximate) False Discovery Rate cutoff for output in significant genes table. Default is 0.2.

Details

DRW-GM uses directed random walk to evaluate the topological importance of each gene in reconstructed gene-metabolite graph through integrating information from matched gene expression profiles and metabolomic profiles. The topological importance of genes are used to weight the genes for inferring robust DRW-GM-based pathway activities. Then the pathway activities are selected to train the classifier.

Value

Fitted "DRWPClassGM" model object.

model

Fitted "fitModel" model object.

AUC

The performance (AUC) of the classifier on feature selection set.

Accuracy

The performance (Accuracy) of the classifier on feature selection set.

pathFeatures

The selected pathway features to build the classifier.

geneFeatures

The genes used to infer the pathways in pathFeatures

tScore

The t statistic and p-value of each gene in xG

vertexWeight

The topological weights of vertexes in globalGraph

pathSet

The pathways used to construct the global directed gene-metabolite graph.

globalGraph

An igraph R object. The global directed gene-metabolite pathway graph.

testStatistic

The test method used to identify differential genes.

classifier

The method to train classifiers.

nFolds

The number of folds to split xG.

numTops

The number of pathway features used for feature selection.

iter

The number of runs to split xG for optimal classifier selection.

Gamma

The restart probability in directed random walk.

Alpha

The proportional coefficient to balance the initial weights of genes and metabolites.

Author(s)

Wei Liu

References

Liu, W., et al., Topologically inferring risk-active pathways toward precise cancer classification by directed random walk. Bioinformatics, 2013. 29(17): p. 2169-77.

See Also

predict.DRWPClassGM

Examples

1
2
3
4
5
6
7
8
	data(GProf8511)
	data(MProf)
	data(pathSet)
	data(dGMGraph)
	fit <- fit.DRWPClassGM(xG=GProf8511$mRNA_matrix, yG.class1=GProf8511$normal, yG.class2=GProf8511$PCA,
                           xM=MProf$Meta_matrix, yM.class1=MProf$normal, yM.class2=MProf$PCA, DEBUG=TRUE, 
                           pathSet=pathSet, globalGraph=dGMGraph, testStatistic="t-test", classifier = "Logistic", 
                           normalize = TRUE, nFolds = 5, numTops=50, iter = 1, Gamma=0.7, Alpha = 0.5)

cuihaibo1/DRWPClass documentation built on May 12, 2017, 2:21 p.m.