survJamda (version 1.1.4)

iter.crossval: Performance assessment of gene signatures by cross-validation.

Description

Assess the performance of a gene signature derived from a single or merged data set by ten iterations of cross validation.

Usage

iter.crossval(data, surv, censor, ngroup = 10, plot.roc = 0, method = "none", 
zscore = 0, gn.nb = 100, gn.nb.display = 0)

Arguments

data

Matrix of gene expression data.

surv

Vector of survival time.

censor

Vector of censoring status. 1 = event occurred, 0 = censored.

ngroup

An integer specifying the number of cross-validation folds.

plot.roc

An integer specifying whether the ROC curves should be plotted or not (1 or 0).

method

A character string specifying the feature selection method: "none" for top-ranking or one of the adjusting methods specified by the p.adjust function

zscore

An integer specifying whether Z-score normalization should be applied or not (1 or 0). 1 if data is a merged data set or 0 if data is a single data set.

gn.nb

An integer specifying the number of genes to select for gene signature.

gn.nb.display

An integer specifying the number of selected genes to display.

Value

Mean of AUC +/- standard deviation of AUC, geometric mean of HR(CI).

Details

The p.adjust function in the R stats package is used and all adjusted p-values not greater than 0.05 are retained if method != "none".

If the user wants to apply his own feature selection method, he should define his function with the same number of parameters as the defined feature selection function of the package, i.e. featureselection.

ROC curves are the plots of the mean of true positives (sensitivity) and the mean of false positives (1-specificity) over ngroup folds of cross-validation.

References

Yasrebi H, Sperisen P, Praz V, Bucher P, 2009 Can Survival Prediction Be Improved By Merging Gene Expression Data Sets?. PLoS ONE 4(10): e7431. doi:10.1371/journal.pone.0007431.

See Also

iter.crossval.combat,

Examples

## Single data set
data(gse4335)
data(gse4335pheno)

#And run the following script:
#iter.crossval(gse4335, gse4335pheno[,6], gse4335pheno[,5])

## To observe the frequency of the CYB5D1 gene selection, run the following script:
#iter.crossval(gse4335, gse4335pheno[,6], gse4335pheno[,5], gn.nb =1, gn.nb.display = 1)

## Merged data set
data(gse4335)
data(gse4335pheno)

data(gse1992)
data(gse1992pheno)

common.gene = intersect(colnames(gse4335), colnames(gse1992))

data = rbind(gse4335[,common.gene], gse1992[,common.gene])
surv = c(gse4335pheno[,6],gse1992pheno[,19])
censor = c(gse4335pheno[,5],gse1992pheno[,18])

#And run the following script:
#iter.crossval(data, surv,censor, zscore=1)

## The function is currently defined as
function(data,surv,censor,ngroup=10,plot.roc=0,method="none",zscore=0,gn.nb=100,gn.nb.display=0){

         require(survival)
         require(survivalROC)

         res = NULL

         file.name=deparse(substitute(data)) 
         if (plot.roc)
	         init.plot(file.name)

         data =data[!is.na(surv),]
         censor= censor[!is.na(surv)]
         surv= surv[!is.na(surv)]

         cat ("Iteration\tAUC\tHR(CI)\t\tP-val\n")
         for (i in 1:ngroup){
                  new.lst = cross.val.surv(data, surv, censor,ngroup, i, method, zscore,
                  gn.nb,gn.nb.display,plot.roc,p.list)
                  res = rbind (res, new.lst)
         }

         if(ngroup != length(surv)){
                  cat ("Avg AUC+/-SD\tHR(CI)\n")
                  if (plot.roc)
                           legend (0.55,0.1, legend = paste("AUC+/-SD =", sprintf("%.2f",
                           as.numeric(mean(res[,1],na.rm = T))), "+/-", sprintf("%.2f",
                           sd (res[,1],na.rm = T)), sep = " "),bty = "n")
                  cat (sprintf("%.2f",as.numeric(mean(res[,1], na.rm = T))),  "+/-",
                  sprintf("%.2f",sd (res[,1],na.rm = T)), "\t", gm(res[,2]), 
                  "(", sprintf("%.2f",ci.gm(res[,2])[1]), "-", sprintf("%.2f",
                  ci.gm(res[,2])[2]), ")\n", sep = "")
         }        
}