importance: Compute an index of importance of the genes in the signature.

In geneSignatureFinder: A Gene-signatures finder tools

Description

An index measuring the importance of each gene in the signature is computed. The more the index is close to 1, the most the corresponding gene is important to separate the survival curves. When the value is about 0, the gene has a low contribution in separating the survival curves. If it happens that the index is negative, the deletion of the gene from the signature improves the distance of the survival curves when all but it are in the signature.

Usage

importance(aSignatureFinder, deep = FALSE, cpuCluster = NULL, stopCpuCluster = TRUE)

Arguments

aSignatureFinder	Structure as results from the function signatureFinder()
deep	flag for future use
cpuCluster	structure as result from the NCPUS() function
stopCpuCluster	flag to control if the channel to the cpu-cluster has to be closed.

Details

The importance index associated to each gene of a signature is computed according to the leave one out strategy. Given a gene, it is removed from the signature and with the other genes in the sequence first a classification is performed and then thet test-value of the log-rank test is evaluated. These test-values are collected in the list L1GeneOutTV. All these values are compared to test-value of the signature: importance = 1 - L1GeneOutTV/aSignatureFinder$tValue.

Value

The function returns the same variable in the input signature structure and

L1GeneOutTV	a list of real values with the test-values of the log-rank test applied to all but one gene in the signature (leave-one-out strategy)
importance	A list of real values measuring the importance of each gene in the signature

Author(s)

Stefano M. Pagnotta and Michele Ceccarelli

See Also

signatureFinder.

Examples

data(geNSCLC)
geData <- geNSCLC
data(stNSCLC)
stData <- stNSCLC
aMakeCluster <- makeCluster(2)
geneSeed <- which(colnames(geData) == "SELP")
signature <- signatureFinder(geneSeed, 
  cpuCluster = aMakeCluster, stopCpuCluster = FALSE)
signature <- importance(signature, cpuCluster = aMakeCluster)
signature
signature$importance
#################################
barplot(signature$importance, main = "Importance based on L1GeneOut",
        sub = paste("Signature starting from:", signature$startingSignature))

geneSignatureFinder documentation built on May 2, 2019, 2:32 p.m.