This function selects, for each chromosome, the most relevant markers according to a response.
markerSelection(
dataSetName,
dataResponse,
chromosome = 1:22,
signal = c("CN", "fracB"),
normalTumorArray,
onlySNP = FALSE,
nbFolds = 10,
loss = c("logistic", "linear"),
plot = TRUE,
pkg = c("HDPenReg", "spikeslab"),
...
)The name of the data-set folder.
A csv files or a data.frame with 2 columns : "files" and "response". The column "files" contains the filename to extract and the second column the response associated to the file.
A vector containing the number of the chromosomes for the SNPs selection.
either "CN" or "fracB". corresponding to which signal will be analyzed (default="CN").
Only in the case of normal-tumor study. A csv file or a data.frame containing the mapping between normal and tumor files. The first column contains the name of normal files and the second the names of associated tumor files.
(only if signal="CN"). If TRUE, only the SNPs probes are used (default=FALSE).
number of folds in the cross validation (default=10).
either "logistic" (binary response) or "linear" (quantitative response), default is "logistic"
If TRUE, cross-validation mean squared error is plotted (default=TRUE).
Either "HDPenReg" or "spikeslab". Ued package in linear case.
Other parameters for HDlars, glmnet or spikeslab function.
a list containing length(chromosme) elements. Each element is a list containing
chromosome corresponding to the signal.
A vector containing the index of all selected markers.
A vector containing the position of all selected markers.
A vector containing the names of all selected markers.
A vector containing the coefficients of all selected markers.
Intercept of the model.
This function requires to use the aroma folder architecture. In your working directory, there must have the rawData folder and totalAndFracBData folder. This function launches the lars algorithm on the CN or fracB data and uses a cross-validation to select the most appropriate solution.
HDPenReg, glmnet, spikeslab