glmSparseNet: Calculate GLM model with network-based regularization

In glmSparseNet: Network Centrality Metrics for Elastic-Net Regularized Models

Description

network parameter accepts:

Usage

glmSparseNet(
  xdata,
  ydata,
  network,
  network.options = networkOptions(),
  experiment.name = NULL,
  ...
)

Arguments

xdata	input data, can be a matrix or MultiAssayExperiment
ydata	response data compatible with glmnet
network	type of network, see below
network.options	options to calculate network
experiment.name	name of experiment to use as input in MultiAssayExperiment object (only if xdata is an object of this class)
...	parameters that glmnet accepts

Details

* string to calculate network based on data (correlation, covariance) * matrix representing the network * vector with already calculated penalty weights (can also be used directly with glmnet)

Value

an object just as glmnet

Examples

xdata <- matrix(rnorm(100), ncol = 20)
glmSparseNet(xdata, rnorm(nrow(xdata)), 'correlation', family = 'gaussian')
glmSparseNet(xdata, rnorm(nrow(xdata)), 'covariance', family = 'gaussian')

#
#
# Using MultiAssayExperiment
# load data
xdata <- MultiAssayExperiment::miniACC
# TODO aking out x indivudals missing values
# build valid data with days of last follow up or to event
event.ix <- which(!is.na(xdata$days_to_death))
cens.ix <- which(!is.na(xdata$days_to_last_followup))
xdata$surv_event_time <- array(NA, nrow(colData(xdata)))
xdata$surv_event_time[event.ix] <- xdata$days_to_death[event.ix]
xdata$surv_event_time[cens.ix] <- xdata$days_to_last_followup[cens.ix]
# Keep only valid individuals
valid.ix <- as.vector(!is.na(xdata$surv_event_time) &
                      !is.na(xdata$vital_status) &
                      xdata$surv_event_time > 0)
xdata.valid <- xdata[, rownames(colData(xdata))[valid.ix]]
ydata.valid <- colData(xdata.valid)[,c('surv_event_time', 'vital_status')]
colnames(ydata.valid) <- c('time', 'status')
glmSparseNet(xdata.valid,
             ydata.valid,
             family          = 'cox',
             network         = 'correlation',
             experiment.name = 'RNASeq2GeneNorm')

glmSparseNet documentation built on April 14, 2021, 6 p.m.