glmSparseNet: Calculate GLM model with network-based regularization

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

Calculate GLM model with network-based regularization

Description

network parameter accepts:

• 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)

Usage

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

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

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

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

Arguments

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

Value

an object just as glmnet

Functions

• glmDegree(): penalizes nodes with small degree (inversion penalization h(x) = 1 / x).

• glmHub(): Penalizes nodes with small degree (normalized heuristic that promotes nodes with many edges).

• glmOrphan(): Penalizes nodes with high degree (normalized heuristic that promotes nodes with few edges).

See Also

Cross-validation functions cv.glmSparseNet().

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
library(MultiAssayExperiment)
data("miniACC", package = "MultiAssayExperiment")

xdata <- miniACC
# TODO aking out x individuals 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 = "RNASeq2GeneNorm"
)

# Degree penalization

xdata <- matrix(rnorm(100), ncol = 5)
glmDegree(
    xdata,
    rnorm(nrow(xdata)),
    "correlation",
    family = "gaussian",
    options = networkOptions(minDegree = .2)
)
xdata <- matrix(rnorm(100), ncol = 5)
glmHub(
    xdata,
    rnorm(nrow(xdata)),
    "correlation",
    family = "gaussian",
    options = networkOptions(minDegree = .2)
)
# Orphan penalization

xdata <- matrix(rnorm(100), ncol = 5)
glmOrphan(
    xdata,
    rnorm(nrow(xdata)),
    "correlation",
    family = "gaussian",
    options = networkOptions(minDegree = .2)
)

sysbiomed/glmSparseNet documentation built on Feb. 17, 2024, 1:38 p.m.