Fit ridge regression model given a defined set of active regulators.

Share:

Description

Given a most likely configuration of active regulators identified by biRte, this method fits a conventional ridge regression model to explain gene expression. This function is required, if one would like to use MAP based prediction of gene expression instead of Bayesian predictions (see birtePredict). To fit the ridge regression model the R-package ridge is employed, which provides an efficient tuning of the regularization hyperparameter.

Usage

1
birteFitRidge(model, mRNA.train, ref.cond=1)

Arguments

model

output of birteRun

mRNA.train

vector of gene expression values

ref.cond

condition to consider

Details

In order to make predictions with the fitted ridge regression model (birtePredict) store it into a slot "fit.ridge" of the object returned by birteRun and birteLimma, respectively.

Value

an object of class "cv.glmnet" (see cv.glmnet)

Author(s)

Holger Froehlich

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
# artificial data
data(humanNetworkSimul)
sim = simulateData(affinities2)
limmamRNA = limmaAnalysis(sim$dat.mRNA, design=NULL, "treated - control")

# burnin and sampling size is much too small in reality
result = birteLimma(dat.mRNA=sim$dat.mRNA, data.regulators=NULL, 
limmamRNA=limmamRNA, 
affinities=affinities2, niter=100, nburnin=100, thin=2)

fit.ridge = birteFitRidge(result, sim$dat.mRNA[,1])