rankFactorsbyProfile: Function to help selecting candidate epigenetic factors based...

In chroGPS: chroGPS2: Generation, visualization and differential analysis of epigenome maps

Description

Function to help selecting candidate epigenetic factors based on their predictive capabilities. This function evaluates how information from a certain set of epigenetic factors can be used to accurately predict information from the rest using linear and logistic regression.

Usage

rankFactorsbyProfile(x,minFactors=5,ranktype='glm',glm.threshold=0.5,verbose=TRUE,maxIter=ncol(x),mc.cores=1)

Arguments

x	A genes x factors table as used in the distGPS function for generation of epigenetic gene maps.
minFactors	Minimum set of 'core' factors to retain. Factors are removed based on how accurately they can be predicted by others.
ranktype	Either 'lm' for using linear regression, or 'glm' for logistic.
glm.threshold	For 'glm', threshold to round values obtained from the predict function so that they are compared with the real values. Defaults to 0.5.
verbose	Informs about progression of the linear or logistic regression steps.
maxIter	Deprecated.
mc.cores	Cores to use in calls to parallel::mclapply.

Value

A data frame with the epigenetic factors in the order they are removed and their prediction accuracy.

Author(s)

Oscar Reina.

See Also

distGPS, domainDist, rankFactorsbyDomain

Examples

## Not run

data(s2)

# Unknown domains
# Perform computation
glm.rank <- rankFactorsbyProfile(s2.tab,ranktype='glm',glm.threshold=0.75,mc.cores=1)

# Returned objects are lists named by the factor with highest prediction accuracy in each iteration
names(glm.rank)

chroGPS documentation built on Oct. 31, 2019, 4:52 a.m.