EnsembleSolver: Create a Solver class object using an ensemble of solvers

In PriceLab/trena-until-01mar2018: Fit transcriptional regulatory networks using gene expression, priors, machine learning

Description

Create a Solver class object using an ensemble of solvers

Usage

EnsembleSolver(mtx.assay = matrix(), targetGene, candidateRegulators,
  solverNames = c("lasso", "lassopv", "pearson", "randomForest", "ridge",
  "spearman"), geneCutoff = 0.1, alpha.lasso = 0.9, alpha.ridge = 0,
  lambda.lasso = numeric(0), lambda.ridge = numeric(0),
  lambda.sqrt = numeric(0), nCores.sqrt = 4, nOrderings.bayes = 10,
  quiet = TRUE)

Arguments

mtx.assay	An assay matrix of gene expression data
targetGene	A designated target gene that should be part of the mtx.assay data
candidateRegulators	The designated set of transcription factors that could be associated with the target gene
solverNames	A character vector of strings denoting
geneCutoff	A fraction (0-1) of the supplied candidate regulators to be included in the fetaures output by the solver (default = 0.1)
alpha.lasso	A fraction (0-1) denoting the LASSO-Ridge balance of the 'glmnet' solver used by the LASSO method (default = 0.9)
alpha.ridge	A fraction (0-1) denoting the LASSO-Ridge balance of the 'glmnet' solver used by the Ridge method (default = 0)
lambda.lasso	The penalty parameter for LASSO, used to determine how strictly to penalize the regression coefficients. If none is supplied, this will be determined via permutation testing (default = NULL).
lambda.ridge	The penalty parameter for Ridge, used to determine how strictly to penalize the regression coefficients. If none is supplied, this will be determined via permutation testing (default = NULL).
lambda.sqrt	The penalty parameter for square root LASSO, used to determine how strictly to penalize the regression coefficients. If none is supplied, this will be determined via permutation testing (default = NULL).
nCores.sqrt	An integer denoting the number of computational cores to devote to the square root LASSO solver, which is the slowest of the solvers (default = 4)
nOrderings.bayes	An integer denoting the number of random starts to use for the Bayes Spike method (default = 10)
quiet	A logical denoting whether or not the solver should print output

Value

A Solver class object with Ensemble as the solver

See Also

solve.Ensemble, getAssayData

Other Solver class objects: BayesSpikeSolver, HumanDHSFilter-class, LassoPVSolver, LassoSolver, PearsonSolver, RandomForestSolver, RidgeSolver, Solver-class, SpearmanSolver, SqrtLassoSolver

Examples

load(system.file(package="trena", "extdata/ampAD.154genes.mef2cTFs.278samples.RData"))
target.gene <- "MEF2C"
tfs <- setdiff(rownames(mtx.sub), target.gene)
ensemble.solver <- EnsembleSolver(mtx.sub, target.gene, tfs)

PriceLab/trena-until-01mar2018 documentation built on May 25, 2019, 1:22 p.m.