elastic_net(x,
candidates = 300,
maxreg = 10,
l1.ratio = 0.7,
alphas = 20,
mutual = FALSE,
TF = NULL,
priority = NULL,
regulators = NULL,
standardize = FALSE,
min.abs.r = 0,
gpu = FALSE,
block.screener = FALSE,
partial.correlation = FALSE,
partial.cond = 5,
negative.quota = -1,
max.row.sum = 0.9,
priority.budget = 0,
prior = NULL,
prior.min.confidence = 0.9,
prior.penalty = 0.3,
prior.sign = "flip",
threads = -1);
x
the gene expression matrix, rows are genes and columns are samples. [as Matrix]
candidates
the max number of the candidate regulators for each target gene, the candidates are picked by the absolute pearson correlation with the target gene. [as integer]
maxreg
the max number of the non-zero regulation edges that each target gene can keep. [as integer]
l1.ratio
the L1 mixing ratio of the elastic net, in the range [0, 1]. [as double]
alphas
the number of the alpha steps in the regularization path. [as double]
mutual
only keep the reversible (mutual) regulation pairs. [as boolean]
TF
an optional gene id vector of the transcript factors, these genes are always included in the candidate regulator pool (they are not affected by the candidates truncation). Non TF genes are still able to be the regulators, so that the knockdown of a non TF gene still has downstream responses.
standardize
z-score the expression vectors before the fitting. [as boolean]
min.abs.r
the minimal absolute pearson correlation for a gene to be a candidate regulator. [as double]
prior
value from the bnlearn::prior_network function.