aracne_options(
estimator = "copula",
threshold = "permutation",
mi.cutoff = 0,
quantile = 0.95,
permutations = 0,
dpi = "auto",
epsilon = 0.1,
dpi.passes = 5,
direction = "tf",
bins = 4,
max.neighbors = 0,
max.genes = 0,
min.nonzero = 0,
log2 = FALSE,
rank = FALSE,
standardize = FALSE,
prior.min.confidence = 0.45,
prior.factor = 0.5,
prior.protection = "graded",
prior.strength = 0.5,
prior.tf.augment = TRUE,
threads = -1,
seed = 1234567,
silent = TRUE);
estimator
the mutual information estimator: copula (default, O(n) per gene pair), adaptive (the adaptive partitioning of the ARACNE paper), histogram (equal frequency bins with the Miller-Madow correction) or kernel (gaussian kernel density, O(n^2)). [as string]
threshold
permutation (default, the quantile of the null distribution that is created by shuffling the samples of each gene) or fixed (use mi_cutoff). [as string]
mi.cutoff
the fixed mutual information threshold; a positive value forces the fixed mode even when threshold is permutation. [as double]
quantile
the quantile of the permutation null distribution, default 0.95. [as double]
permutations
the number of the gene pairs that are sampled for the null distribution, a value that is less than or equals to 0 means "auto". [as integer]
dpi
auto (TF guided when a TF list is given, otherwise standard), standard, tf (TF guided) or off. [as string]
epsilon
the DPI tolerance, default 0.1. This is the main switch of the network density: a larger value keeps more near-tie edges (an edge is only removed when it is weaker than both competitors by more than epsilon), while a smaller value removes the indirect edges more aggressively. [as double]
dpi.passes
the max number of the DPI iteration passes, default 5. Every pass removes the edges that are condemned by the current adjacency, so a smaller value keeps more edges. [as integer]
direction
tf (default) or undirected. [as string]
bins
the bin number of the histogram estimator, default 4. [as integer]
max.neighbors
the max number of the edges that each regulator can keep (top-K pruning), a value that is less than or equals to 0 means "no pruning". [as integer]
max.genes
keep only the top variant genes; the genes of the TF list and of the prior network are always kept. [as integer]
min.nonzero
the minimal ratio of the non-zero samples for a gene to be kept. [as double]
log2
apply a log2(x + 1) transformation before the inference. [as boolean]
rank
use the ranks (spearman) instead of the values (pearson). [as boolean]
standardize
z-score the expression vectors before the inference. [as boolean]
prior.min.confidence
the eligibility threshold of the curated prior edges, default 0.45 (i.e. Low and above). [as double]
prior.factor
the threshold relaxation factor at the full confidence, default 0.5; the prior edge of the confidence c uses tau * (1 - (1 - factor) * c). [as double]
prior.protection
graded (default: the DPI protection window is scaled by the confidence), sampled (a prior edge is fully protected with the probability p = c) or off. [as string]
prior.strength
the strength of the graded protection, default 0.5. [as double]
prior.tf.augment
append the regulators of the prior network into the TF set, so that a incomplete TF annotation does not lose the regulation edges. [as boolean]
threads
the max degree of parallelism, a value that is less than or equals to 0 means "auto". [as integer]
seed
the random seed of the permutation sampling and of the sampled protection. [as integer]
silent
do not print the progress messages. [as boolean]