nem: Implementation of the original NEM

In cbg-ethz/mnem: Mixture Nested Effects Models

Implementation of the original NEM

Description

Infers a signalling pathway from perturbation experiments.

Usage

nem(
  D,
  search = "greedy",
  start = NULL,
  method = "llr",
  marginal = FALSE,
  parallel = NULL,
  reduce = FALSE,
  weights = NULL,
  runs = 1,
  verbose = FALSE,
  redSpace = NULL,
  trans.close = TRUE,
  subtopo = NULL,
  prior = NULL,
  ratio = TRUE,
  domean = TRUE,
  modulesize = 5,
  fpfn = c(0.1, 0.1),
  Rho = NULL,
  logtype = 2,
  modified = FALSE,
  tree = FALSE,
  learnRates = FALSE,
  stepSize = 0.01,
  ...
)

Arguments

D	data matrix with observed genes as rows and knock-down experiments as columns
search	either "greedy", "modules" or "exhaustive" (not recommended for more than five S-genes)
start	either NULL ("null") or a specific network to start the greedy
method	"llr" for log odds or p-values densities or "disc" for binary data
marginal	logical to compute the marginal likelihood (TRUE)
parallel	NULL for no parallel optimization or an integer for the number of threads
reduce	reduce search space (TRUE) for exhaustive search
weights	a numeric vector of weights for the columns of D
runs	the number of runs for the greedy search
verbose	for verbose output (TRUE)
redSpace	reduced search space for exhaustive search; see result of exhaustive search with reduce = TRUE
trans.close	if TRUE uses the transitive closure of adj
subtopo	optional matrix with the subtopology theta as adjacency matrix
prior	a prior network matrix for adj
ratio	if FALSE uses alternative distance for the model score
domean	if TRUE summarizes duplicate columns
modulesize	the max number of S-genes included in one module for search = "modules"
fpfn	numeric vector of length two with false positive and false negative rates
Rho	optional perturbation matrix
logtype	log base of the log odds
modified	if TRUE, assumes a preprocessed data matrix
tree	if TRUE forces tree; does not allow converging edges
learnRates	if TRUE learns rates for false positives/negatives
stepSize	numerical step size for learning rates
...	optional parameters for future search methods

Value

transitively closed matrix or graphNEL

Author(s)

Martin Pirkl

Examples

D <- matrix(rnorm(100*3), 100, 3)
colnames(D) <- 1:3
rownames(D) <- 1:100
adj <- diag(3)
colnames(adj) <- rownames(adj) <- 1:3
scoreAdj(D, adj)

cbg-ethz/mnem documentation built on Nov. 7, 2024, 7:35 p.m.