Run Monte Carlo simulations for one or more patients

Usage

runSim(
  Lmatrix,
  Rmatrix,
  Cmatrix,
  LRmatrix,
  cells,
  communication_type,
  pats = "all",
  N = 10000,
  itNo = 100,
  av = 20,
  output_folder = NULL,
  file.name = NULL,
  norm = FALSE,
  patient_idx = NULL,
  ncores = 1
)

Arguments

Lmatrix: Cell-by-ligand compatibility matrix.
Rmatrix: Cell-by-receptor compatibility matrix.
Cmatrix: Patient-by-cell-type abundance matrix.
LRmatrix: Ligand-receptor-by-patient tensor.
cells: Character vector of cell-type names.
communication_type: Feature family to simulate ("D", "W", "TT", "CT", or "GSCC"), or a character vector of several of these. When more than one is given, every requested type is extracted from the same simulated graphs (via countAllTypes()) instead of re-simulating a fresh set of graphs per type, and one .out file per type is written (file names suffixed with the type, e.g. <file.name>_D.out). With a single type, output naming is unchanged from previous versions (<file.name>.out, no suffix).
pats: Number of patients to process, or "all".
N: Number of cells per graph.
itNo: Number of Monte Carlo iterations.
av: Target average degree.
output_folder: Directory used to write the .out files.
file.name: Output filename stem.
norm: Logical; if TRUE, use a uniformized LR baseline.
patient_idx: Optional single patient index to simulate.
ncores: Number of cores to compute patients on in parallel, via parallel::makeCluster() + doParallel::registerDoParallel() + foreach::foreach(...) %dopar% {...} (the same backend used throughout pipeML), so it works identically on Windows/macOS/Unix. ncores = 1 (the default) runs sequentially via lapply() and skips cluster setup entirely. Patients are independent of each other, so this parallelizes near-linearly. File writing always happens sequentially afterward, in patient order, to keep the on-disk format unchanged. Because cluster workers are separate R processes (not forks), each one loads the installed RaCInG package rather than inheriting the calling session's state – if you are iterating on package source via source() instead of library(RaCInG), reinstall the package first so workers see your latest changes.

Value

Invisibly writes the simulation outputs to disk.