runMeDeCom: runMeDeCom
In lutsik/MeDeCom: Decomposition of heterogeneous methylomes

runMeDeCom

R Documentation

runMeDeCom

Description

Perform a MeDeCom experiment

Usage

runMeDeCom(
  data,
  Ks,
  lambdas,
  opt.method = "MeDeCom.cppTAfact",
  cg_subsets = NULL,
  sample_subset = NULL,
  startT = NULL,
  startA = NULL,
  trueT = NULL,
  trueA = NULL,
  fixed_T_cols = NULL,
  NINIT = 100,
  ITERMAX = 1000,
  NFOLDS = 10,
  N_COMP_LAMBDA = 4,
  NCORES = 1,
  random.seed = NULL,
  num.tol = 1e-08,
  analysis.name = NULL,
  use.ff = FALSE,
  cluster.settings = NULL,
  temp.dir = NULL,
  cleanup = TRUE,
  verbosity = 1L,
  time.stamps = FALSE
)

Arguments

`data`	DNA methylation dataset as a `numeric` matrix (methylation sites vs samples) or an ojbect of class `RnBeadSet`
`Ks`	values of parameter `k` to be tested, vector of type `integer`
`lambdas`	values of parameter `\lambda` to be tested, vector of type `numeric`
`opt.method`	optimization method used. Currently supported values are `"MeDeCom.quadPen"` and `"MeDeCom.cppTAfact"`.
`cg_subsets`	a `list` of `integer` vectors specifying row indices to include into the analysis
`sample_subset`	samples to include into the analysis
`startT`	a `list` of length equal to `length(Ks)` or a `matrix` with `max(Ks)` columns
`startA`	a `list` of length equal to `length(Ks)` or a `matrix` with `max(Ks)` rows
`trueT`	a numeric matrix with as many rows as there are methylation sites in `data`
`trueA`	a numeric matrix with as many columns as there are methylation sites in `data`
`fixed_T_cols`	columsn of T which are known (to be implemented)
`NINIT`	number of random initializations
`ITERMAX`	maximal number of iterations of the alternating optimization scheme
`NFOLDS`	number of cross-validation folds
`N_COMP_LAMBDA`	the number of solutions to compare in the "smoothing" step
`NCORES`	number of cores to be used in the parallelized steps (at best a divisor of NINIT)
`random.seed`	seed for random number generation
`num.tol`	some small parameter
`analysis.name`	a deliberate name of the analysis as a `character` singleton
`use.ff`	use `ff` package functionality for memory optimization
`cluster.settings`	a list with parameters for an HPC cluster
`temp.dir`	a temporary directory for the cluster-based analysis available on all nodes
`cleanup`	if `TRUE` the temporary directory will be removed on completion
`verbosity`	verbosity level, `0` for quiet execution
`time.stamps`	add timestamps to the diagnostic output