score.cnmtf: Calculate Omega matrix

In lgl15/cnmtf: Prioritisation of single nucleotide variants using Corrected non-negative matrix tri-factorisation

Description

Function to calculate the Omega matrix (SNV scores per cluster of patients)

Usage

score.cnmtf(R, out = NULL, pop = NULL, log.file = NULL, name.exp = NULL,
  name.init = NULL, work.dat = NULL, define.k = c("user", "NMTF", "PCA"),
  k1 = NULL, k2 = NULL, estimate.par = TRUE, wparameters = NULL,
  save.parameters = TRUE, run.t.exp = NULL, run.t.par = 4,
  range.parameters = NULL, sequential.estimation = FALSE, max.try0 = 4,
  tao.sc = 4, snps.known = NULL, calcObj = NULL, calcObj2 = NULL,
  init = NULL, parallel.opt = FALSE, n.cores = 2, iters = NULL,
  do.U = FALSE, display.iters = TRUE, score.pvalues = c(TRUE, FALSE,
  "only"), randomisations = 50, random.parallel = FALSE,
  file.Gu = file.Gu, norm.Lu = FALSE)

Arguments

R	Relationship matrix
out	Categorical outcome variable
pop	Population variable
log.file	Log file to track progress of the function
name.exp	Name of experiment to save files
name.init	Name of workspace with initialisations of U and V
work.dat	Folder to save and load workspaces
define.k	Method to define k1. It can be: c("user","method","PCA")
k1	Number of SNV clusters
k2	Number of patient clusters equals number of levels in the outcome
estimate.par	Logical. Estimate penalisation parameters. If FALSE then provide the parameters in the argument "wparameters". If TRUE then provide preliminar weights to the parameters in the argument "wparameters".
wparameters	Either the Penalization parameters (if estimate.par == FALSE) or Weights of penalisation terms to be computed (if estimate.par == TRUE)
save.parameters	Logical. Save parameters to file
run.t.exp	Number of repetitions for the experiment
run.t.par	Number of repetitions for parameters fitting
range.parameters	Range of parameters to be evaluated (if estimate.par == FALSE)
sequential.estimation	Set the parameters in a specific order and carry the optimals
max.try0	Maximum number of tries to fit the parameter
tao.sc	Trehsold of standard deviations for the SNV score
snps.known	List of known associations
calcObj, calcObj2	Number of iterations to check convergency. Check convergency each calcObj number of iterations after first calcObj2 iterations
init	Type of seeding/initialisation of matrices in the algorithm
parallel.opt	Run some instances of the algorithm in parallel
n.cores	Number of cores to use in the parallel processing
iters	Number of iterations
do.U	Perform clustering of SNPs
display.iters	Display the iterations of function cnmtf
score.pvalues	Estimate p-values for the scores c(TRUE, FALSE, "only").
randomisations	Number of randomisations
random.parallel	Logical. Run the radomisations in parallel
file.Gu	Workspace with adjancency matrix for the SNV-SNV network.

Value

The function internally generates:

• Estimation of number of SNP clusters (k1)

• Estimation of penalisation parameters (gamma1, gamma3, gamma2)

The function prints the following objects in workspaces:

• Initialisation of matrices U.init and V.init

• res.cnmtf: results of factorisations

• lcnmtf.ran: results of algorithm applied on randomisations of R

Author(s)

Luis G. Leal, [e-mail]

See Also

Other Factorisation functions: clus.membership, cnmtf, consensus.clust, hierarchical.clust, initialise.UV, neg.constrain, parameters.cnmtf, plot.parameter, pos.constrain, psvd.init, regression.snps, synthetic.gwas

lgl15/cnmtf documentation built on May 28, 2019, 6:33 p.m.