RWR_CV: RWR Cross Validation

In dkainer/RWRtoolkit: Random Walk with Restart Tool Suite

RWR Cross Validation

Description

RWR_CV RWR Cross Validation performs K-fold cross validation on a single gene set, finding the RWR rank of the left-out genes. Can choose: (1) leave-one-out (loo) to leave only one gene from the gene set out and find its rank, (2) cross-validation (kfold) to run k-fold cross-validation for a specified value of k, or (3) singletons (singletons) to use a single gene as a seed and find the rank of all remaining genes.

Usage

RWR_CV(
  data = NULL,
  geneset_path = NULL,
  method = "kfold",
  folds = 5,
  restart = 0.7,
  tau = 1,
  numranked = 1,
  outdir = NULL,
  modname = "default",
  plot = FALSE,
  out_full_ranks = NULL,
  out_mean_ranks = NULL,
  threads = 1,
  verbose = FALSE,
  write_to_file = FALSE
)

Arguments

data	The path to the .Rdata file containing your multiplexed functional networks. This file is produced by RWR_make_multiplex. Default NULL
geneset_path	The path to the gene set file. It must have the following first two columns with no headers tab-delimited: <setid> <gene> <weight>. Default NULL
method	Cross-validation method. Choice of: 'kfold', 'loo', or 'singletons'. Default 'kfold'
folds	Number (k) of folds to use in k-fold CV. Default 5
restart	Set the restart parameter [0,1). Higher value means the walker will jump back to seed node more often. Default 0.7
tau	Comma-separated list of values between that MUST add up to the number of network layers in the .Rdata file. One value per network layer that determines the probability that the random walker will restart in that layer. e.g. if there are three layers (A,B,C) in your multiplex network, then –tau '0.2,1.3,1.5' will mean that layer A is less likely to be walked on after a restart than layers B or C. Default 1.0
numranked	Proportion of ranked genes to return [0,1]. e.g. 0.1 will return the top 10%. Default 1.0
outdir	Path to the output directory. Both 'fullranks' and 'medianranks' will be saved with auto-generated filenames. Can be overridden by specifically setting 'out_full_ranks' and 'out_mean_ranks' parameters. No defined path will output within the same directory from which the original code was run. Default NULL
modname	String to include in output file name. Default "default"
plot	Output plots of ROC, PRC, etc. to file. Default FALSE
out_full_ranks	Specify the full path for the full results. Ignores outdir and modName, using this path instead. Default NULL
out_mean_ranks	Specify the full path for the mean results. Ignores outdir and modName, using this path instead. Default NULL
threads	Specify the number of threads to use. Default for your system is all cores - 1.
verbose	Verbose mode. Default FALSE
write_to_file	Also write the result to a file. Default FALSE, however, if output paths are included, the boolean is switched to true.

Value

Returns a list of four data tables: fullranks, medianranks, metrics, and summary.

Examples

# An example of Running RWR CV
# Loads a 10 layer multiplex and does not write to file:
extdata.dir <- system.file("example_data", package = "RWRtoolkit")
multiplex_object_filepath <- paste(extdata.dir,
                                  "/string_interactions.Rdata",
                                  sep = "")
geneset_filepath <- paste(extdata.dir, "/geneset1.tsv", sep = "")
outdir <- "./rwr_cv"


cv_examples <- RWR_CV(
  data = multiplex_object_filepath,
  tau = "1.0,1.0,1.0,1.0,1.0,1.0,1.0,1.0,1.0",
  geneset_path = geneset_filepath,
  outdir = outdir,
  method = "kfold",
  folds = 3
)

# An example of Running RWR CV with non-default method and writing to file
# Loads a 10 layer multiplex and does not write to file:
cv_examples <- RWR_CV(
  data = multiplex_object_filepath,
  tau = "1.0,1.0,1.0,1.0,1.0,1.0,1.0,1.0,1.0",
  geneset_path = geneset_filepath,
  outdir = outdir,
  method = "singletons",
  write_to_file = TRUE
)

dkainer/RWRtoolkit documentation built on Jan. 11, 2025, 3:26 a.m.