CausalR (version 1.4.3)

RankTheHypotheses: rank the hypotheses

Description

Rank the hypotheses in the causal network. This function can be run with parallelisation using the doParallel flag.

Usage

RankTheHypotheses(network, experimentalData, delta, epsilon = 1e-05, useCubicAlgorithm = TRUE, use1bAlgorithm = TRUE, symmetricCCG = TRUE, listOfNodes = NULL, correctPredictionsThreshold = -Inf, quiet = FALSE, doParallel = FALSE, numCores = NULL)

Arguments

network

Computational Causal Graph, as an igraph.

experimentalData

The experimental data read in using ReadExperimentalData. The results is an n x 2 matrix; where the first column contains the node ids of the nodes in the network that the results refer to. The second column contains values indicating the direction of regulation in the results - (+)1 for up, -1 for down and 0 for insignificant amounts of regulation.

delta

Distance to search within the causal graph.

epsilon

The threshold that is used when calculating the p-value using the cubic algorithm (see 'Assessing statistical significance in causal graphs').

useCubicAlgorithm

An indicator specifying which algorithm will be used to calculate the p-value. The default is set as useCubicAlgorithm = TRUE which uses the cubic algorithm. If this value is set as FALSE, the algorithm will use the much slower quartic algorithm which does compute the exact answer, as opposed to using approximations like the cubic algorithm.

use1bAlgorithm

An indicator specifying whether the 1a or 1b (default, faster) variant of the cubic algorithm described in Chindelevitch's paper will be used to calculate the p-value.

symmetricCCG

This flag specifies whether the CCG is assumed to be symmetric. The value is set as TRUE as a default. If this is the case the running time of the algorithm is reduced since the bottom half of the table can be filled in using the results of calculations performed earlier.

listOfNodes

A list of nodes specified by the user. The algorithm will only calculate and store the results for the nodes in the specified list. The default value is NULL; here the algorithm will calculate and store results for all the nodes in the network.

correctPredictionsThreshold

A threshold on the number of correct predictions for a given hypothesis. If a hypothesis produces fewer correct predictions than predictionsThreshold then the algorithm will not calculate the two p-values. Instead 'NA' will be displayed in the final two columns of the corresponding row of the results table. As a default correctPredictionsThreshold is set as -Inf, so that the p-values are calculated for all specified hypotheses.

quiet

a flag to supress progress output. FALSE by default.

doParallel

A flag for running RankTheHypothesis in parallel mode.

numCores

Number of cores to use if using parallel mode. If the default value of NULL is used, it will attempt to detect the number of cores available and use all of them bar one.

Value

A data frame containing the results of the algorithm.

References

L Chindelevitch et al. Assessing statistical significance in causal graphs. BMC Bioinformatics, 13(35), 2012.

Examples

#get path to example network file
network <- system.file(package='CausalR', 'extdata', 'testNetwork.sif')
#create ccg
ccg <- CreateCCG(network)
#get path to example experimental data
fileName<- system.file(package='CausalR', 'extdata', 'testData.txt')
#read in experimetal data
expData<- ReadExperimentalData(fileName, ccg)
#run in single threaded mode
RankTheHypotheses(ccg, expData, 2)
#run in parallel mode
RankTheHypotheses(ccg, expData, 2, doParallel=TRUE, numCores=2)