testCliqueAsGroup: Tests whether a given group is mutually exclusive
In cbg-ethz/TiMEx: Finding Mutually Exclusive Groups of Alterations in Cancer Datasets

Description Usage Arguments Details Value Author(s) References See Also Examples

testCliqueAsGroup tests whether a group, given as gene indices, is mutually exclusive.

1	testCliqueAsGroup(geneIdx, mat, lo)

`geneIdx`	vector of indices in the input matrix of the genes to be tested
`mat`	binary alteration matrix, with rows representing patients and columns representing genes
`lo`	rate of observation time. Default is 1.

For deciding whether a group is mutually exclusive, the group of genes is fitted to both the Null (Conditional Independence) and the Mutual Exclusivity models. Parameters under the two models are estimated, and, since they are nested, a likelihood ratio test is performed between the corresponding log likelihoods, in order to test whether mu (the intensity of mutual exclusivity) is different from 0. For computing the likelihood of the data under both models, an exhaustive enumeration of all possible orders of the input alterations needs to be performed. Therefore, the complexity of the test is exponential in the number of genes to be tested, which makes it unfeasible for large number of genes (usually more than 10).

lo (the rate of observation time) is by default set to 1, as both models are otherwise unindentifiable. We recommend leaving the value of this parameter unchanged, as otherwise the estimated waiting time rates of the genes require additional interpretation.

For more details on the TiMEx procedure, as well as on the underlying mathematical model, see Constantinescu et al.: TiMEx: A Waiting Time Model for Mutually Exclusive Cancer Alterations. Bioinformatics (2015).

List consisting of:

opMu list as returned by optim. The field par is a vector of n+1 positions (n being the number of genes) containing the estimates of the waiting time rates (lambda) for the n genes under the mutual exclusivity model, followed by the estimate for mu.
opNull list as returned by optim. The field par is a vector of n positions (n being the number of genes) containing the estimates of the waiting time rates (lambda) for the n genes under the null model.
countsVec the contingency table of the input genes, as a vector. The first element is the count of the samples where no gene was altered, the next n elements are the counts of the samples where exactly one gene was altered (starting with gene 1 and ending with gene n), the next n elements are the counts of the samples where exactly two genes were altered (starting with genes 1 and 2, continuing with genes 1 and 3, and ending with genes n-1 and n), and so on.
genes subset of the input binary matrix, corresponding to the genes to be tested
LRT log likelihood ratio (LRT)
pvalueLRT the p-value corresponding to the LRT

Simona Cristea, scristea@jimmy.harvard.edu

Constantinescu et al.: TiMEx: A Waiting Time Model for Mutually Exclusive Cancer Alterations. Bioinformatics (2015).

the wrapper function TiMEx for identifying mutually exclusive groups in a binary dataset with the TiMEx model.

# Tests for mutual exclusivity the group of genes with indices 13, 204, and 
# 310 in the ovarian cancer dataset
data(ovarian)
testGroup<-testCliqueAsGroup(c(13,204,310),ovarian)