optGroups: Locally optimizes pathway assignment

In cbg-ethz/pathTiMEx: Joint Inference of Mutually Exclusive Driver Pathways and their Progression Dynamics

Description

optGroups optimizes the assignment of genes to pathways, for given fixed structure.

Usage

optGroups(Datamat, groupys, gamma, numsave, skipsteps, adjmat)

Arguments

Datamat	binary alteration matrix where rows are samples and columns are genes and an 1 on position [i,j] means that gene j is altered in sample i.
groupys	list with as many elements as mutually exclusive groups, each element consists of the genes assigned to the respective pathway.
gamma	parameter used in optimization to tune whether the MCMC sampler: a larger value encourages the chain to explore each local minimum more thoroughly, while a lower value allows the chain to escape each local area more easily and explore more of the global space. Usually set to 0.5.
numsave	number of minimal scores attained by the chains, to bs used for plotting.
skipsteps	number representing once in how many iterations to save the minimal score for plotting
adjmat	binary square matrix, representing the optimal structure which is kept fixed. The matrix contains one dummy column and row at the end, identically 0, hence its dimension is number of groups +1.

Details

The optimization is done by via an MCMC chain, by minimizing the number of contradictions due to both mutual exclusivity and progression, as the number of ones that would need to be changed to zeroes to ensure consistency of the data with the clusters and structure. The number of iterations in each chain equals numsave*skipsteps. For each run, opening a new mutually exclusive group is allowed.

Value

list consisting of

• mingroupys optimized clustering solution, which minimizes the number of contradictions due to both mutual exclusivity and progression: list with as many elements as mutually exclusive groups, each element consists of the genes assigned to the respective pathway.

• mingroupysscore minimal score, namely the number of contradictions due to both mutual exclusivity and progression.

• accepted number of times the proposed move was accepted.

• allscores the scores attained at each skipsteps iterations.

Author(s)

Simona Constantinescu, [e-mail]

cbg-ethz/pathTiMEx documentation built on May 13, 2019, 2:03 p.m.