mcl: Markov Clustering

Description

mcl implements the Markov Clustering algorithm (van Dongen, 2000) with a fixed expansion parameter (=2).

Usage

mcl(m, infl, iter = 1000, remove.self.loops = FALSE, prune = FALSE,  thresh = 1e-06, pruning.prob = 1e-06, use.sparse = NULL, verbose = FALSE)

Arguments

A numeric matrix, given as input to the Markov Clustering algorithm.

infl

numeric. The inflation parameter for the Markov Clustering algorithm.

iter

integer giving the maximal number of iterations for the Markov Clustering algorithm, set to 1000 be default (in practice the algorithm is shown to converge after 10-100 iterations).

remove.self.loops

boolean indicating whether to remove self loops (i.e. set diagonal entries to 0), set to FALSE by default.

prune

boolean indicating whether to prune small probabilities (i.e. set to 0) in the transition matrix, set to FALSE by default.

thresh

a numeric giving the difference threshold below which the transition matrix is considered to have converged.

pruning.prob

numeric giving the threshold below which pruning should be applied, when prune is TRUE. Set to 1e-06 be default.

use.sparse

a boolean indicating whether to use sparse matrices. By default this value is set to NULL, so that sparse matrices are used only if the transition matrix is sparse enough to justify this representation (50% sparsity and above). When set to TRUE (FALSE), use.sparse will force (disable) the use of sparse matrices.

verbose

boolean indicating whether to print the number of iterations before convergence was achieved.

Value

mcl returns a vector whose i-th entry is the cluster identifier of the i-th node, and two nodes are in the same cluster iff they have the same cluster identifier.

Details

mcl is called from hbm to build a hierarchical block matrix from an association matrix, typically a chromatin contact map.

References