birewire.sampler.bipartite: Efficient generation of a null model for a given bipartite...

In BiRewire: High-performing routines for the randomization of a bipartite graph (or a binary event matrix), undirected and directed signed graph preserving degree distribution (or marginal totals)

Description

The routine samples correctly from the null model of a given bipartite graph creating a set of randomized version of the initial bipartite graph.

Usage

birewire.sampler.bipartite(incidence,K,path,max.iter="n", accuracy=0.00005,
	verbose=TRUE,MAXITER_MUL=10,exact=FALSE,write.sparse=TRUE)

Arguments

incidence	Incidence matrix of the initial bipartite graph. Since 3.6.0 this matrix can contain also NAs and the position of such entries will be preserved by the SA;
K	The number of networks that has to be generated;
path	The directory in which the routine stores the outputs;
max.iter	"n" (default) the number of switching steps to be performed (or if exact==TRUE the number of successful switching steps). If equal to "n" then this number is considered equal to the analytically derived lower bound presented in Gobbi et al. (see References): N={e}/{2(1-d)} \ln{((e-de)/δ)} if exact is FALSE, N={e(1-d)}/{2} \ln{((e-de)/δ)} otherwise , where e is the number of edges of g and d its edge density . This bound is much lower than the empirical one proposed in Milo et al. 2003 (see References);
accuracy	0.00005 (default) is the desired level of accuracy reflecting the average distance between the Jaccard index at the N-th step and its analytically derived fixed point in terms of fracion of common edges;
verbose	TRUE (default). When TRUE a progression bar is printed during computation.
MAXITER_MUL	10 (default). If exact==TRUE in order to prevent a possible infinite loop the program stops anyway after MAXITER_MUL*max.iter iterations;
exact	FALSE (default). If TRUE the program performs max.iter swithcing steps, otherwise the program will count also the not-performed swithcing steps;
write.sparse	TRUE (default). If FALSE the table is written as an R data.frame (long time and more space needed)

Details

The routine creates, starting from the given path, different subfolders in order to have maximum 1000 files for folder . Moreover the incidence matrices are saved using write_stm_CLUTO (sparse matrices) that can be loaded using read_stm_CLUTO. The set is generated calling birewire.rewire.bipartite on the last generated graph starting from the input graph.

Author(s)

Andrea Gobbi: <[e-mail]>

References

Gobbi, A. and Iorio, F. and Dawson, K. J. and Wedge, D. C. and Tamborero, D. and Alexandrov, L. B. and Lopez-Bigas, N. and Garnett, M. J. and Jurman, G. and Saez-Rodriguez, J. (2014) Fast randomization of large genomic datasets while preserving alteration counts Bioinformatics 2014 30 (17): i617-i623 doi: 10.1093/bioinformatics/btu474.

Iorio, F. and and Bernardo-Faura, M. and Gobbi, A. and Cokelaer, T.and Jurman, G.and Saez-Rodriguez, J. (2016) Efficient randomization of biologicalnetworks while preserving functionalcharacterization of individual nodes Bioinformatics 2016 1 (17):542 doi: 10.1186/s12859-016-1402-1.

R. Milo, N. Kashtan, S. Itzkovitz, M. E. J. Newman, U. Alon (2003), On the uniform generation of random graphs with prescribed degree sequences, eprint arXiv:cond-mat/0312028

BiRewire documentation built on Nov. 8, 2020, 8:09 p.m.