dpr: Functions for Network Randomization using different...

In VertexSort: Network Hierarchical Structure and Randomization

Description

These functions can be run in the parallel mode to reduce the time of excution. They randomize a network according to the following five algorithms:

1. degree preserving randomization (dpr).
   

2. similar degree preserving randomization (sdpr).
   

3. degree non-preserving randomization (dnpr).
   

4. in-degree preserving randomization (idpr).
   

5. out-degree preserving randomization (odpr).

Usage

  dpr(vgraph, viteration_no, vparallel = FALSE, vcpus = 1)
  sdpr(vgraph, viteration_no, vparallel = FALSE, vcpus = 1)
  dnpr(vgraph, viteration_no, vparallel = FALSE, vcpus = 1)
  idpr(vgraph, viteration_no, vparallel = FALSE, vcpus = 1)
  odpr(vgraph, viteration_no, vparallel = FALSE, vcpus = 1)

Arguments

vgraph	the graph to be randomized. The graph should be of class igraph.
viteration_no	an integer scalar indicating the desired number of random graphs to be generated.
vparallel	a logical scalar indicating whether to use the parallel programming feature to reduce the process time. The default value is FALSE.
vcpus	an integer scalar determining the number of CPUs to be used when parallel is TRUE. The default value is 1.

Details

The dpr() function randomizes the input graph by randomly selecting two of its edges and exchanging their ends. Multiple edges having the same direction between two nodes are then removed by switching each of them with randomly selected edges. These steps are repeated 10 times the number of edges of the input graph. For more information see references (Abd-Rabbo et al. 2017).

The sdpr() function randomizes the input graph using the matching algorithm, see references (Milo et al. 2004) and examples.

The dnpr() function randomly selects two nodes with replacement to create edges.

The idpr() function randomizes the “actor_id” column of the edges data frame by randomly selecting nodes with replacement from this column. The edges data frame contains the edges making the network and contains two columns: “actor_id” and “target_id”.

The odpr() function randomizes the “target_id” column of the edges data frame by randomly selecting nodes with replacement from this column. The edges data frame contains the edges making the network and contains two columns: “actor_id” and “target_id”.

A log file having the name of each of these functions (e.g. dpr_log.txt, idpr_log.txt) will be created in the working directory and could be opened in order to follow the progression of long randomizations. This log file will be deleted at the end of the task.

Value

graph

list of the randomized graphs of class igraph.

Author(s)

Diala Abd-Rabbo [e-mail]

References

Abd-Rabbo, D. and Michnick, S.W. 2017 BMC Syst Biol 11.

Milo, R., Kashtan, N., Itzkovitz, S. 2004 cond-mat.stat-mech arXiv:cond-mat/0312028v2.

See Also

igraph, snowfall.

Examples

  ## generate a random graph of the kinase-phosphatase network by using
  ## each of the five algorithms

  ## load the VertexSort library
  library(VertexSort)

  ## load interactions of the kinase-phosphatase network (kp-net)
  data("interactions")
  vs_kp_net <- vertex.sort(interactions)
  kp_net    <- vs_kp_net$graph

  ## dpr function: randomize a network with preserving its node degrees
  ## notice the difference in execution time when using and
  ## not using the parallel programning mode
  ptm <- proc.time()
  rand_g <- dpr(kp_net, 4)            # without parallel mode
  proc.time() - ptm

  ptm <- proc.time()
  rand_g <- dpr(kp_net, 4, TRUE, 4)   # with parallel mode
  proc.time() - ptm

  ## verify that rand_g have the same in- and out-degrees as those
  ## of kp_net should obtain TRUE in both commands.
  all(degree(kp_net, V(kp_net), "in")==degree(rand_g[[1]], V(rand_g[[1]]), "in"))
  all(degree(kp_net, V(kp_net), "out")==degree(rand_g[[1]], V(rand_g[[1]]), "out"))

  ## sdpr function: randomize a network with preserving similar node degrees
  rand_g <- sdpr(kp_net, 1)
  ## verify that rand_g have similar in- and out-degrees to those of
  ## kp_net. Should be -1, 0 or 1
  sort(unique(degree(kp_net, V(kp_net), "in")-degree(rand_g[[1]], V(rand_g[[1]]), "in")))
  sort(unique(degree(kp_net, V(kp_net), "out")-degree(rand_g[[1]], V(rand_g[[1]]), "out")))

  ## dnpr function: randomize a network without preserving its node degrees
  rand_g <- dnpr(kp_net, 1)
  ## verify that rand_g have different in- and out-degrees of those of
  ## kp_net. Should get FALSE in both commands
  all(degree(kp_net, V(kp_net), "in")==degree(rand_g[[1]], V(rand_g[[1]]), "in"))
  all(degree(kp_net, V(kp_net), "out")==degree(rand_g[[1]], V(rand_g[[1]]), "out"))

  ## idpr function: randomize a network with preserving its node in-degrees
  rand_g <- idpr(kp_net, 1)
  ## verify that rand_g have same in-degrees and different out-degrees as
  ## those of kp_net. Should get TRUE and FALSE respectively.
  all(degree(kp_net, V(kp_net), "in")==degree(rand_g[[1]], V(rand_g[[1]]), "in"))
  all(degree(kp_net, V(kp_net), "out")==degree(rand_g[[1]], V(rand_g[[1]]), "out"))

  ## odpr function: randomize a network with preserving its node out-degrees
  rand_g <- odpr(kp_net, 1)
  ## verify that rand_g have same out-degrees and different in-degrees as
  ## those of kp_net. Should get FALSE and TRUE respectively.
  all(degree(kp_net, V(kp_net), "in")==degree(rand_g[[1]], V(rand_g[[1]]), "in"))
  all(degree(kp_net, V(kp_net), "out")==degree(rand_g[[1]], V(rand_g[[1]]), "out"))

VertexSort documentation built on May 2, 2019, 9:12 a.m.