testNetworkProperties {BoolNet} | R Documentation |

## Test properties of networks by comparing them to random networks

### Description

This is a general function designed to determine unique properties of biological networks by comparing them to a set of randomly generated networks with similar structure.

### Usage

```
testNetworkProperties(network,
numRandomNets = 100,
testFunction = "testIndegree",
testFunctionParams = list(),
accumulation = c("characteristic", "kullback_leibler"),
alternative=c("greater","less"),
sign.level = 0.05,
drawSignificanceLevel = TRUE,
klBins,
klMinVal = 1e-05,
linkage = c("uniform", "lattice"),
functionGeneration = c("uniform", "biased"),
validationFunction, failureIterations=10000,
simplify = FALSE,
noIrrelevantGenes = TRUE,
d_lattice = 1,
zeroBias = 0.5,
title = "",
xlab,
xlim,
breaks = 30,
...)
```

### Arguments

`network` |
A network structure of class |

`numRandomNets` |
The number of random networks to generate for comparison |

`testFunction` |
The name of a function that calculates characteristic values that describe properties of the network. There are two built-in functions: "testIndegree" calculates the in-degrees of states in the network, and "testAttractorRobustness" counts the occurrences of attractors in perturbed copies. It is possible to supply user-defined functions here. See Details. |

`testFunctionParams` |
A list of parameters to |

`accumulation` |
If "characteristic" is chosen, the test function is required to return a single value that describes the network. In this case, a histogram of these values in random networks is plotted, and the value of the original network is inserted as a vertical line. If "kullback_leibler" is chosen, the test function can return a vector of values which is regarded as a sample from a probability distribution. In this case, the Kullback-Leibler distances of the distributions from the original network and each of the random networks are calculated and plotted in a histogram. The Kullback-Leibler distance measures the difference between two probability distributions. In this case, the resulting histogram shows the distribution of differences between the original network and randomly generated networks. |

`alternative` |
If |

`sign.level` |
If If If |

`drawSignificanceLevel` |
If |

`linkage` , `functionGeneration` , `validationFunction` , `failureIterations` , `simplify` , `noIrrelevantGenes` , `d_lattice` , `zeroBias` |
The corresponding parameters of |

`klBins` |
If |

`klMinVal` |
If |

`title` |
The title of the plots. This is empty by default. |

`xlab` |
Customizes label of the x axis of the histogram. For the built-in test functions, the x axis label is set automatically. |

`xlim` |
Customizes the limits of the x axis of the histogram. For the built-in test functions, suitable values are chosen automatically. |

`breaks` |
Customizes the number of breaks in the |

`...` |
Further graphical parameters for |

### Details

This function generically compares properties of biological networks to a set of random networks. It can be extended by supplying custom functions to the parameter `testFunction`

. Such a function must have the signature

```
function(network,
accumulate=TRUE,
params)
```

- network
This is the network to test. In the process of the comparison, both the original network and the random networks are passed to the function

- accumulate
If

`accumulate=TRUE`

, the function must return a single value quantifying the examined property of the network. If`accumulate=FALSE`

, the function can return a vector of values (e.g., one value for each gene/state etc.)- params
A list of further parameters for the function supplied by the user in

`testFunctionParams`

(see above). This can contain any type of information your test function needs.

Three built-in functions for synchronous Boolean networks already exist:

- testIndegree
This function is based on the observation that, often, in biological networks, many state transitions lead to the same states. In other words, there is a small number of "hub" states. In the state graph, this means that the in-degree of some states (i.e., the number of transitions leading to it) is high, while the in-degree of many other states is 0. We observed that random networks do not show this behaviour, thus it may be a distinct property of biological networks. For this function, the parameter

`alternative`

of`testNetworkProperties`

should be set to "greater".The function does not require any parameter entries in

`params`

. If`accumulate=FALSE`

, it returns the in-degrees of all synchronous states in the network. If`accumulate=TRUE`

, the Gini index of the in-degrees is returned as a characteristic value of the network. The Gini index is a measure of inequality. If all states have an in-degree of 1, the Gini index is 0. If all state transitions lead to one single state, the Gini index is 1.This function requires the igraph package for the analysis of the in-degrees.

- testAttractorRobustness
This function tests the robustness of attractors in a network to noise. We expect attractors in a real network to be less susceptible to noise than attractors in randomly generated networks, as biological processes can be assumed to be comparatively stable. There are modes of generating noise: Either the functions of the network can be perturbed, or the state trajectories can be perturbed in a simulation of the network. If

`perturb="functions"`

or`perturb="transitions"`

, the function generates a number of perturbed copies of the network using`perturbNetwork`

and checks whether the original attractors can still be found in the network. If`perturb="trajectories"`

, the network itself is not perturbed. Instead, a set of random initial states is generated, and a set of perturbed states is generated from these initial states by flipping one or more bits. Then, the function tests whether the attractors are the same for the initial states and the corresponding perturbed states. This corresponds to calling`perturbTrajectories`

with`measure="attractor"`

.`params`

can hold a number of parameters:- numSamples
If

`perturb="trajectories"`

, the number of randomly generated state pairs to generate. Otherwise the number of perturbed networks that are generated.- perturb
Specifies the type of perturbation to be applied (possible values:

`"functions"`

,`"transitions"`

and`"trajectories"`

– see above).- method, simplify, readableFunctions, excludeFixed, maxNumBits, numStates
If

`perturb="functions"`

or`perturb="transitions"`

, these are the corresponding parameters of`perturbNetwork`

that influence the way the network is perturbed.- flipBits
If

`perturb="trajectories"`

, the are the corresponding parameters of`perturbTrajectories`

that defines how many bits are flipped.

If perturb="functions" or

`perturb="transitions"`

and`accumulate=FALSE`

, the function returns a vector of percentages of original attractors found in each of the perturbed copies of the original network. If`accumulate=TRUE`

, the function returns the overall percentage of original attractors found in all perturbed copies.If

`perturb="trajectories"`

and`accumulate=FALSE`

, the function returns a logical vector of length`numSamples`

specifying whether the attractor was the same for each initial state and the corresponding perturbed state. If`accumulate=TRUE`

, the function returns the percentage of pairs of initial states and perturbed states for which the attractors were the same.For this function, the parameter

`alternative`

of`testNetworkProperties`

should be set to "greater".- testTransitionRobustness
This function calls

`perturbTrajectories`

with`measure="hamming"`

to measure the average Hamming distance between successor states of randomly generated initial states and perturbed copies of these states.codeparams can hold parameters

`numSamples, flipBits`

corresponding to the parameters of`perturbTrajectories`

that define how many initial states are drawn and how many bits are flipped.If

`accumulate=FALSE`

, the function returns a numeric vector of length`numSamples`

with the normalized Hamming distances of all pairs of initial states and perturbed copies. If`accumulate=TRUE`

, the mean normalized Hamming distance over all pairs is returned.For this function, the parameter

`alternative`

of`testNetworkProperties`

should be set to "less".

### Value

The function returns a list with the following elements

`hist` |
The histogram that was plotted. The type of histogram depends on the parameter |

`pval` |
If |

`significant` |
If |

### See Also

`generateRandomNKNetwork`

, `perturbNetwork`

, `perturbTrajectories`

, `plotStateGraph`

, `getAttractors`

### Examples

```
## Not run:
# load mammalian cell cycle network
data(cellcycle)
if (interactive())
# do not run these examples in the package check, as they take some time
{
# compare the in-degrees of the states in the
# cell cycle network to random networks
testNetworkProperties(cellcycle, testFunction="testIndegree", alternative="greater")
# compare the in-degrees of the states in the
# cell cycle network to random networks,
# and plot the Kullback-Leibler distances of the 100 experiments
testNetworkProperties(cellcycle, testFunction="testIndegree",
accumulation = "kullback_leibler")
# compare the robustness of attractors in the cell cycle network
# to random networks by perturbing the networks
testNetworkProperties(cellcycle, testFunction="testAttractorRobustness",
testFunctionParams=list(perturb="functions", numSamples=10),
alternative="greater")
# compare the robustness of attractors in the cell cycle network
# to random networks by perturbing the state trajectories
testNetworkProperties(cellcycle, testFunction="testAttractorRobustness",
testFunctionParams=list(perturb="trajectories", numSamples=10),
alternative="greater")
# compare the robustness of single state transitions in the cell cycle network
testNetworkProperties(cellcycle, testFunction="testTransitionRobustness",
testFunctionParams=list(numSamples=10),
alternative="less")
}
## End(Not run)
```

*BoolNet*version 2.1.9 Index]