292

additional analyses, e.g. network and functional analysis (Cytoscape allows different

formats to be loaded, for example .sif). The two plugins BiNGO (GO annotation) and

ClueGO (GO annotation and other databases such as KEGG) are particularly helpful

for detecting the associated functions and signalling pathways for all proteins. From

this, you can in turn create a subnetwork of all proteins for a particular process or

signaling pathway and examine it in detail. To examine the network topology, the

plugin NetworkAnalyzer is suitable. Important topological network parameters are,

for example, the average number of interaction partners, network centrality (provides

information on robustness) and heterogeneity (provides information on organization/

distribution), which you should examine in any case. With the help of the analyses in

Cytoscape, you can get a general overview of the network topology, but you are also

able to detect important hubs (potential targets), which you can further investigate by

means of mathematical modeling c) and then validate experimentally. For step b),

please also refer to our two papers (https://www.ncbi.nlm.nih.gov/pubmed/24558299;

https://www.ncbi.nlm.nih.gov/pubmed/28265997). Another option is provided by our

paper miRNAs (https://pubmed.ncbi.nlm.nih.gov/30421407/). There is also a very

helpful online tutorial on Cytoscape where you can learn about features, plugins, etc.

In addition, other tools for functional analysis and visualization of omics data exist,

such as g:Profiler, GSEA and EnrichmentMap (a nice overview is shown in the paper

[https://www.nature.com/articles/s41596-­018-­0103-­9]).

(c) The mathematical modelling of regulatory networks is a widespread field of applica­

tion in bioinformatics. This enables us to analyze the behavior of a network over time

in order to validate experimental data or to simulate experiments in silico in advance.

For the mathematical modeling of regulatory networks, there are the Boolean, quanti­

tative and semiquantitative methods. In principle, these methods consider the nodes

(proteins) of a network according to their state between 0 and 1, i.e. either activated

(On; maximally activated = 1) or inhibited (Off; maximally inhibited = 0). According

to the initial state (how much is the node activated/deactivated), the further time

course, i.e. how does the state of the node change over time, is calculated for each

individual node of the network. In this way, the behavior or the network interconnec­

tion can be examined in more detail, whereby corresponding network effects, i.e. the

respective effect of a node, also become clear. Boolean modeling always considers the

On/Off -(0/1-) state of a system, i.e. the node is either activated (On; 1) or inhibited

(Off; 0). Quantitative modeling is useful for kinetic data, such as Michaelis–Menten

kinetics. Here one can look at the system state of a network in the interval between 0

and 1, but this requires information about the kinetics. A combination of both methods

is semiquantitative modeling, whereby one is able to consider the system state in the

interval between 0 and 1 as well, but this can be done without knowledge about the

kinetics. An example software for semiquantitative modeling is SQUAD, where the

system state of a network is first represented using a discrete system (Boolean sys­

tem), identifying all steady-state states, which is then transformed into a dynamic

system (differential equation, exponential function). SQUAD identifies all steady

19  Tutorial: An Overview of Important Databases and Programs