63

laborious task for bioinformatics. But nevertheless, one would subsequently set up again a

semi-quantitative or also an exactly quantitative model of the system in order to describe

it more precisely, exactly as just illustrated.

5.2

Generalization: How to Build a Systems Biology Model?

Can we generalize our approach? Yes, it is one way to describe the systems biological

relationships,

1. First gather all the components (consider the “topology” of the model, that is, its

structure).

2. Then I take this structure and now simulate the system behavior in the computer

(“simulation”).

First: Gather All Components of the System (Its Network or Topology)

Each of the steps has its own challenges. For step 1 to work, I need to write my network

with software so that the output is an image that a machine can read. For example, the

CellDesigner and Cytoscape software allow such network descriptions. Their output for­

mat, an XML format (i.e. all parts of the image are marked in a computer-readable way),

can be read by a program. The two softwares differ only in minor details. In any case, how­

ever, they work by connecting proteins (“nodes”) with each other (“edges”). This creates a

network. In order for this network to predict what will actually happen in the cell, Boolean

logic is important in linking, i.e. which protein is linked to another protein and how (activat­

ing or inhibiting). In addition, it is important to pay attention to “And”, “Or” and “Not”, i.e.

whether, for example, activation only occurs when two proteins jointly activate a third (an

“And”), whether one of the two is sufficient for this (then corresponds to an “Or” linkage)

or whether one must not be there (“Not” as well as with SQUAD, “Nor”, “And not”).

In practice, this requires taking into account many sources about the biological system

as well as collecting missing information from databases or determining it from one’s own

domain and sequence analyses (consider phosphorylation sites, function of the proteins

involved, existing interaction domains or known substrate-enzyme relations).

In addition, it is important to be clear about how to decide which article or finding is

most likely in the case of conflicting sources (see Sect. 6.2). It is best to store this informa­

tion in a separate table so that you can later prove on the basis of which data the model was

arrived at.

Second: Simulate the Network and Its Dynamics

It is now possible to follow Boolean logic and thus make statements about signal chains.

One way to do this is to construct a Petri net (Li et al. 2011; Schlatter et al. 2012) from the

network using appropriate software and thus reproduce the signal cascade in a first form.

5.2  Generalization: How to Build a Systems Biology Model?