190

Of course, one can also look the other way around, that a molecule works well as a

drug to influence a receptor, or a protein is itself used as a drug (e.g. the tissue plasmin

activator, TPA, after effect extension by removing an inactivation loop). This type of

protein engineering, which aims at a drug effect, is therefore called drug design. The

first step is to look in detail at the three-dimensional protein structure in order to mark

promising changes in the protein or to see how well a drug (or, more generally, a molec­

ular ligand) fits into the protein structure. We can practice this with programs like

RasMol or PyMOL by using this software to turn the three-dimensional structural coor­

dinates into a colorful, spatial image of the protein structure and look at how well ligand

molecules fit in here (Fig. 1.2). On the other hand, if you want to test a large number of

molecules, you can do this automatically using a substance database with the computer.

In this way, thousands or even millions of compounds are pretested in the computer

(technical term in silico screening) in order to then test the best molecules in the screen

(those with the lowest binding energy) pharmacologically in the experiment. Other cri­

teria (e.g. whether the substance is easy to synthesise) are usually taken into account

when selecting substances. Such an in silico screen is bioinformatically a work of sev­

eral weeks or longer. In addition, molecular dynamics can be considered: The computer

starts from the protein and ligand structure and now systematically samples how the

interaction between the two changes over time. This too now allows accurate predictions

of the effect and how it is best achieved on the protein by the ligand molecule. However,

we can recreate such complex calculations in an introductory way by querying a data­

base where many such results are systematically stored. The DrumPID database (https://

drumpid.bioapps.biozentrum.uni-­wuerzburg.de/compounds/index.php), for example,

thus gives numerous substance suggestions for a protein for which one is looking for

pharmaceuticals and also takes into account filter criteria such as tolerability (e.g.

Lipinski’s rule of five) and whether already approved by the FDA or experimental (Kunz

et al. 2016; a tutorial here also explains its use in more detail).

One step higher, one would like to link different protein components with each other to

form new networks and circuits. This is made possible by our GoSynthetic Database,

which then also directly compares natural networks with technical processes.

https://gosyn.bioapps.biozentrum.uni-­wuerzburg.de

https://www.bakerlab.org

13  Life Invents Ever New Levels of Language