Although interactions between peptides or proteins and surfaces are ubiquitous and offer huge potential for medicine, diagnostics and biotechnology, little research has been done to date. The idea is to systematically understand the interactions of individual amino acids with different materials from the bottom up and use computer models to design complex peptides that can be used for specific technical purposes. The project is primarily concerned with peptides that give new functionality to target molecules as so-called labels. Under certain environmental conditions, for example, these peptides can be specifically and very selectively bound to the surfaces and released again.

While in many industries, such as the automotive industry and the development of electronic components, new products are developed almost entirely with the help of modelling and simulation, these approaches to problems in biotechnology are still at the beginning of development. Transferable structure-property models have to be developed, but due to the complexity of the peptides and the heterogeneous materials in a mostly saline aqueous solution, they are not trivial and only poorly understood. Due to the large to almost infinite parameter space, experimental work is only a drop in the ocean. It has to be supported by ever-improving computer models in order to make a robust prediction possible.

The number of possible biomolecules increases astronomically with the size of the molecules used. The number of possible peptides with a length of 20 amino acids exceeds the number of atoms in the universe. Since the available computing power is constantly increasing, models and simulations can significantly contribute to the systematic investigation of a larger number of biomolecules. As a result, only the really promising molecules have to be produced and tested experimentally with comparably greater effort. In addition, computer models can help to establish scientific models for the functionalisation of surfaces, which further develop our understanding of the interaction between biomolecules and surfaces beyond a specific application.

By repeating experimental investigations and simulations, we were able to develop peptides that bind specifically to magnetic nanoparticles and can be separated under certain environmental conditions. By combining experiment and computer models, a large number of environmental parameters could be systematically varied in order to find the optimal biomolecules for specific, industrially relevant applications. 

Artificial intelligence methods can help to further accelerate existing simulation-based computer models. Complex simulations based on physical laws are replaced by neural networks, which learn the output of these simulations based on existing data. Another important application of artificial intelligence methods is not to completely replace the simulations, but to set the direction in the high-dimensional search space in which the method searches fully automatically for optimized biomolecules.