Advancing data-driven studies of supramolecular polymer systems
Stef Jansen successfully defended his PhD thesis at the Department of Chemical Engineering and Chemistry on February 26th.
Many plastic materials consist of small molecules that are connected by strong chemical bonds to form long chains called polymers. In nature, however, large molecular structures do not always form via strong bonds; instead, they are often created by the stacking of molecular building blocks. In this way, supramolecular polymers can be formed: long chains of individual molecules held together by weak forces. These weak interactions make supramolecular polymers dynamic, allowing them to adapt to their environment, which is essential for cellular function.
Scientists can design artificial molecules that behave in a similar way. Although these structures are much less involved than those in cells, they hold great promise for applications such as biomaterials. By adding additional molecules, the properties of supramolecular polymers can be tailored for specific purposes, such as enabling interactions with certain types of cells. These mixtures of different molecular components are called supramolecular polymer systems.
The possibilities for mixing different components are virtually endless, offering enormous opportunities to design innovative materials. At the same time, this makes it challenging to find the ideal composition. Until now, this has mostly been done by testing many combinations, guided by the experience and intuition of experts. His research focuses on accelerating and improving this process using data-driven techniques such as computational models, machine learning algorithms, and artificial intelligence.
In this dissertation, Stef develops thermodynamic models to help us understand how molecules behave and interact. Combined with experiments, these models can explain and even predict the behavior of supramolecular polymer systems in detail. Because these models require prior knowledge and are less scalable to systems with many components, I use Bayesian optimization for more complex challenges. This machine learning algorithm navigates intelligently and efficiently through possible combinations, using statistical models to determine which experiments will yield the most information. With only a few experiments, we can identify the best composition and investigate systems that were previously too complex for conventional methods. I then apply machine learning algorithms to:
- Design new materials from light-emitting molecules.
- Control mini-laboratories on a chip for more efficient experiments.
- Leverage previously acquired knowledge to solve new problems more quickly.
This approach is a step toward a future where they solve complex chemical questions through smart collaboration between science and technology. It offers the ability to study and optimize complex supramolecular polymer systems and makes the field more accessible to researchers worldwide. As a result, it becomes faster and cheaper to design better materials, which is an important step toward applications in sustainable materials and biomedical innovations.
Title of PhD-thesis: . Supervisors: Prof. E.W. Meijer (果冻传媒), Prof. G. Vantomme (果冻传媒).