The Algorithmic Cheminformatics group develops computational methods for understanding, analysing and designing complex chemical and biochemical systems. Our work is motivated by applications in biotechnology and the life sciences, including metabolic and biosynthetic pathway design, secondary-metabolite production, mass spectrometry, stable-isotope tracing, microbiome analysis and astrobiology. A common challenge is to explore molecular transformations beyond reactions already documented in databases while retaining chemical detail down to individual atoms.
Previous and Current Research
Our research connects biochemical applications with mathematical and algorithmic foundations. In biotechnology, we investigate how metabolic and biosynthetic systems can be analysed and redesigned computationally, including secondary-metabolite production, modular polyketide synthases and alternative pathways through complex reaction networks. In analytical chemistry, we model molecular fragmentation and the movement of labelled atoms. In microbiome research, we study metabolic interactions between organisms and their consequences for microbial communities.
To address such questions, we develop formal representations of molecules, reactions and reaction networks. A central framework is graph transformation, also known as graph rewriting. Molecules are represented as labelled graphs and chemical reactions as transformations of these graphs, providing a generic and executable description of chemical change at atomistic resolution.
Category-theoretical concepts provide mathematical foundations for treating reactions as composable operators and for understanding how local transformations combine into larger transformations and reaction pathways. At the network level, we use Petri nets, directed hypergraphs and integer hyperflows to represent interactions between molecules and reactions. Boolean networks provide a complementary qualitative framework for studying states and dynamics in biological and microbial systems.
A defining feature of our approach is that it does not require a fixed reaction network. Starting from formal reaction rules and a set of initial molecules, we generate and explore possible chemical spaces. Atom mappings and molecular transformations remain explicit throughout this process, making it possible to investigate hypothetical compounds, reactions and pathways alongside experimentally documented ones.
Because the resulting chemical spaces can become very large, we develop graph algorithms, combinatorial optimisation methods and integer-linear programming approaches to identify pathways, transformation patterns and feasible network behaviours under chemical constraints. This work combines advances in mathematical theory with their implementation in practical cheminformatics software.
Future Projects and Aims
Future work will strengthen the connection between these foundations and experimental applications while continuing the development of the underlying mathematical and algorithmic methods.
A central focus will be computational mass spectrometry. We use formal reaction rules to model molecular fragmentation and connect tandem-MS data with candidate fragment structures and plausible fragmentation pathways. Because this approach does not depend exclusively on spectral databases, it can also support the investigation of previously uncharacterised compounds.
We will further develop methods for stable-isotope tracing. Atom-resolved models can predict how labelled atoms propagate through metabolic and biosynthetic reaction networks. Combined with experimental isotope-labelling data, these predictions can help distinguish alternative pathways and support experimental design.
In biotechnology, we will apply our methods to metabolic engineering, modular polyketide synthases and secondary-metabolite biosynthesis. We investigate how changes to enzymes, modules and pathways affect the compounds that can be produced, with the aim of supporting the computational design of efficient and sustainable bioproduction systems.
Microbiome research is pursued particularly through MATOMIC—Mathematical Modelling for Microbial Community Induced Metabolic Diseases—which combines mathematical modelling with experimental approaches to investigate metabolic interactions in the human gut microbiome. Daniel Merkle leads MATOMIC and coordinates TACsy—Training Alliance for Computational Systems Chemistry. TACsy brings together computer science, chemistry, mathematics and machine learning to develop computational methods for large chemical reaction systems.
In astrobiology, generative chemical-space models allow us to investigate prebiotic chemistry, chemical evolution and possible routes from simple molecular building blocks to complex reaction networks.
Across these applications, our goal is to turn rigorous mathematical models and algorithms into practical tools for biotechnology, analytical chemistry, biomedical research and the study of the origins of life.