Research Areas
Our research field encompasses a variety of topics including, but not limited to, statistical inference for dynamical systems, multi-scale models, metabolic models, statistical analysis, and scientific machine learning. Thus, our group is split into smaller teams dedicated to one overarching research topic each. Generally, we aim to provide insights into biological processes and make predictions through analyses, while also considering the complexities of the underlying mechanisms.
This page provides an overview of the different teams and research areas in the Hasenauer Group.
Metabolic Modeling
Metabolic dysfunctions are the driver of many diseases. Metabolism is governed by intricate regulatory mechanisms, and therefore, intuitively understanding dynamical metabolic responses to perturbations is hardly possible. Therefore, we combine dynamical models of metabolic processes with various types of data to get a mechanistic and quantitative understanding of metabolism in health and disease.
Multi-Scale Modeling
Biology is complex, with processes interacting on various spatial and temporal scales, requiring complex mechanistic descriptions and efficient, accurate inference approaches. In this subgroup, we develop such approaches that scale up to complex problems on multiple fronts, combining methods from classical statistical inference and machine learning. Specifically, we work on likelihood-free inference, amortized neural density estimation, surrogate modeling, non-linear mixed effects modeling, multi-cellular modeling, and differential equations. We develop novel methods, and software tools that make these broadly accessible.
Scientific Machine Learning
Our team aims to efficiently combine the advantages of interpretable mechanistic models with data-driven machine learning methods. Mechanistic models like differential equations benefit from decades of research in modeling complex biological processes from first principles. These models achieve interpretability and uncertainty analysis, which are crucial tools for hypothesis testing and validating biological assumptions. In contrast, data-driven approaches from machine learning, particularly neural networks, extract information unbiasedly and require minimal knowledge of scientific laws. This team aims to combine both approaches to construct efficient, data-driven, and interpretable modeling strategies for various biological problems.
Statistical Analysis
We are restructuring our group and will update the following text soon.
The COVID-19 pandemic has dramatically brought to attention the importance of improving our preparedness against the appearance of novel infectious diseases. Mathematical modeling can play a vital role in this effort,bringing to light the fundamental mechanisms underlying the data gathered by clinicians. The two main directions of this subgroup are: epidemiological modeling to understand the spread of the disease and evaluate government actions, and federated analysis of clinical data to combine multiple studies while preserving privacy to assess the evolution of antibodies in transplant patients or risk factors for long COVID.
Statistical Inference
Statistical inference plays a crucial role in the application of mathematical modeling to real-world problems, by providing a framework that facilitates identifiability analysis, uncertainty quantification, comparison of models, and more. In this subgroup, we generally work on applications that arise from collaborations with biological and medical researchers, from both academia and industry. We develop models to describe the real-world processes, and we develop methods to do this better.