Modeling as a tool of modern data science – from heterogeneous data to reliable model predictions in biology and medicine

Abstract

The amount, size and complexity of data has enormously increased in the last decade, opening up new possibilities, especially in the life sciences and in biological and medical contexts. At the same time, modern data science approaches are rapidly developing as computational capabilities increase and have long since found their way into scientific routines and workflows. 

On the other hand, computer models provide a framework for the integration of multi-modal data and enable the inclusion of expert knowledge about the respective system. With a good model, hypotheses can be made and predictions can be made that go beyond the possibilities of machine learning and AI tools, which are often based on data alone. 

In my talk, I will show exemplary results of modeling approaches using different case studies. First, I will present a model for synthetic epigenetic memory circuits based on a methylation-dependent positive feedback loop in Escherichia coli that is able to generate hypotheses about the behavior of individual cells from bulk data. This is work in collaboration with Prof. Albert Jeltsch (University of Stuttgart). Second, I discuss a statistical Bayesian workflow as a tool for processing sparse and highly variable time series data related to liver resection in animals and human patients. This project is part of the research unit FOR 5151 – Quantifying Liver Perfusion-Function Relationship in Complex Resection. 

Vortragende