Abstract: Narrowing the gap between statistical learning and mechanistic modelling to illuminate complex biology
Given enough data, machine and statistical learning methods are often surprisingly successful in predicting the behaviour of biological systems. However, they often provide only limited insights into causal or mechanistic relationships. Mechanistic models are better suited to design intervention and control strategies, but developing a useful mechanistic model is often elusive, in particular when the knowledge about the system interactions is incomplete or uncertain.
In this talk, examples from cancer and microbial systems biology will be used to discuss the gap between and learning algorithms and mechanistic models. In particular, we will explore new ideas for combining both approaches to facilitate mechanistic insights even with incomplete models and data.