Decoding Emergence: from tracks to rules.

 Abstract :

The emergence of complexity is a fascinating feature of living matter. Oomycete plant pathogens, for instance, produce flagellated spores that explore their environment in search of nutrient sources, ultimately sensing the rhizosphere and reaching preferred infection sites in roots. While the behaviour of an individual zoospore may appear erratic, population-level exploration follows reproducible rules that can be investigated by combining high-throughput live imaging, quantitative trajectory analysis and agent-based modelling. In this seminar, I will show how experimental data can infer local motility rules directly, and that their repeated application is sufficient to reconstruct emergent exploration patterns in a simple environment. Building on this initial modelling effort, I will then present our ongoing attempt to extend this framework from data-driven modelling to AI-assisted model discovery. Teams of specialised AI agents, combining biological reasoning, mathematics, programming, and model criticism, are given experimental trajectories and challenged to formulate, implement, falsify and refine alternative behavioural models. Rather than using AI simply to analyse data or write code, we are exploring whether it can participate in the iterative scientific cycle itself: observation, hypothesis generation, modelling, prediction and criticism. This journey not only refines our previously developed models, but also opens the way to increasingly complex zoospore-environment interactions. Ultimately, such approaches could help us decipher how these pathogens navigate heterogeneous soil environments and interact with the rhizosphere.