Identifiability, Sensitivity, and Genetic Algorithms in Bacterial Biofilm Selection Models
Which experiments can actually measure a biofilm selection model's parameters? Optimized sampling beats more data.
Update (2026): now published in the Bulletin of Mathematical Biology.
This paper applies modern tools, identifiability and sensitivity analysis, to models of how bacterial populations are selected for biofilm formation. The central question: what data would we need to measure the model’s parameters accurately?
Using a genetic algorithm to choose sampling times, we found that over a 24-hour experiment, 11 well-chosen time points estimate the biofilm-formation rate more accurately than 51 evenly spaced ones.
The work grew out of a project with Daravuth Cheam, who is running the matching experiments in Vibrio fischeri in the Nishiguchi lab.
Williams, S., Cheam, D., Nishiguchi, M. K., Sindi, S. S., Khatri, S., Rutter, E. M. Bull. Math. Biol. 88(7), 132 (2026).
More on this project: Vibrio fischeri.