Presenter: Victor Chardès, Laboratoire de Physique de l’École
Topic: Inference and modeling of biological networks, from active matter to the immune repertoire
Active matter and the immune repertoire are paradigmatic examples of biological networks. In both cases, their components actively interact to assemble essential functions. Accurate modeling of the interactions at play is key to making predictions about these networks’ organization and function. I will present two contributions to the modeling of these distinct systems. Recent advances in time-resolved measurements of bird flocks supply high-throughput data that allows us to infer the equation of motions that rule the collective motion of birds. To this end, we developed a novel Bayesian inference approach to learn the parameters of stochastic models from discrete, finite-length trajectories. I will show how to derive maximum likelihood estimators for the model parameters in equilibrium and non-equilibrium second order Langevin dynamics. The multi-scale structure of the immune system poses considerable difficulties to the predictive
modeling of its response to infections. We address this challenge by defining long-term optimal immune strategies that maximize the speed of response against sequential infections by an evolving pathogen. I will discuss the resulting trade-off between immune protection against novel viral strains and the necessary reorganization of the repertoire. I will show the implications of the optimality hypothesis on the architecture of the B-cell memory repertoire.