Modeling emergent phenomena in adaptive immunity: from stereotyped immunodominance to ‘implicit regularization’ in antibody evolution: Federica Ferretti ( NITMB, Chicago)
Abstract: The immune system is a complex, self-organized, very dynamic biological system, which must operate reliably and reproducibly to protect organisms from threats. This robustness is mediated by the statistical and biophysical aspects of the various processes of interest: in this sense, it can be seen as an emergent property. In this talk, I will focus on two aspects of antibody evolution where quantitative approaches based on statistical and dynamical modeling can provide a useful description, and help establish a link between microscopic mechanisms and emergent biological function.
I will first introduce the problem of antibody immunodominance, and propose a new quantitative framework to characterize the phenomenon. This framework is based on a simple population dynamics model, which can be mapped to a parabolic Anderson model on structured landscapes.
I will then briefly discuss the theoretical study of a mechanism of `implicit regularization’ in population dynamics, inspired by antibody evolution in heterogeneous lymph-node environments. This mechanism is abstracted into an ‘annealed population heterogeneity’, which can be thought of as a population-wide implementation of the mini-batching strategy used in machine learning. Like in ML, this produces an inductive bias that favors the selection of generalist antibodies—a major goal in the design of immunization strategies and development of therapeutic antibodies.