Many of the most awe-inspiring feats of biological information processing emerge from the collective dynamics of a large number of specialized cells. We are interested in the physics of such emergent multicellular information processing, with a particular focus on the adaptive immune system as an experimental model.

We are building towards quantitative theories, which give insight into cellular mechanisms of regulation and link them to computational properties at the population scale. The principal questions that motivate our research are the following: How is the organization of biological systems shaped by considerations of efficient information transduction? How can complex information processing algorithms be implemented in self-organized cellular dynamics? More practically, our research is guided by questions such as: Are there simple phenomenological laws that describe collective behavior on the population scale despite cellular scale complexity and diversity? Can we leverage ideas and tools from statistical physics to bridge scales? What is the right language of description which allows our models to make contact with experiments?

Common to our different projects is the use of a set of mathematical and computational tools from statistical physics, nonlinear dynamics, and information theory to bioinformatics and machine learning.


Collective computation and regulation in the immune system

To combat infection by diverse and ever-evolving pathogens vertebrates have evolved an intricate defense machinery consisting of a large population of highly specialized cells – the adaptive immune system. To provide efficient defense this system adapts dynamically to the pathogens it encounters. We are interested in understanding the collective mechanisms that regulate this process and the computational principles they implement. In the past, we have made succesful headway on question such as: How does a regulated response to a pathogenic challenge arise from the proliferation of individual cells? And how should a well-adapting immune system best adjust to a changing pathogenic environment?

Reading the T cell receptor code

The specificity of adaptive immune responses relies on the binding of hyper-variable receptors to diverse ligands. Advances in the depth at which the hyper-variable receptor loci can be sequenced provide unprecedented resolution into the many-to-many mapping between receptors and ligands. What does this data reveal about the sequence determinants of specific binding? To address this question we are combining ideas from population genetics, molecular biophysics, and machine learning. By applying them to increasingly abundant data, we are beginning to decipher the probabilistic rules of the degenerate molecular binding code of the immune system. Ultimately, we envisage that an ability to read the T cell receptor code will change the way we track immune responses to infection, vaccinations, and cancer.

High-dimensional ecology and long-term immune repertoire dynamics

New high throughput sequencing techniques make the study of populations with a very large number of distinct species accessible. As the number of species increases so does the number of possible interactions between species and with different environmental factors. This poses a challenge to classical ecological theory. A possible modeling framework is provided by statistical physics: we can hope to understand the collective behavior of such populations using effective models, where stochastic forces account for the net effect of myriad interactions. We have applied this Langevin approach to the dynamics of the human T cell repertoire, which provides an experimentally accessible model system of great practical importance. We have pioneered an approach that links pseudotemporal snapshots of immune composition in people of different ages to underlying dynamics, and have provided evidence that dynamics early in life leave a life-long imprint on repertoire organization.