Overview

The Dinner group develops and applies theoretical methods for relating cellular behavior to molecular properties. We are particularly interested in how proteins regulate access to genes in the context of the development of the immune system. Understanding how such complex behavior arises from physical and chemical features is a problem in fundamental statistical mechanics, but its solution has direct implications for treating autoimmune pathologies and improving gene therapy and vaccination strategies.
One feature that makes theoretical studies of cellular behavior challenging is that the relevant dynamics span a hierarchy of time and length scales ranging from Angstroms and femtoseconds to micrometers and minutes. Experiments are now beginning to bridge gaps in spatial and temporal resolution, and models are vital for design and interpretation of such measurements. Our research thus blends atomic-resolution simulations with coarse-grained numerical and analytical approaches, often in collaboration with experimental groups.

Dynamics of DNA binding at atomic-resolution

DNA transcription, recombination, replication, and repair are all regulated by proteins that bind specific sites on DNA. Despite small copy numbers in cells, such proteins can locate target elements among billions of base pairs thousands of times faster than allowed by a three-dimensional random walk. To objectively evaluate how putative search mechanisms arise in specific molecular situations, which is essential to ultimately be able to make defined interventions, atomic-resolution simulations based on transferable potentials are required. Building on the transition path sampling framework introduced by David Chandler and co-workers, we have introduced novel means for studying dynamics in complex systems and applied them to a DNA repair system.

Gene regulation during development of the immune system

Computational approaches are important for understanding how cell-fate decisions are made because the cooperative nature of the dynamics hinders intuition of responses to experimental probes. We are thus integrating experimental data to construct phenomenological models for the gene regulatory networks that control lineage specification of myeloid and lymphoid cells. Presently, we are exploring various approaches beyond the mean-field to better be able to link antigen receptor specificity with development.

Automated reaction network discovery

Due to the availability of high throughput experimental methods for measuring molecular populations, interpretation of data is now often the bottleneck in assembling gene regulatory and signaling networks. Computational methods can address this problem by identifying molecular circuits consistent with measurements. These networks can in turn be used to design further experiments to discriminate between the suggested mechanisms. The methods that we use to integrate the rate laws that govern gene expression are quite similar in form to those that we use to propagate the positions of atoms in our atomic-resolution studies of DNA-binding proteins. We are seeking to exploit this fact, together with advances in simulation methods for treating complex systems, to improve procedures for fitting experimental data.

Spatial control of reactions within cells

The gene regulatory and signaling network studies above provide examples of all-or-none responses to smooth variations in levels of molecular species. The fact that quantitative differences in protein levels can give rise to qualitative changes in cellular behavior allows developmental programs to be controlled simply through spatial localization of proteins. Because localization is often mediated by the cytoskeleton, we are developing computational algorithms for treating nucleation, growth, and reorganization of actin filament networks.