The Zuckerman group develops computer simulation methods for studying biophysical problems. A main interest is proteins that undergo conformational transitions and essentially function as molecular machines. Another interest is developing software for drug design.
Although it sounds incredible, modern computers are much too slow to perform the calculations that biophysicists would design for them – and, yes, that includes supercomputers. The basic problem is that chemical (i.e., atomistic) detail is critical in determining biomolecular behavior, and at the same time, the main processes of interest (e.g., conformational changes) are much too slow to be observed in simulations.
Modern bio-computations therefore must be designed with great care. The group’s approach to two problems is sketched below.
Slow timescales/conformational transitions. Bimolecular processes such as conformational transitions, protein folding, and inter-molecular binding typically are too slow to be observed using conventional simulations, even on the fastest supercomputers. This motivates the need for ultra-high-performance parallel computing methods. The Zuckerman group has helped to advance the “weighted ensemble” (WE) approach, which orchestrates a large number of parallel trajectories. Remarkably, the WE strategy can yield estimates of key observables such as transition rates faster than conventional parallel simulation using the same resources – and sometimes by orders of magnitude. WE’s effectiveness stems from the way it distributes computing resources to ‘rare event’ processes that normally would be undersampled or perhaps not seen at all. See the WE section of the group’s publications.
The WESTPA simulation package, which was developed largely in the group of our collaborator, Prof. Lillian Chong of the University of Pittsburgh Department of Chemistry, implements a number of approaches developed by our group and others. WESTPA includes state-of-the-art non-Markovian analysis tools we developed. Critically, WESTPA easily interfaces with existing software packages, such as molecular dynamics engines and also simulation packages at other scales, including for systems biology.
Mixed-resolution modeling for drug design. The traditional computational strategy to search for novel drug-candidate molecules is to “dock” (i.e., fit, based on shape and chemistry) a large number of small molecules into a largely rigid protein target. However, some of the most important drug targets, such as the ligand-binding domain of the estrogen receptor, exhibit significant conformational flexibility and thus frustrate conventional docking. The group has therefore developed a simulation platform that permits assessment of small-molecule binding using a more realistic model of protein flexibility. The software uses a mixed-resolution approach in which the binding site is represented in all-atom detail and is coupled to a simplified, but still flexible model of the remainder of the protein.