Large scale, realistic simulations of complex biological and chemical phenomena at the atomic level of resolution level present a grand challenge for molecular simulation. Effective sampling of conformational space may require large numbers of computationally intensive simulations which are coupled to one another. Enhanced conformational sampling algorithms based on the application of biasing forces and replica exchange generalized ensembles, whereby a large number of replicas of the system are simulated in parallel, among the most powerful methods to study a wide variety of physicochemical processes. Uncoupled methods currently in use are very slowly convergent and often of dubious reliability as the independent simulations are not in equilibrium with one another. The key aspect of replica exchange (RE) algorithms is that replicas of the system periodically exchange their state parameters allowing them to rapidly traverse conformational space and to enhance equilibration. Current synchronous formulations of the RE method in wide use, however, are highly limited in terms of scalability and control when many exchanging replicas are involved. This limitation precludes the use of RE simulations to new application areas that require the calculation of high-dimensional free energy surfaces, and necessitate the dynamic control of 103-104 replicas as the landscape evolves. This project involves the development of a robust adaptive force biasing procedure coupled with an asynchronous replica exchange method. The research team is developing a novel infrastructure, the Replica Exchange Frame work (REFW) to enable the execution of very large scale RE simulations on a broad range of production computational resources, including but not limited to NSF TeraGrid (and its successor XD), cloud and campus-level cluster environments, as well as the forthcoming Blue Waters supercomputer. The REFW is being applied to applications that present multiple levels of complexity, such as coupled ligand binding, conformational change and catalysis in the glmS ribozyme/riboswitch that were hitherto not possible.
The cyberinfrastructure created by this research team enables realistic simulations of important biological processes that have relevance in many areas of biology, biophysics, medicinal chemistry, and biophysics with the potential to impact human health. Additionally, the REWF may be applied in many other scientific areas that increasingly rely on realistic simulation including catalysis, earthquake prediction and petroleum engineering. The project is also training the next generation of computational scientists to apply these methods to solve high-impact interdisciplinary research problems. The resulting technology and training enables the study of a host of new reactive chemical problems of unprecedented complexity, and greatly facilitates innovation and discovery through advanced computation.
This is a Cyber-Enabled Discovery and Innovation Program award and is co-funded by the Division of Chemistry and the Division of Physics in the Directorate for Mathematical and Physical Sciences.
