Integrating multi-level molecular simulations across heterogeneous resources

Biomolecular simulations play a key role in the study of complex biological processes at microscopic levels in which macromolecules such as proteins are involved. The simulations are usually computationally demanding and no single method can achieve all levels of details. Thus, the simulations at different levels need to be integrated to jointly manifest atomic insights into these processes. This paper presents a Grid-based simulation framework to support the integration of multi-level simulations by means of dynamic coupling, automated workflow management, resource-dependent job distribution, and XML-based data representation. The framework provides an e-Science infrastructure to support biomolecular simulations on Grids. A biomolecular simulation markup language called BioSimML is developed to provide a formatted data representation to the multi-level simulations. Experimental simulations have shown flexible integration and high performance enhancement achieved in molecular simulations based on our framework. © 2007 IEEE.