"The long-term goal of our research is to advance the state-of-the-art in molecular simulation algorithms by 4-5 orders of magnitude, particularly in the context of the GROMACS software we are developing. This is an immense challenge, but with huge potential rewards: it will be an amazing virtual microscope for basic chemistry, polymer and material science research; it could help us understand the molecular basis of diseases such as Creutzfeldt-Jacob, and it would enable rational design rather than random screening for future drugs. To realize it, we will focus on four critical topics: • ALGORITHMS FOR SIMULATION ON GRAPHICS AND OTHER STREAMING PROCESSORS: Graphics cards and the test Intel 80-core chip are not only the most powerful processors available, but this type of streaming architectures will power many supercomputers in 3-5 years, and it is thus critical that we design new “streamable” MD algorithms. • MULTISCALE MODELING: We will develop virtual-site-based methods to bridge atomic and mesoscopic dynamics, QM/MM, and mixed explicit/implicit solvent models with water layers around macromolecules. • MULTI-LEVEL PARALLEL & DISTRIBUTED SIMULATION: Distributed computing provides virtually infinite computer power, but has been limited to small systems. We will address this by combining SMP parallelization and Markov State Models that partition phase space into transition/local dynamics to enable distributed simulation of arbitrary systems. • EFFICIENT FREE ENERGY CALCULATIONS: We will design algorithms for multi-conformational parallel sampling, implement Bennett Acceptance Ratios in Gromacs, correction terms for PME lattice sums, and combine standard force fields with polarization/multipoles, e.g. Amoeba. We have a very strong track record of converting methodological advances into applications, and the results will have impact on a wide range of fields from biomolecules and polymer science through material simulations and nanotechnology."