Specific aims and approach

In summary, we intend to develop and make available these technologies:

1. Multiscale algorithms for molecular dynamics that broaden the time scales for which atomistic simulations are possible:
   1. More stable multiple time stepping algorithms, that allow anywhere from a twofold to an eightfold asymptotic speedup over the Verlet-I/r-RESPA method using the mollified impulse method, and much larger speedups using mild Langevin damping that does not destroy dynamic properties. This will be accomplished by overcoming nonlinear instabilities present in Verlet-I/r-RESPA, and by a better splitting of time scales that allows one to most efficiently exploit smooth approximations to the forces in MD at different length and time scales.

   2. Multiscale Semi-Implicit Constrained MD (MUSICO). This method overcomes the time scale limit by using a symplectic splitting into an implicit scheme for nearly linear forces and short-range nonbonded forces, and an explicit scheme for long-range electrostatics. Also, this will make the overall program more scalable because non-scalable fast electrostatics methods will be evaluated less often.

2. Scalable biased hybrid Monte Carlo algorithm statistical sampling. Using a cheap, high-order approximation to the Hamiltonian that is exactly integrated by the MD integrator, we get a nearly linear scaling of HMC with system size. This method introduces a systematic error or bias. We remove this bias by modifying the acceptance rule. All of these algorithms will be tested against results in the literature, and then applied to the study of challenging ion channel dynamics and function problems.
3. Dissemination of results. All new methods will be made available in PROTOMOL, our open source, object-oriented, generic parallel-component framework for MD and HMC. This does not exclude incorporating them into other programs. PROTOMOL will be extended to also include:
   1. Intuitive user interface, supporting for example haptic devices, and file compatibility withAMBER, CHARMM, andNAMD.

   2. Simplified simulation protocols, since nearly optimal simulation parameters will be determined at runtime.

   3. Scalable support of Windows and Macintosh personal computers, in addition to the traditional UNIX supercomputers.

   4. Learning modules: Tutorials on different aspects of dynamics and function of ion channels will be incorporated into PROTOMOL. These will be accessible to students of all ages and backgrounds, and they will be tested in a state-of-the-art facility, the Notre Dame Learning Center, in courses taught by the PI at the undergraduate and graduate level. These will eventually be made accessible to the whole community.

4. Collaborative applications. Potassium channel MD simulations for permeation and gating using approximate atomistic models for specific biomolecular processes will be simulated with our multiscale methodology, and compared to other approximate models for the biomolecules, for the permeation and gating of potassium channels. Also, a study of the computational effectiveness of anti breast cancer drugs will be attempted.