We have continued our development of next-generation simulation methods, particularly the AMOEBA force field. The AMOEBA force field improves the accuracy of classical simulations, by introducing high-order fixed multipole moments and induced dipoles. Permanent multipoles provide a realistic approximation to the local electron density around each atom, which is essential for describing the directionality of hydrogen bonds and solvation spheres around a solute molecule; our previous work led to a much more efficient approach to evaluating multipole interactions. More recently we have turned our attention to the induced dipoles, which are the bottleneck in a simulations using the AMOEBA force field. The induced dipoles allow atoms to respond to their local environment, which is important when describing systems containing a combination of hydrophobic and hydrophilic zones, as most proteins do. The induced dipoles are defined as a response to the presence of the permanent multipole moments, as well as induced dipoles on other centers; the latter term means that they must be solved for in an iterative scheme. The iterations are computationally costly, and we have demonstrated that converging the equations loosely, to make them more tractable, has disastrous consequences for a simulation. To address these shortcomings, we developed a completely new scheme that embodies all of the physics of the iterative approach, but has an analytic form that is guaranteed to give stable simulations. The preliminary version of our method has shown great promise as a much more efficient way to treat the induced dipoles in the AMOEBA force field and we are working with the creator of that force field, Jay Ponder of Washington University in St Louis, to rigorously test the theory and ensure its adoption by the simulation community. AMOEBA implementation in CHARMM We have implemented static multipole interactions through hexadecapole, direct approximation of polarizable dipole interactions (iAMOEBA), iterative solution of dipoles (AMOEBA) and the perturbation approximation to dipoles. We have also implemented the Halgren buffered 14-7 potential used by AMOEBA. Virials for static multipoles, direct polarizable dipoles and iterated dipoles are also implemented. Remaining work for this project is proper accounting of 1-5 scaling, AMOEBA bonded terms and the perturbation virial. Softcore potentials in MPOLE, free energy calculations with polarizable force fields. We are implementing CHARMM's PERT/PSSP functionality in the MPOLE module of CHARMM. We implemented this for the static multipoles and van der Waals potentials so far. We are also working to combine our theory developments with state-of-the-art hardware, to increase the scope of problems that can be solved using modern force field methods. Graphical processing units (GPUs) have become ade factostandard coprocessor on modern supercomputers, but require custom code in order to operate. Working with Peter Eastman and Vijay Pande (Stanford University), we are porting our new methods into the OpenMM simulation package, which will make the tools developed in our lab freely available to all researchers. For example, speed increases of roughly 35% have been found with our new quasi-internal frame approach with spherical harmonics. Recently, code has been developed to allow the use of Hamiltonian replica exchange with CHARMMs DOMDEC functionality, which provides significantly improved scalability when compared with the old, atom-decomposition code. This effort is significant because it allows substantially improved simulation time scales to be reached using Hamiltonian replica exchange, yielding much more accurate sampling for a given amount of computer time. This new code is also compatible with DOMDEC-GPU, which means that GPUs may now be used for Hamiltonian replica exchange simulations. We are testing this new code on various systems where we need to calculate accurate free energies between different charge states to support constant pH calculations. In recent years, this lab has developed a series new computational methods, such as the self-guided Langevin dynamics for efficient conformational searching and sampling, the isotropic periodic sum method for accurate and efficient calculation of long-range interactions, and the map-based modeling tool, EMAP, for electron microscropy studies. Implementation of these new methods enables researchers to tackle difficult problems. These methods have been implemented into CHARMM to expand its capability in molecular simulation, conformational search, and structure prediction. These methods are all available in CHARMM version 40. These methods are also been implemented into another widely used simulation package, AMBER, to extend the user scope to access these methods. The SGLD, IPS, and EMAP methods are available in AMBER version 14. We did not add substantial new compute resources to LoBoS this year, however we did provision a new, more reliable and higher performing storage system to host user home directories. This new storage architecture creates larger storage pools, which is necessary as the amount of data generated by the lab each year increases. LoBoS now has 600 TB of network attached storage and over 1 PB of total storage when back-up and administrative volumes are considered.
Showing the most recent 10 out of 15 publications
