168
views
0
recommends
+1 Recommend
0 collections
    0
    shares
      • Record: found
      • Abstract: found
      • Article: not found

      Routine Microsecond Molecular Dynamics Simulations with AMBER on GPUs. 1. Generalized Born

      research-article

      Read this article at

      ScienceOpenPublisherPMC
      Bookmark
          There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.

          Abstract

          We present an implementation of generalized Born implicit solvent all-atom classical molecular dynamics (MD) within the AMBER program package that runs entirely on CUDA enabled NVIDIA graphics processing units (GPUs). We discuss the algorithms that are used to exploit the processing power of the GPUs and show the performance that can be achieved in comparison to simulations on conventional CPU clusters. The implementation supports three different precision models in which the contributions to the forces are calculated in single precision floating point arithmetic but accumulated in double precision (SPDP), or everything is computed in single precision (SPSP) or double precision (DPDP). In addition to performance, we have focused on understanding the implications of the different precision models on the outcome of implicit solvent MD simulations. We show results for a range of tests including the accuracy of single point force evaluations and energy conservation as well as structural properties pertainining to protein dynamics. The numerical noise due to rounding errors within the SPSP precision model is sufficiently large to lead to an accumulation of errors which can result in unphysical trajectories for long time scale simulations. We recommend the use of the mixed-precision SPDP model since the numerical results obtained are comparable with those of the full double precision DPDP model and the reference double precision CPU implementation but at significantly reduced computational cost. Our implementation provides performance for GB simulations on a single desktop that is on par with, and in some cases exceeds, that of traditional supercomputers.

          Related collections

          Most cited references47

          • Record: found
          • Abstract: found
          • Article: found
          Is Open Access

          Rapid planetesimal formation in turbulent circumstellar discs

          The initial stages of planet formation in circumstellar gas discs proceed via dust grains that collide and build up larger and larger bodies (Safronov 1969). How this process continues from metre-sized boulders to kilometre-scale planetesimals is a major unsolved problem (Dominik et al. 2007): boulders stick together poorly (Benz 2000), and spiral into the protostar in a few hundred orbits due to a head wind from the slower rotating gas (Weidenschilling 1977). Gravitational collapse of the solid component has been suggested to overcome this barrier (Safronov 1969, Goldreich & Ward 1973, Youdin & Shu 2002). Even low levels of turbulence, however, inhibit sedimentation of solids to a sufficiently dense midplane layer (Weidenschilling & Cuzzi 1993, Dominik et al. 2007), but turbulence must be present to explain observed gas accretion in protostellar discs (Hartmann 1998). Here we report the discovery of efficient gravitational collapse of boulders in locally overdense regions in the midplane. The boulders concentrate initially in transient high pressures in the turbulent gas (Johansen, Klahr, & Henning 2006), and these concentrations are augmented a further order of magnitude by a streaming instability (Youdin & Goodman 2005, Johansen, Henning, & Klahr 2006, Johansen & Youdin 2007) driven by the relative flow of gas and solids. We find that gravitationally bound clusters form with masses comparable to dwarf planets and containing a distribution of boulder sizes. Gravitational collapse happens much faster than radial drift, offering a possible path to planetesimal formation in accreting circumstellar discs.
            Bookmark
            • Record: found
            • Abstract: found
            • Article: found
            Is Open Access

            Why do ultrasoft repulsive particles cluster and crystallize? Analytical results from density functional theory

            We demonstrate the accuracy of the hypernetted chain closure and of the mean-field approximation for the calculation of the fluid-state properties of systems interacting by means of bounded and positive-definite pair potentials with oscillating Fourier transforms. Subsequently, we prove the validity of a bilinear, random-phase density functional for arbitrary inhomogeneous phases of the same systems. On the basis of this functional, we calculate analytically the freezing parameters of the latter. We demonstrate explicitly that the stable crystals feature a lattice constant that is independent of density and whose value is dictated by the position of the negative minimum of the Fourier transform of the pair potential. This property is equivalent with the existence of clusters, whose population scales proportionally to the density. We establish that regardless of the form of the interaction potential and of the location on the freezing line, all cluster crystals have a universal Lindemann ratio L = 0.189 at freezing. We further make an explicit link between the aforementioned density functional and the harmonic theory of crystals. This allows us to establish an equivalence between the emergence of clusters and the existence of negative Fourier components of the interaction potential. Finally, we make a connection between the class of models at hand and the system of infinite-dimensional hard spheres, when the limits of interaction steepness and space dimension are both taken to infinity in a particularly described fashion.
              Bookmark
              • Record: found
              • Abstract: found
              • Article: not found

              Hyperfine-Shifted 13C and 15N NMR Signals from Clostridium pasteurianum Rubredoxin: Extensive Assignments and Quantum Chemical Verification

              Stable isotope-labeling methods, coupled with novel techniques for detecting fast-relaxing NMR signals, now permit detailed investigations of paramagnetic centers of metalloproteins. We have utilized these advances to carry out comprehensive assignments of the hyperfine-shifted 13C and 15N signals of the rubredoxin from Clostridium pasteurianum (CpRd) in both its oxidized and reduced states. We used residue-specific labeling (by chemical synthesis) and residue-type-selective labeling (by biosynthesis) to assign signals detected by one-dimensional 15N NMR spectroscopy, to nitrogen atoms near the iron center. We refined and extended these 15N assignments to the adjacent carbonyl carbons by means of one-dimensional 13C[15N] decoupling difference experiments. We collected paramagnetic-optimized SuperWEFT 13C[13C] constant time COSY (SW-CT-COSY) data to complete the assignment of 13C signals of reduced CpRd. By following these 13C signals as the protein was gradually oxidized, we transferred these assignments to carbons in the oxidized state. We have compared these assignments with hyperfine chemical shifts calculated from available X-ray structures of CpRd in its oxidized and reduced forms. The results allow the evaluation of the X-ray structural models as representative of the solution structure of the protein, and they provide a framework for future investigation of the active site of this protein. The methods developed here should be applicable to other proteins that contain a paramagnetic center with high spin and slow electron exchange.
                Bookmark

                Author and article information

                Journal
                J Chem Theory Comput
                J Chem Theory Comput
                ct
                jctcce
                Journal of Chemical Theory and Computation
                American Chemical Society
                1549-9618
                1549-9626
                26 March 2012
                08 May 2012
                : 8
                : 5
                : 1542-1555
                Affiliations
                []San Diego Supercomputer Center, simpleUniversity of California San Diego , 9500 Gilman Drive MC0505, La Jolla, California 92093, United States
                []simpleNVIDIA Corporation , 2701 San Tomas Expressway, Santa Clara, California 95050, United States
                [§ ]Department of Chemistry and Biochemistry, simpleUniversity of California San Diego , 9500 Gilman Drive MC0505, La Jolla, California 92093, United States
                Author notes
                Article
                10.1021/ct200909j
                3348677
                22582031
                fa324d78-b45d-4dc3-8269-8390e4fd0a5b
                Copyright © 2012 American Chemical Society

                This is an open-access article distributed under the ACS AuthorChoice Terms & Conditions. Any use of this article, must conform to the terms of that license which are available at http://pubs.acs.org.

                History
                : 20 December 2011
                : 15 April 2012
                : 08 May 2012
                : 26 March 2012
                Categories
                Article
                Custom metadata
                ct200909j
                ct-2011-00909j

                Computational chemistry & Modeling
                Computational chemistry & Modeling

                Comments

                Comment on this article