William Humphrey

NAMD: a Parallel, Object-Oriented Molecular Dynamics Program

NAMD is a molecular dynamics program designed for high performance simulations of large biomolecular systems on parallel computers. An object-oriented design imple mented using C++ facilitates the incorporation of new algorithms into the program. NAMD uses spatial decom position coupled with a multithreaded, message-driven design, which is shown to scale efficiently to multiple processors. Also, NAMD incorporates the distributed par allel multipole tree algorithm for full electrostatic force evaluation in O(N) time. NAMD can be connected via a communication system to a molecular graphics program in order to provide an interactive modeling tool for viewing and modifying a running simulation. The application of NAMD to a protein-water system of 32,867 atoms illus trates the performance of NAMD.

Efficient coupling of parallel applications using PAWS

PAWS (Parallel Application WorkSpace) is a software infrastructure for use in connecting separate parallel applications within a component-like model. A central PAWS Controller coordinates the linking of serial or parallel applications across a network to allow them to share parallel data structures such as multidimensional arrays. Applications use the PAWS API to indicate which data structures are to be shared and at what points the data is ready to be sent or received. PAWS implements a general parallel data descriptor and automatically carries out parallel layout remapping when necessary. Connections can be dynamically established and dropped, and can use multiple data transfer pathways between applications. PAWS uses the NEXUS communication library and is independent of the applications parallel communication mechanism.

MDScope — A Visual Computing Environment for Structural Biology

We describe the program MDScope, an integrated set of computational tools which function as an interactive visual computing environment for the simulation and study of biopoly-mers. This environment consists of three parts: (1) vmd, a molecular visualization program for interactive display of molecular systems; (2) namd, a molecular dynamics program designed for performance, scalability, modularity, and portability, which runs in parallel on a variety of computer platforms; (3) MDCOMM, a protocol and library which functions as the unifying communication agent between the visualization and simulation components of MDScope. Modularity in both vmd and namd is accomplished through an object-oriented design, which facilitates the addition of features and new algorithms.

Three Electronic State Model of the Primary Phototransformation of Bacteriorhodopsin

The primary all-trans --> 13-cis photoisomerization of retinal in bacteriorhodopsin has been investigated by means of quantum chemical and combined classical/quantum mechanical simulations employing the density matrix evolution method. Ab initio calculations on an analog of a protonated Schiff base of retinal in vacuo reveal two excited states S1 and S2, the potential surfaces of which intersect along the reaction coordinate through an avoided crossing, and then exhibit a second, weakly avoided, crossing or a conical intersection with the ground state surface. The dynamics governed by the three potential surfaces, scaled to match the in situ level spacings and represented through analytical functions, are described by a combined classical/quantum mechanical simulation. For a choice of nonadiabatic coupling constants close to the quantum chemistry calculation results, the simulations reproduce the observed photoisomerization quantum yield and predict the time needed to pass the avoided crossing region between S1 and S2 states at tau1 = 330 fs and the S1 --> ground state crossing at tau2 = 460 fs after light absorption. The first crossing follows after a 30 degrees torsion on a flat S1 surface, and the second crossing follows after a rapid torsion by a further 60 degrees. tau1 matches the observed fluorescence lifetime of S1. Adjusting the three energy levels to the spectral shift of D85N and D212N mutants of bacteriorhodospin changes the crossing region of S1 and S2 and leads to an increase in tau1 by factors 17 and 10, respectively, in qualitative agreement with the observed increase in fluorescent lifetimes.

Molecular dynamics study of the 13-cis form (bR548) of bacteriorhodopsin and its photocycle.

The structure and the photocycle of bacteriorhodopsin (bR) containing 13-cis,15-syn retinal, so-called bR548, has been studied by means of molecular dynamics simulations performed on the complete protein. The simulated structure of bR548 was obtained through isomerization of in situ retinal around both its C13-C14 and its C15-N bond starting from the simulated structure of bR568 described previously, containing all-trans,15-anti retinal. After a 50-ps equilibration, the resulting structure of bR548 was examined by replacing retinal by analogues with modified beta-ionone rings and comparing with respective observations. The photocycle of bR548 was simulated by inducing a rapid 13-cis,15-anti-->all-trans,15-syn isomerization through a 1-ps application of a potential that destabilizes the 13-cis isomer. The simulation resulted in structures consistent with the J, K, and L intermediates observed in the photocycle of bR548. The results offer an explanation of why an unprotonated retinal Schiff base intermediate, i.e., an M state, is not formed in the bR548 photocycle. The Schiff base nitrogen after photoisomerization of bR548 points to the intracellular rather than to the extracellular site. The simulations suggest also that leakage from the bR548 to the bR568 cycle arises due to an initial 13-cis,15-anti-->all-trans,15-anti photoisomerization.

Photoproducts of bacteriorhodopsin mutants: a molecular dynamics study.

Molecular dynamics simulations of wild-type bacteriorhodopsin (bR) and of its D85N, D85T, D212N, and Y57F mutants have been carried out to investigate possible differences in the photoproducts of these proteins. For each mutant, a series of 50 molecular dynamics simulations of the photoisomerization and subsequent relaxation process were completed. The photoproducts can be classified into four distinct classes: 1) 13-cis retinal, with the retinal N-H+ bond oriented toward Asp-96; 2) 13-cis retinal, with the N-H+ oriented toward Asp-85 and hydrogen-bonded to a water molecule; 3) 13,14-di-cis retinal; 4) all-trans retinal. Simulations of wild-type bR and of its Y57F mutant resulted mainly in class 1 and class 2 products; simulations of D85N, D85T, and D212N mutants resulted almost entirely in class 1 products. The results support the suggestion that only class 2 products initiate a functional pump cycle. The formation of class 1 products for the D85N, D85T, and D212N mutants can explain the reversal of proton pumping under illumination by blue and yellow light.

Generic Programming in POOMA and PETE

POOMA is a C++ framework for developing portable scientific applications for serial and parallel computers using high-level physical abstractions. PETE is an general-purpose expression-template library employed by POOMA to implement expression evaluation. This paper discusses generic programming techniques that are used to achieve flexibility and high performance in both POOMA and PETE. POOMAs array class factors the data representation and look-up into a generic engine concept. PETEs expression templates are used to build and operate efficiently on expressions. PETE is implemented using generic techniques that allow it to adapt to a variety of client-class interfaces, and to provide a powerful and flexible compile-time expression-tree-traversal mechanism.

Optimization of Data-Parallel Field Expressions in the POOMA Framework

The POOMA framework is a C++ class library for the development of large-scale parallel scientific applications. POOMAs Field class implements a templated, multidimensional, data-parallel array that partitions data in a simulation domain into sub-blocks. These subdomain blocks are used on a parallel computer in data-parallel Field expressions. In this paper we describe the design of Fields, their implementation in the POOMA framework, and their performance on a Silicon Graphics Inc. Origin 2000. We focus on the aspects of the Field implementation which relate to efficient memory use and improvement of run-time performance: reducing the number of temporaries through expression templates, reducing the total memory used by compressing constant regions, and performing calculations on sparsely populated Fields by using sparse index lists.

OOPS: an object-oriented particle simulation class library for distributed architectures

Abstract A general purpose, object-oriented particle simulation (OOPS) library has been developed for use on a variety of system architectures with a uniform high-level interface. This includes the development of library implementations for the CM5, PVM clusters, and the CRI T3D. Codes written on any of these platforms can be ported to other platforms without modifications by utilizing the high-level library. The general character of the library allows application to such diverse areas as plasma physics, suspension flows, vortex simulations, porous media, and materials science.

Particle Beam Dynamics Simulations Using the POOMA Framework

A program for simulation of the dynamics of high intensity charged particle beams in linear particle accelerators has been developed in C++ using the POOMA Framework, for use on serial and parallel architectures. The code models the trajectories of charged particles through a sequence of different accelerator beamline elements such as drift chambers, quadrupole magnets, or RF cavities. An FFT-based particle-in-cell algorithm is used to solve the Poisson equation that models the Coulomb interactions of the particles. The code employs an object-oriented design with software abstractions for the particle beam, accelerator beamline, and beamline elements, using C++ templates to efficiently support both 2D and 3D capabilities in the same code base. The POOMA Framework, which encapsulates much of the effort required for parallel execution, provides particle and field classes, particle-field interaction capabilities, and parallel FFT algorithms. The performance of this application running serially and in parallel is compared to an existing HPF implementation, with the POOMA version seen to run four times faster than the HPF code.