Skip to content


    Skip to content

    MOLECULAR SCIENCE | Contents | Next


    Greater Computing Power and Collaboration Help Advance Molecular Dynamics Simulations

    PROJECT LEADER
    J. Andrew McCammon
    UC San Diego

    Scalable parallel computation has enabled molecular biologists and biochemists to do more and longer simulations, using more accurate methods to approximate molecular interactions. To take best advantage of the new intensity and quickening pace of computational work, researchers must collaborate more efficiently while processing massive sets of data.

    "We're a far-flung crowd," said J. Andrew McCammon, who leads a group of computational biochemists at UC San Diego. "We have collaborators from Washington to Texas to Maryland, and co-workers in labs abroad. We must share our simulations across the Net, and we need to generalize the environments for our codes so various combinations of hardware and interfaces can be used interchangeably."

    McCammon, the Joseph Mayer Professor of Theoretical Chemistry in the Department of Chemistry and Biochemistry and the Department of Pharmacology at UC San Diego, is no stranger to the art of extending the range and scope of computational technology. He and his group have used SDSC resources since 1986 and have originated a number of widely used molecular simulation and dynamics programs.

    As participants in the NPACI Molecular Science thrust area, the McCammon group exemplifies the state of the art in biomolecular computation. This summary of some of their current projects indicates the directions that the computational infrastructure is taking to advance molecular science.

    ACETYLCHOLINESTERASE
    MOUSE ACHE
    HIV INTEGRASE
    GREEN FLUORESCENT PROTEIN

    ACETYLCHOLINESTERASE

    A main focus for the group is acetylcholinesterase (AChE), an enzyme of fundamental interest for its biological activity and its special biophysical properties. It acts very rapidly to stop neurotransmission at cholinergic synapses like those found in the brain and at neuromuscular junctions--consistent with the need for speedy responses in the neuromuscular system. AChE has practical importance in medicine as a target for drugs for the management of Alzheimer's and other diseases and in agriculture as a target for pesticides.

    The first AChE structure, determined by X-ray crystallography, was that of the electric ray Torpedo californica. "But the structure presented a puzzle," said Jim Briggs, an adjunct assistant professor of Pharmacology at UC San Diego who collaborates with the McCammon group.

    AChE had only one apparent route of access to the active site--a 20-angstrom-long channel from the surface of the enzyme. But the center of the channel was too narrow (as seen in the crystal structure) to allow binding of the substrate acetylcholine or expulsion of the products choline and acetic acid. This, and the lack of pathways for solvent water to exit as the substrate entered, suggested to the group that dynamic openings and closings of the channel must occur if the enzyme were to function properly.

    Top | Contents | Next


    B68-cmykFigure 1: AChE "Side Door"

    Some of the largest AChE simulations to date have been conducted by McCammon group researcher Stanislaw Wlodek at the University of Houston's Texas Center for Advanced Molecular Computation. An open "side door" in one of the AChE subunits was recorded at 152 picoseconds of one such molecular dynamics simulation.


    Brief molecular dynamics simulations, in which only a portion of the enzyme and the surrounding water was allowed to move, displayed the expected fluctuations in the channel. "The first simulations also showed the transient opening of a secondary passageway to the active site," Briggs said, "a kind of back door through which the substrate might enter, propelled or 'steered' by the electrostatic field of the molecule." The work was published in Science in 1994.

    Some of the largest simulations of AChE to date have been conducted by McCammon group researcher Stanislaw Wlodek at the University of Houston's Texas Center for Advanced Molecular Computation, an NPACI partner. "We began this work on the Intel Paragon at SDSC, where we did a very long simulation--750 picoseconds," Wlodek said. "With 128 processors, a 5-picosecond chunk of the simulation took about 10 hours." The group is now working on the CRAY T3E and the IBM SP machines, which are much faster, and performing longer simulations.

    The simulations watch the dynamic fluctuations in the channel. The channel walls can have openings large enough to admit solvent molecules, while the channel itself can open widely enough to admit substrate. "Even though the active site looks inaccessible, it isn't," Wlodek said (Figure 1). "It's more like Swiss cheese, in communication with the outside across the walls, which extends the original 'back door' picture."


    open_closed-cmykFigure 2: AChE Solvent-accessible Surfaces

    Simulations by the McCammon group showed the solvent-accessible surfaces of the reaction gorge in one of the subunits of acetylcholinesterase (AChE) illustrating extreme cases of closed and fully open principal entry to the active site as revealed by molecular dynamics simulation. The substrate acetylcholine (ACh) is shown for comparison.


    MOUSE ACHE

    In another recent AChE development, McCammon colleague Palmer Taylor and his co-workers in the Pharmacology Department at UC San Diego have now solved the structure of AChE from a mouse. "Mouse AChE is of interest in part because of its greater homology to human AChE and relevance to human medicine," McCammon said. This is the structure the group is using for investigations of AChE inhibitors, including huperzine, a candidate anti-Alzheimer's drug. The new simulations make use of a scalable parallel molecular dynamics code, nwARGOS, part of the NWChem suite of programs from the Pacific Northwest National Laboratory (PNNL) in Washington. The nwARGOS software was developed by former McCammon group member and collaborator Tjerk Straatsma now at PNNL, which is also an NPACI partner.

    Taylor's group has data on the binding of an inhibitor protein called fasciculin-2 to the surface of mouse AChE. This binding is also very fast, diffusion-controlled, and accelerated by electrostatic steering. "We have started a large program of simulations of fasciculin-2, because the interaction is a model for many other protein-protein reactions," McCammon said.

    Top | Contents | Next

    HIV INTEGRASE

    Briggs directs the group's new work on HIV integrase, the "third target" for AIDS drug intervention. The first target was the virus's reverse transcriptase, inhibited by the drug AZT; then the protease; and now the integrase, which inserts viral DNA into the DNA of normal cells. The integrase step recruits the host cell machinery to make new viral particles.

    "This is a large project involving several groups," Briggs said. The McCammon group is conducting simulations. A group at the Salk Institute led by Senyon Choe is improving on the original crystallographic structure published in 1994 by a group at NIH, and another group at Salk led by Rick Bushman is working on virology. Organic synthesis of potential inhibitors is being carried out in the lab of Jay Siegel at UC San Diego.

    Top | Contents | Next

    GREEN FLUORESCENT PROTEIN

    McCammon group postdoctoral researcher Volkhard Helms has been studying green fluorescent protein (GFP). This protein is found in the bioluminescent jellyfish Aequorea victoriana and similar organisms, and it is becoming established as a valuable "reporter molecule" for imaging and monitoring the location of proteins and their role in gene expression.

    Helms is investigating GFP with its chromophore, the central portion of the molecule responsible for the fluorescence, either neutral or anionic, corresponding to different optical behaviors. "To simulate the dynamics of the proteins in solvent in atomic detail during one nanosecond takes about 10 CPU-days on 32 nodes of the T3E," Helms said. The simulations show very little motion in the protein, consistent with the rigidity of its beta barrel structure and with the overall stability of the molecule. "This would also explain the ability of GFP to protect its fluorescence signal against environmental effects," Helms said.

    Summarizing their work, the scientists all feel that biomolecular computation has reached a threshold. "Certainly all our work would benefit from much faster network communication," McCammon said. "The growth in the size of computation that can be performed and the amount of visualization necessary to understand complex protein dynamics just underlines our need for environments in which we can collaborate, animate results in real time, present those results, and teach students to appreciate the intricate structure and beauty of molecular interactions." --MM

    Top | Contents | Next