Highlights of our Work

2017 | 2016 | 2015 | 2014 | 2013 | 2012 | 2011 | 2010 | 2009 | 2008 | 2007 | 2006 | 2005 | 2004 | 2003 | 2002 | 2001

Crashing Wave

image size: 573.0KB
Olga Svinarski and VMD

Water is ubiquitous; living cells and the cells' biomolecules are greatly influenced by water. Accounting for water in computational modeling, a key access route to imaging cell processes, comes at a stiff price: up to 90% of the simulation volume needs to be a water bath embedding simulated biomolecular machinery. The sheer number of water molecules arising here, sometimes millions, and the strong hydrodynamic drag resisting biomolecular dynamics, slows down simulations tremendously. Computational biologists had invented a means to replace water by a kind of continuous ether, so-called implicit solvent, that accounts for key electrostatic features of water around biomolecules, but does not slow down simulations. Computational hinderance through explicit water is getting worse with increasing size of biomolecular machines, but so far the implicit solvent description could not be effectively employed to model large systems using supercomputers. In a recent study, the situation has been repaired and the use of an implicit solvent in molecular dynamics simulations of systems as large as the whole ribosome (containing 300,000 ribosome atoms and eliminating 2,700,000 water atoms) have been documented. In a further study, the implicit solvent description has been extended to account also for hydrophobic interactions at biomolecular surfaces; the extension took advantage of delegating calculations both to GPU and CPU as the latter is more suitable for the evaluation of surface interactions. More information regarding NAMD's parallel computer-ready implicit solvent can be obtained from the NAMD website.
DNA translocation through graphene nanopore

image size: 649.7KB
made with VMD
movie: 2.2MB

Threading DNA electrically through nanometer-sized pores, so-called nanopores, holds promise for detecting and sequencing DNA (see Nov 2005 and Oct 2004 highlights). Nanopore measurements tend to be the more sensitive the smaller the pores are. The material graphene, which is just one atom thick and looks like a two-dimensional ``honeycomb" made up of carbon atoms, offers the ultimate physical resolution for measuring DNA (the stacking distance between base-pairs in DNA is about 0.35 nm). As reported recently, molecular dynamics simulations using NAMD revealed the motion of DNA being threaded through graphene nanopores at atomic level resolution. Simulations not only agree qualitatively with previous experiments on DNA translocation through graphene nanopores, but go one step further than the experiments and suggest how individual base pairs can be discriminated. The recent computational study is one further example for the guidance that molecular dynamics simulations provide in nanosensor development (see a recent review). More information can be found on our graphene nanopore website.
Immersive Out-of-Core Visualization with VMD

image size: 314.7KB
made with VMD

Computer simulations of the biomolecular processes in human cells guide better understanding of health and disease as well as development of dietary supplements and pharmacological treatments. Such simulations are extremely demanding and, in fact, all too often still limited by technological feasibility. However, technological advances are being brought to bear on computer simulations in biomedicine through highly dedicated biomedical engineers, who have often speerheaded uses of new computer technologies such that they became available in biomedicine much sooner than in other fields. A case in point is solid state disk (SSD) technology that can serve as extremely fast and large computer memory. Conventional RAM (random access memory) is fast, but limited in size due to cost; the well known hard disks (HDs) can hold large data sets at an affordable price, but are slow. The new SSDs are in the middle ground, faster than HDs, slower than RAM, offering large data storage at an affordable price. Modern uses of SSDs in smart phones and tablets attest to the usefulness of SSDs. The biomolecular visualization and analysis software VMD in its next release (VMD 1.9.1) makes the power of SSDs as a huge, yet fast storage medium available to biomedical researchers. As reported, this will allow them to view and analyze through VMD on the fly Gigabytes-to-Terabytes of simulation data, that are being raked into a computer at the rate of up to 4 Gigabytes per second (one high definition video of a long movie per second!). For more on this and other revolutionary features of VMD 1.9.1 see our VMD web site.
NAMD Scaling on Jaguar XT5
The NAMD molecular dynamics program excels at simulating, in atomic detail, the complex molecular machinery of living cells. NAMD now enables hundred-million-atom simulations using the full capabilities of the nation's fastest supercomputers due to parallel programming innovations reported in a paper at the SC11 supercomputing conference. While parallel supercomputers have massive amounts of memory in aggregate, any part of the machine can directly access only a small fraction. To fit the molecular blueprint of cellular structures such as the chromatophore in the memory of such machines, a compressed data format was developed that exploits the repetitive nature of proteins, nucleic acids, lipid membranes, and solvent. This format requires only a small amount of data to be copied to all parts of the supercomputer, while the remainder is read from disks in parallel and distributed across the machine. Simulation output and the balancing of work loads are similarly distributed. The Charm++ parallel programming system, on which NAMD is built, was also extended to allow the molecular structure and other data to be shared among the increasing number of cores in modern processors. A hundred-million-atom simulation can now run on supercomputers with only 2 gigabytes of memory per node, such as the IBM BlueGene/P. The enhancements are available to computational biologists world-wide in special memory-optimized versions of NAMD 2.8.
Helically symmetric nitrilase

image size: 536.0KB
made with VMD
movie: 3.3MB

Inorganic nature brings about symmetry that one can admire, for example when one sees a polished diamond. Living nature, too, brings about symmetric structures; for example, many processes in living cells are carried out by highly symmetric protein complexes. In case of living cells symmetry comes about partially out of physical or geometrical necessity, but also partially out of its usefulness for the intended purpose. Hence, understanding symmetry in biology goes beyond related studies of the beauty of symmetry in the physical and mathematical sciences as one also asks: Does symmetry help? Put it another way: Symmetry in living cells is beautiful and useful! The proteins in symmetric complexes are often imaged by electron microscopy (EM), but unfortunately not at a resolution high enough to recognize chemical detail, as is needed in most cases, e.g., in case of the study of a symmetric multi-enzyme system. Computational biologists have developed a method to solve the resolution problem, using high resolution images obtained through X-ray scattering of related molecules and matching them through molecular dynamics using NAMD to the image seen in EM. This method, called molecular dynamics flexible fitting (MDFF) has been highlighted previously and is described on our MDFF website. As reported recently, MDFF has now been extended to determine the atomic-level structure of symmetric multi-protein systems. MDFF has been applied successfully to three highly symmetric multi-protein systems: (i) GroEL-GroES, a protein complex that assists proteins to fold properly into their native conformation; (ii) a nitrilase multi-enzyme system that converts massive amounts of molecules into forms more suitable for a bacterial cell; (iii) Mm-cpn, a protein complex supposedly involved in assisting protein folding in so-called archaebacteria. In all cases the symmetry of the protein complexes plays a key role. For more details, see the Method section of our MDFF website.
IG6

image size: 284.7KB
made with VMD

Bacteria contain the simplest photosynthetic machineries found in nature. Higher organisms like algae and plants practice photosynthesis in a more elaborate but principally similar manner as bacteria. But even for its simplicity, the bacterial photosynthetic unit is not without its unsolved mysteries. Take, for example, the crucial photosynthetic core complex, which performs light absorption and the initial processing of the light energy. In certain bacterial species, the core complex are each ring-shaped singlets, but in some other bacterias they can be 8-shaped doublets, formed by the association of two ring-shaped core complexes. The question arises: What holds doublets together. It turned out that doublets arise when bacteria bacterial contain two copies, i.e., twins, of an additional small protein called PufX. Recently, computational biologists and biochemists have come together to study PufX from different bacteria. They found an interesting trend showing that bacterial species with PufX that associates with another PufX protein also contain core complexes that form 8-shaped doublets. On the other hand, bacteria whose PufX is incapable of associating with another PufX have only singlets of ring-shaped core complexes. These new results resonate with a proposed role of PufX as the protein that holds two core complexes together to form 8-shaped doublets. Bacteria with PufX that cannot perform this role therefore only have singlets of ring-shaped core complexes. The needed simulations were done with NAMD. More details can be found on our photosynthetic core complex website and in a previous highlight on PufX.
Methylated DNA

image size: 94.3KB
made with VMD
movie: 771.9KB

The genes of organisms, like plants and animals, offer the blueprint, not only to build the organism anew from a seed or fertilized egg cell, but also to adapt the living organism to its habitat and life experience. For example, a type of tree growing in an arid or wet region will adapt expression of its genes for root growth optimal to circumstances. A child living on a scarce or abundant diet or with little or much physical activity will adapt body growth accordingly. The needed, life time adaptation requires an individual's change in gene expression. This change is the subject of epigenetics. One control element in epigenetics is that cytosine bases of an organism's DNA become methylated in a chemical reaction in which a hydrogen atom is replaced by a methyl group (CH3). Another control element involves hydroxymethylation of cytosine bases where the H atom is replaced by a hydroxymethyl group (CH2OH); hydroxymethylation arises mainly in brain tissue. DNA methylation and hydroxymethylation patterns depend on an organism's individual history; aberrant patterns can be the cause of diseases, for example, of certain cancers. It is known that the proteins involved in gene expression can recognize methylated sites of DNA and, thereby, direct gene expression; DNA methylation also affects the packing of DNA in the chromosomes. However, methylation and hydroxymethylation may also affect gene expression directly; indeed, experiment and computational modeling with NAMD suggest this as an intriguing third way for methylation and hydroxymethylation to regulate gene expression. Methylation is shown, as reported recently, to make it more difficult to separate the two strands of DNA, as is necessary during gene expression. An earlier experimental-computational study had revealed already that methylated DNA can pass narrow synthetic nanopores more readily than unmethylated DNA can (see the Feb 2009 highlight). A recent experimental-computational study has shown that also in the case of hydroxymethylation mechanical properties of DNA become altered, namely, the forces needed to separate the two strands of DNA are strongly affected. The three experimental-computational findings advance our understanding of methylation and hydroxymethylation-based epigenetics and of how our body adapts to our life style and environment. More on our DNA methylation and hydroxymethylation website.
One monomer of the FMO trimer showing 8 bacteriochlorophylls.

image size: 218.3KB
made with VMD

Green sulfur bacteria are life forms that use sun light as a main food source. They harvest the light by absorbing the sun's photons with their chlorosome system, an assembly of thousands of chlorophyll molecules. Absorption produces electronic excitation of the chlorophyll molecules, but the excitation is only short-lived (lifetime of about 1 nanosecond) and needs to be turned quickly enough into a more stable form of energy. The latter is achieved in a protein complex called the reaction center, but for this purpose the electronic excitation needs to travel from the chlorosome to the reaction center through a protein complex called the Fenna-Matthews-Olson (FMO) protein. FMO is a crucial bottleneck, acting as an energy faucet, through which the short-lived excitations need to flow in a fraction of a nanosecond. Electronic excitations are quantum phenomena highly sensitive to thermal noise. Biophysicists spend much effort to measure thermal effects on FMO electronic excitation flow. Now researchers have complemented measurements through calculations and have shown why FMO function is actually robust against thermal noise. Using a combination of classical and quantum mechanical calculations they quantified the thermal noise present in FMO (reported recently here and here), and determined that thermal noise greatly reduces quantum coherent excitation in the transport through FMO (reported here), but that does not seem to be detrimental to excitation flow. More information can be found here.
Osmolyte passage times
through MscS Cytoplasmic domain.

image size: 324.1KB
made with VMD
movie: 3.5MB

Bacterial cells enclose themselves with a cell membrane to maintain an optimal interior ion concentration and interior electrical potential. The potential is about -0.1 V inside versus 0 V outside, the potential difference amounting to an important energy reservoir for the cell. Closing itself off from the outside can be dangerous, though, for a cell, namely when sudden changes outside the cell can bring a cell to burst. This can happen when ions outside the cell are washed away suddenly: outside water pushes itself then into the cell as the water prefers to be near ions found then only inside the cell, a behavior called osmosis. The osmotic push can be so hard that the cell can burst. To prevent such burst, cells evolved safety valves, one being called mechanosensitive channel of small conductance or MscS. Under pressure, the valve, i.e., MscS, opens and enough ions leave the cell to keep water from pushing in too hard (see the Mar 2008 highlight, "Observation and Simulation depict Cell's Safety Valve", Feb 2007 highlight, "Observing and Modeling a crucial Membrane Channel", May 2006 highlight, "Electrical Safety Valve", and the Nov 2004 highlight, "Japanese Lantern Protein"). But since the electrical potential inside is negative, mainly negative ions would leave the cell, discharging its potential and leaving the cell without energy. A theoretical and computational study, the latter carried out using NAMD, reports now that MscS developed apparently an ingenious solution: ions going through the MscS valve must pass a balloon-like filter that manages to mix positive and negative ions so that only a 1:1 mixture leaves a cell under osmotic shock, thereby providing protection against the shock without compromising the cell's electrical potential. More information can be found on our MscS website.
Myosin VI

image size: 266.3KB

Motor proteins transport cargos from one place in living cells to another, for example transport cell components along the long axons of nerve cells. A motor protein, such as myosin VI, has to "walk" or "run" along the cellular highway of actin filaments to perform the transport. In the case of myosin VI, snapshots from crystallography revealed that the protein's "legs" are too short to explain the step size taken. Computational and experimental biophysicists have now solved the mystery of how myosin VI dimers realize their large step size despite their short legs. The investigation, based on the program NAMD and reported recently, demonstrates that the answer lies in the flexibility of the legs. Myosin VI is able to triple the length of each leg, made of a short bundle of up-down-up connected alpha-helices, by extending the bundle to a stretched-out down-down-down geometry of segments, like turning a letter z into a a single long line. In the telescoping process, myosin VI also gets help from its well-known binding partners, namely calmodulins. The calmodulins direct the telescoping of the protein legs as well as strengthen the extended legs. Together with an earlier study of the "neck" region of the molecule (see December 2010 highlight on Opposites Attract in a Motor Protein), the scientists have established how walking myosin VI achieves its wide stride. More information can be found on our motor protein website.
Protein in Implicit Solvent

image size: 347.6KB
made with VMD

Water is the essential solvent that shapes protein structure and function, but for researchers using molecular dynamics flexible fitting (MDFF) to fit large biomolecular models to data from cryo-electron microscopy, such as fitting the classical ribosome into the ratcheted map, it was a mixed blessing. Since the network of hydrogen bonds that gives water its unique properties must rearrange as the solute moves, water molecules not only increase the size of the simulation but slow the fitting process. Leaving water out completely was a common practice, relying on the MDFF fitting potential to prevent the dehydrated protein from shriveling. The 2.8 release of NAMD provides a better option: a new implementation of the generalized Born implicit solvent model that scales to thousands of cores for large biomolecular aggregates thanks to NAMD's unique parallel structure and measurement-based load balancing system. By eliminating explicit water molecules from the simulation, an implicit solvent model helps shape protein structure while adapting immediately to new conformations. With this best-of-both-worlds option now available in NAMD, biomedical researchers using MDFF need fear water no longer.
Flagellin Translocation in Flagellum

image size: 638.3KB
made with VMD
movie: 8.1MB

Bacterial cells can swim and use for this purpose one or more flagella, whiplike appendages that exceed the length of the cell severalfold. The flagella are made of many thousand copies of a protein called flagellin, arranged in a helical fashion such that the flagella are hollow inside, forming a very long channel. When the flagella are rotated by the cell counter-clockwise, the cell swims straight; when they are rotated clockwise, the cell turns to a new direction. Through swimming and turning the cell searches its habitat for food and avoids trouble. But sometimes a flagellum breaks and needs to grow back. At this point starts an amazing process: the cell makes new flagellin and pumps the unfolded protein into the flagellar channel, extending its length. This is like squeezing toothpaste out of a tube, except in reverse, like pumping toothpaste into the tube at the toothpaste factory, and the tube is extremely long. Now researchers have described the process that makes flagella grow step-by-step through a combination of mathematics, physics, and molecular modeling using NAMD. As reported, the researchers reproduce the time course of growth as well as the length of the growth and also explain how friction of the protein paste is kept extremely low to make the flagella grow many times the length of the cell itself. More information here.
Two pigment molecules (BChls) vibrating

image size: 231.6KB
made with VMD

Photosynthetic life forms bottle the energy of sun light. How do they do it? Fast! Indeed, the first nanosecond after absorption of sun light is crucial in photosynthetic light harvesting. During this time absorbed solar energy is in its least stable form, that of electronically excited molecules which decay by re-emitting a photon (fluorescence) at a rate of 1 every nanosecond (1 every 0.000000001 seconds). Fluorescence would be wasteful to the organism and to avoid it the molecular excitation energy is transported over tens to hundreds of nanometers through an energy transfer network to so-called photosynthetic reaction centers where it is converted into a more stable form of energy (see our recent review on light harvesting). The fast transport is achieved by transferring excitation energy between clusters of strongly interacting pigment molecules that act as stepping stones and as a result the excitation energy is used in about 0.1 nanoseconds, i.e., within 10% of the fluorescence decay time, thus bottling sun light with an efficiency of 90%. The thermal motion of the pigment molecules and their protein scaffold greatly influences the excitation transport. A recent study showed that correlated thermal fluctuations that arise in pigment clusters affect the excitation transfer particularly strongly, typically slowing transfer it down. Pigment clusters that avoid correlated thermal motion increase the efficiency of light harvesting. More information can be found here.
VMD 1.9

image size: 1.9MB
Olga Svinarski and VMD

The molecular graphics program VMD has evolved to version 1.9, featuring new and improved tools for molecular cell biology. A key new tool in VMD permits visual diagnosis of the long-time dynamics of large structures. Improved computer power permits now simulation of processes like protein folding that stretch over microseconds to milliseconds. While short in human time, the process measured in terms of basic molecular motion like bond vibration is seemingly endless, involving hundreds of gigabytes of data and long hours of human attention, if not automated. VMD offers now with the Timeline tool, a convenient way of scanning such data for key "events" that signal successful process steps. Timeline can likewise assist in monitoring the dynamics of large cellular machines involving millions of atoms. VMD 1.9 sports other new tools like ParseFEP for analyzing so-called free energy perturbation calculations determining the energy arising in reaction processes as calculated with VMD's sister program NAMD. VMD includes now also a lightening fast tool to detect spatial regularities in the arrangement of small molecules in order to detect if they constitute disordered or ordered arrangements. By calculating the so-called radial distribution function the tool could identify plug formation in nanopore sensors due to precipitates that go undetected by density or visual inspection. Pleasing to the eye, molecular graphics features in VMD have been enhanced, enabling faster ray tracing, new shading features, X3D molecular scene export for display in WebGL-capable browsers including Chrome, Firefox, and Safari. See more on the VMD 1.9 release page.
IG6

image size: 60.3KB
made with VMD

A cell that is alive can be recognized typically under a microscope through cell motion, streaming of inner cell fluids, and many other mechanical cell activities. These activities are the result of mechanical forces that arise in the cell, but even though the effect of the forces on cells are tremendous, the magnitude of the forces when measured by physical instruments, are extremely small, 1/1000000000000 of the force felt when lifting a Kilogramm. Sensing such small force requires extremely high precision in research techniques, and, for more than a decade, computational biologists, using NAMD, have exploited the all-atom resolution of molecular dynamics modeling to explain very successfully the molecular level effect of small forces in cells (for example, see our previous highlights on muscle protein, blood clot protein, and other successful applications). But simulations, too, are challenged by the small magnitude of cellular forces, mainly because precise simulation requires extremely extensive simulations, that cover a duration as close as possible to the biological time scale. Advances in computational biology have now permitted in the case of a the muscle protein titin a simulation of unprecedented accuracy that resolved clearly the relationship between molecular structure of titin and its ability to sustain the forces that arise in muscle function. For this purpose a key element of titin was stretched at a velocity slow enough that hydrodynamic drag, that came about as an unwanted byproduct of earlier simulations, was negligible compared to titin's intrinsic force bearing properties. The new simulations opened an unveiled view on muscle elasticity. More information can be found on our titin website, and in a recent review .
Drude water

image size: 118.5KB
made with VMD
movie (YouTube)

Molecular modeling simulates the motion of cellular biomolecules at the atomic level. To make the simulations faithful, the physical forces acting between atoms need to be described accurately. Electric field effects between atoms, so-called atomic polarizabilities, are especially difficult to model well in a computationally cost effective way. There is an ongoing effort in the molecular modeling community to develop cost effective models that more faithfully represent the microscopic properties of biomolecules due to the ambient electric field effects. Recent development work has added support in the simulation program NAMD for one of these advanced modeling efforts. As reported, the new algorithms used in NAMD achieve good parallel computing performance, with a cost that is not more than twice that of the standard model, not accounting for atomic polarizabilities. The new model is demonstrated to reproduce many physical properties better than the standard model, including more accurate bulk density and surface tension at the interface between liquids and more accurate diffusive behavior of ions in a solution. More on our research webpage.
inserting transmembrane helix

image size: 397.6KB
made with VMD
movie: 9.1MB

Constructing and correctly placing new proteins is a complicated task for living cells. Starting with nothing more than a sequence of DNA, the cell has to translate the genetic code, stitch together the constituent amino acids, and then place the newly made protein where its function is needed. To accomplish these feats, the cell using tools such as the ribosome, a protein factory (see the Dec. 2009 highlight Managing the Protein Assembly Line), and the protein-conducting channel, a switching station within the membrane (see the Nov. 2008 highlight Patching a Leaky Channel). All instructions for making a nascent protein and localizing it, e.g., to the watery cytoplasm or the oily membrane, are carried within its DNA sequence, but how it is read and then executed has long been unclear. Now, two recent research advances picture these processes in astonishing detail. The first advance (reported here), from a collaboration between cryo-electron microscopists and computational biologists using MDFF, shows an atomic level structure that caught the ribosome in the act of inserting a protein into a membrane. The structure reveals the newly forming protein transiting from within the ribosome to the channel and then through an open gate in the protein-conducting channel into the membrane. The second advance (reported here), accomplished with the simulation program NAMD, explained how the ribosome and the protein-conducting channel manage to pay the energy price of inserting a new protein, one amino acid a time, into the membrane. For more information see our protein translocation website.
GPU Programming Workshop
A recent review explores how the Graphics Processing Units (GPUs) found in commodity high-end video cards are increasingly being used not only for interactive molecular graphics, but also for molecular simulation and analysis. NAMD and VMD both support GPU-acceleration using NVIDIA CUDA, enabling computationally demanding simulation, visualization and analysis tasks (e.g., of the electrostatics of Tamiflu binding) to be run with shorter turnaround on modestly priced GPU clusters, desktop, and laptop computers. This affordable computational power is particularly compelling for interactive modeling, with a recent report detailing how interactive molecular dynamics simulations with haptic feedback are now possible on GPU-accelerated desktop computers. Three recent book chapters detail the application of GPU computing techniques to the calculation of electrostatic potentials, interactive display of molecular orbitals, and more general molecular modeling algorithms. In August 2010 the Resource held its first workshop on GPU programming for molecular modeling to bring the benefits of GPU computing to a broader range of molecular modeling tools and magnify the impact of our GPU computing research.

2017 | 2016 | 2015 | 2014 | 2013 | 2012 | 2011 | 2010 | 2009 | 2008 | 2007 | 2006 | 2005 | 2004 | 2003 | 2002 | 2001