|
 |
|
 |
|
 |
|
 |
|
|
|
|
|
|
|
|
|
|
|||||||||
|
|
|||||||||||||||||
| ||||||||||||||||||||||
|
From Computational Biophysics to Systems Biology 2007 | ||||||||||||||||||||||
|
|
List of AbstractsInvited Talks
Contributed Talks
Posters
Invited TalksStretching to understand proteinsMechanical stretching of single proteins has been studied experimentally for about 50 proteins yielding a variety of force patterns and peak forces. Here, we perform a theoretical survey of 7749 proteins of known native structure and map out the landscape of possible dynamical behaviors under stretching at constant speed. The model used is constructed based on the native geometry. It is solved by methods of molecular dynamics and validated by comparing the theoretical predictions to experimental results. We characterize the distribution of peak forces and on correlations with the system size and with the structure classification as characterized by the CATH scheme. We identify proteins with the biggest forces and show that they belong to few topology classes. We determine which protein segments act as mechanical clamps and show that, in most cases, they correspond to long stretches of parallel beta-strands, but other mechanisms are also possible. We then consider stretching by fluid flows. We show that unfolding induced by a uniform flow shows a richer behavior than that in the force clamp. The dynamics of unfolding is found to depend strongly on the selection of the amino acid, usually one of the termini, which is anchored. These features offer potentially wider diagnostic tools to investigate structure of proteins compared to experiments based on the atomic force microscopy. Advances In De Novo Protein DesignThe primary objective in de novo protein design is to determine the
amino acid sequences which are compatible with existing or postulated
template backbone structures that may be rigid or flexible. The de
novo protein design problem is of fundamental importance since it
addresses the mapping of the space of amino acid sequences to known
protein folds or postulated/putative protein folds. It is also of
significant practical importance since it can lead to the improved
design of inhibitors, design of novel sequences with better stability,
design of catalytic sites of enzymes, and drug discovery.
Chromatin dynamics in silicioThe packing of the genomic DNA in the living cell is essential for its biological function. While
individual aspects of the genome architecture, such as DNA and nucleosome structure or the arrangement
of chromosome territories are well studied, much information is missing for a unified description of
cellular DNA at all its structural levels. Computer modeling can contribute to such a description.
Causality and Correlation Analyses of Molecular Dynamics Simulation DataComplex conformational transitions as well as proton or electron transfer processes in (bio)molecular
systems require rational interpretation, for example in terms of time-ordered events and causal relations
between them. Therefore in post-processing of classical or quantum-classical MD simulations data we are
interested in extracting correlations among various events, and if possible, in describing their mutual
influence and temporal ordering. Such strategy is quite general and refers not only to dynamical properties
of biomolecular systems but also to a variety of complex systems in natural sciences and in economy. One
should note that in 2003 Granger got the Nobel prize for his contribution to the causality analysis in
economy [1]. The method analyzes signals (time series) generated by complex systems affected by stochastic
fluctuations, assuming linear relations between a present variable and its values in preceding time-steps.
An extended, multi channel Granger-type analysis was also applied in causality analyses of EEG signals [2].
The similar strategy based on a Multi-Variate Autoregressive Model (MVAR) and the Directed Transfer
Function (DTF) was applied by us [3] in the analysis of MD simulation data [4,5]. Input signals are
trajectories or their combinations, and thermal interactions with the environment are responsible for
the stochastic perturbations. DTF is equivalent to the Granger analysis, but its computational
implementation is simpler. The MVAR-DTF method filters out the thermal noise and detects causal relations
among selected degrees of freedom treated as signals. It is capable to detect correlations between motions
of proton transfer processes (hoppings) and motions of hydrogen bond donors and hydrogen bond acceptors, or
to detect correlated conformational motions in molecular subunits. Since causality relations cannot be
easily derived from other conventional analyses therefore we think that our approach [3] defines the
promising strategy for future research.
Mesoscopic dynamics with the UNRES force field – a tool for studying the kinetics and thermodynamics of protein foldingAll atom simulations of protein folding starting from arbitrary structures are currently possible only for small proteins and, consequently, united-residue models of polypeptide chains are used in the field. The UNRES model and the respective force field developed in our laboratory belong to this class. In the UNRES model a polypeptide chain is represented as a sequence of alpha-carbon atoms with attached united side chains and united peptide groups located in the middle between the consecutive alpha-carbon atoms. The force field has been derived as a potential of mean force corresponding to given coarse-grained geometry but some of its terms were until recently still defined as statistical potentials. Recently we implemented mesoscopic molecular dynamics in the UNRES force field; compared to all-atom molecular dynamics with explicit solvent the simulation time is reduced by a factor of 4,000, which enables real-time in silico simulations of protein folding from scratch. Implementation of replica-exchange molecular dynamics and multiplexing replica-exchange molecular dynamics extended the application of the method to the thermodynamics of protein folding. In this talk new feature of the UNRES force field, such as the replacement of statistical local-interaction potentials with physics-based potentials and introduction of variable topology of disulfide bridges, as well as the extension of the treatment to study multi-chain proteins will be introduced. Applications of the method to thermodynamics-based protein-structure prediction, including its performance in the recent Community Wide Experiment on the Critical Assessment of Techniques for Protein Structure Prediction, CASP7, and derivation of kinetic equations of folding will be discussed. Mechanisms of Protein Assembly and Folding: Lessons from Minimalist ModelsKey words: protein folding funnel, protein recognition, minimal frustration
Mixed Quantum Mechanics/Classical Mechanics methods for the study of enzymatic reactionsI will present a general view of QM/MM methods and their use in biomolecular simulations. In particular, I will show the implementation of the SCC-DFTB (Self-consistent charge-Density Functional Tight Binding) method into the Amber 9 suite of programs, and some applications with this new method. We are focused in the study of the enzyme trans-sialidase of the parasitic T.cruzi. This protozoan is the causing agent of Human American Trypanosomiasis, or Chagas disease. Inhibition of its trans-sialidase has been proven to effectively prevent infection to humans. We will present a mixed QM/MM study of the mechanism of this enzyme and show how knowledge of the details can help on inhibitor design. Pushing details into the interactomeInteraction networks badly need details in order to be useful for
systems
biology. Many experiments suggest that pairs of proteins are
involved in physical
interactions, though few give any insights as to the details of how
they
are mediated. We have worked on inferrring details at various
levels from
interaction networks. I will discuss our attempts to: infer
details of
interaction strength from purification data (and use this to deduce
complexes, ref. 1), model interactions within complexes using
three-dimensional structures (2,3), and identify new modes of
domain/peptide recognition involved in mediating interactions (4).
Evolution of Protein SimulationOur efforts in protein simulation originated from our early theoretical
considerations of the effect of hydrogen bonds on protein reactivity (1),
and of the physical origin of hydrophobic interactions (2) and their effect
on protein reactivity. Applications of these ideas to experimental studies
of bovine pancreatic ribonuclease A led to the identification of three
tyrosyl...aspartate interactions (3) which serve as distance constraints on the
folding of the backbone chain, and thus, with the aid of an empirical
potential function, help to simulate the three-dimensional structure.
This motivated our initial development (4), and subsequent improvement,
of an all-atom force field and methods for its global optimization to
compute protein structure. Initial results from applications of this
approach led to structures of the cyclic decapeptide gramicidin S,
models of the fibrous protein collagen, and the three-helix bundle of
the 46-residue protein A (5), all of which were verified by experiment.
These developments, and the need to extend the all-atom model to a
united-residue (UNRES) one, to be able to apply simulation methods to
proteins larger than protein A, will be discussed, together with results
with UNRES in CASP blind tests.
Prediction of Protein Structure, Function and Druggability on a Proteomic ScaleA novel method for the prediction of protein structure, function and druggability based on the sequence-to-structure-to-function paradigm has been developed. We first show recent results that suggest that for compact single domain proteins, the PDB is most likely complete and that the completeness can be explained by the packing of compact, hydrogen bonded, secondary structural elements. We next present results from the application of our structure prediction algorithm, TASSER to all GPCRs in the human genome. Based on confidence criteria, 90% should have approximately correct structures, and clustering shows that structurally similar GPCRs have similar function even when their sequences are diverse. We then describe our multimeric structure prediction algorithm, m-TASSER, and its application to the prediction of protein-protein interactions. Next, a newly developed method for the accurate inference of protein biochemical function is presented and results of the comprehensive analysis of all sequenced genomes and the automated assignment of proteins to metabolic pathways are shown. Finally, we combine these approaches into a pathway-based method for the prediction of druggable protein targets and apply the resulting methodology to the human genome. The E-Cell Project and Challenges in Computational Systems BiologySummarizing our experience in launching and running the E-Cell Project in the past ten years, I will talk about some of major challenges we believe we will face in the next ten years of cell and systems biological simulation, including the following two; (1) Undeniably the last ten years of computational systems biology has been (re)discovery of the biggest bottleneck in biochemical modeling; the lack of high-throughput and reliable means of obtaining reaction rate coefficients. Computational aids in determination of reaction rate coefficients will be one area in which fruitful interactions between molecular biology, biophysical chemistry, and supercomputing is highly expected. (2) Macromolecular crowding is ubiquitous and found in all types of cellular organisms on the earth, which can, when coupled with localization and diffusion, alter biochemical dynamics, change equilibrium points, slow down and change the manner how big molecules diffuse, and amplify intrinsic noise. It is also a suspected physico-chemical factor behind the emergence of eukaryotic organisms. Development of formal treatment for crowded intracellular media and scalable computational means for it will be some of the most important tasks left for computational biologists. Measuring Energy Landscape Parameters of Biomolecules and their complexesRecent single molecule experiments on proteins, RNA, and protein-protein complexes provide unprecedented glimpse of their energy landscapes. I will describe theoretical and computational methods to extract the parameters governing their dynamics from single molecule force-unfolding and force-quench refolding. Applications of these ideas to proteins, RNA, and cell-adhesion complexes will be given. Bridging from molecular simulation to biochemical networksHow can we make the connection between the 3D structures of individual proteins and understanding how complex biological systems involving many proteins work? Here, I will discuss two approaches that we have been developing: (1) multiscale techniques using molecular and Brownian dynamics simulation to permit the spatial and temporal properties of large systems to be simulated using atomic detail structures; and (2) qPIPSA (quantitative Protein Interaction Property Similarity Analysis) to estimate kinetic parameters for mathematical modeling of biochemical pathways by using protein structural information. Contributed TalksResidual Entropy of Ice I from Multicanonical SimulationsWe introduce two simple models with nearest neighbor interactions on 3D hexagonal lattices. Each model allows one to calculate the residual entropy of ice I (ordinary ice) by means of multicanonical simulations. This gives the correction to the residual entropy derived by Linus Pauling in 1935. Our estimate is found to be within less than 0.1% of an analytical approximation by Nagle which improved on Pauling’s result. In biological applications at room temperature small, ice-like clusters are of importance, whose entropy could be computed by the same method. A molecular dynamics study of the basic fibroblast growth factor – fibroblast growth factor receptor complex.The growth of cells is a tightly regulated process. A normal cell stays in a
resting state for days, weeks, or years, until the balance of
growth-stimulatory and inhibitory signals from outside the cell initiates the
division process. This regulation occurs via protein interactions that
constitute the control system that drives and coordinates the cell cycle.
In multicellular organisms, this cycle is controlled by highly specific
proteins, namely the growth factors.
Photosensory proteins as vehicles to bridge computational biophysics and systems biology to initiate the field of synthetic biologyThe systems biology approach to understand basic questions related to health,
food and the environment can be applied while incorporating different levels
of molecular detail. Occasionally it is claimed that the approach chosen is
based on ’first principles’. Actually it will still take several
decades before such an approach may become feasible (that is if ’first
principles’ is assumed to have is regular meaning in chemistry).
Currently, at best, systems descriptions are provided that take enzyme-kinetic
characteristics (i.e. KM and Vmax) as the most detailed
elements of the models generated.
Comparing Semi-Empirical versus Classic Charge Assignments in BioMolecules and their Effect on Electrostatic PotentialsThe program LocalSCF is used to consider a subset of protein structures at the semi-empirical QM level of theory. Model Hamiltonians include AM1, MNDO, PM3 and PM5. All biomolecules are also studied with classic charge assignments using the AMBER force field. An important aspect of classic biomolecular simulation can thus be addressed, namely to what extent the usual concept of a single set of static atomic partial charges per type of amino acid will hold in general for the entire global protein structure. Semi-empirical charges will vary with different chemical neighbourhood inside the protein and the question remains how severely these alterations will influence global electrostatic properties of the protein. In order to probe this effect we use grid maps of electrostatic potentials obtained after solution of the Poisson-Boltzmann equation. Source charges to these solutions will either have been the classic AMBER ones or some set of semi-empirical charges from the list of models mentioned above. In comparing different models to each other we hope to recognize systematic trends as well as to identify a recommended way of doing proper charge assignments in proteins. Microcanonical Analyses of Peptide Aggregation ProcessesWe propose the use of microcanonical analyses for numerical studies of peptide aggregation
transitions. Performing multicanonical Monte Carlo simulations of a simple hydrophobic-polar
continuum model for interacting heteropolymers of finite length, we find that the
microcanonical entropy behaves convex in the transition region, leading to a negative
microcanonical specific heat. As this effect is also seen in first-order-like transitions of
other finite systems, our results provide clear evidence for recent hints that the
characterisation of phase separation in first-order-like transitions of finite systems profits
from this microcanonical view.
DNA packaging and electrostatic interactionsIn cell nucleus, long DNA (roughly 2 m in humans) is packed about 400,000 times by nuclear proteins forming compact but dynamic state called chromatin. DNA is highly negatively charged polyelectrolyte and therefore huge repulsive force between the DNA molecules must be overcome to form chromatin. The first level of DNA packaging is the nucleosome core particle (NCP), composed of 150 bp of DNA wrapped around a histone octamer core and flexible positively N-terminal domains (histone tails) protruding out from the NCP. Linear arrays of the NCP are further folded into higher level chromatin structures with the histone tails playing important roles in their formation. Theoretical models from Poisson Boltzmann (PB) approximation to all-atom molecular dynamics simulations are used to describe chromatin/nucleosome statics and dynamics. Applying the PB theory we show that free energy of NCP formation is extremely favorable that challenges established opinion about marginal stability of the NCP. To describe the influence of the histone tails on aggregation and dynamics of the NCPs, we carried out molecular dynamics simulations with coarse-grained approximation of the NCP. Our results are in good agreement with experimental neutron scattering and X-ray diffraction data. To reveal molecular details of the histone tail-DNA binding and dynamics, all-atom MD simulations were undertaken in a system comprising several DNA oligomers and fragments of the histone tails. Correlation between DNA-DNA distance and binding the histone tails to DNA is clearly observed. At the same time, binding of the tails does not restrict internal dynamics of the DNA. Anisotropic internucleosome interactions and geometrical constraints favour the two-start helical structure of chromatinThe structures of chromatin at the high density characteristic of the silent phase are strongly influenced by the packing of nucleosome core particles (NCPs), the anisotropic attractive interactions between two of them and constraints, such as the DNA bending, imposed to the wrapped and linker DNA segments. In this work, coarse--grained models of chromatin are studied. For a pair of NCPs, a simple single-site anisotropic potential energy function is designed on the basis of the experimental data reported for the ordered phases of NCPs. This potential energy function is employed in random-walks of chromatin models where the NCP DNA wrapping is modulated in length, while the linker segments are modulated in both length and curvature. These models support the two--start helical topology for chromatin in the absence of linker histones. The geometry of two-start helical configurations is characterized by poorly bent linkers and by a moderate reduction of wrapped DNA in the NCP. Mechanism of fibril formation of short peptidesThe mechanism of oligomerization of Alzheimer’s
Aβ16-22 peptides is studied by the
all-atom simulations with the GROMOS96 force field 43a1 in
explicit water. The time to get three peptides into an anti-parallel
arrangement was found to be
≈ 200 ns. This value was estimated directly from the kinetic
data as well as from the free energy landscape theory using the
"liquid crystal" order parameter P2 as a reaction coordinate.
The routes to the ordered state of the tetramer
(N=4), pentamer (N=5) and hexamer (N=6) were studied by adding
one peptide to the preformed anti-parallel conformations of (N-1) peptides.
Dihedral angle principal component analysis is employed to study the
conformations of peptides in detail. Our results suggest that the size of
the critical
nucleus of Aβ16-22 peptides is larger than 6. Importantly,
our study reveals that
the oligomer growth obeys the
two-stage "dock-lock" mechanism [1]. We have also developed a toy lattice
model [2] to further support this mechanism.
Multiple beta-sheet molecular dynamics of two Abl-SH3 domain peptidesTo simulate amyloid formation six ten-strand antiparallel beta-sheet stacks were constructed for 1) DLSFMKGE peptide (10x6xMK) , 2) DLSFKKGE peptide (10x6xKK). Both systems were covered with 10 A layer of explicit water and simulated by molecular dynamics (MD), Amber 8.0 force field, NTP, starting from 10K and gradually heating up the system by 10 degrees till 300 K, and then continuing MD simulation at 300 K. The role of metals in misfolding and aggregation processes X-ray spectroscopy and numerical simulationsAmyloidosis is a family of pathologies caused by the transition of endogenous
proteins and peptides from the physiological globular configuration to a pathological
fibrillar state (misfolding). The term describes a heterogeneous group of diseases
(more than 20), which are characterized by extra-cellular deposition of fibrillar
material [1]. Among them we focus here on the Alzheimer’s disease (AD), a progressive
and devastating neurodegenerative pathology affecting an important fraction of the aged
population in the developed world [2].
Channel Transport & Molecular Motors without Brownian RatchetsBrownian rotors are a large and important class of molecular motors in biological systems. Usually the function of these molecular machines is described as being a Brownian ratchet. However, Brownian ratchets need non-equilibrium fluctuations of an asymmetric potential for its function. We demonstrate that the Brownian ratchet paradigm does not apply to Brownian rotors. Rather, recent analytical results on facilitated transport through biological channels (W.R. Bauer & W. Nadler, PNAS (2006) 103, 11446) can be employed to understand also these purely concentration gradient-driven systems. The zinc-finger motif of T.thermophilus ribosomal protein S14 and the functionality of E.coli ribosomeThe structure of protein S14 (TthS14) of the 30S ribosomal subunit from
Thermus thermophilus contains a CXXC-X12-CXXC motif that coordinates a zinc
ion with the widely conserved cysteine 24 at its first position. The structural
and functional importance of cysteine 24, was studied by its replacement
with serine, and by incorporating the resulting mutant into Escherichia coli
ribosomes and by Molecular Dynamics Simulations (MDS).
All-atom protein folding and structure prediction in a transferable universal free-energy force-field.Exploiting Anfinsen’s thermodynamic hypothesis, all-atom free-energy
force-fields offer a promising alternative to kinetic molecular mechanics
simulations of protein folding and association. Here we report an accurate,
transferable all-atom biophysical force-field (PFF02)1 that stabilizes the
native conformation of a wide range of proteins as the global optimum of the
free-energy landscape. For 32 proteins of the ROSETTA decoy set and 6 proteins
that we have previously folded with PFF01 we find near native conformations
with an average backbone RMSD of 2.14 to the native conformation and a
average z-score of -3.46 to the corresponding decoy set2. Generating
continuous folding trajectories starting from completely extended
conformations we predicatively and reproducibly fold three non-homologous
hairpin-peptides, a three-stranded beta sheet, the all-helical 40 amino-acid
HIV accessory protein and a zinc-finger motif to near-native conformations.
In addition, we demonstrate all-atom folding of the 54 amino-acid engrailed
homeodomain and 56 amino-acid E domain of protein A in about 24 hours using a
massively parallel evolutionary algorithm3 on a distributed computational
architecture. These data demonstrate the viability and efficiency of the
free-energy approach for de-novo protein folding and offer perspectives for
rational force-field evolution and protein structure prediction.
Knots in Proteins, DNA and Polymers: Statistics, Function and EvolutionAlthough globular homopolymers display an abundance of knots (Virnau et al, J. Am. Chem. Soc. 127, 15102 (2005)), there are few knots in proteins and DNA. Can this absence of entanglement be explained in terms of statistical mechanics or is there an evolutionary bias? Do knots in proteins serve a purpose and how do they actually fold? After a brief discussion of knots in polymers and DNA, we will present an overview of knotted proteins in the current version of the Protein Data Bank (Virnau et al, PLOS Comp Biol 2, e122 (2006)). We will also discuss some particularly intriguing examples of this set and the evolutionary context in which knots appear. PostersConformational study of Amyloid beta (ABeta) peptide.Protein aggregation is a widespread phenomenon that arises from early folding intermediates through kinetic competition between proper folding and misfolding. Failure of the protein to fold correctly is associated with the malfunction of biological systems, leading to a broad range of diseases e.g. Spongiform encephalopathy disease, Huntington’s, Alzheimer’s disease (AD). ALZHEIMER’S DISEASE is the most common cause of senile dementia. Pathological hallmarks of the disease include senile plaques and neurofibrillary tangles. The amyloidal plaques found in AD are composed of a ~ 4.2 kD peptide called Aβ protein. The plaques result from the transformation of soluble Aβ protein monomer into insoluble aggregates. Determination of the molecular structure of Aβ fibrils by using molecular dynamic simulation softwares provide an insight into the precise arrangement of monomers that would allow targeting the critical steps in fibrilogenesis process. This background provides us with a path to proceed towards the designing of new targets for the development of aggregation modifiers that could potentially limit the toxicity of ABeta. Global persistence exponent of the helix-coil transition in polypeptidesIn this work we investigated the behavior of the global persistence exponent of the helix-coil transition in polypeptides. We define our order parameter as the number of helical residues qH max((2 <nH(T)> /(N-2) - 1), 0). Here we define a residue as helical if its backbone dihedral angles (φ,ψ) take values in the range (-70°±30°,-37°±30°) and the residue is hydrogen bonded. The normalization factor N-2 is chosen instead of N, the number of residues, because the flexible terminal residues are usually not part of an α-helix. Our definition ensures that 0 ≤ qH ≤ 1 and qh(Tc)0. The dynamical exponent θg that governs the behavior of the global persistence probability is obtained through the short-time Monte Carlo simulations of the helix-coil transition are based on a detailed, all-atom representation of the molecules and an implicit solvation model to approximate the interaction with the surrounding solvent. Our results obtained for the polyalanine and the 34-residue human parathyroid fragment PTH(1-34)are in good agreement each other, and indicate universality of the helix-coil transition in proteinlike molecules. Coarse-grained lattice model for molecular recognitionEquilibrium aspects of molecular recognition are investigated using coarse-grained models for the recognition process of two rigid biomolecules. To this end, a two-stage approach is adopted. First, the structure of the target molecule is fixed and learned by a probe molecule resulting in an ensemble of probe sequences. In a second step the recognition ability of the designed probe ensemble with respect to the chosen target sequence is tested by comparing the free energy of association with the previously fixed target structure and a different competing structure. Particular attention is paid to the influence of cooperative effects accompanying the association of the target biomolecule and the probe molecules. Cooperativity is found to enhance selectivity. In addition it is discussed how correlated hydrophobicity distributions affect the recognition ability. REMD simulations of Aβ16-22 peptides aggregation in aqueous solutionReplica Exchange Molecular Dynamics (REMD) simulation is an efficient way for equilibration and simulation of complex molecular systems at different temperatures. Experimental studies show that the short peptide fragements Aβ16-22 form fibrils, as it is known from the full length β-amyloid peptide. This fibril growth is strongly temperature dependent. We report here a study of the temperature dependence of the Aβ16-22 fibril nucleation and elongation in explicit water. We simulated ten Aβ16-22 peptides with 5900 SPC/E water in a 5.8 nm cubic box and used 76 replicas (with 20 ns simulation time per replica) distributed over a temperature range from 285.0 to 606.3 K. Electrostatic interactions play a critical role in initializing the Aβ16-22 aggregation and in stabilizing the aggregates at low temperatures. At higher temperatures the Aβ16-22 aggregation is mainly driven by the formation of hydrophobic environments that exclude water. Our data demonstrate that amphiphilicity is critical in determining the temperature dependent structural organization of β-sheets in the amyloid fibril. The twisted nature of the amyloid fibrils results from the stabilization of the protofilaments by hydrophobic interactions. Dimensionality Reduction Techniques for Protein Folding TrajectoriesIn our work we analyze large and highdimensional data from protein folding simulations. The main goals are to extract the underlying dimensionality, to find a small number of features that describe the data with high accuracy and to find interesting clusters in the data. Thus, we handle a problem of dimensionality reduction. Dimensionality reduction aims to find a mapping of the original space into a space of a few interesting dimensions, the user then can use for interpretation and analysis. We study modern dimensionality reduction techniques and combine them with promising distance measures, suitable to explain dissimilarities between the data points generated by ProFASi - a protein folding and aggregation simulator. Chemical space of auxin, its multi-phenomenology and plural protein interactionThe irregular net of physiological processes in plants has been the medium for false intuitive theories (hypothesis) of structure-activity relationships of auxins, whose chemical definition still being unclear. Molecular Quantum Similarity Measures (MQSM) scored a conceptual framework to uncover such phenomenon. The quantum objects (auxins) were classified by clusters methods based on a feed-back with a biological consensus variable. Next, standardized bioassays at a multidimensional scaling level with parallel screening of different auxins were carried out. Our approach, on the one side, found a new active compound (2, 6-dibromo-phenol) and shown hardness (η) as a common variable to predict the auxin´s biological activities. On the other side, it suggest two sets of molecular properties able to split different physiological functions of these compounds, as well as, fit both structural binding pockets (known until now: ABP1 and TIR1) of the proteins disclosed as putative auxin receptors. Crystal water molecules and solvation effects on Protein-ligand dockingIn docking simulations the crystal water molecules are usually neglected or are considered
as a rigid part of the protein. Recently, we integrated fully movable water molecules
into our docking program FlexScreen.
The water molecule has translational and rotational degrees of freedoms. If its presence
is very unfavorable at a protein-ligand conformation, it can also be pushed aside to
non-existence. Now, at the same time as the best protein-ligand conformations is searched
for, the water molecules are optimized to the each updated conformation.
Semiautomatic workflow for fold recognition - results from the CASP 2006 competitionIn order to test combinations of physics-based simulation techniques and sequence-based prediction methods our group participated in summer 2006 at the "Critical Assessment of Techniques for Protein Structure Prediction" (CASP) competition. As a first-time participant our goal was to establish a semi-automatic workflow by combining existing methods for fold recognition with our algorithms, and testing heuristics for the selection at each step. We give an overview of the workflow and the results of an in-depth statistical analysis of our results. In particular we assess the significance of measured performance differences between the prediction methods. Analyzing our workflow we try to find the critical points where alternative decisions lead to a significant change in the results. Our aim is to obtain rules that guide the decision process in the workflow to optimize our predictions. Influence of Normalization and Local Smoothing Procedures on Causality Time-series Analyses of Molecular Dynamics DataDetecting causality relationships between conformational changes in biomolecular systems
simulated with molecular dynamics (MD) methods is of crucial importance for describing their
mechanisms and understanding the logic of their functioning. An attempt to approach this
problem was presented in our recent study [1]. We followed the Granger causality methodology
[2] and applied a Multi-Variate Autoregressive Model (MVAR) with Directed Transfer Function,
which was used successfully in EEG time-series analyses [3]. However, the method still
requires some tuning, and in this presentation we deal with the two following problems. Interaction of Biological Matter with Nanomaterials: A First-Principles ApproachMotivation for this project stems from recent experimental and molecular dynamics investigations which suggest that DNA can assist in the separation process of carbon nanotubes according to the structural and electronic properties of the nanotubes. Sensor-related applications of DNA-CNT systems have also been emerging recently. As a specific case, the interaction of single-stranded DNA with different types of single-walled carbon nano-tubes is being investigated using Density Functional Theory (DFT). Computational Reconstruction of Macromolecular AssembliesCryo electron microscopy (cryo EM) has proven to be an effective technique to
gather low resolution information of macromolecular structures. X-ray crystallography
and NMR spectroscopy on the other hand are able to supply atomic detail structures
but tend to fail in the investigation of large specimen. We present a method for
elucidating macromolecular structures by combining atomic detail structures of
subunits with overall shape information gathered by cryo EM - a problem for which
only few computational methods have been published so far.
Verification of Protein-Protein Interactions by Use of Docking TechniquesFor the understanding of large macromolecular complexes such as ribosomes the analysis of
protein-protein interactions is essential. These intermolecular interactions are strongly
dependent on the three-dimensional structures of the corresponding molecules. In case that
the structures are known they can be directly used while in many other cases homology
modeling techniques can be applied. We have developed a novel algorithm (1) for this
purpose that allows the combination with additional experimental data (2) to further
improve the structural models. Currently we are developing tools based on a data driven
docking approach and the 3D structures of the individual molecules to investigate whether
proposed intermolecular interactions can be verified or falsified. In this contribution
we will show first results to demonstrate the principal applicability of our approach.
Boundary Element Method (BEM) with parametrically defined surfacesThis poster describes a new Boundary Element Method (BEM) implementation for biomolecular sovation with parametrically defines surfaces. First, multi-scale volumetric synthetic electron density maps are constructed from parsed atomic location data of biomolecules, using Gaussian isotropic kernels. An appropriate parameter selection is made for constructing an error bounded implicit solvation surface approximation to the Lee-Richards molecular surface. Next, four different methods are used to extract triangular meshes for the molecular surface. They are: marching cubes, marching tetrahedra, marching cubes with dual contouring and marching tetrahedra with dual contouring. Then generated meshes are used in BEM electrostatic calculation. In this work we study: 1) calculation time and accuracy for muiltipole and direct electrostatic solvers; 2) energy convergence and calculation time with the density of boundary points for 4 different meshing algorithms; 3) energy convergence for different iterative linear solvers. Parameterization of the Potential Energy Surface of the Double Proton Transfer in Porphyrin and PorpheceneDouble proton transfer processes (DPTPs) are commonly observed, from small systems to
large ones including proteins. For better understanding of DPTPs and coupled correlated
atomic motions, we plan to use quantum-classical molecular dynamics (QCMD) simulations.
The main problem of the future QCMD simulations is lack of reliable and fast generators of
the potential energy surfaces (PES) for the atomic motions. Based on our previous studies
for the formic acid dimer, we focused our research on larger and well experimentally studied
molecules: porphyrin and porphycene.
Our aim is to construct a fast and precise PES generator in an analytic form, to be used
in QCMD simulations for motions of the two inner hydrogens (protons) and for vibrations of
the molecular skeleton in the ground or a metastable excited Born-Oppenheimer states.
Such generator is being parametrized based on ab initio and/or DFT calculations with
stationary geometries (the minima and the saddle points of the PES), and using Hessians and
the PES scans for the proton motions in a
fixed-skeleton frame. The PES for the skeletons and for
the inner protons are parametrized in an analytic form consisting of multi-body interatomic
potentials. We will present novel analytic formulae of the PES for the double proton transfer,
along with resulting parametrized
energy profiles for the correlated motions of the two protons.
A statistical approach to deriving and analyzing a propensity scale for predicting exposed Transmembrane Beta-Barrel residues from protein sequenceIn the current study, we present an algorithm to analytically derive a novel propensity scale for the Transmembrane residues to be exposed to the lipid bilayer. Since it is very difficult to experimentally determine the 3D structures of membrane Beta-barrel proteins and given the fact that membrane Beta-barrel proteins perform several important functions in cell proteome of both gram-negative bacteria and eukaryotes, it is imperative to develop in silico methods for the modeling of their 3D structure. Our method takes into account the evolutionary conservation and frequency profile to derive a positional score for a given Transmembrane residue. The scale is derived such that the positional score for a given residue is maximally correlated with its relative solvent-accessible surface area (rSASA) value. A leave-one-out test with the known structures demonstrates the correlation coefficient between the observed and predicted rSASA values to be around 0.57. Analysis of the derived scale provides interesting insights into structural aspects of Beta-barrel residues. Along with experimental data on protein folding and protein biosynthesis, burial status of Transmembrane residues will later be used as constraints for ab-initio assembly of Transmembrane proteins. A web-service will also be shortly made available. Efficient Combination of Wang-Landau and Transition Matrix Monte Carlo MethodsAn efficient combination of the Wang-Landau and transition matrix Monte Carlo methods for protein and peptide simulations is described. At the initial stage of simulation the algorithm behaves like the Wang-Landau algorithm, allowing to sample the entire interval of energies, and at the later stages, it behaves like transition matrix Monte Carlo method and has significantly lower statistical errors. This combination allows to achieve fast convergence to the correct values of density of states. We propose that the violation of TTT identities may serve as a qualitative criterion to check the convergence of density of states. The simulation process can be parallelized by cutting the entire interval of simulation into subintervals. The violation of ergodicity in this case is discussed. We test the algorithm on a set of peptides of different lengths and observe good statistical convergent properties for the density of states. We believe that the method is of general nature and can be used for simulations of other systems with either discrete or continuous energy spectrum. Conformational Studies of UDP-GlcNAc in Environments of Increasing ComplexityThe "effective" dynamics of a biomolecular system can often be described by
means of a Markov chain describing "flipping dynamics" between metastable
(geometrical large scale) molecule conformations on "long" time scales, while,
on shorter time scales, the flexibility within conformations can be modelled
by stochastic differential equations. One Step Ahead: Role of Filopodia in Adhesion Formation During Epithelial Cell MigrationCell adhesion is one of the most essential prerequisites for cell function and movement. It depends strongly on focal adhesion complexes connecting the extracellular matrix to the actin cytoskeleton. Especially in moving cells focal adhesions are highly dynamic structures and believed to be formed closely behind the lamellipodia leading edge. We expand this model by the function of filopodia as characteristic feature of many motile cells. Filopodia were thought to act just as guiding cues to direct cell migration. Here, we show for keratinocytes that the place of adhesion site formation is located in filopodia and not in lamellipodia. Stable adhesion site formation depends strongly on the temporal persistence and spatial location of filopodia. These structures form small but fully assembled micro-focal adhesions upon movement. Such complexes contain all early adhesion site markers but also marker proteins for mature adhesions. Micro-focal adhesions when reached by the lamellipodium are just increased in size resulting in classical focal adhesion sites. Study of Protein Structural Descriptors: towards similarity and classificationWe have investigated the ability of structural descriptors in describing protein structural similarity and classification on 77 proteins extracted from SCOP. Using paired protein profiles we have trained a support vector classifier to map protein pairs to an appropriate level in the SCOP hierarchy. These profiles are composed of structural descriptors derived from the geometric properties of secondary structure elements. At the protein level, test sets were created from 511 protein pairs which were not used to train the classifier. The 11-fold cross-validation has correctly classified 97%, 37%, 62% and 82% of protein pairs belonging to the same class, fold, super family and family, respectively. These preliminary results show the applicability of the method towards the identification of novel protein folds and families. Continuing work towards classification of larger set of protein pairs and to improve classification to the same fold and super family is underway. Algorithmic Refinements to an Enhanced Poisson-Boltzmann Approach used in BioMolecular SimulationIn a series of recent publications we have introduced a very general description of solvation for biomolecular structure. It was shown that the enhanced continuum electrostatics approach could be decomposed into a series of individual terms, each of them representing its own portion of distinct physical interaction. Care has been taken to operate the model at conditions that guaranteed a maximun level of numerical accuracy. However, a number of internal parameters might still need further optimization in order to speed up the procedure. Among these factors are i) the exact value of the exit criterion used to terminate the calculation, ii) the array dimension regulating the allowed number of consecutive DIIS steps, iii) the switch criterion used to move from the pre-DIIS stage to the DIIS stage, iv) the dependence on system size of the number of necessary iterations to achieve convergence, v) the dependence on renormalization factors applied to the net sum of polarization charges, and vi) the influence of very small-sized boundary elements, or the introduced change when merging these very small-sized elements to larger ones from the neighbourhood. Therefore in this present study we want to address these points and examine their consequences on run-time performance with regard to a series of test proteins of increasing size which has been used previously. Aggregation of the beta-amyloid proteins: Monte Carlo optimization studyThe Alzheimer’s disease is associated with misfolding of the amyloid protein
[1]. In the functional native form, this small protein, of 42 amino acids, has
globular structure with two alpha-helices. However, under certain conditions,
it can form aggregates of beta sheets that, on a larger scale, are arranged as
long insoluble fibrils. These fibrils, being toxic, are found in the brain
extracellular space of patients with the Alzheimer’s disease. The amyloid
protein in this form is known as Abeta peptide. It has been an object of
several experimental and computational studies. The most particular interest
was concentrated on the fragment 16-22 responsible for the beta sheets
formation [2,3].
Analysis and optimization of the Flex-Screen docking approach using DUD benchmarking databaseScreening performance of the all-atom Flex-Screen docking approach including receptor flexibility is investigated by using of a directory of useful decoys, DUD (http://blaster.docking.org/dud). DUD is a bias-corrected benchmarking database based on 40 different target proteins, where each receptor is associated with a native ligand, a set of annotated ligands (2950 in total), and a set of decoy molecules (95 316 in total) that are physically similar but chemically distinct from ligands so that they are unlikely to be binders. The docking performance is evaluated using two criteria: 1) geometrical fidelity of the docked poses compared to those of the experimental structures; 2) enrichment of annotated ligands and decoys, which shows the ability of the docking calculations to distinct between true positives and nonbinder molecules with the same physical properties. Based on these results the scoring function and the receptor side-chain rearrangement procedure used in Flex-Screen are optimized. Protein folding and structure prediction of proteins containing disulfide bridgesIn this contribution, we present a computational study of helical and beta-sheet proteins containing up to three disulfide bridges. We used the force fields PFF01/02 to calculate the internal free energy of the protein and search for the global free-energy minimum using of a stochastic optimization methods, in particular the basin hopping method. Recently, it has been shown that this approach can correctly predict the tertiary structure of a number of proteins. In this work, a constraining Morse potential was adopted to describe the covalent disulfide bonds between cysteine residues. For all proteins studied, inclusion of the constraining potential resulted in improved RMSDb values compared to constraint-free simulations. We report folding simulations for 1KVG, 1PG1, 1WQE and 1HD6, which were folded correctly from extended conformations with inclusion of the constraining potential. Prediction accuracy of the results was assessed by comparing the simulated structures with experimental NMR structures. Water percolation governs polymorphic transition and conductivity of DNAWe report on the first computer simulation studies of the percolation
transition of water at the surface of the DNA double helix. At low hydrations,
only small water clusters are attached to the DNA surface, whereas, at high
hydrations, it is homogeneously covered by a spanning water network. The
spanning water network is formed via a percolation transition at an
intermediate hydration number of about 15 water molecules per nucleotide,
which is very close to the midpoint of polymorphic transitions between A-
and B-forms of the double helix. Formation of spanning water networks results
in sigmoid like acceleration of long-range ion transport in good agreement
with experiment.
Molecular Dynamics in Excited States: Landau – Zener Model of Nitric Oxide Geminate Recombination to Nitrile HydrataseMolecular dynamics simulations on two coupled electronic surfaces are employed to
investigate the geminate recombination of nitric oxide to photoactive
biotechnological enzyme nitrile hydratase (NHase). NHase enzymatic activity is
triggered by photodissociation of NO molecule. This molecule is bound to the
non-heme ferric catalytic centre in dark conditions and dissociates after light
absorption [1, 2]. The crossing between the ground and the excited state surfaces
is treated using the Landau – Zener model [3, 4]. The NO recombination curves
and recombination rates for fully protonated and double – deprotonated models
of the enzyme active site were calculated. Residues critical for
dissociation/recombination processes are pointed out. Our results strongly suggest
that the recombination of NO ligand to the NHase active center is a picosecond
time-scale process.
Exploring conformational space and dynamics of RNA hairpins by MD simulations: structure-function correlation of HIV-1 genome regulatory elementsElucidating high-resolution RNA structure of dynamically inter-converting
conformational substates poses significant challenges to NMR spectroscopy
and X-ray crystallography, which remain limited in applicability to
"static" average structure determination of well folded conformations. The
functionally active RNA conformations may not always be the most populated
in solution. Transiently populated conformational sub-states may be
captured during protein recognition and stabilized by binding divalent ions
or other biomolecular complex components.
Multiple stepwise refolding upon force quench, mechanical and thermal unfolding pathways of proteinsSingle molecule AFM experiments have generated mechanical folding trajectories
for polyproteins starting from initially stretched conformations. Refolding,
monitored by the end-to-end distance, occurs in distinct multiple stages [1].
Inspired by these experiments we have probed the folding dynamics of the
single domain I27 from the muscle protein titin and ubiquitin (Ub) using a
coarse-grained model. The refolding upon force quench starts from an ensemble
of extended structures. Folding pathways are primarily determined by the
ensemble of initial structures, from which folding commences. Folding
initiated with random coil and stretched conformations follows distinct
pathways determined by initial unfolded ensemble. The extended initial
structures increase folding and collapse time scales and decouple both kinetic
processes. Furthermore, the folding from stretched ensemble introduces
additional relaxation stages associated with rapid chain contraction.
Surprisingly, a quenched background force does not change the nature of
folding kinetics, but merely increases the height of free energy folding
barrier [2]. We show that the dependence of refolding times on quench force
follows the Bell formula. The locationtransition state along the reaction
coordinate, given by end-to-end distance, was determined. Our result is
consistent with the experiments on Ub [1]. Our simulations describe, at a
molecular level, the distinct mechanisms of relaxation and refolding of Ub
and titin domain after muscle stretching. The mechanical and thermal unfolding
pathways of Ub have been studied and compared with experiments in detail [3].
Parallelization of ECEPP/3 in SMMPThe power consumption of modern processors makes it difficult to increase
their clock speed further. Even in the PC market CPU manufacturer now include
multiple compute cores on a single chip to improve performance and keep up
with Moore’s law, a trend likely to continue. This is even more important in
high performance computing, where cooling and electricity bills are becoming
a large issue. The compactness and low power consumption of an IBM BlueGene,
e.g, is only possible because of CPUs with moderate clock rates.
Folding and aggregation of proteins with Monte Carlo simulationsWe present Monte Carlo studies of several small proteins of up to 23 residues, using an implicit water all-atom model. The interaction potential does not make use of any information about the known experimental structures of the molecules. The model describes the folding behaviour of all the molecules displayed on this poster for exactly the same set of parameters of the potential. It is also interesting that for the molecules of this size, which the model successfully folds, the calculated thermodynamic behaviour is very often in agreement with experiments. Examination of Protein Dynamics at Focal Adhesion Sites by FRAP-AnalysisCell adhesion is an essential process for tissue integrity and cell movement.
The adhesion process itself depends on clustered protein complexes called
focal adhesion sites. These adhesion sites form a connection between the
extracellular matrix which is surrounding the cell and the cell actin
cytoskeleton. Focal adhesions are characterized by a specific set of proteins
as integrins spanning the plasma membrane, regulatory kinases or proteins
like vinculin, zyxin or VASP bridging the integrins to actin fibers. In
addition, focal adhesion sites can adapt in size and shape to cellular
growth conditions. Thus, formation and release of focal adhesion sites is
highly dynamic in moving cells but barely detectable when a cell is
stationary. It is highly unknown, if various proteins additionally exchange
in stable adhesion sites, and if such putative protein exchange dynamics
goes along with the variable formation dynamics of whole adhesion sites.
Aggregate size and shape distributions in amyloid-β peptide solutionsA peptide with 42 amino acid residues (Aβ(1--42)) plays a key role in
the pathogenesis of the Alzheimer’s disease. It is highly prone to self
aggregation leading to the formation of fibrils which are deposited in
so-called amyloid plaques in the brain of affected individuals (1,2).
PSO@Autodock : A novel bio-algorithm based fast flexible docking tool for virtual screeningVirtual screening techniques applying computational molecular docking methods have proven to
be a viable alternative to experimental High Throughput Screening. Given the 3-D structure
of the protein target, molecular docking methods allow the screening of large libraries of
compounds and provide detailed information on the protein-ligand interactions. Thus, molecular
docking methods contribute significantly to the understanding of activity of candidate
molecules at the molecular level.
Properties of Thick PolymersWe explore the applicability of simple polymer models for the modelling of certain properties of biologic matter. In particulular we calculate the ground state configurations in precise Monte Carlo simulations and classify the shapes of thick polymers. We discuss, whether there exists some kind of geometrical universality in the shape of ground state configurations. The universal shapes, which are discovered are circles, twisted circles, helices, hypercubic face centered crystals, and with the inclusion of dipole-dipole interactions, sheets which are 2-dimensional in nature. We also present first results from a simulation of thick heteropolymers. A Network-based approach to Biomolecular DynamicsThe dynamics of biomolecular processes, including protein folding,
binding and polymerization, is dynamically metastable: The system
remains for long times within a given metastable state ("conformation"),
before rapidly switching to another. Recent computational methods allow
the metastable states to be computed from molecular dynamics simulations
and the system’s kinetic and thermodynamic properties to be obtained from
the resulting network of metastable states.
Steered Molecular Dynamics as a Virtual Atomic Force MicroscopeComputer modeling based on simple laws of classical mechanics are successfully
used in interpretation of single molecule atomic force microscopy experiments
[1-4]. In this presentation new examples of applications of the steered molecular
dynamics method (SMD) to stretching single proteins will be given. Mechanical
properties of oncogenic protein gankyrin were studied on 100 ns time-scale using
CHARMM27 force field and NAMD code. After initial stretching of the whole polymer
by en external force a sequential unfolding of gankyrin units was observed. SMD was
also used to estimate forces during forced 2 ns dissociation of ligands in an
enzyme-ligand model system. Photoactive enzyme nitrile hydratase and nitrile and
amide ligands were investigated [5, 6]. Some biological aspects of protein
architecture may be discovered in such physical computer experiments, however,
a critical problem of fast time scale prevents quantitative prediction of molecular
forces. Possible strategies of avoiding this bottleneck will be discussed.
Optimization of the UNRES force field for prediction of protein thermodynamic stabilityDuring the last several years, we have been developing the coarse-grained UNRES force field in which each amino-acid residue is represented as two interaction sites: the united side chain and the united peptide group. The respective energy function has been defined as a restricted free energy (RFE) function of a polypeptide chain plus the surrounding solvent, which was later broken down into factors, each corresponding to interactions between a number of coarse-grained sites. We have developed a mesoscopic dynamics treatment to simulate protein folding with UNRES and enhanced it by introducing a replica-exchange algorithm to search the conformational space more efficiently. We also developed a method for the optimization of UNRES force field which takes into account the folding temperature. Results of optimization of the UNRES force field with 1EM7 as a training protein and performance of the optimized force field to reproduce the difference in the folding temperatures and thermodynamic stability of its mutants will be presented. Simulation of linker histone in chromatin fiberThe dynamics and interaction between protein-protein and protein-DNA molecules appearing on different time and length scales in the cell are of fundamental interest. In chromatin, linker histone binds to the highly charged nucleosome particle driven by electrostatic attraction. However, its position and orientation with respect to the nucleosome core as well as its binding sites are unresolved problems. Using implicit representation of the surrounding molecules, rigid-body docking of linker histone to the nucleosome particle (SDA software) has been performed revealing how the protein binds to DNA on the nucleosome. Simulation of diffusional motion of both molecules for computation of the association rates and encounter complex formation is under study with SDA. The results will be integrated in higher-order simulation of chromatin fiber. The Locally Enhanced Sampling Study of Large Ligands Diffusion inside Enzyme. Acrylonitrile and Acrylamide Journey in Nitrile HydrataseNitrile Hydratese (NHase) is a metaloenzyme with non-standard active site
containing noncorin Co(+3). In industry it is used for a large scale conversion
of toxic nitriles into useful amides. In this research NHase from Pseudonocardia
Thermophila JCM 3095 (1IRE) is investigated. In order to understand its excellent
catalytic activity the possible transport routes of substrates and products have
to be determined [1]. Transient states are not easily elucidated using experimental
techniques, so computer modeling of molecular dynamics (MD) helps a lot. The main
goals of finding cavities inside the protein and entry/exit pathways for a
substrate (acrylonitrile) and the product (acrylamide) have been achieved. In our
opinion a very convenient tool for that type of study is MD with the Locally
Enhanced Sampling (LES) Hamiltonian [2]. The LES method, by multiplying
non-interacting ligand copies, allows for better probing of the conformational
space than the standard MD method, despite known problems with the energy
equipartition [3, 4].
Receptor specific forcefield: improving classical forcefields with quantum mechanical calculationsWe demonstrate how the inclusion of QM-calculations of receptor-ligand complexes with the Fragment Molecular Orbital Method (FMO) can be used to improve a classical forcefield. In comparing this QM-forcefield for protein and ligand with a standard ab-initio forcefield (ESFF) we can demonstrate a performance gain (an increased correlation to an experimentally measured or QM calculated binding energy). Additional we investigate the accuracy and reliability of our docking tool FlexScreen in predicting the experimental binding mode for 83 receptor-ligand complexes from the high-resolution subset of the ASTEX/CCDC data set. De novo Folding of Two-Helix Potassium Channel Blockers with Free-Energy Models and Molecular DynamicsWe report the predictive de novo folding of three two-helix proteins using the free-energy protein forcefield PFF01. Starting from random initial conformations 40-90% of the members of the simulated ensembles converge to near-native conformations. The free energy landscapes of these proteins are very simple, suggesting them as candidates for all-atom molecular dynamics simulations. In several independent MD simulations we find the formation of the correct secondary structure and several folding events into the tertiary structure. Kinetic clustering analysis of PINWW unfolding trajectoriesPCCA (Perron-Cluster-Cluster-Analysis) is a tool to investigate the kinetics of a complex system in simulation. We have applied PCCA to unfolding trajectories of the PINWW-domain to find states on the ns-timescale. DNATagger, colors for codonsMultiple sequence alignment is an important tool to study the evolution of related proteins. Coloring the
characters representing nucleotide or amino acid residues make their visualization more comfortable and
understandable for human beings.
Brownian dynamics simulations of protein cluster dynamicsMost proteins in the cell are active in complexes with two to several hundreds of components. Because only very small assemblies can be studied in an all-atom framework, coarse-grained approaches are required to model the association and dissociation dynamics of larger protein assemblies. We model proteins as spherical particles with specific binding sites. Their motion is simulated with Brownian dynamics. The diffusion of the non-spherical clusters is treated using bead models for hydrodynamics in the viscous regime. Using computer simulations, we measure first passage times and related quantities for the cluster assembly. For simple cases, our results can be compared to exact first passage time calculations. Aggregation of a fragment of the islet amyloid polypeptide treated as phase transition: a cluster analysisAmyloid polypeptides are weakly soluble in water. Even at low concentrations, their
aqueous solutions undergo a phase separation into two phases with an organic-rich phase
being an ordered solid (fibrillar) phase. Theoretical studies of fibril formation can be
performed by simulations of peptide-water mixtures with peptide concentrations deeply
inside the two-phase region. In such states, various system properties, including
clustering (aggregation) of molecules, is extremely sensitive to the system size [1]. We
performed MD simulations of 12 amyloidogenic fragments of IAPP (residues 15-19) in
liquid water, starting from different random configurations. Analysis of peptide clustering
and aggregation evidences features typical for a phase separation. The formation of a
stable aggregate is hampered by the small system size, however. We propose to use a
cluster analysis to select the configurations relevant for larger systems and, therefore, for
comparison with experimental data.
Selection of the optimal Go modelAll-atom simulations are often found impractical to use when studying large conformational changes along multiple trajectories of big molecules, their complexes, and larger bio-structures. A coarse grained description offers a reasonable alternative in this case. The Go-like models are the simplest implementation of such an approach. These models are defined phenomenologically by the requirement that the ground state of the model coincides with the experimentally determined native structure of a protein or other biomolecule. However, there are many possible choices of a Hamiltonian that are consistent with the requirement. It is then relevant to ask which variant is the most adequate physically. Here, we propose to use experimental data on protein stretching as providing a selection criterion. Specifically, we consider more than 20 different versions of the Go model, determine characteristic scales of a force of resistance to stretching in a set of 22 proteins, and select an optimal model that matches the experimental findings the best. Most of the models considered have Lennard-Jones contact potentials between the C-alpha atoms but there are choices related to the definition of the contact map and amplitudes of the contact potentials. We considered uniform and nonuniform amplitudes. In the latter case, the strength of the nonunoformity was governed either by specificity or by properties related to the geometry of the side groups. The optimal choice turned out to be the simplest variant: uniform couplings and no attractive i,i+2 contacts. A Load Balanced Force-Domain Decomposition Algorithm for Parallel Molecular Dynamics SimulationsA force-domain decomposition algorithm is presented, which is based on a work-weighted partitioning of the interaction matrix onto processors. Asynchronous communications efficiently hide communication between processors. Communication is reduced by i) exchanging only non-redundant information between processors and ii) sorting particles according to a space-filling Hilbert curve, which guarantees localization between interaction partners in physical space as well as in memory. For short range interactions, loadbalancing is achieved i) by distributing equal number of interactions on each processor or ii) by balancing the time spent in the force routine. Results show good parallel scaling behavior and it is shown that the proposed algorithm works especially well for non-homogenous systems. Hydrodynamic Interactions and Protein UnfoldingIt has been widely recognized that the water environment affects the energy landscape and functionality of biomolecules in a profound way. There is, however, another solvent-related effect that is considered less frequently: the hydrodynamic interactions between individual segments of a biomolecule that moves. These interactions may affect the dynamics of conformational changes because motion of one segment of a molecule generates a local fluid flow which influences another segment. In this study, we consider mechanical stretching of proteins and study the relevance of hydrodynamic interactions to the process. The stretching can be accomplished in several ways and we discuss three of them: at a constant speed, at a constant force, and through a fluid flow. A coarse-grained, Go-type model of a protein is used, constructed based on the knowledge of its native state. The results of molecular dynamic simulation show that hydrodynamic interactions have significant impact on both protein unfolding times and unfolding forces. Steered Classical and Quantum Path-Integral Molecular Dynamics Simulations of Strongly Coupled Protons Motions in PorphyceneMD simulations of many body quantum-classical systems is of big importance for description
of functioning of many (bio)molecular and nanosystems. Porphycene containing two strongly
coupled quantum protons in a classical molecular cave is a model system which is used by us
to develop novel methodological QCMD approaches. The system is also of practical importance
for molecular nanotechnologies. Energy profiles for the proton motions in porhycene have
already been studied by us using a SCC-DFTB approach [1], however, classical MD simulations
were not sufficiently precise to determine effective barriers for the proton transfer. Sine
the quantum dynamical motions of the protons have considerable impact on the motions of the
classical atoms and vice versa, in the current approach the whole system is simulated using
a steered Path Integral Molecular Dynamics (PIMD) method with an ”on the fly” Car-Parrinello DFT propagation scheme for all nuclei, and with the adiabatic QD of the
electrons [2].
Sidechain Ordering in HomopolymersIn order to study the relation between backbone and side chain ordering in proteins, we have performed multicanonical simulations of deka-peptide chains with various side groups. Glu10, Gln10, Asp10, Asn10, and Lys10 were selected to cover a wide variety of possible interactions between the side chains of the monomers. All homopolymers undergo helix-coil transitions. We found that peptides with long side chains that are capable of hydrogen bonding, i.e. Glu10, and Gln10, exhibit a second transition at lower temperatures connected with side chain ordering. This occurs in gas phase as well as in solvent, although the character of the side chain structure is different in each case. However, in polymers with short side chains capable of hydrogen bonding, i.e. Asp10 and Asn10, side chain ordering takes place over a wide temperature range and exhibits no phase transition like character. Moreover, non-backbone hydrogen bonds show enhanced formation and fluctuations already at the helix-coil transition temperature, indicating competition between side chain and backbone hydrogen bond formation. Again, these results are qualitatively independent of the environment. Side chain ordering in Lys10, whose side groups are long and polar, also takes place over a wide temperature range and exhibits no phase transition like character in both environments. Dihedral pattern in coil regions of protein structuresMost secondary structure prediction programs predict the membership of a
protein residue in one of three classes: (alpha)-helix, extended (beta sheet)
and (random)-coil. While the term "helix" roughly describes a repetitive
pattern of dihedral angles and the same is true for the extended parts of
beta sheets, the term "coil" does not imply any geometrical structure, but
merely the absence of repetitive dihedral angle pattern.
|
||||||||||||||||||||
|
|
|
|
|
|
|
|
|
|
 |
Forschungszentrum Jülich D-52425 Jülich |
 |
John von Neumann Institute for Computing (NIC) |
|
 Life Science
|
|
Imprint |
|
25. May 2012 
Martina Kamps
|
