Bojan Zagrovic

Simulation of Folding of a Small Alpha-helical Protein in Atomistic Detail using Worldwide- distributed Computing

By employing thousands of PCs and new worldwide-distributed computing techniques, we have simulated in atomistic detail the folding of a fast-folding 36-residue alpha-helical protein from the villin headpiece. The total simulated time exceeds 300 micros, orders of magnitude more than previous simulations of a molecule of this size. Starting from an extended state, we obtained an ensemble of folded structures, which is on average 1.7A and 1.9A away from the native state in C(alpha) distance-based root-mean-square deviation (dRMS) and C(beta) dRMS sense, respectively. The folding mechanism of villin is most consistent with the hydrophobic collapse view of folding: the molecule collapses non-specifically very quickly ( approximately 20ns), which greatly reduces the size of the conformational space that needs to be explored in search of the native state. The conformational search in the collapsed state appears to be rate-limited by the formation of the aromatic core: in a significant fraction of our simulations, the C-terminal phenylalanine residue packs improperly with the rest of the hydrophobic core. We suggest that the breaking of this interaction may be the rate-determining step in the course of folding. On the basis of our simulations we estimate the folding rate of villin to be approximately 5micros. By analyzing the average features of the folded ensemble obtained by simulation, we see that the mean folded structure is more similar to the native fold than any individual folded structure. This finding highlights the need for simulating ensembles of molecules and averaging the results in an experiment-like fashion if meaningful comparison between simulation and experiment is to be attempted. Moreover, our results demonstrate that (1) the computational methodology exists to simulate the multi-microsecond regime using distributed computing and (2) that potential sets used to describe interatomic interactions may be sufficiently accurate to reach the folded state, at least for small proteins. We conclude with a comparison between our results and current protein-folding theory.

Native-like Mean Structure in the Unfolded Ensemble of Small Proteins

The nature of the unfolded state plays a great role in our understanding of proteins. However, accurately studying the unfolded state with computer simulation is difficult, due to its complexity and the great deal of sampling required. Using a supercluster of over 10,000 processors we have performed close to 800 micros of molecular dynamics simulation in atomistic detail of the folded and unfolded states of three polypeptides from a range of structural classes: the all-alpha villin headpiece molecule, the beta hairpin tryptophan zipper, and a designed alpha-beta zinc finger mimic. A comparison between the folded and the unfolded ensembles reveals that, even though virtually none of the individual members of the unfolded ensemble exhibits native-like features, the mean unfolded structure (averaged over the entire unfolded ensemble) has a native-like geometry. This suggests several novel implications for protein folding and structure prediction as well as new interpretations for experiments which find structure in ensemble-averaged measurements.

Conformational selection and induced fit mechanism underlie specificity in noncovalent interactions with ubiquitin

Noncovalent binding interactions between proteins are the central physicochemical phenomenon underlying biological signaling and functional control on the molecular level. Here, we perform an extensive structural analysis of a large set of bound and unbound ubiquitin conformers and study the level of residual induced fit after conformational selection in the binding process. We show that the region surrounding the binding site in ubiquitin undergoes conformational changes that are significantly more pronounced compared with the whole molecule on average. We demonstrate that these induced-fit structural adjustments are comparable in magnitude to conformational selection. Our final model of ubiquitin binding blends conformational selection with the subsequent induced fit and provides a quantitative measure of their respective contributions.

Solvent viscosity dependence of the folding rate of a small protein: Distributed computing study

By using distributed computing techniques and a supercluster of more than 20,000 processors we simulated folding of a 20‐residue Trp Cage miniprotein in atomistic detail with implicit GB/SA solvent at a variety of solvent viscosities (γ). This allowed us to analyze the dependence of folding rates on viscosity. In particular, we focused on the low‐viscosity regime (values below the viscosity of water). In accordance with Kramers theory, we observe approximately linear dependence of the folding rate on 1/γ for values from 1–10−1× that of water viscosity. However, for the regime between 10−4–10−1× that of water viscosity we observe power‐law dependence of the form k ∼ γ−1/5. These results suggest that estimating folding rates from molecular simulations run at low viscosity under the assumption of linear dependence of rate on inverse viscosity may lead to erroneous results.

Comparing Atomistic Simulation Data With the NMR Experiment: How Much Can NOEs Actually Tell Us?

Simulated molecular dynamics trajectories of proteins and nucleic acids are often compared with nuclear magnetic resonance (NMR) data for the purposes of assessing the quality of the force field used or, equally important, trying to interpret ambiguous experimental data. In particular, nuclear Overhauser enhancement (NOE) intensities or atom–atom distances derived from them are frequently calculated from the simulated ensembles because the distance restraints derived from NOEs are the key ingredient in NMR‐based protein structure determination. In this study, we ask how diverse and nonnative‐like an ensemble of structures can be and still match the experimental NOE distance upper bounds well. We present two examples in which simulated ensembles of highly nonnative polypeptide structures (an unfolded state ensemble of the villin headpiece and a high‐temperature denatured ensemble of lysozyme) are shown to match fairly well the experimental NOE distance upper bounds from which the corresponding native structures were derived. For example, the unfolded ensemble of villin headpiece, which is on average 0.90 ± 0.13 nm root‐mean‐square deviation away from the native NMR structure, deviates from the experimental restraints by only 0.027 nm on average. However, this artificially good agreement is largely a consequence of 1) the highly nonlinear effects of r−6 (or r−3) averaging and 2) focusing only on the experimentally observed set of NOE bounds. Namely, in addition to the experimentally observed NOEs, both simulated ensembles (especially the villin ensemble) also predict a large number of NOEs, which are not seen in the experiment. If these are taken into account, the agreement between simulation and experiment gets markedly worse, as it should, given the nonnative nature of the underlying simulated ensembles. In light of the examples given, we conclude that comparing experimental NOE distance restraints with large simulated ensembles provides just by itself only limited information about the quality of simulation. Proteins 2006.

Determination of Ensemble-Average Pairwise Root Mean-Square Deviation from Experimental B-Factors

Root mean-square deviation (RMSD) after roto-translational least-squares fitting is a measure of global structural similarity of macromolecules used commonly. On the other hand, experimental x-ray B-factors are used frequently to study local structural heterogeneity and dynamics in macromolecules by providing direct information about root mean-square fluctuations (RMSF) that can also be calculated from molecular dynamics simulations. We provide a mathematical derivation showing that, given a set of conservative assumptions, a root mean-square ensemble-average of an all-against-all distribution of pairwise RMSD for a single molecular species, (1/2), is directly related to average B-factors () and (1/2). We show this relationship and explore its limits of validity on a heterogeneous ensemble of structures taken from molecular dynamics simulations of villin headpiece generated using distributed-computing techniques and the Folding@Home cluster. Our results provide a basis for quantifying global structural diversity of macromolecules in crystals directly from x-ray experiments, and we show this on a large set of structures taken from the Protein Data Bank. In particular, we show that the ensemble-average pairwise backbone RMSD for a microscopic ensemble underlying a typical protein x-ray structure is approximately 1.1 A, under the assumption that the principal contribution to experimental B-factors is conformational variability.

Structural correspondence between the α-helix and the random-flight chain resolves how unfolded proteins can have native-like properties

Recently, we have proposed that, on average, the structure of the unfolded state of small, mostly α-helical proteins may be similar to the native structure (the mean-structure hypothesis). After examining thousands of simulations of both the folded and the unfolded states of five polypeptides in atomistic detail at room temperature, we report here a result that seems at odds with the mean-structure hypothesis. Specifically, the average inter-residue distances in the collapsed unfolded structures agree well with the statistics of the ideal random-flight chain with link length of 3.8 Å (the length of one amino acid). A possible resolution of this apparent contradiction is offered by the observation that the inter-residue distances in a typical α-helix over short stretches are close to the average distances in an ideal random-flight chain.

Conformational averaging in structural biology: issues, challenges and computational solutions

Most experimental methods in structural biology provide time- and ensemble-averaged signals and, consequently, molecular structures based on such signals often exhibit only idealized, average features. Second, most experimental signals are only indirectly related to real, molecular geometries, and solving a structure typically involves a complicated procedure, which may not always result in a unique solution. To what extent do such conformationally-averaged, non-linear experimental signals and structural models derived from them accurately represent the underlying microscopic reality? Are there some structural motifs that are actually artificially more likely to be seen in an experiment simply due to the averaging artifact? Finally, what are the practical consequences of ignoring the averaging effects when it comes to functional and mechanistic implications that we try to glean from experimentally-based structural models? In this review, we critically address the work that has been aimed at studying such questions. We summarize the details of experimental methods typically used in structural biology (most notably nuclear magnetic resonance, X-ray crystallography and different types of spectroscopy), discuss their individual susceptibility to conformational (motional) averaging, and review several theoretical approaches, most importantly molecular dynamics simulations that are increasingly being used to aid experimentalists in interpreting structural biology experiments.

Folding probabilities: A novel approach to folding transitions and the two-dimensional Ising-model

The theoretical concept of folding probability, p(fold), has proven to be a useful means to characterize the kinetics of protein folding. Here, we illustrate the practical importance of p(fold) and demonstrate how it can be determined theoretically. We derive a general analytical expression for p(fold) and show how it can be estimated from simulations for systems where the transition rates between the relevant microstates are not known. By analyzing the Ising model we are able to determine the scaling behavior of the numerical error in the p(fold) estimate as function of the number of analyzed Monte Carlo runs. We apply our method to a simple, newly developed protein folding model for the formation of alpha helices. It is demonstrated that our technique highly parallelizes the calculation of p(fold) and that it is orders of magnitude more efficient than conventional approaches.

Structure and dynamics of two β-peptides in solution from molecular dynamics simulations validated against experiment

We have studied two different β-peptides in methanol using explicit solvent molecular dynamics simulations and the GROMOS 53A6 force field: a heptapeptide (peptide 1) expected to form a left-handed 314-helix, and a hexapeptide (peptide 2) expected to form a β-hairpin in solution. Our analysis has focused on identifying and analyzing the stability of the dominant secondary structure conformations adopted by the peptides, as well as on comparing the experimental NOE distance upper bounds and 3J-coupling values with their counterparts calculated on the basis of the simulated ensembles. Moreover, we have critically compared the present results with the analogous results obtained with the GROMOS 45A3 (peptide 1) and 43A1 (peptide 2) force fields. We conclude that within the limits of conformational sampling employed here, the GROMOS 53A6 force field satisfactorily reproduces experimental findings regarding the behavior of short β-peptides, with accuracy that is comparable to but not exceeding that of the previous versions of the force field.GCE legendConformational clustering analysis of the simulated ensemble of a ß-hexapeptide with two different simulation setups (a and b). The central members of all of the clusters populating more than 5% of all of the structures are shown, together with the most dominant hydrogen bonds and the corresponding percentages of cluster members containing them