Using motion planning to map protein folding landscapes and analyze folding kinetics of known native structures.

We investigate a novel approach for studying the kinetics of protein folding. Our framework has evolved from robotics motion planning techniques called probabilistic roadmap methods (PRMs) that have been applied in many diverse fields with great success. In our previous work, we presented our PRM-based technique and obtained encouraging results studying protein folding pathways for several small proteins. In this paper, we describe how our motion planning framework can be used to study protein folding kinetics. In particular, we present a refined version of our PRM-based framework and describe how it can be used to produce potential energy landscapes, free energy landscapes, and many folding pathways all from a single roadmap which is computed in a few hours on a desktop PC. Results are presented for 14 proteins. Our ability to produce large sets of unrelated folding pathways may potentially provide crucial insight into some aspects of folding kinetics, such as proteins that exhibit both two-state and three-state kinetics that are not captured by other theoretical techniques.     

Using motion planning to study protein folding pathways.

We present a framework for studying protein folding pathways and potential landscapes which is based on techniques recently developed in the robotics motion planning community. Our focus in this work is to study the protein folding mechanism assuming we know the native fold. That is, instead of performing fold prediction, we aim to study issues related to the folding process, such as the formation of secondary and tertiary structure, and the dependence of the folding pathway on the initial denatured conformation. Our work uses probabilistic roadmap (PRM) motion planning techniques which have proven successful for problems involving high-dimensional configuration spaces. A strength of these methods is their efficiency in rapidly covering the planning space without becoming trapped in local minima. We have applied our PRM technique to several small proteins (~60 residues) and validated the pathways computed by comparing the secondary structure formation order on our paths to known hydrogen exchange experimental results. An advantage of the PRM framework over other simulation methods is that it enables one to easily and efficiently compute folding pathways from any denatured starting state to the (known) native fold. This aspect makes our approach ideal for studying global properties of the protein's potential landscape, most of which are difficult to simulate and study with other methods. For example, in the proteins we study, the folding pathways starting from different denatured states sometimes share common portions when they are close to the native fold, and moreover, the formation order of the secondary structure appears largely independent of the starting denatured conformation. Another feature of our technique is that the distribution of the sampled conformations is correlated with the formation of secondary structure and, in particular, appears to differentiate situations in which secondary structure clearly forms first and those in which the tertiary structure is obtained more directly. Overall, our results applying PRM techniques are very encouraging and indicate the promise of our approach for studying proteins for which experimental results are not available.     

Using motion planning to study RNA folding kinetics.

We propose a novel, motion planning based approach to approximately map the energy landscape of an RNA molecule. A key feature of our method is that it provides a sparse map that captures the main features of the energy landscape which can be analyzed to compute folding kinetics. Our method is based on probabilistic roadmap motion planners that we have previously successfully applied to protein folding. In this paper, we provide evidence that this approach is also well suited to RNA. We compute population kinetics and transition rates on our roadmaps using the master equation for a few moderately sized RNA and show that our results compare favorably with results of other existing methods.     

Kinetics analysis methods for approximate folding landscapes

MOTIVATION: Protein motions play an essential role in many biochemical processes. Lab studies often quantify these motions in terms of their kinetics such as the speed at which a protein folds or the population of certain interesting states like the native state. Kinetic metrics give quantifiable measurements of the folding process that can be compared across a group of proteins such as a wild-type protein and its mutants. RESULTS: We present two new techniques, map-based master equation solution and map-based Monte Carlo simulation, to study protein kinetics through folding rates and population kinetics from approximate folding landscapes, models called maps. From these two new techniques, interesting metrics that describe the folding process, such as reaction coordinates, can also be studied. In this article we focus on two metrics, formation of helices and structure formation around tryptophan residues. These two metrics are often studied in the lab through circular dichroism (CD) spectra analysis and tryptophan fluorescence experiments, respectively. The approximated landscape models we use here are the maps of protein conformations and their associated transitions that we have presented and validated previously. In contrast to other methods such as the traditional master equation and Monte Carlo simulation, our techniques are both fast and can easily be computed for full-length detailed protein models. We validate our map-based kinetics techniques by comparing folding rates to known experimental results. We also look in depth at the population kinetics, helix formation and structure near tryptophan residues for a variety of proteins. AVAILABILITY: We invite the community to help us enrich our publicly available database of motions and kinetics analysis by submitting to our server: http://parasol.tamu.edu/foldingserver/.     

Efficient traversal of beta-sheet protein folding pathways using ensemble models

Molecular dynamics (MD) simulations can now predict ms-timescale folding processes of small proteins; however, this presently requires hundreds of thousands of CPU hours and is primarily applicable to short peptides with few long-range interactions. Larger and slower-folding proteins, such as many with extended β-sheet structure, would require orders of magnitude more time and computing resources. Furthermore, when the objective is to determine only which folding events are necessary and limiting, atomistic detail MD simulations can prove unnecessary. Here, we introduce the program tFolder as an efficient method for modelling the folding process of large β-sheet proteins using sequence data alone. To do so, we extend existing ensemble β-sheet prediction techniques, which permitted only a fixed anti-parallel β-barrel shape, with a method that predicts arbitrary β-strand/β-strand orientations and strand-order permutations. By accounting for all partial and final structural states, we can then model the transition from random coil to native state as a Markov process, using a master equation to simulate population dynamics of folding over time. Thus, all putative folding pathways can be energetically scored, including which transitions present the greatest barriers. Since correct folding pathway prediction is likely determined by the accuracy of contact prediction, we demonstrate the accuracy of tFolder to be comparable with state-of-the-art methods designed specifically for the contact prediction problem alone. We validate our method for dynamics prediction by applying it to the folding pathway of the well-studied Protein G. With relatively very little computation time, tFolder is able to reveal critical features of the folding pathways which were only previously observed through time-consuming MD simulations and experimental studies. Such a result greatly expands the number of proteins whose folding pathways can be studied, while the algorithmic integration of ensemble prediction with Markovian dynamics can be applied to many other problems.     

On the relation between native geometry and conformational plasticity

In protein folding the term plasticity refers to the number of alternative folding pathways encountered in response to free energy perturbations such as those induced by mutation. Here we explore the relation between folding plasticity and a gross, generic feature of the native geometry, namely, the relative number of local and non-local native contacts. The results from our study, which is based on Monte Carlo simulations of simple lattice proteins, show that folding to a structure that is rich in local contacts is considerably more plastic than folding to a native geometry characterized by having a very large number of long-range contacts (i.e., contacts between amino acids that are separated by more than 12 units of backbone distance). The smaller folding plasticity of native geometries is probably a direct consequence of their higher folding cooperativity that renders the folding reaction more robust against single- and multiple-point mutations.     

Protein folding: Optimized sequences obtained by simulated breeding in a minimalist model

For a minimalist model of protein folding, which we introduced recently, we investigate various methods to obtain folding sequences. A detailed study of random sequences shows that, for this model, such sequences usually do not fold to their ground states during simulations. Straightforward techniques for the construction of folding sequences, based solely on the target structure, fail. We describe in detail an optimization algorithm, based on genetic algorithms, for the “simulated breeding” of folding sequences in this model. We find that, for any target structure studied, there is not only a single folding sequence but a patch of sequences in sequence space that fold to this structure. In addition, we show that, much as in real proteins, nonhomologous sequences may fold to the same target structure. © 1997 John Wiley & Sons, Inc.     

Simulating protein motions with rigidity analysis.

Protein motions, ranging from molecular flexibility to large-scale conformational change, play an essential role in many biochemical processes. Despite the explosion in our knowledge of structural and functional data, our understanding of protein movement is still very limited. In previous work, we developed and validated a motion planning based method for mapping protein folding pathways from unstructured conformations to the native state. In this paper, we propose a novel method based on rigidity theory to sample conformation space more effectively, and we describe extensions of our framework to automate the process and to map transitions between specified conformations. Our results show that these additions both improve the accuracy of our maps and enable us to study a broader range of motions for larger proteins. For example, we show that rigidity-based sampling results in maps that capture subtle folding differences between protein G and its mutants, NuG1 and NuG2, and we illustrate how our technique can be used to study large-scale conformational changes in calmodulin, a 148 residue signaling protein known to undergo conformational changes when binding to Ca(2+). Finally, we announce our web-based protein folding server which includes a publicly available archive of protein motions: (http://parasol.tamu.edu/foldingserver/).     

Computational investigation of the effect of thermal perturbation on the mechanical unfolding of titin I27

The emergence of single-molecule force measurement experiments has facilitated a better understanding of protein folding pathways and the thermodynamics involved. Computational methods such as steered molecular dynamics (SMD) simulations are helpful in providing atomistic level information on the unfolding pathways. Recent experimental studies have showed that combinations of single-molecule experiments with traditional methods such as chemical and/or thermal denaturation yield additional insights into the folding phenomenon. In this study, we report results from extensive computations (a total of about 60 SMD simulations with a total length of about 0.4 μs) that address the effect of thermal perturbation on the mechanical stability of the I27 domain of the protein titin. A wide range of temperatures (280-340 K) were considered for the pulling, which was done at both constant velocity and constant force using SMD simulations. Good agreement with experimental data, such as for the trends in changes in average force and the maximum force with respect to the temperature, was obtained. This study identifies two competing pathways for the mechanical unfolding of I27, and illustrates the significance of combining various techniques to examine protein folding.     

Multiple parallel-pathway folding of proline-free Staphylococcal nuclease

When a protein exhibits complex kinetics of refolding, we often ascribe the complexity to slow isomerization events in the denatured protein, such as cis/trans isomerization of peptidyl prolyl bonds. Does the complex folding kinetics arise only from this well-known reason? Here, we have investigated the refolding of a proline-free variant of staphylococcal nuclease by stopped-flow, double-jump techniques, to examine the folding reactions without the slow prolyl isomerizations. As a result, the protein folds into the native state along at least two accessible parallel pathways, starting from a macroscopically single denatured-state ensemble. The presence of intermediates on the individual folding pathways has revealed the existence of multiple parallel pathways, and is characterized by multi-exponential folding kinetics with a lag phase. Therefore, a "single" amino acid sequence can fold along the multiple parallel pathways. This observation in staphylococcal nuclease suggests that the multiple folding may be more general than we have expected, because the multiple parallel-pathway folding cannot be excluded from proteins that show simpler kinetics.     

Lattice simulations of cotranslational folding of single domain proteins

We use lattice protein models and Monte Carlo simulations to study cotranslational folding of small single domain proteins. We show that the assembly of native structure begins during late extrusion stages, but final formation of native state occurs during de novo folding, when all residues are extruded. There are three main results in our study. First, for the sequences displaying two-state refolding mechanism de novo cotranslational folding pathway differs from that sampled in in vitro refolding. The change in folding pathways is due to partial assembly of native interactions during extrusion that results in different starting conditions for in vitro refolding and for de novo cotranslational folding. For small single domain proteins cotranslational folding is slower than in vitro refolding, but is generally fast enough to be completed before the release from a ribosome. Second, we found that until final stages of biosynthesis cotranslational folding is essentially equilibrium. This observation is explained by low stability of structured states for partially extruded chains. Finally, our data suggest that the proteins, which refold in vitro slowly via intermediates, complete their de novo folding after the release from a ribosome. Comparison of our lattice cotranslational simulations with recent experimental and computational studies is discussed.     

Analyses of protein sequences using inter-residue average distance statistics to study folding processes and the significance of their partial sequences

One of the goals of molecular bioinformatics is decoding amino acid sequences to extract information on the principles of protein folding. However, this is difficult to perform with standard bioinformatics techniques such as multiple sequence alignment and so on. Thus, we propose a technique based on inter-residue average distance statistics to make predictions regarding the protein folding mechanisms of amino acid sequences. Our method involves constructing a kind of predicted contact map called an Average Distance Map (ADM) based on average distance statistics to pinpoint regions of possible folding nuclei for proteins. Only information on the amino acid sequence of a given protein is required for the present method. In this article, we summarize the results of studies using our method to analyze how specific protein sequences affect folding properties. In particular, we present studies on proteins in the phage lysozyme, such as the globin, fatty acid binding protein-like, and the cupredoxin-like fold families. In the present review, we characterize the 3D architectures of these proteins through the properties of the protein ADMs. Furthermore, we combine the information on the conserved residues within the regions predicted by the ADMs with our results obtained so far. Such information may help identify the folding characteristics of each protein. We discuss this possibility in the present review.     

Predicting protein folding cores by empirical potential functions

Theoretical and in vitro experiments suggest that protein folding cores form early in the process of folding, and that proteins may have evolved to optimize both folding speed and native-state stability. In our previous work (Chen et al., Structure, 14 (2006) 1401), we developed a set of empirical potential functions and used them to analyze interaction energies among secondary-structure elements in two β-sandwich proteins. Our work on this group of proteins demonstrated that the predicted folding core also harbors residues that form native-like interactions early in the folding reaction. In the current work, we have tested our empirical potential functions on structurally-different proteins for which the folding cores have been revealed by protein hydrogen-deuterium exchange experiments. Using a set of 29 unrelated proteins, which have been extensively studied in the literature, we demonstrate that the average prediction result from our method is significantly better than predictions based on other computational methods. Our study is an important step towards the ultimate goal of understanding the correlation between folding cores and native structures.     

The tetrameric α-helical membrane protein GlpF unfolds via a dimeric folding intermediate

Many membrane proteins appear to be present and functional in higher-order oligomeric states. While few studies have analyzed the thermodynamic stability of α-helical transmembrane (TM) proteins under equilibrium conditions in the past, oligomerization of larger polytopic monomers has essentially not yet been studied. However, it is vital to study the folding of oligomeric membrane proteins to improve our understanding of the general mechanisms and pathways of TM protein folding. To investigate the folding and stability of the aquaglyceroporin GlpF from Escherichia coli, unfolding of the protein in mixed micelles was monitored by steady-state fluorescence and circular dichroism spectroscopy as well as by seminative sodium dodecyl sulfate-polyacrylamide gel electrophoresis analyses. On the basis of our results, it appears most likely that GlpF unfolds in a two-step process, involving the equilibrium of tetrameric, dimeric, and monomeric GlpF species. A kinetic analysis also indicates an intermediate along the kinetic GlpF unfolding pathway, and thus, two phases are involved in GlpF unfolding. While three-state unfolding pathways and a dimeric folding intermediate are not uncommon for water-soluble proteins, a stable (un)folding intermediate with a decreased oligomeric structure has not been detected or reported for any α-helical membrane protein.     

Complex folding pathways in a simple beta-hairpin

The determination of the folding mechanisms of proteins is critical to understand the topological change that can propagate Alzheimer and Creutzfeld-Jakobs diseases, among others. The computational community has paid considerable attention to this problem; however, the associated time scale, typically on the order of milliseconds or more, represents a formidable challenge. Ab initio protein folding from long molecular dynamics simulations or ensemble dynamics is not feasible with ordinary computing facilities and new techniques must be introduced. Here we present a detailed study of the folding of a 16-residue beta-hairpin, described by a generic energy model and using the activation-relaxation technique. From a total of 90 trajectories at 300 K, three folding pathways emerge. All involve a simultaneous optimization of the complete hydrophobic and hydrogen bonding interactions. The first two pathways follow closely those observed by previous theoretical studies (folding starting at the turn or by interactions between the termini). The third pathway, never observed by previous all-atom folding, unfolding, and equilibrium simulations, can be described as a reptation move of one strand of the beta-sheet with respect to the other. This reptation move indicates that non-native interactions can play a dominant role in the folding of secondary structures. Furthermore, such a mechanism mediated by non-native hydrogen bonds is not available for study by unfolding and Gō model simulations. The exact folding path followed by a given beta-hairpin is likely to be influenced by its sequence and the solvent conditions. Taken together, these results point to a more complex folding picture than expected for a simple beta-hairpin.     

On prediction of folding nuclei in globular proteins

The approach described in this paper on the prediction of folding nuclei in globular proteins with known three dimensional structures is based on a search of the lowest saddle points through the barrier separating the unfolded state from the native structure on the free-energy landscape of protein chain. This search is performed by a dynamic programming method. Comparison of theoretical results with experimental data on the folding nuclei of two dozen of proteins shows that our model provides good phi value predictions for proteins whose structures have been determined by X-ray analysis, with a less limited success for proteins whose structures have been determined by NMR techniques only. Consideration of a full ensemble of transition states results in more successful prediction than consideration of only the transition states with the minimal free energy. In conclusion we have predicted the localization of folding nuclei for three dimensional protein structures for which kinetics of folding is studied now but the localization of folding nuclei is still unknown.     

Local secondary structure content predicts folding rates for simple,two-state proteins

Many single-domain proteins exhibit two-state folding kinetics, with folding rates that span more than six orders of magnitude. A quantity of much recent interest for such proteins is their contact order, the average separation in sequence between contacting residue pairs. Numerous studies have reached the surprising conclusion that contact order is well-correlated with the logarithm of the folding rate for these small, well-characterized molecules. Here, we investigate the physico-chemical basis for this finding by asking whether contact order is actually a composite number that measures the fraction of local secondary structure in the protein; viz. turns, helices, and hairpins. To pursue this question, we calculated the secondary structure content for 24 two-state proteins and obtained coefficients that predict their folding rates. The predicted rates correlate strongly with experimentally determined rates, comparable to the correlation with contact order. Further, these predicted folding rates are correlated strongly with contact order. Our results suggest that the folding rate of two-state proteins is a function of their local secondary structure content, consistent with the hierarchic model of protein folding. Accordingly, it should be possible to utilize secondary structure prediction methods to predict folding rates from sequence alone.     

Outlining folding nuclei in globular proteins

Our theoretical approach for prediction of folding/unfolding nuclei in three-dimensional protein structures is based on a search for free energy saddle points on networks of protein unfolding pathways. Under some approximations, this search is performed rapidly by dynamic programming and results in prediction of Phi values, which can be compared with those found experimentally. In this study, we optimize some details of the model (specifically, hydrogen atoms are taken into account in addition to heavy atoms), and compare the theoretically obtained and experimental Phi values (which characterize involvement of residues in folding nuclei) for all 17 proteins, where Phi values are now known for many residues. We show that the model provides good Phi value predictions for proteins whose structures have been determined by X-ray analysis (the average correlation coefficient is 0.65), with a more limited success for proteins whose structures have been determined by NMR techniques only (the average correlation coefficient is 0.34), and that the transition state free energies computed from the same model are in a good anticorrelation with logarithms of experimentally measured folding rates at mid-transition (the correlation coefficient is -0.73).