Folding Molecular Dynamics Simulations Research Articles

Critical assessment of popular biomolecular force fields for molecular dynamics simulations of folding and enzymatic activity of main protease of coronavirus SARS-CoV-2

The main cysteine protease (Mpro) of coronavirus SARS-CoV-2 has become a promising target for computational development in anti-COVID-19 treatments. Here, we benchmarked the performance of six biomolecular molecular dynamics (MD) force fields (OPLS-AA, CHARMM27, CHARMM36, AMBER03, AMBER14SB and GROMOS G54A7) and three water models (TIP3P, TIP4P and SPC) for reproducing the native fold and the enzymatic activity of Mpro as monomeric and dimeric units. The MD sampling up to 1 μs suggested that the proper choice of the force fields and water models plays an essential role in reproducing the tertiary structure and the inter-residue distance between the catalytic dyad His41-Cys145. We found that while most benchmarked all-atom force fields reproduce well the native fold of Mpro, the CHARMM27/TIP3P and OPLS-AA/TIP4P setups revealed a good performance in reproducing the structure of the catalytic domain. In addition, these FF setups were also well-adopted for MD sampling of Mpro at the physiologic conditions by mimicking the presence of 100 mM NaCl and the elevated temperature of 310 K. Finally, both FFs were also performed well in reproducing the native fold of Mpro in a dimeric form. Therefore, comparing the preservation of the native fold of Mpro and the stability of its catalytic site architecture, our MD benchmarking suggests that the OPLS-AA/TIP4P and CHARMM27/TIP3P MD setups at the physiologic conditions may be well-suited for rapid in silico screening and developing broad-spectrum anti-coronaviral therapeutic agents.

A proof-of-concept study of the secondary structure of influenza A, B M2 and MERS- and SARS-CoV E transmembrane peptides using folding molecular dynamics simulations in a membrane mimetic solvent.

Here, we have carried out a proof-of-concept molecular dynamics (MD) simulation with adaptive tempering in a membrane mimetic environment to study the folding of single-pass membrane peptides. We tested the influenza A M2 viroporin, influenza B M2 viroporin, and protein E from coronaviruses MERS-Cov-2 and SARS-CoV-2 peptides with known experimental secondary structures in membrane bilayers. The two influenza-derived peptides are significantly different in the peptide sequence and secondary structure and more polar than the two coronavirus-derived peptides. Through a total of more than 50 μs of simulation time that could be accomplished in trifluoroethanol (TFE), as a membrane model, we characterized comparatively the folding behavior, helical stability, and helical propensity of these transmembrane peptides that match perfectly their experimental secondary structures, and we identified common motifs that reflect their quaternary organization and known (or not) biochemical function. We showed that BM2 is organized into two structurally distinct parts: a significantly more stable N-terminal half, and a fast-converting C-terminal half that continuously folds and unfolds between α-helical structures and non-canonical structures, which are mostly turns. In AM2, both the N-terminal half and C-terminal half are very flexible. In contrast, the two coronavirus-derived transmembrane peptides are much more stable and fast helix-formers when compared with the influenza ones. In particular, the SARS-derived peptide E appears to be the fastest and most stable helix-former of all the four viral peptides studied, with a helical structure that persists almost without disruption for the whole of its 10 μs simulation. By comparing the results with experimental observations, we benchmarked TFE in studying the conformation of membrane and hydrophobic peptides. This work provided accurate results suggesting a methodology to run long MD simulations and predict structural properties of biologically important membrane peptides.

A molecular dynamics simulation study on the propensity of Asn-Gly-containing heptapeptides towards β-turn structures: Comparison with ab initio quantum mechanical calculations.

Both molecular mechanical and quantum mechanical calculations play an important role in describing the behavior and structure of molecules. In this work, we compare for the same peptide systems the results obtained from folding molecular dynamics simulations with previously reported results from quantum mechanical calculations. More specifically, three molecular dynamics simulations of 5 μs each in explicit water solvent were carried out for three Asn-Gly-containing heptapeptides, in order to study their folding and dynamics. Previous data, based on quantum mechanical calculations within the DFT framework have shown that these peptides adopt β-turn structures in aqueous solution, with type I’ β-turn being the most preferred motif. The results from our analyses indicate that at least for the given systems, force field and simulation protocol, the two methods diverge in their predictions. The possibility of a force field-dependent deficiency is examined as a possible source of the observed discrepancy.

The balance between side-chain and backbone-driven association in folding of the α-helical influenza A transmembrane peptide.

The correct balance between attractive, repulsive and peptide hydrogen bonding interactions must be attained for proteins to fold correctly. To investigate these important contributors, we sought a comparison of the folding between two 25-residues peptides, the influenza A M2 protein transmembrane domain (M2TM) and the 25-Ala (Ala25 ). M2TM forms a stable α-helix as is shown by circular dichroism (CD) experiments. Molecular dynamics (MD) simulations with adaptive tempering show that M2TM monomer is more dynamic in nature and quickly interconverts between an ensemble of various α-helical structures, and less frequently turns and coils, compared to one α-helix for Ala25 . DFT calculations suggest that folding from the extended structure to the α-helical structure is favored for M2TM compared with Ala25 . This is due to CH⋯O attractive interactions which favor folding to the M2TM α-helix, and cannot be described accurately with a force field. Using natural bond orbital (NBO) analysis and quantum theory atoms in molecules (QTAIM) calculations, 26 CH⋯O interactions and 22 NH⋯O hydrogen bonds are calculated for M2TM. The calculations show that CH⋯O hydrogen bonds, although individually weaker, have a cumulative effect that cannot be ignored and may contribute as much as half of the total hydrogen bonding energy, when compared to NH⋯O, to the stabilization of the α-helix in M2TM. Further, a strengthening of NH⋯O hydrogen bonding interactions is calculated for M2TM compared to Ala25 . Additionally, these weak CH⋯O interactions can dissociate and associate easily leading to the ensemble of folded structures for M2TM observed in folding MD simulations.

Universal and Accessible Entropy Estimation Using a Compression Algorithm.

Entropy and free-energy estimation are key in thermodynamic characterization of simulated systems ranging from spin models through polymers, colloids, protein structure, and drug design. Current techniques suffer from being model specific, requiring abundant computation resources and simulation at conditions far from the studied realization. Here, we present a universal scheme to calculate entropy using lossless-compression algorithms and validate it on simulated systems of increasing complexity. Our results show accurate entropy values compared to benchmark calculations while being computationally effective. In molecular-dynamics simulations of protein folding, we exhibit unmatched detection capability of the folded states by measuring previously undetectable entropy fluctuations along the simulation timeline. Such entropy evaluation opens a new window onto the dynamics of complex systems and allows efficient free-energy calculations.

A network of conformational transitions in an unfolding process of HP-35 revealed by high-temperature MD simulation and a Markov state model**Project supported by the National Natural Science Foundation of China (Grant Nos. 11175068 and 11474117) and the Self-determined Research Funds of CCNU from the Colleges Basic Research and Operation of MOE, China (Grant No. 230-20205170054).

An understanding of protein folding/unfolding processes has important implications for all biological processes, including protein degradation, protein translocation, aging, and diseases. All-atom molecular dynamics (MD) simulations are uniquely suitable for it because of their atomic level resolution and accuracy. However, limited by computational capabilities, nowadays even for small and fast-folding proteins, all-atom MD simulations of protein folding still presents a great challenge. An alternative way is to study unfolding process using MD simulations at high temperature. High temperature provides more energy to overcome energetic barriers to unfolding, and information obtained from studying unfolding can shed light on the mechanism of folding. In the present study, a 1000-ns MD simulation at high temperature (500 K) was performed to investigate the unfolding process of a small protein, chicken villin headpiece (HP-35). To infer the folding mechanism, a Markov state model was also built from our simulation, which maps out six macrostates during the folding/unfolding process as well as critical transitions between them, revealing the folding mechanism unambiguously.

Folding Simulations of a Nuclear Receptor Box-Containing Peptide Demonstrate the Structural Persistence of the LxxLL Motif Even in the Absence of Its Cognate Receptor.

Regulation of nuclear receptors by their coactivators involves the recognition and binding of a specific sequence motif contained in the coactivator sequence. This motif is known as the nuclear receptor (NR) box and contains a conserved LxxLL subsequence, where L is leucine and x is any amino acid residue. Crystallographic studies have shown that the LxxLL motifs adopt an α-helical conformation when bound to their cognate nuclear receptors. Here we use an extensive set of folding molecular dynamics simulations to examine whether the α-helical conformation demonstrated by the LxxLL motifs in the bound state may represent a persistent structural preference of these peptides even in the absence of their cognate receptors. To this end, we have performed a grand total of 35 μs of adaptive tempering folding simulations of an NR-box-containing peptide derived from Drosophila's fushi tarazu segmentation gene product. Our simulations-performed using full electrostatics and an explicit representation of two different solvents (water and a TFE/water mixture)-clearly indicate the presence of a persistent helical preference of the LxxLL motif with a concomitant native-like structure and contacts between the motif's leucine residues. To lend further support to our findings, we compare the simulation-derived peptide dynamics with experimental NMR-derived nuclear Overhauser effect (NOE) measurements that had been previously obtained for the same peptide in the same two solvents. The comparison demonstrates a quantitative agreement between simulation and experiment with average upper bound NOE violations of less than 0.084 Å, thus independently validating our main conclusion concerning the intrinsic preference of NR-box motifs to form helical structures even in the absence of their cognate receptors.

Sensitivity of Folding Molecular Dynamics Simulations to Even Minor Force Field Changes.

We examine the sensitivity of folding molecular dynamics simulations on the choice between three variants of the same force field (the AMBER99SB force field and its ILDN, NMR-ILDN, and STAR-ILDN variants). Using two different peptide systems (a marginally stable helical peptide and a β-hairpin) and a grand total of more than 20 μs of simulation time we show that even relatively minor force field changes can lead to appreciable differences in the peptide folding behavior.

Folding molecular dynamics simulations accurately predict the effect of mutations on the stability and structure of a vammin-derived peptide.

Folding molecular dynamics simulations amounting to a grand total of 4 μs of simulation time were performed on two peptides (with native and mutated sequences) derived from loop 3 of the vammin protein and the results compared with the experimentally known peptide stabilities and structures. The simulations faithfully and accurately reproduce the major experimental findings and show that (a) the native peptide is mostly disordered in solution, (b) the mutant peptide has a well-defined and stable structure, and (c) the structure of the mutant is an irregular β-hairpin with a non-glycine β-bulge, in excellent agreement with the peptide's known NMR structure. Additionally, the simulations also predict the presence of a very small β-hairpin-like population for the native peptide but surprisingly indicate that this population is structurally more similar to the structure of the native peptide as observed in the vammin protein than to the NMR structure of the isolated mutant peptide. We conclude that, at least for the given system, force field, and simulation protocol, folding molecular dynamics simulations appear to be successful in reproducing the experimentally accessible physical reality to a satisfactory level of detail and accuracy.

As good as it gets? Folding molecular dynamics simulations of the LytA choline-binding peptide result to an exceptionally accurate model of the peptide structure

Folding simulations of a choline-binding peptide derived from the Streptococcus pneumoniae LytA protein converged to a model of the peptide's folded state structure which is in outstanding agreement with the experimentally-determined structures, reaching values for the root mean squared deviation as low as 0.24Å for the peptide's backbone atoms and 0.65Å for all non-hydrogen atoms.

Molecular Dynamics Simulations of Folding of Supported Graphene

The controlled folding of graphene structures driven by molecular interactions with water nanodroplets is analyzed taking into account interactions with supported substrates such as silicon dioxide (SiO2), hexamethyldisilazane (HMDS), and isopropyl alcohol (IPA) on SiO2. The interaction between unsupported graphene and a water nanodroplet is strong enough to induce folding on graphene nanostructures. However, when the graphene is supported on SiO2, the attraction between graphene and the substrate prevents graphene from folding but if the substrate is a hydrophobic surface such as HMDS or a solvent such as IPA, the interaction between graphene and the substrate is weak, and depending on the geometry of the graphene structure, folding is possible. The selection of an acceptable substrate and graphene geometry opens the possibility of a controlled fabrication of graphene-based capsules, scrolls, sandwiches, and rings able to carry a payload.

Computational analysis of folding and mutation properties of C5 domain of myosin binding protein C

Thermal folding molecular dynamics simulations of the domain C5 of Myosin binding protein C were performed using a native-centric model to study the role of three mutations related to Familial Hypertrophic Cardiomyopathy. Mutation of Asn755 causes the largest shift of the folding temperature, and the residue is located in the CFGA' beta-sheet featuring the highest phi-values. The mutation thus appears to reduce the thermodynamic stability in agreement with experimental data. The mutations on Arg654 and Arg668, conversely, cause little change in the folding temperature and they reside in the low phi-value BDE beta-sheet, so that their pathological role cannot be related to impairment of the folding process but possibly to the binding with target molecules. As the typical signature of Domain C5 is the presence of a longer and destibilizing CD-loop with respect to the other Ig-like domains, we completed the work with a bioinformatic analysis of this loop showing a high density of negative charge and low hydrophobicity. This indicates the CD-loop as a natively unfolded sequence with a likely coupling between folding and ligand binding.

Data Mining in Bioinformatics (BIOKDD)

Data Mining is the process of automatic discovery of novel and understandable models and patterns from large amounts of data. Bioinformatics is the science of storing, analyzing, and utilizing information from biological data such as sequences, molecules, gene expressions, and pathways. Development of novel data mining methods will play a fundamental role in understanding these rapidly expanding sources of biological data. Data mining approaches seem ideally suited for bioinformatics, which is data-rich, but lacks a comprehensive theory of life's organization at the molecular level. The extensive databases of biological information create both challenges and opportunities for developing novel data mining methods. The 6th Workshop on Data Mining in Bioinformatics (BIOKDD) was held on August 20th, 2006, Philadelphia, PA, USA, in conjunction with the 12th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. The goal of the workshop was to encourage KDD researchers to take on the numerous challenges that Bioinformatics offers. The BIOKDD workshops have been held annually in conjunction with the ACM SIGKDD Conferences, since 2001. Additional information about BIOKDD can be obtained online [1]. Five revised and expanded papers were selected from the BIOKDD workshop, out of a total of 18 submissions, to appear in Algorithms for Molecular Biology (AMB). These papers underwent another round of external reviewing prior to being accepted for AMB. An overview of each paper is given below. In the paper titled Automatic Layout and Visualization of Biclusters, Gregory A. Grothaus, Adeel Mufti and T. M. Murali [2], present a novel method to display biclusters mined from gene expression data. The approach allows querying and visual exploration of the clusters/sub-matrices. The software is also available as open-source. In ExMotif: Efficient Structured Motif Extraction, Yongqiang Zhang and Mohammed J. Zaki [3], describe a new algorithm called EXMOTIF to extract frequent motifs from DNA sequences. The method can mine structured motifs and profiles which have variable gaps between different elements. The demonstrate the efficiency of the method compared to state-of-the-art methods, and also demonstrate an application in mining composite transcription factor binding sites. In the paper Refining Motifs by Improving Information Content Scores using Neighborhood Profile Search, Chandan K. Reddy, Yao-Chung Weng and Hsiao-Dong Chiang [4], show how one can refine the profile motifs discovered via Expectation Maximization and Gibbs Sampling based methods. They search the neighborhood regions of the initial alignments to obtain locally optimal solutions, which improve the information content of the discovered profiles. In their paper, A Novel Functional Module Detection Algorithm for Protein-Protein Interaction Networks, Woochang Hwang, Young-Rae Cho, Aidong Zhang and Murali Ramanathan [5], describe the unexpected properties of the protein-protein interaction (PPI) networks and their use in a clustering method to detect biologically relevant functional modules. They propose a new method called STM (signal transduction model) to detect the PPI modules, and compare it with previous approaches to demonstrate its effectiveness in discovering large and arbitrary shaped clusters. In A Spatio-temporal Mining Approach towards Summarizing and Analyzing Protein Folding Trajectories, Hui Yang, Srinivasan Parthasarathy and Duygu Ucar [6], describe a method to mine protein folding molecular dynamics simulations datasets. They describe a spatio-temporal association discovery approach to mine protein folding trajectories, to identify critical events and common pathways.

Protein folding: Web alert

Various genome projects are identifying thousands of new protein sequences every year, which has led to an expanded interest in protein folding. There are two parts to the protein folding problem: the first is predicting the three-dimensional structure of a protein from its amino-acid sequence; and the second — the focus of this Web alert — is to understand how proteins fold. Increased interest in protein folding has been fuelled by the belief that many human diseases (Alzheimer’s and Parkinson’s, for example) are caused by aggregation or misfolding of proteins. The FASEB Society has a valuable essay (aimed at the general public) that describes known diseases related to protein (mis)folding (www.faseb.org/opar/protfold/protein.html). It is possible that the selective pressures that act during evolution function to preserve amino-acid sequences that not only form structures that adequately perform specific biological functions, but also to preserve those that fold quickly enough to avoid aggregation or misfolded states. Although a search for ‘protein folding’ using a search engine like AltaVista has more than 13,600 hits, the number of useful protein-folding resources is limited. A good place to start, to get acquainted with the people in the field, is the comprehensive list of protein-folding groups compiled by Heinrich Roder’s laboratory at Fox Chase Cancer Center (http://www.fccc.edu/research/labs/roder/folding_groups.html), which contains links to more than 80 research teams working on various aspects of protein folding. Several of the scientists listed here have their own web pages, some of which are very creative, and include links to other useful sites. Protein folding is quite different from small-molecule chemical reactions (in which covalent bonds are formed or broken in well-defined transition states), because of the many degrees of freedom allowed a polypeptide chain. In order to understand the complexity of protein-folding processes at the microscopic level, a statistical approach is required. The current theoretical and computational frontline, with respect to protein-folding dynamics, is well described in the Pittsburgh Supercomputing Center’s interactive collection of articles from 1998 (http://www.psc.edu/science/lifeproc.html). In particular, the papers by C.L. Brooks III and P. Kollman are admirable. Vijay Pande’s website (http://www.stanford.edu/group/pandegroup/#research) is also excellent, and includes movies of all-atom molecular-dynamics simulations of protein folding. The Database of Macromolecular Movements (http://bioinfo.mbb.yale.edu/MolMovDB/) is another amusing site to visit. It contains short videos of motions in proteins (including a few examples of folding and unfolding). This site also includes free software tools for protein geometric analyses. In order to place protein folding in its realistic in vivo context, there are additional factors to address. Around 5% of all proteins made in cells (often larger, multidomain proteins) require help from ‘chaperonin’ proteins to fold properly. To learn more about these reactions, visit the Chaperonin Home Page (http://bioc09.uthscsa.edu/∼seale/Chap/chap.html) located at the University of Texas Health Science Center. This web page links to researchers in the field and lists thermodynamic data for a large number of mutants of the most famous chaperonin pair GroEL/GroES. The site also includes a protein-structure gallery and a recent paper alert. 20–30% of all proteins function within cellular membranes. It has not yet been resolved (from either a thermodynamic or a kinetic viewpoint) how such proteins are synthesized and then inserted into the membranes. An excellent resource for this topic is the web pages of Stephen White’s laboratory at UC Irvine (http://blanco.biomol.uci.edu/mp_assembly.html). The general principles of membrane-protein structure and stability, the group’s own research approach, a wealth of interesting links and downloadable key research papers are all available here. In order to perform their biological functions, many proteins coordinate small cofactors such as metal ions. The recently created PROMISE database (http://bioinf.leeds.ac.uk/promise/) provides a wealth of information on these types of proteins (although there is not much on the proteins’ folding properties yet). In addition, this site links to an impressive number of bioinorganic web resources. The field of protein folding is growing at a fast pace, as is the number of researchers that need to keep up with the current literature and news. A perfect way to do this is to attend the Johns Hopkins Protein Folding Meeting, which has taken place each spring for the past five years. This meeting attracts the finest people in protein folding, and the conference covers all aspects of this fascinating field. The program for the 5th Meeting (taking place this month) can be viewed at http://www.jhu.edu/∼ipmb/protein.html. Protein folding is a rapidly expanding field that involves researchers from many disciplines. We expect new web resources to be developed that continue to facilitate discussion and collaboration.

Identifying the protein folding nucleus using molecular dynamics

Molecular dynamics simulations of folding in an off-lattice protein model reveal a nucleation scenario, in which a few well-defined contacts are formed with high probability in the transition state ensemble of conformations. Their appearance determines folding cooperativity and drives the model protein into its folded conformation. Amino acid residues participating in those contacts may serve as “accelerator pedals” used by molecular evolution to control protein folding rate.

Molecular Dynamics Simulations of Folding in Cyclic Alkanes

Molecular Dynamics Simulations of Folding in Cyclic Alkanes

Molecular dynamics simulations of synthetic peptide folding.

Because the time scale of protein folding is much greater than that of the widely used simulations of native structures, a detailed report of molecular dynamics simulations of folding has not been available. In this study, we included the average solvent effect in the potential functions to simplify the calculation of the solvent effect and carried out long molecular dynamics simulations of the alanine-based synthetic peptides at 274 K. From either an extended or a randomly generated conformation, the simulations approached a helix-coil equilibrium in about 3 ns. The multiple minima problem did not prevent helix folding. The calculated helical ratio of Ac-AAQ-AAAAQAAAAQAAY-NH2 was 47%, in good agreement with the circular dichroism measurement (about 50%). A helical segment with frayed ends was the most stable conformation, but the hydrophobic interaction favored the compact, distorted helix-turn-helix conformations. The transition between the two types of conformations occurred in a much larger time scale than helix propagation. The transient hydrogen bonds between the glutamine side chain and the backbone carbonyl group could reduce the free energy barrier of helix folding and unfolding. The substitution of a single alanine residue in the middle of the peptide with valine or glycine decreased the average helical ratio significantly, in agreement with experimental observations.