Fast tree search for a triangular lattice model of protein folding

Using a triangular lattice model to study the designability of protein folding, we overcame the parity problem of previous cubic lattice model and enumerated all the sequences and compact structures on a simple two-dimensional triangular lattice model of size 4 5 6 5 4. We used two types of amino acids, hydrophobic and polar, to make up the sequences, and achieved 223W212 different sequences excluding the reverse symmetry sequences. The total string number of distinct compact structures was 219,093, excluding reflection symmetry in the self-avoiding path of length 24 triangular lattice model. Based on this model, we applied a fast search algorithm by constructing a cluster tree. The algorithm decreased the computation by computing the objective energy of non-leaf nodes. The parallel experiments proved that the fast tree search algorithm yielded an exponential speed-up in the model of size 4 5 6 5 4. Designability analysis was performed to understand the search result.     

Exploring the conformational space of protein loops using a mean field technique with MOLS sampling

We have recently developed a computational technique that uses mutually orthogonal Latin square sampling to explore the conformational space of oligopeptides in an exhaustive manner. In this article, we report its use to analyze the conformational spaces of 120 protein loop sequences in proteins, culled from the PDB, having the length ranging from 5 to 10 residues. The force field used did not have any information regarding the sequences or structures that flanked the loop. The results of the analyses show that the native structure of the loop, as found in the PDB falls at one of the low energy points in the conformational landscape of the sequences. Thus, a large portion of the structural determinants of the loop may be considered intrinsic to the sequence, regardless of either adjacent sequences or structures, or the interactions that the atoms of the loop make with other residues in the protein or in neighboring proteins.     

Effects of salt bridges on protein structure and design.

Theoretical calculations (Hendsch ZS & Tidor B, 1994, Protein Sci 3:211-226) and experiments (Waldburger CD et al., 1995, Nat Struct Biol 2:122-128; Wimley WC et al., 1996, Proc Natl Acad Sci USA 93:2985-2990) suggest that hydrophobic interactions are more stabilizing than salt bridges in protein folding. The lack of apparent stability benefit for many salt bridges requires an alternative explanation for their occurrence within proteins. To examine the effect of salt bridges on protein structure and stability in more detail, we have developed an energy function for simple cubic lattice polymers based on continuum electrostatic calculations of a representative selection of salt bridges found in known protein crystal structures. There are only three types of residues in the model, with charges of -1, 0, or + 1. We have exhaustively enumerated conformational space and significant regions of sequence space for three-dimensional cubic lattice polymers of length 16. The results demonstrate that, while the more highly charged sequences are less stable, the loss of stability is accompanied by a substantial reduction in the degeneracy of the lowest-energy state. Moreover, the reduction in degeneracy is greater due to charges that pair than for lone charges that remain relatively exposed to solvent. We have also explored and illustrated the use of ion-pairing strategies for rational structural design using model lattice studies.     

A fast conformational search strategy for finding low energy structures of model proteins.

We describe a new computer algorithm for finding low-energy conformations of proteins. It is a chain-growth method that uses a heuristic bias function to help assemble a hydrophobic core. We call it the Core-directed chain Growth method (CG). We test the CG method on several well-known literature examples of HP lattice model proteins [in which proteins are modeled as sequences of hydrophobic (H) and polar (P) monomers], ranging from 20-64 monomers in two dimensions, and up to 88-mers in three dimensions. Previous nonexhaustive methods--Monte Carlo, a Genetic Algorithm, Hydrophobic Zippers, and Contact Interactions--have been tried on these same model sequences. CG is substantially better at finding the global optima, and avoiding local optima, and it does so in comparable or shorter times. CG finds the global minimum energy of the longest HP lattice model chain for which the global optimum is known, a 3D 88-mer that has only been reachable before by the CHCC complete search method. CG has the potential advantage that it should have nonexponential scaling with chain length. We believe this is a promising method for conformational searching in protein folding algorithms.     

Sequence-structure signals of 3D domain swapping in proteins

Three-dimensional domain swapping occurs when two or more identical proteins exchange identical parts of their structure to generate an oligomeric unit. It affects proteins with diverse sequences and structures, and is expected to play important roles in evolution, functional regulation and even conformational diseases. Here, we search for traces of domain swapping in the protein sequence, by means of algorithms that predict the structure and stability of proteins using database-derived potentials. Regions whose sequences are not optimal with regard to the stability of the native structure, or showing marked intrinsic preferences for non-native conformations in absence of tertiary interactions are detected in most domain-swapping proteins. These regions are often located in areas crucial in the swapping process and are likely to influence it on a kinetic or thermodynamic level. In addition, cation-pi interactions are frequently observed to zip up the edges of the interface between intertwined chains or to involve hinge loop residues, thereby modulating stability. We end by proposing a set of mutations altering the swapping propensities, whose experimental characterization would contribute to refine our in silico derived hypotheses.     

Why do proteins divide into domains? Insights from lattice model simulations

It is known that larger globular proteins are built from domains, relatively independent structural units. A domain size seems to be limited, and a single domain consists of from few tens to a couple of hundred amino acids. Based on Monte Carlo simulations of a reduced protein model restricted to the face centered simple cubic lattice, with a minimal set of short-range and long-range interactions, we have shown that some model sequences upon the folding transition spontaneously divide into separate domains. The observed domain sizes closely correspond to the sizes of real protein domains. Short chains with a proper sequence pattern of the hydrophobic and polar residues undergo a two-state folding transition to the structurally ordered globular state, while similar longer sequences follow a multistate transition. Homopolymeric (uniformly hydrophobic) chains and random heteropolymers undergo a continuous collapse transition into a single globule, and the globular state is much less ordered. Thus, the factors responsible for the multidomain structure of proteins are sufficiently long polypeptide chain and characteristic, protein-like, sequence patterns. These findings provide some hints for the analysis of real sequences aimed at prediction of the domain structure of large proteins.     

Importance of context in protein folding: secondary structural propensities versus tertiary contact-assisted secondary structure formation

Molecular dynamics simulations can be used to reveal the detailed conformational behaviors of peptides and proteins. By comparing fragment and full-length protein simulations, we can investigate the role of each peptide segment in the folding process. Here, we take advantage of information regarding the helix formation process from our previous simulations of barnase and protein A as well as new simulations of four helical fragments from these proteins at three different temperatures, starting with both helical and extended structures. Segments with high helical propensity began the folding process by tethering the chain through side chain interactions involving either polar interactions, such as salt bridges, or hydrophobic staples. These tethers were frequently nonnative (i.e., not i --> i + 4 spacing) and provided a scaffold for other residues, thereby limiting the conformational search. The helical structure then propagated on both sides of the tether. Segments with low stability and propensity formed later in the folding process and utilized contacts with other portions of the protein when folding. These helices formed via a tertiary contact-assisted mechanism, primarily via hydrophobic contacts between residues distant in sequence. Thus, segments with different helical propensities appear to play different roles during protein folding. Furthermore, the active role of nonlocal side chains in helix formation highlights why we must move beyond simple hierarchical models of protein folding.     

Local and nonlocal interactions in globular proteins and mechanisms of alcohol denaturation.

How important are helical propensities in determining the conformations of globular proteins? Using the two-dimensional lattice model and two monomer types, H (hydrophobic) and P (polar), we explore both nonlocal interactions, through an HH contact energy, as developed in earlier work, and local interactions, through a helix energy, σ. By computer enumeration, the partition functions for short chains are obtained without approximation for the full range of both types of energy. When nonlocal interactions dominate, some sequences undergo coil-globule collapse to a unique native structure. When local interactions dominate, all sequences undergo helix–coil transitions. For two different conformational properties, the closest correspondence between the lattice model and proteins in the Protein Data Bank is obtained if the model local interactions are made small compared to the HH contact interaction, suggesting that helical propensities may be only weak determinants of globular protein structures in water. For some HP sequences, varying σ/ leads to additional sharp transitions (sometimes several) and to “conformational switching” between unique conformations. This behavior resembles the transitions of globular proteins in water to helical states in alcohols. In particular, comparison with experiments shows that whereas urea as a denaturant is best modeled as weakening both local and nonlocal interactions, trifluoroethanol is best modeled as mainly weakening HH interactions and slightly enhancing local helical interactions.     

Folding protein alpha-carbon chains into compact forms by Monte Carlo methods.

A method is presented for generating folded chains of specific amino acid sequences on a simple cubic lattice. Monte Carlo simulations are used to transform extended geometries of simplified alpha-carbon chains for eight small monomeric globular proteins into folded states. Permitted chain transitions are limited to a few types of moves, all restricted to occur on the lattice. Crude residue-residue potentials derived from statistical structure data are used to describe the energies for each conformer. The low resolution structures obtained by this procedure contain many of the correct gross features of the native folded architectures with respect to average residue energy per nonbonded contact, segment density, and location of surface loops and disulfide pairs. Rms deviations between these and the native X-ray structures and percentage of native long-range contacts found in these final folded structures are 7.6 +/- 0.7 A and 48 +/- 3%, respectively. This procedure can be useful for predicting approximate tertiary interactions from amino acid sequence.     

Enhanced hybrid search algorithm for protein structure prediction using the 3D-HP lattice model

The problem of protein structure prediction in the hydrophobic-polar (HP) lattice model is the prediction of protein tertiary structure. This problem is usually referred to as the protein folding problem. This paper presents a method for the application of an enhanced hybrid search algorithm to the problem of protein folding prediction, using the three dimensional (3D) HP lattice model. The enhanced hybrid search algorithm is a combination of the particle swarm optimizer (PSO) and tabu search (TS) algorithms. Since the PSO algorithm entraps local minimum in later evolution extremely easily, we combined PSO with the TS algorithm, which has properties of global optimization. Since the technologies of crossover and mutation are applied many times to PSO and TS algorithms, so enhanced hybrid search algorithm is called the MCMPSO-TS (multiple crossover and mutation PSO-TS) algorithm. Experimental results show that the MCMPSO-TS algorithm can find the best solutions so far for the listed benchmarks, which will help comparison with any future paper approach. Moreover, real protein sequences and Fibonacci sequences are verified in the 3D HP lattice model for the first time. Compared with the previous evolutionary algorithms, the new hybrid search algorithm is novel, and can be used effectively to predict 3D protein folding structure. With continuous development and changes in amino acids sequences, the new algorithm will also make a contribution to the study of new protein sequences.     

Refined Genetic Algorithm Simulations to Model Proteins

The advent of completely sequenced genomes is leading to an unprecedented growth of sequence information while adequate structure information is often lacking. Genetic algorithm simulations have been refined and applied as a helpful tool for this question. Modified strategies are tested first on simple lattice protein models. This includes consideration of entropy (protein adjacent water shell) and improved search strategies (pioneer search +14%, systematic recombination +50% in search efficiency). Next, extension to grid free simulations of proteins in full main chain representation is examined. Our protein main chain simulations are further refined by independent criteria such as fitness per residue to judge predicted structures obtained at the end of a simulation. Protein families and protein interactions predicted from the complete H. pylori genomic sequence demonstrate how the full main chain simulations are then applied to model new protein sequences and protein families apparent from genome analysis.     

Elucidation of interactions regulating conformational stability and dynamics of SARS-CoV-2 S-protein

The ongoing COVID-19 pandemic caused by the new coronavirus, SARS-CoV-2, calls for urgent developments of vaccines and antiviral drugs. The spike protein of SARS-CoV-2 (S-protein), which consists of trimeric polypeptide chains with glycosylated residues on the surface, triggers the virus entry into a host cell. Extensive structural and functional studies on this protein have rapidly advanced our understanding of the S-protein structure at atomic resolutions, although most of these structural studies overlook the effect of glycans attached to the S-protein on the conformational stability and functional motions between the inactive down and active up forms. Here, we performed all-atom molecular dynamics simulations of both down and up forms of a fully glycosylated S-protein in solution as well as targeted molecular dynamics simulations between them to elucidate key interdomain interactions for stabilizing each form and inducing the large-scale conformational transitions. The residue-level interaction analysis of the simulation trajectories detects distinct amino acid residues and N-glycans as determinants on conformational stability of each form. During the conformational transitions between them, interdomain interactions mediated by glycosylated residues are switched to play key roles on the stabilization of another form. Electrostatic interactions, as well as hydrogen bonds between the three receptor binding domains, work as driving forces to initiate the conformational transitions toward the active form. This study sheds light on the mechanisms underlying conformational stability and functional motions of the S-protein, which are relevant for vaccine and antiviral drug developments.     

Ideal architecture of residue packing and its observation in protein structures.

A simple model of sphere packing has been investigated as an ideal model for long-range interactions for the packing of non-bonded residues in protein structures. By superposing all residues, the geometry of packing around a central residue is investigated. It is found that all residues conform almost perfectly to this lattice model for sphere packing when a radius of 6.5 A is used to define non-bonded (virtual) interacting residues. Side-chain positions with respect to sequential backbone segments are relatively regular as well. This lattice can readily be used in conformation simulations to reduce the conformational space.     

Solvent entropy-driven searching for protein modeling examined and tested in simplified models.

Solvent entropy is a force to consider in protein folding and protein design but is difficult to model. It is investigated here in the context of the hp model: Two types of residues, hydrophobic and hydrophilic, are modeled on a lattice. Nine chains and two- and three-dimensional simulations are compared. We show that considering solvent entropy alone, efficient folding of lattice chains (identification of the native fold) can be achieved by an entropy-driven simulation on its own. Moreover, in a detailed comparison over a wide range of parameters, entropy-guided searching outperforms an energy-driven search in the model. The combination of energy- and entropy-driven search yields the most efficient searching. It is compared in detail with the above results, indicating also how this solvent shell model may advantageously be implemented in more complex protein modeling simulations.     

LMProt: an efficient algorithm for Monte Carlo sampling of protein conformational space

A new and efficient Monte Carlo algorithm for sampling protein configurations in the continuous space is presented; the efficiency of this algorithm, named Local Moves for Proteins (LMProt), was compared to other alternative algorithms. For this purpose, we used an intrachain interaction energy function that is proportional to the root mean square deviation (rmsd) with respect to alpha-carbons from native structures of real proteins. For phantom chains, the LMProt method is approximately 10(4) and 20 times faster than the algorithms Thrashing (no local moves) and Sevenfold Way (local moves), respectively. Additionally, the LMProt was tested for real chains (excluded-volume all-atoms model); proteins 5NLL (138 residues) and 1BFF (129 residues) were used to determine the folding success xi as a function of the number eta of residues involved in the chain movements, and as a function of the maximum amplitude of atomic displacement delta r(max). Our results indicate that multiple local moves associated with relative chain flexibility, controlled by appropriate adjustments for eta and delta r(max), are essential for configurational search efficiency.     

A theoretical model for the association of amphiphilic transmembrane peptides in lipid bilayers

A theoretical model is proposed for the association of trans-bilayer peptides in lipid bilayers. The model is based on a lattice model for the pure lipid bilayer, which accounts accurately for the most important conformational states of the lipids and their mutual interactions and statistics. Within the lattice formulation the bilayer is formed by two independent monolayers, each represented by a triangular lattice, on which sites the lipid chains are arrayed. The peptides are represented by regular objects, with no internal flexibility, and with a projected area on the bilayer plane corresponding to a hexagon with seven lattice sites. In addition, it is assumed that each peptide surface at the interface with the lipid chains is partially hydrophilic, and therefore interacts with the surrounding lipid matrix via selective anisotropic forces. The peptides would therefore assemble in order to shield their hydrophilic residues from the hydrophobic surroundings. The model describes the self-association of peptides in lipid bilayers via lateral and rotational diffusion, anisotropic lipid-peptide interactions, and peptide-peptide interactions involving the peptide hydrophilic regions. The intent of this model study is to analyse the conditions under which the association of trans-bilayer and partially hydrophilic peptides (or their dispersion in the lipid matrix) is lipid-mediated, and to what extent it is induced by direct interactions between the hydrophilic regions of the peptides. The model properties are calculated by a Monte Carlo computer simulation technique within the canonical ensemble. The results from the model study indicate that direct interactions between the hydrophilic regions of the peptides are necessary to induce peptide association in the lipid bilayer in the fluid phase. Furthermore, peptides within each aggregate are oriented in such a way as to shield their hydrophilic regions from the hydrophobic environment. The average number of peptides present in the aggregates formed depends on the degree of mismatch between the peptide hydrophobic length and the lipid bilayer hydrophobic thickness: The lower the degree of mismatch is the higher this number is. Received: 30 December 1996 / Accepted: 9 May 1997     

TOUCHSTONE II: a new approach to ab initio protein structure prediction

We have developed a new combined approach for ab initio protein structure prediction. The protein conformation is described as a lattice chain connecting C(alpha) atoms, with attached C(beta) atoms and side-chain centers of mass. The model force field includes various short-range and long-range knowledge-based potentials derived from a statistical analysis of the regularities of protein structures. The combination of these energy terms is optimized through the maximization of correlation for 30 x 60,000 decoys between the root mean square deviation (RMSD) to native and energies, as well as the energy gap between native and the decoy ensemble. To accelerate the conformational search, a newly developed parallel hyperbolic sampling algorithm with a composite movement set is used in the Monte Carlo simulation processes. We exploit this strategy to successfully fold 41/100 small proteins (36 approximately 120 residues) with predicted structures having a RMSD from native below 6.5 A in the top five cluster centroids. To fold larger-size proteins as well as to improve the folding yield of small proteins, we incorporate into the basic force field side-chain contact predictions from our threading program PROSPECTOR where homologous proteins were excluded from the data base. With these threading-based restraints, the program can fold 83/125 test proteins (36 approximately 174 residues) with structures having a RMSD to native below 6.5 A in the top five cluster centroids. This shows the significant improvement of folding by using predicted tertiary restraints, especially when the accuracy of side-chain contact prediction is >20%. For native fold selection, we introduce quantities dependent on the cluster density and the combination of energy and free energy, which show a higher discriminative power to select the native structure than the previously used cluster energy or cluster size, and which can be used in native structure identification in blind simulations. These procedures are readily automated and are being implemented on a genomic scale.     

Reconstruction of protein backbones from the BriX collection of canonical protein fragments

As modeling of changes in backbone conformation still lacks a computationally efficient solution, we developed a discretisation of the conformational states accessible to the protein backbone similar to the successful rotamer approach in side chains. The BriX fragment database, consisting of fragments from 4 to 14 residues long, was realized through identification of recurrent backbone fragments from a non-redundant set of high-resolution protein structures. BriX contains an alphabet of more than 1,000 frequently observed conformations per peptide length for 6 different variation levels. Analysis of the performance of BriX revealed an average structural coverage of protein structures of more than 99% within a root mean square distance (RMSD) of 1 Angstrom. Globally, we are able to reconstruct protein structures with an average accuracy of 0.48 Angstrom RMSD. As expected, regular structures are well covered, but, interestingly, many loop regions that appear irregular at first glance are also found to form a recurrent structural motif, albeit with lower frequency of occurrence than regular secondary structures. Larger loop regions could be completely reconstructed from smaller recurrent elements, between 4 and 8 residues long. Finally, we observed that a significant amount of short sequences tend to display strong structural ambiguity between alpha helix and extended conformations. When the sequence length increases, this so-called sequence plasticity is no longer observed, illustrating the context dependency of polypeptide structures.     

Monte carlo study of the percolation in two-dimensional polymer systems

The structure of a two-dimensional film formed by adsorbed polymer chains was studied by means of Monte Carlo simulations. The polymer chains were represented by linear sequences of lattice beads and positions of these beads were restricted to vertices of a two-dimensional square lattice. Two different Monte Carlo methods were employed to determine the properties of the model system. The first was the random sequential adsorption (RSA) and the second one was based on Monte Carlo simulations with a Verdier-Stockmayer sampling algorithm. The methodology concerning the determination of the percolation thresholds for an infinite chain system was discussed. The influence of the chain length on both thresholds was presented and discussed. It was shown that the RSA method gave considerably lower thresholds for longer chains. This behavior can be explained by a different pool of chain conformations used in the calculations in both methods under consideration.