Computational Structure Prediction Research Articles

Original article Computational structure prediction and analyze active ligand binding site of defense and lytic enzymes of Trichodermaharzianum

Original article Computational structure prediction and analyze active ligand binding site of defense and lytic enzymes of Trichodermaharzianum

Exhaustive Exploration of the Conformational Landscape of Small Cyclic Peptides Using a Robotics Approach.

Small cyclic peptides represent a promising class of therapeutic molecules with unique chemical properties. However, the poor knowledge of their structural characteristics makes their computational design and structure prediction a real challenge. In order to better describe their conformational space, we developed a method, named EGSCyP, for the exhaustive exploration of the energy landscape of small head-to-tail cyclic peptides. The method can be summarized by (i) a global exploration of the conformational space based on a mechanistic representation of the peptide and the use of robotics-based algorithms to deal with the closure constraint and (ii) an all-atom refinement of the obtained conformations. EGSCyP can handle D-form residues and N-methylations. Two strategies for the side-chains placement were implemented and compared. To validate our approach, we applied it to a set of three variants of cyclic RGDFV pentapeptides, including the drug candidate Cilengitide. A comparative analysis was made with respect to replica exchange molecular dynamics simulations in implicit solvent. Its results show that the EGSCyP method provides a very complete characterization of the conformational space of small cyclic pentapeptides.

Homology modeling in drug discovery: Overview, current applications, and future perspectives.

Homology modeling is one of the computational structure prediction methods that are used to determine protein 3D structure from its amino acid sequence. It is considered to be the most accurate of the computational structure prediction methods. It consists of multiple steps that are straightforward and easy to apply. There are many tools and servers that are used for homology modeling. There is no single modeling program or server which is superior in every aspect to others. Since the functionality of the model depends on the quality of the generated protein 3D structure, maximizing the quality of homology modeling is crucial. Homology modeling has many applications in the drug discovery process. Since drugs interact with receptors that consist mainly of proteins, protein 3D structure determination, and thus homology modeling is important in drug discovery. Accordingly, there has been the clarification of protein interactions using 3D structures of proteins that are built with homology modeling. This contributes to the identification of novel drug candidates. Homology modeling plays an important role in making drug discovery faster, easier, cheaper, and more practical. As new modeling methods and combinations are introduced, the scope of its applications widens.

Deep convolutional networks for quality assessment of protein folds.

The computational prediction of a protein structure from its sequence generally relies on a method to assess the quality of protein models. Most assessment methods rank candidate models using heavily engineered structural features, defined as complex functions of the atomic coordinates. However, very few methods have attempted to learn these features directly from the data. We show that deep convolutional networks can be used to predict the ranking of model structures solely on the basis of their raw three-dimensional atomic densities, without any feature tuning. We develop a deep neural network that performs on par with state-of-the-art algorithms from the literature. The network is trained on decoys from the CASP7 to CASP10 datasets and its performance is tested on the CASP11 dataset. Additional testing on decoys from the CASP12, CAMEO and 3DRobot datasets confirms that the network performs consistently well across a variety of protein structures. While the network learns to assess structural decoys globally and does not rely on any predefined features, it can be analyzed to show that it implicitly identifies regions that deviate from the native structure. The code and the datasets are available at https://github.com/lamoureux-lab/3DCNN_MQA. Supplementary data are available at Bioinformatics online.

Knotty: efficient and accurate prediction of complex RNA pseudoknot structures.

The computational prediction of RNA secondary structure by free energy minimization has become an important tool in RNA research. However in practice, energy minimization is mostly limited to pseudoknot-free structures or rather simple pseudoknots, not covering many biologically important structures such as kissing hairpins. Algorithms capable of predicting sufficiently complex pseudoknots (for sequences of length n) used to have extreme complexities, e.g. Pknots has O(n6) time and O(n4) space complexity. The algorithm CCJ dramatically improves the asymptotic run time for predicting complex pseudoknots (handling almost all relevant pseudoknots, while being slightly less general than Pknots), but this came at the cost of large constant factors in space and time, which strongly limited its practical application (∼200 bases already require 256GB space). We present a CCJ-type algorithm, Knotty, that handles the same comprehensive pseudoknot class of structures as CCJ with improved space complexity of Θ(n3+Z)-due to the applied technique of sparsification, the number of 'candidates', Z, appears to grow significantly slower than n4 on our benchmark set (which include pseudoknotted RNAs up to 400nt). In terms of run time over this benchmark, Knotty clearly outperforms Pknots and the original CCJ implementation, CCJ 1.0; Knotty's space consumption fundamentally improves over CCJ 1.0, being on a par with the space-economic Pknots. By comparing to CCJ 2.0, our unsparsified Knotty variant, we demonstrate the isolated effect of sparsification. Moreover, Knotty employs the state-of-the-art energy model of 'HotKnots DP09', which results in superior prediction accuracy over Pknots. Our software is available at https://github.com/HosnaJabbari/Knotty. Supplementary data are available at Bioinformatics online.

RNA secondary structure prediction with pseudoknots: Contribution of algorithm versus energy model.

MotivationRNA is a biopolymer with various applications inside the cell and in biotechnology. Structure of an RNA molecule mainly determines its function and is essential to guide nanostructure design. Since experimental structure determination is time-consuming and expensive, accurate computational prediction of RNA structure is of great importance. Prediction of RNA secondary structure is relatively simpler than its tertiary structure and provides information about its tertiary structure, therefore, RNA secondary structure prediction has received attention in the past decades. Numerous methods with different folding approaches have been developed for RNA secondary structure prediction. While methods for prediction of RNA pseudoknot-free structure (structures with no crossing base pairs) have greatly improved in terms of their accuracy, methods for prediction of RNA pseudoknotted secondary structure (structures with crossing base pairs) still have room for improvement. A long-standing question for improving the prediction accuracy of RNA pseudoknotted secondary structure is whether to focus on the prediction algorithm or the underlying energy model, as there is a trade-off on computational cost of the prediction algorithm versus the generality of the method.ResultsThe aim of this work is to argue when comparing different methods for RNA pseudoknotted structure prediction, the combination of algorithm and energy model should be considered and a method should not be considered superior or inferior to others if they do not use the same scoring model. We demonstrate that while the folding approach is important in structure prediction, it is not the only important factor in prediction accuracy of a given method as the underlying energy model is also as of great value. Therefore we encourage researchers to pay particular attention in comparing methods with different energy models.

Computational Structure Prediction of (4,4)-Connected Copper Paddle-wheel-based MOFs: Influence of Ligand Functionalization on the Topological Preference

The effect of linkers with extended π-system on the topological preference of (4,4)-connected copper paddle-wheel-based metal–organic frameworks (MOFs) was investigated using the reverse topological approach (RTA) in which a genetic algorithm (GA) and the DFT-derived force field MOF-FF were used for ranking and predicting the most stable phase. Three tetracarboxylate linkers bearing different functionality, namely, phenylene (L1), naphthalene (L2), and anthracene (L3) groups, were studied. All potential topologies including nbo-b, ssa, ssb, pts, and lvt-b were considered. The computational results reveal that nbo-b is the most stable topology for all three investigated linkers. However, L2 is also formed in ssb according to experimental findings. Our simulation results show that the CH−π interactions with a Y-shaped configuration between naphthalene moieties of L2 stabilize the ssb framework. Unlike L2, CH−π interactions are not favorable for L1 and L3 because of unsuitable size of the π-system. The resul...

Computational prediction of vaccine potential epitopes and 3-dimensional structure of XAGE-1b for non-small cell lung cancer immunotherapy

BackgroundXAGE-1b is shown to be overexpressed in lung adenocarcinoma and to be a strong immunogenic antigen among non-small cell lung cancer (NSCLC) patients. However, 3D structure of XAGE-1b is not available and its confirmation has not been solved yet. MethodsMultiple sequence alignment was run to select the most reliable templates. Homology modeling technique was performed using computer-based tool to generate 3-dimensional structure models, eight models were generated and assessed on basis of local and global quality. Immune Epitope Database (IEDB) tools were then used to determine potential B-Cell epitopes while NetMHCpan algorithms were used to enhance the determination for potential epitopes of both Cytotoxic T-lymphocytes and T-helper cells. ResultsComputational prediction was performed for B-Cell epitopes, prediction results generated; 3 linear epitopes where XAGE-1b (13-21) possessed the best score of 0.67, 5 discontinuous epitopes where XAGE-1b (40-52) possessed the best score of 0.67 based on the predicted model of the finest quality. For a potential vaccine design, computational prediction yielded potential Human Leukocyte Antigen (HLA) class I epitopes including HLA-B*08:01-restricted XAGE-1b (3-11) epitope which was the best with 0.2 percentile rank. Regarding HLA Class II epitopes, HLA-DRB1*12:01-restricted XAGE-1b (25-33) was the most antigenic epitope with 5.91 IC50 value. IC50 values were compared with experimental values and population coverage percentages of epitopes were computed. ConclusionsThis study predicted a model of XAGE-1b tertiary structure which could explain its antigenic function and facilitate usage of predicted peptides for experimental validation towards designing immunotherapies against NSCLC.

Estimating Energy Parameters for RNA Secondary Structure Predictions Using Both Experimental and Computational Data.

Computational RNA secondary structure prediction depends on a large number of nearest-neighbor free-energy parameters, including 10 parameters for Watson-Crick stacked base pairs that were estimated from experimental measurements of the free energies of 90 RNA duplexes. These experimental data are provided by time-consuming and cost-intensive experiments. In contrast, various modified nucleotides in RNAs, which would affect not only their structures but also functions, have been found, and rapid determination of energy parameters for a such modified nucleotides is needed. To reduce the high cost of determining energy parameters, we propose a novel method to estimate energy parameters from both experimental and computational data, where the computational data are provided by a recently developed molecular dynamics simulation protocol. We evaluate our method for Watson-Crick stacked base pairs, and show that parameters estimated from 10 experimental data items and 10 computational data items can predict RNA secondary structures with accuracy comparable to that using conventional parameters. The results indicate that the combination of experimental free-energy measurements and molecular dynamics simulations is capable of estimating the thermodynamic properties of RNA secondary structures at lower cost.

Discovery of the Tyrobetaine Natural Products and Their Biosynthetic Gene Cluster via Metabologenomics.

Natural products (NPs) are a rich source of medicines, but traditional discovery methods are often unsuccessful due to high rates of rediscovery. Genetic approaches for NP discovery are promising, but progress has been slow due to the difficulty of identifying unique biosynthetic gene clusters (BGCs) and poor gene expression. We previously developed the metabologenomics method, which combines genomic and metabolomic data to discover new NPs and their BGCs. Here, we utilize metabologenomics in combination with molecular networking to discover a novel class of NPs, the tyrobetaines: nonribosomal peptides with an unusual trimethylammonium tyrosine residue. The BGC for this unusual class of compounds was identified using metabologenomics and computational structure prediction data. Heterologous expression confirmed the BGC and suggests an unusual mechanism for trimethylammonium formation. Overall, the discovery of the tyrobetaines shows the great potential of metabologenomics combined with molecular networking and computational structure prediction for identifying interesting biosynthetic reactions and novel NPs.

Statistical modeling of RNA structure profiling experiments enables parsimonious reconstruction of structure landscapes

RNA plays key regulatory roles in diverse cellular processes, where its functionality often derives from folding into and converting between structures. Many RNAs further rely on co-existence of alternative structures, which govern their response to cellular signals. However, characterizing heterogeneous landscapes is difficult, both experimentally and computationally. Recently, structure profiling experiments have emerged as powerful and affordable structure characterization methods, which improve computational structure prediction. To date, efforts have centered on predicting one optimal structure, with much less progress made on multiple-structure prediction. Here, we report a probabilistic modeling approach that predicts a parsimonious set of co-existing structures and estimates their abundances from structure profiling data. We demonstrate robust landscape reconstruction and quantitative insights into structural dynamics by analyzing numerous data sets. This work establishes a framework for data-directed characterization of structure landscapes to aid experimentalists in performing structure-function studies.

Protein Structure Refinement via Molecular Dynamics Simulations

Computational protein structure prediction has long been an important alternative to experimental structure determination. Comparative modeling approaches can often generate approximate models but generating highly accurate models that are comparable to experimental structures has remained a significant challenge. Refinement methods via molecular dynamics simulations are an avenue for bridging this gap. MD-based protocols that provide consistent but still modest improvement of initial models are presented based on results from recent rounds of CASP. Furthermore, new detailed analysis of microsecond simulations combined with Markov state modeling was used to construct energy landscapes that connect initial homology models to native structures. These maps provide insight into the key kinetic barriers and energetics along transition pathways from homology models towards the native state and inform the design of more effective structure refinement protocols that can truly reach the native state from initial comparative models.

Antibody side chain conformations are position-dependent.

Side chain prediction is an integral component of computational antibody design and structure prediction. Current antibody modelling tools use backbone-dependent rotamer libraries with conformations taken from general proteins. Here we present our antibody-specific rotamer library, where rotamers are binned according to their immunogenetics (IMGT) position, rather than their local backbone geometry. We find that for some amino acid types at certain positions, only a restricted number of side chain conformations are ever observed. Using this information, we are able to reduce the breadth of the rotamer sampling space. Based on our rotamer library, we built a side chain predictor, position-dependent antibody rotamer swapper (PEARS). On a blind test set of 95 antibody model structures, PEARS had the highest average χ1 and χ1+2 accuracy (78.7% and 64.8%) compared to three leading backbone-dependent side chain predictors. Our use of IMGT position, rather than backbone ϕ/ψ, meant that PEARS was more robust to errors in the backbone of the model structure. PEARS also achieved the lowest number of side chain-side chain clashes. PEARS is freely available as a web application at http://opig.stats.ox.ac.uk/webapps/pears.

Exploring the effects of sparse restraints on protein structure prediction.

One of the main barriers to accurate computational protein structure prediction is searching the vast space of protein conformations. Distance restraints or inter-residue contacts have been used to reduce this search space, easing the discovery of the correct folded state. It has been suggested that about 1 contact for every 12 residues may be sufficient to predict structure at fold level accuracy. Here, we use coarse-grained structure-based models in conjunction with molecular dynamics simulations to examine this empirical prediction. We generate sparse contact maps for 15 proteins of varying sequence lengths and topologies and find that given perfect secondary-structural information, a small fraction of the native contact map (5%-10%) suffices to fold proteins to their correct native states. We also find that different sparse maps are not equivalent and we make several observations about the type of maps that are successful at such structure prediction. Long range contacts are found to encode more information than shorter range ones, especially for α and αβ-proteins. However, this distinction reduces for β-proteins. Choosing contacts that are a consensus from successful maps gives predictive sparse maps as does choosing contacts that are well spread out over the protein structure. Additionally, the folding of proteins can also be used to choose predictive sparse maps. Overall, we conclude that structure-based models can be used to understand the efficacy of structure-prediction restraints and could, in future, be tuned to include specific force-field interactions, secondary structure errors and noise in the sparse maps.

New Deep Learning Methods for Protein Loop Modeling.

Computational protein structure prediction is a long-standing challenge in bioinformatics. In the process of predicting protein 3D structures, it is common that parts of an experimental structure are missing or parts of a predicted structure need to be remodeled. The process of predicting local protein structures of particular regions is called loop modeling. In this paper, five new loop modeling methods based on machine learning techniques, called NearLooper, ConLooper, ResLooper, HyLooper1, and HyLooper2 are proposed. NearLooper is based on the nearest neighbor technique. ConLooper applies deep convolutional neural networks to predict ${\mathrm{C}}_{{{\alpha }}}$Cα atoms distance matrix as an orientation-independent representation of protein structure. ResLooper uses residual neural networks instead of deep convolutional neural networks. HyLooper1 combines the results of NearLooper and ConLooper while HyLooper2 combines NearLooper and ResLooper. Three commonly used benchmarks for loop modeling are used to compare the performance between these methods and existing state-of-the-art methods. The experiment results show promising performance in which our best method improves existing state-of-the-art methods by 28 and 54 percent of average RMSD on two datasets while being comparable on the other one.

A cyber-linked undergraduate research experience in computational biomolecular structure prediction and design.

Computational biology is an interdisciplinary field, and many computational biology research projects involve distributed teams of scientists. To accomplish their work, these teams must overcome both disciplinary and geographic barriers. Introducing new training paradigms is one way to facilitate research progress in computational biology. Here, we describe a new undergraduate program in biomolecular structure prediction and design in which students conduct research at labs located at geographically-distributed institutions while remaining connected through an online community. This 10-week summer program begins with one week of training on computational biology methods development, transitions to eight weeks of research, and culminates in one week at the Rosetta annual conference. To date, two cohorts of students have participated, tackling research topics including vaccine design, enzyme design, protein-based materials, glycoprotein modeling, crowd-sourced science, RNA processing, hydrogen bond networks, and amyloid formation. Students in the program report outcomes comparable to students who participate in similar in-person programs. These outcomes include the development of a sense of community and increases in their scientific self-efficacy, scientific identity, and science values, all predictors of continuing in a science research career. Furthermore, the program attracted students from diverse backgrounds, which demonstrates the potential of this approach to broaden the participation of young scientists from backgrounds traditionally underrepresented in computational biology.

Three-body interactions improve contact prediction within direct-coupling analysis.

The prediction of residue contacts in a protein solely from sequence information is a promising approach to computational structure prediction. Recent developments use statistical or information theoretic methods to extract contact information from a multiple sequence alignment. Despite good results, accuracy is limited due to usage of two-body interactions within a Potts model. In this paper we generalize this approach and propose a Hamiltonian with an additional three-body interaction term. We derive a mean-field approximation for inference of three-body couplings within a Potts model which is fast enough on modern computers. Finally, we show that our model has a higher accuracy in predicting residue contacts in comparison with the plain two-body-interactionmodel.

An Algorithm for Template-Based Prediction of Secondary Structures of Individual RNA Sequences.

While understanding the structure of RNA molecules is vital for deciphering their functions, determining RNA structures experimentally is exceptionally hard. At the same time, extant approaches to computational RNA structure prediction have limited applicability and reliability. In this paper we provide a method to solve a simpler yet still biologically relevant problem: prediction of secondary RNA structure using structure of different molecules as a template. Our method identifies conserved and unconserved subsequences within an RNA molecule. For conserved subsequences, the template structure is directly transferred into the generated structure and combined with de-novo predicted structure for the unconserved subsequences with low evolutionary conservation. The method also determines, when the generated structure is unreliable. The method is validated using experimentally identified structures. The accuracy of the method exceeds that of classical prediction algorithms and constrained prediction methods. This is demonstrated by comparison using large number of heterogeneous RNAs. The presented method is fast and robust, and useful for various applications requiring knowledge of secondary structures of individual RNA sequences.

LightDock: a new multi-scale approach to protein-protein docking.

Computational prediction of protein-protein complex structure by docking can provide structural and mechanistic insights for protein interactions of biomedical interest. However, current methods struggle with difficult cases, such as those involving flexible proteins, low-affinity complexes or transient interactions. A major challenge is how to efficiently sample the structural and energetic landscape of the association at different resolution levels, given that each scoring function is often highly coupled to a specific type of search method. Thus, new methodologies capable of accommodating multi-scale conformational flexibility and scoring are strongly needed. We describe here a new multi-scale protein-protein docking methodology, LightDock, capable of accommodating conformational flexibility and a variety of scoring functions at different resolution levels. Implicit use of normal modes during the search and atomic/coarse-grained combined scoring functions yielded improved predictive results with respect to state-of-the-art rigid-body docking, especially in flexible cases. The source code of the software and installation instructions are available for download at https://life.bsc.es/pid/lightdock/. juanf@bsc.es. Supplementary data are available at Bioinformatics online.

Computationally-Guided Synthetic Control over Pore Size in Isostructural Porous Organic Cages.

The physical properties of 3-D porous solids are defined by their molecular geometry. Hence, precise control of pore size, pore shape, and pore connectivity are needed to tailor them for specific applications. However, for porous molecular crystals, the modification of pore size by adding pore-blocking groups can also affect crystal packing in an unpredictable way. This precludes strategies adopted for isoreticular metal–organic frameworks, where addition of a small group, such as a methyl group, does not affect the basic framework topology. Here, we narrow the pore size of a cage molecule, CC3, in a systematic way by introducing methyl groups into the cage windows. Computational crystal structure prediction was used to anticipate the packing preferences of two homochiral methylated cages, CC14-R and CC15-R, and to assess the structure–energy landscape of a CC15-R/CC3-S cocrystal, designed such that both component cages could be directed to pack with a 3-D, interconnected pore structure. The experimental gas sorption properties of these three cage systems agree well with physical properties predicted by computational energy–structure–function maps.