A COMPARATIVE STUDY OF PROTEIN TERTIARY STRUCTURE PREDICTION METHODS

Proteins are essential parts of our life and participate in virtually every process within a cell. The understanding of protein structures is vital to determine the function of a protein. Protein structure prediction (PSP) from amino acid sequence is one of the high focus problems in bioinformatics today. This is due to the fact that the biological function of the protein is determined by its three dimensional structure. Thus, protein structure prediction is a fundamental area of computational biology. Its importance is intensed by large amounts of sequence data coming from PDB (Protein Data Bank) and the fact that experimentally methods such as X-ray crystallography or Nuclear Magnetic Resonance (NMR)which are used to determining protein structures remains very expensive and time consuming. For minimizing the time, computational methods are used for protein folding and structure prediction problem. In this paper results of protein p53 are discussed.

Chapter 1 of this thesis provides an overview of the structure and function of proteins as well as their investigation via NMR spectroscopy. Proteins are one of the most important building blocks of life. They are part of every organism and fulfill a variety of functions including, for example, the catalysis of many chemical reactions, which is the reason various processes in an organism become possible in the first place. Proteins are made up of amino acids, which are linked together via peptide bonds during the protein biosynthesis. The 22 proteinogenic amino acids result in an almost unlimited number of sequential combinations that give proteins their versatility. The sequence of amino acids in a protein is also called the primary structure. From this, locally limited structural motifs such as α-helices and β-sheets are formed, representing the secondary structure. Further intramolecular interactions create the tertiary structure of a protein, which can be seen as the overall three-dimensional structure of the protein. Furthermore, several individual proteins can assemble into a larger complex to form a quaternary structure. There are various methods for investigating the structure of proteins, including nuclear magnetic resonance (NMR) spectroscopy. In NMR spectroscopy, atomic nuclei with a non-zero nuclear spin are polarized by a constant magnetic field and then excited using radio frequency pulses. The emitted resonance frequencies are detected in the form of a free induction decay (FID) and converted into an NMR spectrum by a Fourier transformation. Many NMR experiments are available for researching proteins. By default, a two-dimensional 1H15N-HSQC spectrum is measured, which provides signals for each N-H bond in a protein meaning that one signal is obtained for each N-H in the backbone of the amino acids. Using 3D NMR experiments, these signals can then be assigned to the respective amino acids in the protein sequence. This makes it possible to investigate the dynamics of proteins as well as the formation of complexes with a ligand or another biomacromolecule if they interact. Further 3D NMR experiments also make it possible to assign the side-chains of the protein. With the help of the determined resonance assignments and spatial information from various NOESY spectra, a protein structure can then be calculated using a suitable computer program. Recently, the prediction of protein structures using AI (artificial intelligence)-supported programs has also become more prevalent. In particular, the artificial neural network AlphaFold has been extremely successful in predicting correct structures without any experimental data provided by the user. Chapter 2 of this thesis introduces the research field of small proteins. These are proteins with a length of usually less than 100 amino acids, although definitions in literature vary and upper limits of, for example, 50 or 200 amino acids exist as well. Various other names are also used to describe small proteins, such as μ-proteins or sORF-encoded peptides. The field of small proteins is still relatively new, as they were overlooked in research for many years. One reason for this was the lower limit of 100 codons that was set for the identification of open reading frames (ORFs) in genome projects. This served to exclude numerous biologically irrelevant ORFs from genome annotation. However, this meant that functional ORFs coding for small proteins were also ignored. Another reason was that experimental methods were not designed for the analysis of small proteins, meaning they could often not be detected at all. However, new analytical methods such as ribosome profiling and a generally increased focus on the detection of small proteins have led to their detection in all domains of life. In the research group of Prof. Dr. Schwalbe, small proteins of various prokaryotes were structurally characterized as part of a special priority program by the DFG. NMR spectroscopy was primarily used for this purpose, as it is the method of choice for the investigation of small proteins. The results of this research were published in several articles, two of which are part of this dissertation.

ProteinShop Gives Researchers a Hands-On Tool for Manipulating, Visualizing Protein Structures. The Human Genome Project and other biological research efforts are creating an avalanche of new data about the chemical makeup and genetic codes of living organisms. But in order to make sense of this raw data, researchers need software tools which let them explore and model data in a more intuitive fashion. With this in mind, researchers at Lawrence Berkeley National Laboratory and the University of California, Davis, have developed ProteinShop, a visualization and modeling program which allows researchers to manipulate protein structures with pinpoint control, guided in large part by their own biological and experimental instincts. Biologists have spent the last half century trying to unravel the ''protein folding problem,'' which refers to the way chains of amino acids physically fold themselves into three-dimensional proteins. This final shape, which resembles a crumpled ribbon or piece of origami, is what determines how the protein functions and translates genetic information. Understanding and modeling this geometrically complex formation is no easy matter. ProteinShop takes a given sequence of amino acids and uses visualization guides to help generate predictions about the secondary structures, identifying alpha helices and flat beta strands, and the coil regions that bind them. Once secondary structures are in place, researchers can twist and turn these pre-configurations until they come up with a number of possible tertiary structure conformations. In turn, these are fed into a computationally intensive optimization procedure that tries to find the final, three-dimensional protein structure. Most importantly, ProteinShop allows users to add human knowledge and intuition to the protein structure prediction process, thus bypassing bad configurations that would otherwise be fruitless for optimization. This saves compute cycles and accelerates the entire process, so that more and larger problems can be attempted. Currently, the program designers are working to make ProteinShop more applicable and adaptable to different protein folding methodologies. If users could manipulate structures from a biological point of view, and then put them back in the queue for more optimization, the process of experimentation and discovery in protein research could be greatly enhanced. The group is also investigating the use of stereoscopic rendering and three-dimensional input devices to remove the limitations of a two-dimensional interface. Clearly, protein-folding research will have far-reaching ramifications. It could lead to new insights about diseases ranging from Alzheimer's to Cystic fibrosis, which scientists believe are caused by protein folding gone wrong. A better understanding of protein structures could also lead to the engineering of altogether new proteins, and shed light on how drugs bind proteins to alter their structure and function. Above all, ProteinShop is an important tool that will help scientists unravel one of the most challenging problems that theoretical and computational chemistry has to offer.

Knowledge of the protein three dimensional structure gives the most accurate guide to the significance of individual residues to protein function. Three dimensional structural information is essential for the rational design and interpretation of site directed mutagenesis studies aimed at enhancing stability or catalytic activity of a protein. Unfortunately, the determination of protein three dimensional structures by X-ray crystallography and NMR techniques is often difficult, and as a consequence, there are only around 1000 protein structures known to atomic resolution. In contrast, the rapid development of nucleic acid and protein sequencing techniques has led current protein sequence databanks to contain over 40,000 unique entries. When a newly determined sequence shows similarity to a protein of known three dimensional structure, then a tertiary model may be built for the new protein with a fair degree of confidence (e.g. see Blundell et al., 1987, for review). However, until recently, attempts to predict protein structure from the amino acid sequence have had little success, with three state (α-helix, β-strand and loop) secondary structure prediction accuracies of 55–65%.

Summary form only given. The understanding of protein structures is vital to determine the function of a protein and its interaction with DNA, RNA and enzyme. The information about its conformation can provide essential information for drug design and protein engineering. While there are over a million known protein sequences, only a limited number of protein structures are experimentally determined. Hence, prediction of protein structures from protein sequences using computer programs is an important step to unveil proteins' three dimensional conformation and functions. As a result, prediction of protein structures has profound theoretical and practical influence over biological study. In this talk, we would show how to use machine learning methods with various advanced encoding schemes and classifiers improve the accuracy of protein structure prediction. The explanation of how a decision is made is also important for improving protein structure prediction. The reasonable interpretation is not only useful to guide the "wet experiments", but also the extracted rules are helpful to integrate computational intelligence with symbolic AI systems for advanced deduction. Some preliminary results using SVM and decision tree for rule extraction and prediction interpretation would also be presented.

Editorial overview: New protein production tools for structural biology

Understanding protein structures is crucial for various bioinformatics research, including drug discovery, disease diagnosis, and evolutionary studies. Protein structure classification is a critical aspect of structural biology, where supervised machine learning algorithms classify structures based on data from databases such as Protein Data Bank (PDB). However, the challenge lies in designing numerical embeddings for protein structures without losing essential information. Although some effort has been made in the literature, researchers have not effectively and rigorously combined the structural and sequence-based features for efficient protein classification to the best of our knowledge. To this end, we propose numerical embeddings that extract relevant features for protein sequences fetched from PDB structures from popular datasets such as PDB Bind and STCRDAB. The features are physicochemical properties such as aromaticity, instability index, flexibility, Grand Average of Hydropathy (GRAVY), isoelectric point, charge at pH, secondary structure fracture, molar extinction coefficient, and molecular weight. We also incorporate scaling features for the sliding windows (e.g., k-mers), which include Kyte and Doolittle (KD) hydropathy scale, Eisenberg hydrophobicity scale, Hydrophilicity scale, Flexibility of the amino acids, and Hydropathy scale. Multiple-feature selection aims to improve the accuracy of protein classification models. The results showed that the selected features significantly improved the predictive performance of existing embeddings.

Protein sequence information is believed to embed the hint of their structures. To uncover the nature between protein sequence and their structures, this study motivates to inspect the dynamic interactions between various protein sequence features, and identify the sequential differences between the different protein structures. Protein sequence data from all structural classes in CATH and SCOP, and the structural disordered proteins from DisProt, as well as the structural motifs in PROSITE, are analyzed in this study. Betweenness and closeness centrality measures are employed to capture the topology of the networks constructed from amino acid feature interactions, while statistical tests are further implemented to compare the feature series distributions. Key findings suggest that in all structural classes, the features for Ala and α-helix and bend preference property, Ala and side-chain size, Ala and Gly, as well as Met and Leu attain significant interactions between each other, and the feature for Leu, Val, and Asn are acted as the critical sources of feature interactions, whereas Cys, His, Trp, and Met exhibit weak intra-type interactions with other features. These implicate that these feature interactions may have little impact in coding the structural differences. For the α structures, Glu, Pro and side-chain size, hydrophobicity properties exhibit high importance in feature interactions, whereas Gly, Thr and physical properties such as α-helix and bend preference, extended structural preference, pK-C value and surrounding hydrophobicity for β structures, show special high importance in β structures. Both α and β types of structures show Ser as the common sources of feature interactions, while the mixed α and β structures not only show common characters with the α and β types of structures, but also preferred interactions between Met, Lys and double-bend preference property, and between the sequence arrangements of Cys, His, Met, Tyr and amino acid composition features. The intrinsically disordered proteins (IDPs) present high frequency for the repetition patterns of certain amino acids, while the different structural motifs also show special characters. More sequential differences between the structures can also be identified from K-mers statistics and feature series distributions. The new discoveries reveal the nature of amino acid feature interaction mechanics, and show great importance of these interactions in coding the different types of protein structures. The results can not only contribute to future molecular design for protein-based vaccine or drug, but also enlighten the development for new protein structural classifiers.

Protein structure prediction (PSP) problems are a major biocomputing challenge, owing to its scientific intrinsic that assists researchers to understand the relationship between amino acid sequences and protein structures, and to study the function of proteins. Although computational resources increased substantially over the last decade, a complete solution to PSP problems by computational methods has not yet been obtained. Using only one energy function is insufficient to characterize proteins because of their complexity. Diverse protein energy functions and evolutionary computation algorithms have been extensively studied to assist in the prediction of protein structures in different ways. Such algorithms are able to provide a better protein with less computational resources requirement than deep learning methods. For the first time, this study proposes a many-objective PSP (MaOPSP) problem with four types of objectives to alleviate the impact of imprecise energy functions for predicting protein structures. A many-objective evolutionary algorithm (MaOEA) is utilized to solve MaOPSP. The proposed method is compared with existing methods by examining 34 proteins. An analysis of the objectives demonstrates that our generated conformations are more reasonable than those generated by single/multiobjective optimization methods. Experimental results indicate that solving a PSP problem as an MaOPSP problem with four objectives yields better PSPs, in terms of both accuracy and efficiency. The source code of the proposed method can be found at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://toyamaailab.github.io/sourcedata.html</uri> .

One of the main research problems in Structural Bioinformatics is related to the prediction of three-dimensional structures (3-D) of polypeptides or proteins. The rate at which amino acid sequences are identified is increasing faster than the 3-D protein structure determination by experimental methods. Computational prediction methods have been developed during the last years, but the problem still remains challenging because of the complexity and high dimensionality of a protein conformational search space. In this article we present a hybrid genetic algorithm for the Protein Structure Prediction (PSP) Problem. A genetic algorithm is combined with a structured population, and it is hybridized with a path-relinking procedure that helps the algorithm to scape from local minima. We perform a set of experiments and show that the proposed hybrid genetic algorithm is effective in finding good quality solutions for the PSP Problem.

The development and evaluation of functions for protein energetics is an important part of current research aiming at understanding protein structures and functions. Knowledgebase mean force potentials are derived from statistical analysis of interacting groups in experimentally determined protein structures. Current knowledge-based mean force potentials are based on the inverse Boltzmannpsilas law, which calculate the ratio of the observed probability with respect to the probability of the reference state. In this study, a general probability framework is presented with the aim to develop novel energy scores. A class of empirical probability functions is derived by decomposing the joint probability of backbone dihedral angles and amino acid sequences. The neighboring interactions are modeled by conditional probabilities. Such probability functions are based on the strict probability theory and some suitable suppositions for convenience of computation. Experiments are performed on several well-constructed decoy sets and the results show that the empirical probability functions presented here outperform previous statistical potentials based on dihedral angles. Such probability functions will be helpful for protein structure prediction,model quality evaluation, transcription factors identification and other challenging problems in computational biology.

Understanding a protein’s 3D structure, which is linked to its function, is essential for biological processes, identifying disease mechanisms, and drug development. Traditional experimental techniques like X-ray crystallography, nuclear magnetic resonance (NMR), and cryogenic electron microscopy (Cryo-EM) have been used to determine precise protein structures. However, these methods are resource-intensive, resulting in only a few thousand structures being deciphered every year. This is in stark contrast to the extensive number of sequences in the UniProt database. To this end, numerous studies have investigated the possibility of integrating template-based and template-free structure prediction approaches. The most notable advancement was through the use of a hybrid approach in 2021 when DeepMind’s AlphaFold2 (AF2) introduced a protein structure prediction model with unprecedented accuracy using an end-to-end deep learning algorithm. Despite its widespread impact, AF2 continues to face architectural challenges and struggles with limited available data, preventing it from generating structures consistent with those that result from experimental methods. Recently, several techniques, such as smart sampling methods, and hybrid experimental approaches that enhance Cryo-EM maps with AF2 structures have been developed to address these limitations, surpassing AF2 and traditional methods in many instances. As developments of novel core algorithms allow AF2 to model a wider range of proteins with higher accuracy and hybrid AF2-experimental approaches expedite experimental data collection for AF2, a pure computation based solution to protein structure prediction seems increasingly likely. This review will provide an overview of AF2 and identify promising advancements in protein structure prediction since its introduction.

Covalent labeling mass spectrometry experiments are growing in popularity and provide important information regarding protein structure. Information obtained from these experiments correlates with residue solvent exposure within the protein in solution. However, it is impossible to determine protein structure from covalent labeling data alone. Incorporation of sparse covalent labeling data into the protein structure prediction software Rosetta has been shown to improve protein tertiary structure prediction. Here, covalent labeling techniques were analyzed computationally to provide insight into what labeling data is needed to optimize tertiary protein structure prediction in Rosetta. We have successfully implemented a new scoring functionality that provides improved predictions. We developed two new covalent labeling based score terms that use a "cone"-based neighbor count to quantify the relative solvent exposure of each amino acid. To test our method, we used a set of 20 proteins with structures deposited in the Protein Data Bank. Decoy model sets were generated for each of these 20 proteins, and the normalized covalent labeling score versus RMSD distributions were evaluated. On the basis of these distributions, we have determined an optimal subset of residues to use when performing covalent labeling experiments in order to maximize the structure prediction capabilities of the covalent labeling data. We also investigated how much false negative and false positive data can be tolerated without meaningfully impacting protein structure prediction. Using these new covalent labeling score terms, protein models were rescored and the resulting models improved by 3.9 Å RMSD on average. New models were also generated using Rosetta's AbinitioRelax program under the guidance of covalent labeling information, and improvement in model quality was observed.

This chapter discusses protein structure determinations, which involve experimental and computational methods, and combined applications of both. For many years, X-ray crystallography and fibre diffraction were the only methods for determining the positions of individual atoms in macromolecular structures. A companion appeared in the 1980s, when K. Wüthrich, R.R. Ernst, and their coworkers developed methods for solving protein structures by nuclear magnetic resonance (NMR) spectroscopy. With a third technique, cryo-electron microscopy (cryo-EM), it has been possible to determine structures of large aggregates, including viral capsids, large protein complexes, and intact ribosomes. Cryo-EM of large multiprotein complexes plus X-ray crystal structures of individual components has proved a powerful combination. The chapter also considers protein structure prediction and modelling.

Structural genomics provides an important approach for characterizing and understanding systems biology. As a step toward better integrating protein three-dimensional (3D) structural information in cancer systems biology, we have constructed a Human Cancer Pathway Protein Interaction Network (HCPIN) by analysis of several classical cancer-associated signaling pathways and their physical protein-protein interactions. Many well known cancer-associated proteins play central roles as "hubs" or "bottlenecks" in the HCPIN. At least half of HCPIN proteins are either directly associated with or interact with multiple signaling pathways. Although some 45% of residues in these proteins are in sequence segments that meet criteria sufficient for approximate homology modeling (Basic Local Alignment Search Tool (BLAST) E-value <10(-6)), only approximately 20% of residues in these proteins are structurally covered using high accuracy homology modeling criteria (i.e. BLAST E-value <10(-6) and at least 80% sequence identity) or by actual experimental structures. The HCPIN Website provides a comprehensive description of this biomedically important multipathway network together with experimental and homology models of HCPIN proteins useful for cancer biology research. To complement and enrich cancer systems biology, the Northeast Structural Genomics Consortium is targeting >1000 human proteins and protein domains from the HCPIN for sample production and 3D structure determination. The long range goal of this effort is to provide a comprehensive 3D structure-function database for human cancer-associated proteins and protein complexes in the context of their interaction networks. The network-based target selection (BioNet) approach described here is an example of a general strategy for targeting co-functioning proteins by structural genomics projects.