Protein Interactions

The following sections are included:

• Dedication
• Contents
• Preface
• Acknowledgments
• About the Editor

Proteins interact with other proteins to perform vital cellular functions. Understanding various features of protein–protein interfaces is important for their prediction as well as designing inhibitors/activators of protein–protein interactions (PPIs). In this article, the structural and dynamical characteristics of PPIs and their evolutionary conservation have been reviewed. The extent of partner retention and similarity in the quaternary structure and quaternary states of protein–protein complexes (PPCs) depends on the evolutionary divergence among related proteins. Similar interacting modes are observed in protein complexes if the related proteins share high sequence identity, and at low sequence identities, interface residues may differ. Presence of additional structural elements also brings diversity among the evolutionarily related PPCs. These observations suggest caution when predicting interfaces based on homologous complexes with low sequence identities. Binding of two proteins can also lead to change in structure and/or dynamics at the interfaces as well as distant sites in PPCs. These changes often have functional relevance and regulate protein (dis)assembly through allosteric communication. Various examples have been discussed in this article to understand how sequence identities between evolutionarily conserved PPCs affect their interfacial characteristics and how the binding of partner proteins affects their structure and dynamics.

Knowledge of protein–protein interactions (PPIs) facilitates annotation of protein functions and drug development efforts. Computational prediction of PPIs is motivated by a growing gap between the number of known and the number of functionally annotated protein sequences. This chapter focuses on sequence-based predictors of protein-binding residues (PBRs). These methods rely on predictive models that were developed from training data where the native annotations of PBRs were collected either from structures of protein–protein complexes (structure-annotated predictors) or from the intrinsically disordered/unstructured regions (disorder-annotated predictors). This is the first overview that considers both groups of methods. A comprehensive set of 36 predictors including 19 structure-annotated and 17 disorder-annotated tools have been surveyed. The fact that six new methods were published in 2018 alone suggests that this is still an active research area. Their availability, impact, predictive architectures, and outputs have been discussed. These predictors rely on different combinations of five main types of inputs and typically utilize machine learning-derived predictive models. Methods with Web servers are the most highly cited, with the average citation count of 104, compared to the overall average of 74. The lack of solutions that combine structure-annotated and disorder-annotated data to accurately predict PBRs in both structured and disordered regions has been noted.

Most cellular functions are mediated by the interaction of proteins to form dimeric or multimeric complexes as functional elements. Protein–protein complex structure can be predicted by computational protein–protein docking methods. A variety of docking methodologies are available and can be accessed as Web servers or stand-alone applications. The chapter is intended to give an overview on available approaches, on the refinement and scoring of predicted structures and discusses recent progress and future challenges of the protein–protein docking field.

Protein–protein interactions (PPIs) are important for several cellular processes. The principles governing the recognition of protein–protein complexes are investigated using the information on binding-site residues at the interface, specific non-covalent interactions as well as binding affinity. Experimentally, binding affinity of PPIs is studied with isothermal titration calorimetry, surface plasmon resonance and fluorescence- based techniques. On the other hand, computational methods have been developed to understand/predict the binding affinity and the effects of mutations. In this chapter, with a brief introduction on the concept of binding affinity, we outline the experimental techniques to explore the binding affinity of protein–protein complexes along with computational analysis and prediction of binding affinity. Further, we illustrate the utilities and applications of PROXiMATE database for binding affinity of protein–protein complexes and their mutants as well as effects of mutations on binding affinity.

Interatomically, protein’s ability to enact careful and tight interactions with its biomolecular partners is pivotal necessitating proper biological function. Protein–protein interactions (PPIs) are paramount in biological processes as manifested by their tenacious-cellular duties. Protein interfaces are the contact regions between proteins. The interface residues are under constant evolutionary restriction than surface residues. Variations in interface residues which alter binding affinity of proteins may lead to pronounce perturbation or absolute annulment of their functions, potentially resulting in diseases. Hence, wealth of investigations on mutational consequences of PPIs has evolved. The availability of both experimental and computational techniques to assess the effects of mutations on protein–protein binding affinities is essential for varieties of biomedical applications, spurring establishment of various mutational databases. Here, handy experimental and computational methods for detecting/predicting consequences of mutations on PPIs have been explored. Also, protocols and features of proteins utilized by these techniques have been elaborated and updates on mutational databases have been provided. Finally, case studies on mutations emerging at PPI interfaces and their involvements in human-related diseases such as cancer have been provided.

Protein–protein interactions (PPIs) are involved in all critical cellular functions, and mutations in PPIs frequently lead to phenotypic changes that might cause a disease. In this chapter, the various effects that could result from mutations in PPIs such as destabilization/stabilization of a complex, appearance/disappearance of posttranslational modifications, switching on/off of signaling pathways and coevolutionary effects have been summarized. Furthermore, the various computational tools that have been developed by the research community for the prediction of these mutational effects based on protein sequence, structure and PPI network analysis have been described. These computational tools help us obtaining better understanding of how PPIs evolve and how various mutations could lead to serious diseases such as cancer and neurodegenerative disease. Ultimately such predictions facilitate discovery of drugs targeting specific PPIs and PPI-regulated networks.

Protein–nucleic acid (NA) interactions are involved in several cellular processes including gene regulation, replication, packaging and control. The functions of protein–NA complexes are studied with the interactions between proteins and NAs using experimental methods and computational techniques. In this chapter, we survey the databases reported in the literature based on structures, binding and affinity. In addition, the contribution of noncovalent interactions to the binding of protein–NA complexes and disorder-to-order transitions (DOTs) upon the complex formation has been reviewed. Further, we discuss various mechanisms proposed to understand the recognition of protein–NA complexes and the importance of molecular dynamics (MD) simulations to gain deep insights. The role of different features to relate the binding affinity of protein–DNA complexes has also been explored. In essence, this comprehensive review provides ample insights to understand the structure–function relationship in protein–NA complexes.

The information on interactions between proteins and nucleic acids in protein–nucleic acid complexes provides deep insights for understanding their functions. Experimentally, several techniques including X-ray crystallography, isothermal titration calorimetry and DNA/RNA footprinting are used to identify nucleic acid binding proteins and their binding sites. On the other hand, computational tools have been developed for annotating nucleic acid-binding proteins to compensate the requirements of manpower, resources and time for experiments. In this chapter, the methods for identifying DNA- or RNA-binding proteins and features used in these algorithms have been reviewed. Different methods for identifying the binding sites in protein–nucleic acid complex structures are outlined along with computational methods for predicting the residues at the interface. We have also listed a set of tools available in the literature for predicting DNA/RNA-binding proteins and their binding sites along with annotating those proteins in genomic sequences.

Interactions between protein and nucleic acid (DNA or RNA) molecules are essential for a variety of cellular processes. A large amount of interaction data generated by experimental technologies have triggered the development of several computational methods. Most computational methods have focused on predicting binding sites in a sequence or on determining whether an interaction exists between a pair of protein and nucleic acid sequences. Regarding the problem of predicting binding sites between proteins and nucleic acids, many existing computational methods are limited to finding nucleic acid binding sites in proteins instead of protein-binding sites in nucleic acids. Compared to the problem of predicting nucleic acid binding sites in proteins, predicting protein-binding sites in nucleic acids is more challenging for several reasons. In this article, the existing computational methods for predicting protein-binding sites in nucleic acids and future directions in predicting protein-binding sites in nucleic acids have been reviewed.

Molecular docking is a traditional in silico method to obtain the structure of multi-molecule complexes, allowing to get a deeper insight into the functions of the biologically important molecules through their interactions in the complex structure. Although there are a number of highly developed docking methods and algorithms, which were exhaustively tested for the docking of the small molecules into protein, in many cases they cannot be directly applied for docking of oligonucleotides due to the specificity of oligonucleotide molecule construction, flexibility and other aspects. For that reason, a number of original docking algorithms and scoring functions were developed specifically for DNA/RNA docking recently.

Protein–RNA interactions are involved in many important biological processes, and knowing the protein–RNA complex structures can provide significant insights into the mechanisms of these interactions. However, experimentally detecting protein–RNA complex structures is time-consuming and costly. Protein–RNA docking is one of the first steps to predict the three-dimensional (3D) protein–RNA complex structures, which aims to model protein–RNA complexes starting from separate structures. In this review, the latest advances in the methodology development for computational analysis of protein–RNA docking have been discussed. In general, the protein–RNA docking mainly comprises two modules of conformation sampling and scoring function design, and their technical details will be described. Evaluation of the performance of protein–RNA docking requires the benchmark data sets, which have been discussed in this chapter. Finally, the challenges of current protein–RNA docking methods have been briefly discussed, and some future directions for further development in this area have been suggested.

Carbohydrates are considered the molecules of molecular recognition because of their influential roles in interacting with viruses, toxins and cell–cell adhesion phenomena. The recognition phenomena are due to the diversified sequences and three-dimensional structures of carbohydrates and their conformational dynamics, binding site residues of proteins and noncovalent interactions between them. This review provides (1) different databases of carbohydrates, which have vital roles in studying protein–carbohydrate interactions, (2) the role of molecular dynamics (MD) simulations for deducing conformational models of carbohydrates as well as studying the interactions between proteins and carbohydrates and (3) design of inhibitors to avoid pathogenic interactions.

Quantitative structure–activity relationship (QSAR), a 50-year-old technology, has gained prominence till date due to its wide applicability in medicinal chemistry, computer-aided drug design (CADD), toxicology, pharmacology and environmental risk assessment. It is one of the major computational tools employed in medicinal chemistry for the development of reliable statistical prediction models that aid drug discovery. This is mainly by depicting the relationships between physicochemical properties of chemical compounds and their biological activities to predict the activities of new chemical structures. Despite many success stories of the technique, challenges run in parallel pertaining to its reliability and applicability. This chapter is devoted to outline the principles, methodology and lessons learned from theory and application of QSAR in identifying potent drug candidates.

Radical new opportunities for drug discovery are due to the revolution in biology over the past four decades. Major drug targets in the form of molecular components of disease processes have been developed for use in automated assays. After screening natural compounds, chemical databanks or combinatorial libraries to identify the lead compounds, the target macromolecules are used as a starting point for structure-based approaches. To study complex biological and chemical systems, pharmaceutical research incorporates molecular modeling methods and variety of drug discovery programs. For the identification and development of novel promising compounds, the combination of computational and experimental strategies is of great value. In the field of modern drug design, molecular docking methods explore the various ligand conformations adopted within the binding sites of macromolecular targets and select the best conformation. As various docking algorithms are available, an understanding of the advantages and limitations of each method is of great importance. The characterization of protein–ligand binding sites is essential for investigating new functional roles related to major biological research areas like health, food and energy. As the laboratory techniques for drug discovery are very much time-consuming and expensive, computational methods have been developed to assist drug discovery and development, and a number of automated methods have emerged to identify the promising drug candidates.

An understanding of the rules of receptor interactions with ligand molecules is of utmost importance in the area of computer-aided drug discovery (CADD). Current docking algorithms endeavor toward prediction of biologically relevant ligands, which can bind to a specific cavity/active site of biomolecule(s)/receptor(s) inducing the required upregulation or downregulation. These algorithms aim to predict favorable orientation and conformation of a ligand (small molecule) when bound to a target receptor (protein/DNA) to make a stable complex. The effectiveness of such algorithms largely depends on the adopted mathematical model of scoring, which predicts the binding free energy between the receptor and the ligand. The quality of the available force fields and the extent of conformational sampling make binding free energy estimations challenging. However, incorporation of other approaches, such as machine learning, newer force fields and increased exploration in conformational search space, has made current generation scoring functions more promising. This chapter illustrates the broad classification of various available docking and scoring algorithms, their applications and limitations along with their comparative assessment on PDB-bind core data set of 2018 (same as 2016) release comprising 295 protein–ligand complexes. The advancements in the field of docking and scoring have led to a correlation of ~0.8 with experimental data in generic cases, whereas in specific cases a correlation of over 0.98 has also been reported.

The following section is included:

• Index