Current Protein and Peptide Science - Volume 17, Issue 1, 2016

Network theory has become an excellent method of choice through which biological data are smoothly integrated to gain insights into complex biological problems. Understanding protein structure, folding, and function has been an important problem, which is being extensively investigated by the network approach. Since the sequence uniquely determines the structure, this review focuses on the networks of non-covalently connected amino acid side chains in proteins. Questions in structural biology are addressed within the framework of such a formalism. While general applications are mentioned in this review, challenging problems which have demanded the attention of scientific community for a long time, such as allostery and protein folding, are considered in greater detail. Our aim has been to explore these important problems through the eyes of networks. Various methods of constructing protein structure networks (PSN) are consolidated. They include the methods based on geometry, edges weighted by different schemes, and also bipartite network of protein-nucleic acid complexes. A number of network metrics that elegantly capture the general features as well as specific features related to phenomena, such as allostery and protein model validation, are described. Additionally, an integration of network theory with ensembles of equilibrium structures of a single protein or that of a large number of structures from the data bank has been presented to perceive complex phenomena from network perspective. Finally, we discuss briefly the capabilities, limitations, and the scope for further explorations of protein structure networks.

In this work, we present an extensive analysis of protein contact network topology applied to a wide data set. We extended the concept of degree distribution to graphlets, describing local connectivity patterns. We compared results to those derived from artificial networks of the same size (number of nodes), reproducing the average degree of each protein network. The artificial networks resemble the coiling of immaterial cords and we tried to understand if they could catch the protein structure topology upon the sole constraint of backbone (cord). We found a surprisingly similar pattern for local topological descriptors (graphlets distribution) while real proteins and cords differ at large extent in the global topological invariant average shortest path that presumably catches the systemic nature of protein and the non negligible encumbrance of backbone (residues steric hindrance). We demonstrated average shortest path to link polymer length and physical size of the molecule, and its minimization plays the role of 'target function' of folding process.

Protein homodimers pose some intriguing questions about the relation between structure and stability. We approached the problem by means of a topological methodology based on protein contact networks. We correlated local interface descriptors with structure and energy global properties of the systems under analysis. We demonstrated that the graph energy, formerly applied to the analysis of unconjugated hydrocarbons structures, is the bridge between the topological and energetic description of protein complexes. This is a first step for the generation of a “protein structural formula”, analogous to the molecular graphs in organic chemistry.

Proteins can be conveniently represented as networks of interacting residues, thus allowing the study of several network parameters that can shed light onto several of their structural and functional aspects. With respect to the binding of ligands, which are central for the function of many proteins, network analysis may constitute a possible route to assist the identification of binding sites. As the bulk of this review illustrates, this has generally been easier for enzymes than for non-enzyme proteins, perhaps due to the different topological nature of the binding sites of the former over those of the latter. The article also illustrates how network representations of binding sites can be used to search PDB structures in order to identify proteins that bind similar molecules and, lastly, how codifying proteins as networks can assist the analysis of the conformational changes consequent to ligand binding.

Catalytic residues play a significant role in enzyme functions. With the recent accumulation of experimentally determined enzyme 3D structures and network theory on protein structures, the prediction of catalytic residues by amino acid network (AAN, where nodes are residues and links are residue interactions) has gained much interest. Computational methods of identifying catalytic residues are traditionally divided into two groups: sequence-based and structure-based methods. Two new structure- based methods are proposed in current advances: AAN and Elastic Network Model (ENM) of enzyme structures. By concentrating on AAN-based approach, we herein summarized network properties for predictions of catalytic residues. AAN attributes were showed responsible for performance improvement, and therefore the combination of AAN with previous sequence and structural information will be a promising direction for further improvement. Advantages and limitations of AAN-based methods, future perspectives on the application of AAN to the study of protein structure-function relationships are discussed.

A systematic comparison with the Wild-Type (WT) of one-point mutants of bacteriophage T4 lysozyme was carried out using as difference markers the topological parameters of the protein contact networks corresponding to each crystallographic structure. The investigation concerned changes at the resolution level of single residue along the protein sequence. The results were correlated with (reported) changes in functional properties and (observed) changes in the information provided by the energy dissipation algorithm of the “Turbine” software simulation tool. The critical factor leading to significant difference among mutants and WT is in most cases associated to the sensitivity towards mutation of relatively short windows in the amino acidic sequence not necessarily contiguous to the active site.

Protein structures can be conceptualized as context-aware self-organizing systems. One of its emerging properties is a modular architecture. Such modular architecture has been identified as domains and defined as its units of evolution and function. However, this modular architecture is not exclusively defined by domains. Also, the definition of a domain is an ongoing debate. Here we propose differentiating structural, evolutionary and functional domains as distinct concepts. Defining domains or modules is confounded by diverse definitions of the concept, and also by other elements inherent to protein structures. An apparent hierarchy in protein structure architecture is one of these elements, where lower level interactions may create noise for the definition of higher levels. Diverse modularity-molding factors such as folding, function, and selection, can have a misleading effect when trying to define a given type of module. It is thus important to keep in mind this complexity when defining modularity in protein structures and interpreting the outcome modularity inference approaches.

Analyzing protein-protein interaction (PPI) networks has been a crucial prerequisite for understanding the molecular basis for most of the diseases. Although several investigations have been carried out on PPI network analysis, none of them explicitly considered binding affinity as a criterion for the analysis. In this work, we have performed a systematic analysis of protein-protein interaction networks in five organisms based on the binding affinity of interacting partners. We observed that eukaryotes are marginally dominated with high affinity complexes and an opposite trend was observed in prokaryotes. In addition, hub-hub interactions have the highest percentage of “high affinity” interactions followed by hubnonhub and nonhub-nonhub interactions. Further, all organisms contain hubs, which are enriched specifically with high or low affinity complexes irrespective of the dominance of these interactions. Sub network analysis indicates that the closed triad motifs with high and low affinity complexes are more significant than the open motifs. The analysis of clustering coefficient and amino acid properties showed specific preferences in different organisms. These findings deepen the knowledge of PPI networks and provide useful insights for target identification in drug discovery.

Metabolism is a set of fundamental processes that play important roles in a plethora of biological and medical contexts. It is understood that the topological information of reconstructed metabolic networks, such as modular organization, has crucial implications on biological functions. Recent interpretations of modularity in network settings provide a view of multiple network partitions induced by different resolution parameters. Here we ask the question: How do multiple network partitions affect the organization of metabolic networks? Since network motifs are often interpreted as the super families of evolved units, we further investigate their impact under multiple network partitions and investigate how the distribution of network motifs influences the organization of metabolic networks. We studied Homo sapiens, Saccharomyces cerevisiae and Escherichia coli metabolic networks; we analyzed the relationship between different community structures and motif distribution patterns. Further, we quantified the degree to which motifs participate in the modular organization of metabolic networks.