Simulated evolution of protein-protein interaction networks with realistic topology

We model the evolution of eukaryotic protein-protein interaction (PPI) networks. In our model, PPI networks evolve by two known biological mechanisms: (1) Gene duplication, which is followed by rapid diversification of duplicate interactions. (2) Neofunctionalization, in which a mutation leads to a new interaction with some other protein. Since many interactions are due to simple surface compatibility, we hypothesize there is an increased likelihood of interacting with other proteins in the target protein's neighborhood. We find good agreement of the model on 10 different network properties compared to high-confidence experimental PPI networks in yeast, fruit flies, and humans. Key findings are: (1) PPI networks evolve modular structures, with no need to invoke particular selection pressures. (2) Proteins in cells have on average about 6 degrees of separation, similar to some social networks, such as human-communication and actor networks. (3) Unlike social networks, which have a shrinking diameter (degree of maximum separation) over time, PPI networks are predicted to grow in diameter. (4) The model indicates that evolutionarily old proteins should have higher connectivities and be more centrally embedded in their networks. This suggests a way in which present-day proteomics data could provide insights into biological evolution.     

Protein flexibility predictions using graph theory

Techniques from graph theory are applied to analyze the bond networks in proteins and identify the flexible and rigid regions. The bond network consists of distance constraints defined by the covalent and hydrogen bonds and salt bridges in the protein, identified by geometric and energetic criteria. We use an algorithm that counts the degrees of freedom within this constraint network and that identifies all the rigid and flexible substructures in the protein, including overconstrained regions (with more crosslinking bonds than are needed to rigidify the region) and underconstrained or flexible regions, in which dihedral bond rotations can occur. The number of extra constraints or remaining degrees of bond-rotational freedom within a substructure quantifies its relative rigidity/flexibility and provides a flexibility index for each bond in the structure. This novel computational procedure, first used in the analysis of glassy materials, is approximately a million times faster than molecular dynamics simulations and captures the essential conformational flexibility of the protein main and side-chains from analysis of a single, static three-dimensional structure. This approach is demonstrated by comparison with experimental measures of flexibility for three proteins in which hinge and loop motion are essential for biological function: HIV protease, adenylate kinase, and dihydrofolate reductase.     

Protein interaction networks in medicine and disease

The physical interaction of proteins is subject to intense investigation that has revealed that proteins are assembled into large densely connected networks. In this review, we will examine how signaling pathways can be combined to form higher order protein interaction networks. By using network graph theory, these interaction networks can be further analyzed for global organization, which has revealed unique aspects of the relationships between protein networks and complex biological phenotypes. Moreover, several studies have shown that the structure and dynamics of protein networks are disturbed in complex diseases such as cancer progression. These relationships suggest a novel paradigm for treatment of complex multigenic disease where the protein interaction network is the target of therapy more so than individual molecules within the network.     

基于知识的蛋白质结构预测

介绍了近几年基于知识的蛋白质三维结构预测方法及其进展．目前,基于知识的结构预测方法主要有两类,一类是同源蛋白模建,这种技术比较成熟,模建的结果可靠性比较高,但只适用于同源性比较高的目标序列的模建;另一类方法即蛋白质逆折叠技术,主要包括3D profile方法和基于势函数的方法,给出的是目标蛋白质的空间走向,它主要可用于序列同源性比较低的蛋白质的结构预测．     

Integration of protein motions with molecular networks reveals different mechanisms for permanent and transient interactions

The integration of molecular networks with other types of data, such as changing levels of gene expression or protein-structural features, can provide richer information about interactions than the simple node-and-edge representations commonly used in the network community. For example, the mapping of 3D-structural data onto networks enables classification of proteins into singlish- or multi-interface hubs (depending on whether they have >2 interfaces). Similarly, interactions can be classified as permanent or transient, depending on whether their interface is used by only one or by multiple partners. Here, we incorporate an additional dimension into molecular networks: dynamic conformational changes. We parse the entire PDB structural databank for alternate conformations of proteins and map these onto the protein interaction network, to compile a first version of the Dynamic Structural Interaction Network (DynaSIN). We make this network available as a readily downloadable resource file, and we then use it to address a variety of downstream questions. In particular, we show that multi-interface hubs display a greater degree of conformational change than do singlish-interface ones; thus, they show more plasticity which perhaps enables them to utilize more interfaces for interactions. We also find that transient associations involve smaller conformational changes than permanent ones. Although this may appear counterintuitive, it is understandable in the following framework: as proteins involved in transient interactions shuttle between interchangeable associations, they interact with domains that are similar to each other and so do not require drastic structural changes for their activity. We provide evidence for this hypothesis through showing that interfaces involved in transient interactions bind fewer classes of domains than those in a control set.     

Mammalian vitreous humor contains networks of hyaluronan molecules: electron microscopic analysis using the hyaluronan-binding region (G1) of aggrecan and link protein.

Vitreous humor from human, bovine, and chicken eyes was analyzed by rotary shadowing to characterize further the supramolecular organization of the gel-like matrix which forms this tissue. Extensive filamentous networks, distinct from collagen fibrils, were found in both human and bovine vitreous but not in chicken vitreous. The networks consisted of branching structures of various diameters, due to variable numbers of hyaluronan molecules being laterally associated with each other and apparently giving rise to a three-dimensional lattice. These networks could be decorated in a specific and regular manner by the hyaluronan-binding region called G1 purified from bovine nasal septum cartilage. The extent of decoration of hyaluronan was dependent on the relative concentration of G1. In the presence of an excess of G1 the networks were destabilized giving rise to individual unbranched hyaluronan chains of varying length that were saturated with G1. One or more globular proteins, as yet uncharacterized, were seen interacting with the hyaluronan networks, often at branch points. These proteins may serve to stabilize the three-dimensional structure of the matrix although highly ordered networks were also observed without globular proteins. Link protein, which also binds to hyaluronan, bound to the networks in a fashion clearly distinct from G1. Neither G1 nor link protein bound directly to human or bovine vitreous collagen fibrils. However, link protein did bind extensively to the glycosaminoglycan coat of chicken vitreous collagen fibrils described previously (D. W. Wright, and R. Mayne J. Ultrastruct. Mol. Struct. Res. 100, 224-234, 1988), while G1 did not. Digestion of the chicken vitreous collagen fibrils with Streptomyces hyaluronidase did not result in the removal of the glycosaminoglycan coat of the collagen fibrils nor did it affect the binding of G1 or link protein to the fibrils, indicating that hyaluronan is not a component of this structure. These studies demonstrate that proteins with specific binding properties can be used as probes to investigate the structure of the native vitreous humor gel from several species and suggest that this method potentially can be used for structural studies of other connective tissue matrices.     

How do oncoprotein mutations rewire protein–protein interaction networks?

The acquisition of mutations that activate oncogenes or inactivate tumor suppressors is a primary feature of most cancers. Mutations that directly alter protein sequence and structure drive the development of tumors through aberrant expression and modification of proteins, in many cases directly impacting components of signal transduction pathways and cellular architecture. Cancer-associated mutations may have direct or indirect effects on proteins and their interactions and while the effects of mutations on signaling pathways have been widely studied, how mutations alter underlying protein–protein interaction networks is much less well understood. Systematic mapping of oncoprotein protein interactions using proteomics techniques as well as computational network analyses is revealing how oncoprotein mutations perturb protein–protein interaction networks and drive the cancer phenotype.     

PPRODO: prediction of protein domain boundaries using neural networks

Successful prediction of protein domain boundaries provides valuable information not only for the computational structure prediction of multidomain proteins but also for the experimental structure determination. Since protein sequences of multiple domains may contain much information regarding evolutionary processes such as gene-exon shuffling, this information can be detected by analyzing the position-specific scoring matrix (PSSM) generated by PSI-BLAST. We have presented a method, PPRODO (Prediction of PROtein DOmain boundaries) that predicts domain boundaries of proteins from sequence information by a neural network. The network is trained and tested using the values obtained from the PSSM generated by PSI-BLAST. A 10-fold cross-validation technique is performed to obtain the parameters of neural networks using a nonredundant set of 522 proteins containing 2 contiguous domains. PPRODO provides good and consistent results for the prediction of domain boundaries, with accuracy of about 66% using the +/-20 residue criterion. The PPRODO source code, as well as all data sets used in this work, are available from http://gene.kias.re.kr/ approximately jlee/pprodo/.     

Conserved residue clustering and protein structure prediction

Protein residues that are critical for structure and function are expected to be conserved throughout evolution. Here, we investigate the extent to which these conserved residues are clustered in three-dimensional protein structures. In 92% of the proteins in a data set of 79 proteins, the most conserved positions in multiple sequence alignments are significantly more clustered than randomly selected sets of positions. The comparison to random subsets is not necessarily appropriate, however, because the signal could be the result of differences in the amino acid composition of sets of conserved residues compared to random subsets (hydrophobic residues tend to be close together in the protein core), or differences in sequence separation of the residues in the different sets. In order to overcome these limits, we compare the degree of clustering of the conserved positions on the native structure and on alternative conformations generated by the de novo structure prediction method Rosetta. For 65% of the 79 proteins, the conserved residues are significantly more clustered in the native structure than in the alternative conformations, indicating that the clustering of conserved residues in protein structures goes beyond that expected purely from sequence locality and composition effects. The differences in the spatial distribution of conserved residues can be utilized in de novo protein structure prediction: We find that for 79% of the proteins, selection of the Rosetta generated conformations with the greatest clustering of the conserved residues significantly enriches the fraction of close-to-native structures.     

核磁共振波谱法在蛋白质三维结构解析中的应用

蛋白质特定的三维结构与其生物功能密切相关,因此,研究蛋白质的三维结构有助于揭示其生物功能机制。将核磁共振（NMR）波谱法应用于研究溶液状态下蛋白质的三维结构,能够更加准确地揭示蛋白质结构与生物功能之间的关系。本文综述了NMR解析蛋白质三维结构的理论和技术方法,以及NMR结合其他生物物理手段,并辅以分子建模计算法研究蛋白质三维结构的研究进展和最新方法,为精准解析蛋白质的三维结构提供思路及策略。     

Improving the prediction accuracy of residue solvent accessibility and real-value backbone torsion angles of proteins by guided-learning through a two-layer neural network

This article attempts to increase the prediction accuracy of residue solvent accessibility and real-value backbone torsion angles of proteins through improved learning. Most methods developed for improving the backpropagation algorithm of artificial neural networks are limited to small neural networks. Here, we introduce a guided-learning method suitable for networks of any size. The method employs a part of the weights for guiding and the other part for training and optimization. We demonstrate this technique by predicting residue solvent accessibility and real-value backbone torsion angles of proteins. In this application, the guiding factor is designed to satisfy the intuitive condition that for most residues, the contribution of a residue to the structural properties of another residue is smaller for greater separation in the protein-sequence distance between the two residues. We show that the guided-learning method makes a 2-4% reduction in 10-fold cross-validated mean absolute errors (MAE) for predicting residue solvent accessibility and backbone torsion angles, regardless of the size of database, the number of hidden layers and the size of input windows. This together with introduction of two-layer neural network with a bipolar activation function leads to a new method that has a MAE of 0.11 for residue solvent accessibility, 36 degrees for psi, and 22 degrees for phi. The method is available as a Real-SPINE 3.0 server in http://sparks.informatics.iupui.edu.     

Novel protein folds and their nonsequential structural analogs

Newly determined protein structures are classified to belong to a new fold, if the structures are sufficiently dissimilar from all other so far known protein structures. To analyze structural similarities of proteins, structure alignment tools are used. We demonstrate that the usage of nonsequential structure alignment tools, which neglect the polypeptide chain connectivity, can yield structure alignments with significant similarities between proteins of known three-dimensional structure and newly determined protein structures that possess a new fold. The recently introduced protein structure alignment tool, GANGSTA, is specialized to perform nonsequential alignments with proper assignment of the secondary structure types by focusing on helices and strands only. In the new version, GANGSTA+, the underlying algorithms were completely redesigned, yielding enhanced quality of structure alignments, offering alignment against a larger database of protein structures, and being more efficient. We applied DaliLite, TM-align, and GANGSTA+ on three protein crystal structures considered to be novel folds. Applying GANGSTA+ to these novel folds, we find proteins in the ASTRAL40 database, which possess significant structural similarities, albeit the alignments are nonsequential and in some cases involve secondary structure elements aligned in reverse orientation. A web server is available at http://agknapp.chemie.fu-berlin.de/gplus for pairwise alignment, visualization, and database comparison.     

Predicting protein secondary structure content. A tandem neural network approach.

A priori knowledge of secondary structure content can be of great use in theoretical and experimental determination of protein structure. We present a method that uses two computer-simulated neural networks placed in "tandem" to predict the secondary structure content of water-soluble, globular proteins. The first of the two networks, NET1, predicts a protein's helix and strand content given information about the protein's amino acid composition, molecular weight and heme presence. Because NET1 contained more adjustable parameters (network weights) than learning examples, this network experienced problems with memorization, which is the inability to generalize onto new, never-seen-before examples. To overcome this problem, we designed a second network, NET2, which learned to determine when NET1 was in a state of generalization. Together, these two networks produce prediction errors as low as 5.0% and 5.6% for helix and strand content, respectively, on a set of protein crystal structures bearing little homology to those used in network training. A comparison between three other methods including a multiple linear regression analysis, a non-hidden-node network analysis and a secondary structure assignment analysis reveals that our tandem neural network scheme is, indeed, the best method for predicting secondary structure content. The results of our analysis suggest that the knowledge of sequence information is not necessary for highly accurate predictions of protein secondary structure content.     

The identification of similarities between biological networks: application to the metabolome and interactome

The increasing interest in systems biology has resulted in extensive experimental data describing networks of interactions (or associations) between molecules in metabolism, protein-protein interactions and gene regulation. Comparative analysis of these networks is central to understanding biological systems. We report a novel method (PHUNKEE: Pairing subgrapHs Using NetworK Environment Equivalence) by which similar subgraphs in a pair of networks can be identified. Like other methods, PHUNKEE explicitly considers the graphical form of the data and allows for gaps. However, it is novel in that it includes information about the context of the subgraph within the adjacent network. We also explore a new approach to quantifying the statistical significance of matching subgraphs. We report similar subgraphs in metabolic pathways and in protein-protein interaction networks. The most similar metabolic subgraphs were generally found to occur in processes central to all life, such as purine, pyrimidine and amino acid metabolism. The most similar pairs of subgraphs found in the protein-protein interaction networks of Drosophila melanogaster and Saccharomyces cerevisiae also include central processes such as cell division but, interestingly, also include protein sub-networks involved in pre-mRNA processing. The inclusion of network context information in the comparison of protein interaction networks increased the number of similar subgraphs found consisting of proteins involved in the same functional process. This could have implications for the prediction of protein function.     

Structure-function relationships in a bacterial DING protein

A recombinant DING protein from Pseudomonas fluorescens has been previously shown to have a phosphate-binding site, and to be mitogenic for human cells. Here we report the three-dimensional structure of the protein, confirming a close similarity to the "Venus flytrap" structure seen in other human and bacterial phosphate-binding proteins. Site-directed mutagenesis confirms the role of a key residue involved in phosphate binding, and that the mitogenic activity is not dependent on this property. Deletion of one of the two hinged domains that constitute the Venus flytrap also eliminates phosphate binding whilst enhancing mitogenic activity.     

MetaGO: Predicting Gene Ontology of Non-homologous Proteins Through Low-Resolution Protein Structure Prediction and Protein–Protein Network Mapping

Homology-based transferal remains the major approach to computational protein function annotations, but it becomes increasingly unreliable when the sequence identity between query and template decreases below 30%. We propose a novel pipeline, MetaGO, to deduce Gene Ontology attributes of proteins by combining sequence homology-based annotation with low-resolution structure prediction and comparison, and partner's homology-based protein–protein network mapping. The pipeline was tested on a large-scale set of 1000 non-redundant proteins from the CAFA3 experiment. Under the stringent benchmark conditions where templates with > 30% sequence identity to the query are excluded, MetaGO achieves average F-measures of 0.487, 0.408, and 0.598, for Molecular Function, Biological Process, and Cellular Component, respectively, which are significantly higher than those achieved by other state-of-the-art function annotations methods. Detailed data analysis shows that the major advantage of the MetaGO lies in the new functional homolog detections from partner's homology-based network mapping and structure-based local and global structure alignments, the confidence scores of which can be optimally combined through logistic regression. These data demonstrate the power of using a hybrid model incorporating protein structure and interaction networks to deduce new functional insights beyond traditional sequence homology-based referrals, especially for proteins that lack homologous function templates. The MetaGO pipeline is available at http://zhanglab.ccmb.med.umich.edu/MetaGO/.     

Neuro-fuzzy structural classification of proteins for improved protein secondary structure prediction

Fourier transform infrared (FTIR) spectroscopy is a very flexible technique for characterization of protein secondary structure. Measurements can be carried out rapidly in a number of different environments based on only small quantities of proteins. For this technique to become more widely used for protein secondary structure characterization, however, further developments in methods to accurately quantify protein secondary structure are necessary. Here we propose a structural classification of proteins (SCOP) class specialized neural networks architecture combining an adaptive neuro-fuzzy inference system (ANFIS) with SCOP class specialized backpropagation neural networks for improved protein secondary structure prediction. Our study shows that proteins can be accurately classified into two main classes "all alpha proteins" and "all beta proteins" merely based on the amide I band maximum position of their FTIR spectra. ANFIS is employed to perform the classification task to demonstrate the potential of this architecture with moderately complex problems. Based on studies using a reference set of 17 proteins and an evaluation set of 4 proteins, improved predictions were achieved compared to a conventional neural network approach, where structure specialized neural networks are trained based on protein spectra of both "all alpha" and "all beta" proteins. The standard errors of prediction (SEPs) in % structure were improved by 4.05% for helix structure, by 5.91% for sheet structure, by 2.68% for turn structure, and by 2.15% for bend structure. For other structure, an increase of SEP by 2.43% was observed. Those results were confirmed by a "leave-one-out" run with the combined set of 21 FTIR spectra of proteins.