Yeast Protein-protein Interaction Networks Research Articles

Identifying Protein Complexes by Reducing Noise in Interaction Networks

Identifying protein complexes in protein-protein interaction (PPI) networks is a fundamental problem in computational biology. High-throughput experimental techniques have generated large, experimentally detected PPI datasets. These interactions represent a rich source of data that can be used to detect protein complexes; however, such interactions contain much noise. Therefore, these interactions should be validated before they could be applied to detect protein complexes. We propose an efficient measure to estimate PPI reliability (PPIR) and reduce noise level in two different yeast PPI networks. PPIRU, which is a new protein complex clustering algorithm based on PPIR, is introduced. Experiments demonstrated that interactome graph weighting methods incorporating PPIR clearly improve the results of several clustering algorithms. PPIR also outperforms other PPI graph weighting schemes in most cases. We compare PPIRU with several efficient, existing clustering algorithms and reveal that the accuracy values of PPIRU clusters are much higher than those of other algorithms.

Controllability in protein interaction networks.

Recently, the focus of network research shifted to network controllability, prompting us to determine proteins that are important for the control of the underlying interaction webs. In particular, we determined minimum dominating sets of proteins (MDSets) in human and yeast protein interaction networks. Such groups of proteins were defined as optimized subsets where each non-MDSet protein can be reached by an interaction from an MDSet protein. Notably, we found that MDSet proteins were enriched with essential, cancer-related, and virus-targeted genes. Their central position allowed MDSet proteins to connect protein complexes and to have a higher impact on network resilience than hub proteins. As for their involvement in regulatory functions, MDSet proteins were enriched with transcription factors and protein kinases and were significantly involved in bottleneck interactions, regulatory links, phosphorylation events, and genetic interactions.

Modelling the Yeast Interactome

The topology behind biological interaction networks has been studied for over a decade. Yet, there is no definite agreement on the theoretical models which best describe protein-protein interaction (PPI) networks. Such models are critical to quantifying the significance of any empirical observation regarding those networks. Here, we perform a comprehensive analysis of yeast PPI networks in order to gain insights into their topology and its dependency on interaction-screening technology. We find that: (1) interaction-detection technology has little effect on the topology of PPI networks; (2) topology of these interaction networks differs in organisms with different cellular complexity (human and yeast); (3) clear topological difference is present between PPI networks, their functional sub-modules, and their inter-functional “linkers”; (4) high confidence PPI networks have more “geometrical” topology compared to predicted, incomplete, or noisy PPI networks; and (5) inter-functional “linker” proteins serve as mediators in signal transduction, transport, regulation and organisational cellular processes.

Predicting Essential Proteins Based on Weighted Degree Centrality.

Essential proteins are vital for an organism's viability under a variety of conditions. There are many experimental and computational methods developed to identify essential proteins. Computational prediction of essential proteins based on the global protein-protein interaction (PPI) network is severely restricted because of the insufficiency of the PPI data, but fortunately the gene expression profiles help to make up the deficiency. In this work, Pearson correlation coefficient (PCC) is used to bridge the gap between PPI and gene expression data. Based on PCC and edge clustering coefficient (ECC), a new centrality measure, i.e., the weighted degree centrality (WDC), is developed to achieve the reliable prediction of essential proteins. WDC is employed to identify essential proteins in the yeast PPI and e-Coli networks in order to estimate its performance. For comparison, other prediction technologies are also performed to identify essential proteins. Some evaluation methods are used to analyze the results from various prediction approaches. The prediction results and comparative analyses are shown in the paper. Furthermore, the parameter λ in the method WDC will be analyzed in detail and an optimal λ value will be found. Based on the optimal λ value, the differentiation of WDC and another prediction method PeC is discussed. The analyses prove that WDC outperforms other methods including DC, BC, CC, SC, EC, IC, NC, and PeC. At the same time, the analyses also mean that it is an effective way to predict essential proteins by means of integrating different data sources.

Identification and Evolution of Structurally Dominant Nodes in Protein-Protein Interaction Networks

It is well known that protein-protein interaction (PPI) networks are typical evolving complex networks. Identification of important nodes has been an emerging popular topic in complex networks. Many indexes have been proposed to measure the importance of nodes in complex networks, such as degree, closeness, betweenness, k-shell, clustering coefficient, semi-local centrality, eigenvector centrality. Based on multivariate statistical analysis, through integrating the above indexes and further considering the appearances of nodes in network motifs, this paper aims at developing a new measure to characterize the structurally dominant proteins (SDP) in PPI networks. Moreover, we will further investigate the evolution of the defined dominant nodes in temporal evolving real-world and artificial PPI networks. Our results indicate that the constructed artificial networks have some similar statistical properties as those of the real-world evolving networks. In this case, the artificial PPI networks can be used to further investigate the above evolution characteristics of the real-world evolving networks. Simulation results reveal that SDP in the yeast PPI networks are evolutionary conserved, however, the undominant nodes evolve rapidly. Furthermore, PPI networks are very robust against random mutations, while fragile yet with certain robustness to targeted mutations on SDP. Our investigations shed some light on the future applications of the evolving characteristics of bio-molecular networks, such as reengineering of particular networks for technological, synthetic or pharmacological purposes.

Joint clustering of protein interaction networks through Markov random walk

Biological networks obtained by high-throughput profiling or human curation are typically noisy. For functional module identification, single network clustering algorithms may not yield accurate and robust results. In order to borrow information across multiple sources to alleviate such problems due to data quality, we propose a new joint network clustering algorithm ASModel in this paper. We construct an integrated network to combine network topological information based on protein-protein interaction (PPI) datasets and homological information introduced by constituent similarity between proteins across networks. A novel random walk strategy on the integrated network is developed for joint network clustering and an optimization problem is formulated by searching for low conductance sets defined on the derived transition matrix of the random walk, which fuses both topology and homology information. The optimization problem of joint clustering is solved by a derived spectral clustering algorithm. Network clustering using several state-of-the-art algorithms has been implemented to both PPI networks within the same species (two yeast PPI networks and two human PPI networks) and those from different species (a yeast PPI network and a human PPI network). Experimental results demonstrate that ASModel outperforms the existing single network clustering algorithms as well as another recent joint clustering algorithm in terms of complex prediction and Gene Ontology (GO) enrichment analysis.

The centrality of cancer proteins in human protein-protein interaction network: a revisit.

Topological analysis of protein-protein interaction (PPI) networks has been widely applied to the investigation on cancer mechanisms. However, there is still a debate on whether cancer proteins exhibit more topological centrality compared to the other proteins in the human PPI network. To resolve this debate, we first identified four sets of human proteins, and then mapped these proteins into the yeast PPI network by homologous genes. Finally, we compared these proteins' properties in human and yeast PPI networks. Experiments over two real datasets demonstrated that cancer proteins tend to have higher degree and smaller clustering coefficient than non-cancer proteins. Experimental results also validated that cancer proteins have larger betweenness centrality compared to the other proteins on the STRING dataset. However, on the BioGRID dataset, the average betweenness centrality of cancer proteins is larger than that of disease and control proteins, but smaller than that of essential proteins.

A least square method based model for identifying protein complexes in protein-protein interaction network.

Protein complex formed by a group of physical interacting proteins plays a crucial role in cell activities. Great effort has been made to computationally identify protein complexes from protein-protein interaction (PPI) network. However, the accuracy of the prediction is still far from being satisfactory, because the topological structures of protein complexes in the PPI network are too complicated. This paper proposes a novel optimization framework to detect complexes from PPI network, named PLSMC. The method is on the basis of the fact that if two proteins are in a common complex, they are likely to be interacting. PLSMC employs this relation to determine complexes by a penalized least squares method. PLSMC is applied to several public yeast PPI networks, and compared with several state-of-the-art methods. The results indicate that PLSMC outperforms other methods. In particular, complexes predicted by PLSMC can match known complexes with a higher accuracy than other methods. Furthermore, the predicted complexes have high functional homogeneity.

Network based prediction of protein localisation using diffusion kernel.

We present NetLoc, a novel diffusion Kernel-based Logistic Regression (KLR) algorithm for predicting protein subcellular localisation using four types of protein networks including physical PPI networks, genetic Protein-Protein Interaction (PPI) networks, mixed PPI networks and co-expression networks. NetLoc is applied to yeast protein localisation prediction. The results showed that protein networks can provide rich information for protein localisation prediction, achieving Area Under Curve (AUC) score of 0.93. We also showed that networks with high connectivity and high percentage of co-localised PPI lead to better prediction performance. Investigation showed that NetLoc is a very robust approach which can produce good performance (AUC = 0.75) only using 30% of original interactions and capable of producing overall accuracy greater than 0.5 only with 20% annotation coverage. Compared to the previous network feature based prediction algorithm which achieved AUC scores of 0.49 and 0.52 on the yeast PPI network, NetLoc achieved significantly better overall performance with the AUC of 0.74.

Detecting Non-Uniform Clusters in Large-Scale Interaction Graphs

Graph clustering becomes difficult as the graph size and complexity increase. In particular, in interaction graphs, the clusters are small and the data on the underlying interaction are not only complex, but also noisy due to the lack of information and experimental errors. The graphs representing such data consist of (possibly overlapping) clusters of non-uniform size with some false positive and false negative links. In this article, we propose a new approach, assuming that clusters in the graphs of protein-protein interaction (PPI) networks resemble corrupted cliques. Therefore, the problem can be reduced to looking for clusters only among nodes of approximately similar degrees. This idea was implemented using a soft version of the Farthest-Point-First (FPF) clustering algorithm with the Jaccard distance function modified to perform on slightly overlapping clusters. The StripClust program developed by us was tested on a synthetic network and on the yeast PPI network.

K-Partite cliques of protein interactions: A novel subgraph topology for functional coherence analysis on PPI networks

Many studies are aimed at identifying dense clusters/subgraphs from protein–protein interaction (PPI) networks for protein function prediction. However, the prediction performance based on the dense clusters is actually worse than a simple guilt-by-association method using neighbor counting ideas. This indicates that the local topological structures and properties of PPI networks are still open to new theoretical investigation and empirical exploration. We introduce a novel topological structure called k-partite cliques of protein interactions—a functionally coherent but not-necessarily dense subgraph topology in PPI networks—to study PPI networks. A k-partite protein clique is a maximal k-partite clique comprising two or more nonoverlapping protein subsets between any two of which full interactions are exhibited. In the detection of PPI's maximal k-partite cliques, we propose to transform PPI networks into induced K-partite graphs where edges exist only between the partites. Then, we present a maximal k-partite clique mining (MaCMik) algorithm to enumerate maximal k-partite cliques from K-partite graphs. Our MaCMik algorithm is then applied to a yeast PPI network. We observed interesting and unusually high functional coherence in k-partite protein cliques—the majority of the proteins in k-partite protein cliques, especially those in the same partites, share the same functions, although k-partite protein cliques are not restricted to be dense compared with dense subgraph patterns or (quasi-)cliques. The idea of k-partite protein cliques provides a novel approach of characterizing PPI networks, and so it will help function prediction for unknown proteins.

Identifying Protein Complexes Based on Multiple Topological Structures in PPI Networks

Various computational algorithms are developed to identify protein complexes based on only one of specific topological structures in protein-protein interaction (PPI) networks, such as cliques, dense subgraphs, core-attachment structures and starlike structures. However, protein complexes exhibit intricate connections in a PPI network. They cannot be fully detected by only single topological structure. In this paper, we propose an algorithm based on multiple topological structures to identify protein complexes from PPI networks. In the proposed algorithm, four single topological structure based algorithms are first employed to identify raw predictions with specific topological structures, respectively. Those raw predictions are trimmed according to their topological information or GO annotations. Similar results are carefully merged before generating final predictions. Numerical experiments are conducted on a yeast PPI network of DIP and a human PPI network of HPRD. The predicted results show that the multiple topological structure based algorithm can not only obtain a more number of predictions, but also generate results with high accuracy in terms of f-score, matching with known protein complexes and functional enrichments with GO.

Nodes Having a Major Influence to Break Cooperation Define a Novel Centrality Measure: Game Centrality

Cooperation played a significant role in the self-organization and evolution of living organisms. Both network topology and the initial position of cooperators heavily affect the cooperation of social dilemma games. We developed a novel simulation program package, called ‘NetworGame’, which is able to simulate any type of social dilemma games on any model, or real world networks with any assignment of initial cooperation or defection strategies to network nodes. The ability of initially defecting single nodes to break overall cooperation was called as ‘game centrality’. The efficiency of this measure was verified on well-known social networks, and was extended to ‘protein games’, i.e. the simulation of cooperation between proteins, or their amino acids. Hubs and in particular, party hubs of yeast protein-protein interaction networks had a large influence to convert the cooperation of other nodes to defection. Simulations on methionyl-tRNA synthetase protein structure network indicated an increased influence of nodes belonging to intra-protein signaling pathways on breaking cooperation. The efficiency of single, initially defecting nodes to convert the cooperation of other nodes to defection in social dilemma games may be an important measure to predict the importance of nodes in the integration and regulation of complex systems. Game centrality may help to design more efficient interventions to cellular networks (in forms of drugs), to ecosystems and social networks.

IDENTIFICATION OF ESSENTIAL PROTEINS FROM WEIGHTED PROTEIN–PROTEIN INTERACTION NETWORKS

Identifying essential proteins is very important for understanding the minimal requirements of cellular survival and development. Fast growth in the amount of available protein-protein interactions has produced unprecedented opportunities for detecting protein essentiality on network level. A series of centrality measures have been proposed to discover essential proteins based on network topology. Unfortunately, the protein-protein interactions produced by high-throughput experiments generally have high false positives. Moreover, most of centrality measures based on network topology are sensitive to false positives. We therefore propose a new method for evaluating the confidence of each interaction based on the combination of logistic regression-based model and function similarity. Nine standard centrality measures in weighted network were redefined in this paper. The experimental results on a yeast protein interaction network shows that the weighting method improved the performance of centrality measures considerably. More essential proteins were discovered by the weighted centrality measures than by the original centrality measures used in the unweighted network. Even about 20% improvements were obtained from closeness centrality and subgraph centrality.

Model the evolution of protein interaction network assisted with protein age

Understanding the dynamic evolutionary process of protein interaction network is fundamental to study cell's activity. As an important attribute of proteins, their age can provide heuristic information for studying the evolution of protein interaction network. In this paper an age-dependent duplication-divergence model (ADD) is proposed to model the evolutionary process of protein interaction network, which integrates the age information of proteins into duplication-divergence (DD) model. Moreover, we also reverse protein interaction network with our model, which is named as inverse age-dependent duplication-divergence model (IADD). The results show that the network built by ADD model conforms to the yeast protein interaction network in the interaction density pattern and other topological properties. The protein age groups predicted by IADD model also show a high agreement with existing research results.

Modular Biological Function Is Most Effectively Captured by Combining Molecular Interaction Data Types

Large-scale molecular interaction data sets have the potential to provide a comprehensive, system-wide understanding of biological function. Although individual molecules can be promiscuous in terms of their contribution to function, molecular functions emerge from the specific interactions of molecules giving rise to modular organisation. As functions often derive from a range of mechanisms, we demonstrate that they are best studied using networks derived from different sources. Implementing a graph partitioning algorithm we identify subnetworks in yeast protein-protein interaction (PPI), genetic interaction and gene co-regulation networks. Among these subnetworks we identify cohesive subgraphs that we expect to represent functional modules in the different data types. We demonstrate significant overlap between the subgraphs generated from the different data types and show these overlaps can represent related functions as represented by the Gene Ontology (GO). Next, we investigate the correspondence between our subgraphs and the Gene Ontology. This revealed varying degrees of coverage of the biological process, molecular function and cellular component ontologies, dependent on the data type. For example, subgraphs from the PPI show enrichment for 84%, 58% and 93% of annotated GO terms, respectively. Integrating the interaction data into a combined network increases the coverage of GO. Furthermore, the different annotation types of GO are not predominantly associated with one of the interaction data types. Collectively our results demonstrate that successful capture of functional relationships by network data depends on both the specific biological function being characterised and the type of network data being used. We identify functions that require integrated information to be accurately represented, demonstrating the limitations of individual data types. Combining interaction subnetworks across data types is therefore essential for fully understanding the complex and emergent nature of biological function.

Assessment of high-confidence protein–protein interactome in yeast

The identification of protein–protein interactions (PPIs) and their networks is vitally important to systemically define and understand the roles of proteins in biological systems. In spite of development of numerous experimental systems to detect PPIs and diverse research on assessment of the quality of the obtained data, a consensus – highly reliable, almost complete – interactome of Saccharomyces cerevisiae is not presented yet. In this work, we proposed an unsupervised statistical approach to create a high-confidence yeast PPI network. For this, we assembled databases of interacting protein pairs for yeast and obtained an extremely large PPI dataset which comprises of 135154 non-redundant interactions between 6191 yeast proteins. A scoring scheme considering eight heterogeneous biological features resulted with a broad score distribution and a highly reliable network consisting of 29046 physical interactions with scores higher than the threshold value of 0.85, for which sensitivity, specificity and coverage were 86%, 68%, and 72%, respectively. We evaluated our method by comparing it with other scoring schemes and showed that reducing the noise inherent in experimental PPIs via our scoring scheme further increased the accuracy. Current study is expected to increase the efficiency of the methodologies in biological research which make use of protein interaction networks.

Identifying correlations between chromosomal proximity of genes and distance of their products in protein-protein interaction networks of yeast.

In this article we present evidence for a relationship between chromosome gene loci and the topological properties of the protein-protein interaction network corresponding to the set of genes under consideration. Specifically, for each chromosome of the Saccharomyces cerevisiae genome, the distribution of the intra-chromosome inter-gene distances was analyzed and a positive correlation with the distance among the corresponding proteins of the protein-protein interaction network was found. In order to study this relationship we used concepts based on non-parametric statistics and information theory.We provide statistical evidence that if two genes are closely located, then it is likely that their protein products are closely located in the protein-protein interaction network, or in other words, that they are involved in the same biological process.

Predicting the Binding Patterns of Hub Proteins: A Study Using Yeast Protein Interaction Networks

BackgroundProtein-protein interactions are critical to elucidating the role played by individual proteins in important biological pathways. Of particular interest are hub proteins that can interact with large numbers of partners and often play essential roles in cellular control. Depending on the number of binding sites, protein hubs can be classified at a structural level as singlish-interface hubs (SIH) with one or two binding sites, or multiple-interface hubs (MIH) with three or more binding sites. In terms of kinetics, hub proteins can be classified as date hubs (i.e., interact with different partners at different times or locations) or party hubs (i.e., simultaneously interact with multiple partners).MethodologyOur approach works in 3 phases: Phase I classifies if a protein is likely to bind with another protein. Phase II determines if a protein-binding (PB) protein is a hub. Phase III classifies PB proteins as singlish-interface versus multiple-interface hubs and date versus party hubs. At each stage, we use sequence-based predictors trained using several standard machine learning techniques.ConclusionsOur method is able to predict whether a protein is a protein-binding protein with an accuracy of 94% and a correlation coefficient of 0.87; identify hubs from non-hubs with 100% accuracy for 30% of the data; distinguish date hubs/party hubs with 69% accuracy and area under ROC curve of 0.68; and SIH/MIH with 89% accuracy and area under ROC curve of 0.84. Because our method is based on sequence information alone, it can be used even in settings where reliable protein-protein interaction data or structures of protein-protein complexes are unavailable to obtain useful insights into the functional and evolutionary characteristics of proteins and their interactions.AvailabilityWe provide a web server for our three-phase approach: http://hybsvm.gdcb.iastate.edu.

SPINAL: scalable protein interaction network alignment

Given protein-protein interaction (PPI) networks of a pair of species, a pairwise global alignment corresponds to a one-to-one mapping between their proteins. Based on the presupposition that such a mapping provides pairs of functionally orthologous proteins accurately, the results of the alignment may then be used in comparative systems biology problems such as function prediction/verification or construction of evolutionary relationships. We show that the problem is NP-hard even for the case where the pair of networks are simply paths. We next provide a polynomial time heuristic algorithm, SPINAL, which consists of two main phases. In the first coarse-grained alignment phase, we construct all pairwise initial similarity scores based on pairwise local neighborhood matchings. Using the produced similarity scores, the fine-grained alignment phase produces the final one-to-one mapping by iteratively growing a locally improved solution subset. Both phases make use of the construction of neighborhood bipartite graphs and the contributors as a common primitive. We assess the performance of our algorithm on the PPI networks of yeast, fly, human and worm. We show that based on the accuracy measures used in relevant work, our method outperforms the state-of-the-art algorithms. Furthermore, our algorithm does not suffer from scalability issues, as such accurate results are achieved in reasonable running times as compared with the benchmark algorithms. Supplementary Document, open source codes, useful scripts, all the experimental data and the results are freely available at http://code.google.com/p/spinal/. Supplementary data are available at Bioinformatics online.