Studying protein interaction networks using functional and topological information is important for understanding cellular organization and functionality. This study deals with identifying important proteins in protein interaction networks using SWEMODE (Lubovac, et al, 2006) and analyzing topological and functional properties of these proteins with the help of information derived from modular organization in protein interaction networks as well as information available in public resources, in this case, annotation sources describing the functionality of proteins. Multi-modular proteins are short-listed from the modules generated by SWEMODE. Properties of these short-listed proteins are then analyzed using functional information from SGD Gene Ontology(GO) (Dwight, et al., 2002) and MIPS functional categories (Ruepp, et al., 2004). Topological features such as lethality and centrality of these proteins are also investigated, using graph theoretic properties and information on lethal genes from Yeast Hub (Kei-Hoi, et al., 2005). The findings of the study based on GO terms reveal that these important proteins are mostly involved in the biological process of “organelle organization and biogenesis” and a majority of these proteins belong to MIPS “cellular organization” and “transcription” functional categories. A study of lethality reveals that multi-modular proteins are more likely to be lethal than proteins present only in a single module. An examination of centrality (degree of connectivity of proteins) in the network reveals that the ratio of number of important proteins to number of hubs at different hub sizes increases with the hub size (degree).

<p>Studying protein interaction networks using functional and topological information is important for understanding cellular organization and functionality. This study deals with identifying important proteins in protein interaction networks using SWEMODE (Lubovac, et al, 2006) and analyzing topological and functional properties of these proteins with the help of information derived from modular organization in protein interaction networks as well as information available in public resources, in this case, annotation sources describing the functionality of proteins. Multi-modular proteins are short-listed from the modules generated by SWEMODE. Properties of these short-listed proteins are then analyzed using functional information from SGD Gene Ontology(GO) (Dwight, et al., 2002) and MIPS functional categories (Ruepp, et al., 2004). Topological features such as lethality and centrality of these proteins are also investigated, using graph theoretic properties and information on lethal genes from Yeast Hub (Kei-Hoi, et al., 2005). The findings of the study based on GO terms reveal that these important proteins are mostly involved in the biological process of “organelle organization and biogenesis” and a majority of these proteins belong to MIPS “cellular organization” and “transcription” functional categories. A study of lethality reveals that multi-modular proteins are more likely to be lethal than proteins present only in a single module. An examination of centrality (degree of connectivity of proteins) in the network reveals that the ratio of number of important proteins to number of hubs at different hub sizes increases with the hub size (degree).</p>

No description available.

An important aspect of systems biology is the elucidation of the protein-protein interactions (PPIs) that control important biological processes within a cell and between organisms. In particular, at the cellular and molecular level, interactions between a pathogen and its host play a vital role in initiating infection and a successful pathogenesis. Despite recent successes in the advancement of the systems biology of model organisms to understand complex diseases, the analysis of infectious diseases at the systems-level has not received as much attention. Since pathogen related disease is responsible for millions of deaths and billions of dollars in damage to crops and livestock, understanding the mechanisms employed by pathogens to infect their hosts is critical in the development of new and effective therapeutic strategies. The research presented here is one of the first computational approaches to studying host-pathogen PPI networks. This dissertation has two main aims. First, we discuss analytical tools for studying host-pathogen networks to identify common pathways perturbed and manipulated by pathogens. We present the first global comparison of the host-pathogen PPI networks of 190 different pathogens and their interactions with human proteins. We also present the construction and analysis of three highly infectious human-bacterial PPI networks: <i>Bacillus anthracis</i>, <i>Francislla tularensis</i>, and <i>Yersinia pestis</i>. The second aim of the research presented here is the development of predictive models for identifying PPIs between host and pathogen proteins. We present two methods: (i) a domain-based approach that uses frequency of domain-pairs in intra-species PPIs, and (ii) a supervised machine learning method that is trained on known inter-species PPIs. The techniques developed in this dissertation, along with the informative datasets presented, will serve as a foundation for the field of computational pathosystems biology. / Ph. D.

BACKGROUND:A hub protein is one that interacts with many functional partners. The annotation of hub proteins, or more generally the protein-protein interaction "degree" of each gene, requires quality genome-wide data. Data obtained using yeast two-hybrid methods contain many false positive interactions between proteins that rarely encounter each other in living cells, and such data have fallen out of favor.RESULTS:We find that protein "stickiness", measured as network degree in ostensibly low quality yeast two-hybrid data, is a more predictive genomic metric than the number of functional protein-protein interactions, as assessed by supposedly higher quality high throughput affinity capture mass spectrometry data. In the yeast Saccharomyces cerevisiae, a protein's high stickiness, but not its high number of functional interactions, predicts low stochastic noise in gene expression, low plasticity of gene expression across different environments, and high probability of forming a homo-oligomer. Our results are robust to a multiple regression analysis correcting for other known predictors including protein abundance, presence of a TATA box and whether a gene is essential. Once the higher stickiness of homo-oligomers is controlled for, we find that homo-oligomers have noisier and more plastic gene expression than other proteins, consistent with a role for homo-oligomerization in mediating robustness.CONCLUSIONS:Our work validates use of the number of yeast two-hybrid interactions as a metric for protein stickiness. Sticky proteins exhibit low stochastic noise in gene expression, and low plasticity in expression across different environments.

Redes complexas são utilizadas para representar sistemas complexos, compostos de elementos que interagem uns com os outros. Uma das grandes vantagens de se empregar as redes é a possibilidade de se estudar a topologia presente nos mais diversos sistemas para obtermos informações sobre eles, entendê-los e compará-los. Devido à sua importância para a compreensão de processos intracelulares, desde início do desenvolvimento da área das redes complexas estudou-se a topologia da interação entre proteínas. Entretanto nos últimos anos com o desenvolvimento de novas técnicas de detecção o número de proteínas e interações reportadas cresceu de maneira muito acentuada; além disso, também existem alguns pontos sobre a sua topologia sobre os quais ainda não existe um consenso, como por exemplo qual a distribuição de graus desse tipo de rede. Neste trabalho estudamos as propriedades topológicas de redes de interação entre proteínas, utilizando as informações do banco de dados STRING, com ênfase no comportamento de suas medidas de centralidade e do espectro da matriz Laplaciana normalizada. Tanto a análise das medidas de centralidade e de suas correlações, quanto do espectro da matriz Laplaciana mostram que existem padrões topológicos que são conservados entre as redes dos organismos e que os mesmos também podem ser empregados para sua caracterização. Nossos resultados também mostram que as funções biológicas desempenhadas pelas proteínas podem ser identificadas pelas medidas de centralidade. Especificamente para a centralidade de autovetor, nossas análises indicam que ela está localizada nos maiores K-cores das redes consideradas. Os resultados aqui obtidos ressaltam que muitas informações relevantes podem ser extraídas da topologia das interações entre proteínas, além de indicarem a existência de possíveis estruturas conservadas; entretanto devido a incompletude dessas redes mais estudos precisam ser conduzidos para a avaliação de possíveis mudanças nos resultados aqui apresentados. / Complex networks can be used to model complex systems, composed of main elements that interact with each other. The advantage of using this approach is the possibility to study the topology of a wide range of systems so that we can get more information, understand and compare them. Due to its importance on the understanding of the intracellular biological processes, since the early beginning of the development of the complex networks field protein-protein interaction topologies have been studied. However, new techniques for the detection of proteins and their interactions have been developed recently, which has significantly increased the availability and reliability of the corresponding data over the last few years; moreover, there still are some debate about the topology of protein-protein interaction networks such as the degree distribution of this type of network. Here we will study the topological properties of protein-protein interaction networks created using the information of the STRING database focusing on centrality measures of their nodes, the correlation between them, and the normalized Laplacian matrix spectrum. Our results show the existence of topological patterns conserved between the protein interaction networks of different organisms and that both the correlation of the centrality pairs and the spectrum of the Laplacian matrix can be used for network characterization. Another study indicates that the set of centrality measures of a protein can be used to identify clusters with well defined biological functions. A more detailed look at the eigenvector centrality behavior reveals that this measure is localized on the proteins of the highest k-cores for all networks. These results highlight the importance of the topology on the study of protein-protein interactions and that more studies can lead to a better a more complete understanding of such systems.

Protein-protein interactions are important for the prediction of protein functions since two interacting proteins usually have similar functions in a cell. Available protein interaction networks are incomplete / but, they can be used to predict new interactions in a supervised learning framework. However, in the case that the known protein network includes large number of protein pairs, the training time of the machine learning algorithm becomes quite long. In this thesis work, our aim is to predict protein-protein interactions with a known portion of the interaction network. We used Support Vector Machines (SVM) as the machine learning algoritm and used the already known protein pairs in the network. We chose to use phylogenetic profiles of proteins to form the feature vectors required for the learner since the similarity of two proteins in evolution gives a reasonable rating about whether the two proteins interact or not. For large data sets, the training time of SVM becomes quite long, therefore we reduced the data size in a sensible way while we keep approximately the same prediction accuracy. We applied a number of clustering techniques to extract the most representative data and features in a two categorical framework. Knowing that the training data set is a two dimensional matrix, we applied data reduction methods in both dimensions, i.e., both in data size and in feature vector size. We observed that the data clustered by the k-means clustering technique gave superior results in prediction accuracies compared to another data clustering algorithm which was also developed for reducing data size for SVM training. Still the true positive and false positive rates (TPR-FPR) of the training data sets constructed by the two clustering methods did not give satisfying results about which method outperforms the other. On the other hand, we applied feature selection methods on the feature vectors of training data by selecting the most representative features in biological and in statistical meaning. We used phylogenetic tree of organisms to identify the organisms which are evolutionarily significant. Additionally we applied Fisher&sbquo / &Auml / &ocirc / s test method to select the features which are most representative statistically. The accuracy and TPR-FPR values obtained by feature selection methods could not provide to make a certain decision on the performance comparisons. However it can be mentioned that phylogenetic tree method resulted in acceptable prediction values when compared to Fisher&sbquo / &Auml / &ocirc / s test.

Proteins participate in the majority of cellular processes. To determine the function of a protein it is not sufficient to solely know its sequence, its structure in isolation, or how it works individually. Additionally, we need to know how the protein interacts with other proteins in biological networks. This is because most of the proteins perform their main function through interactions. This thesis sets out to improve the understanding of protein-protein interaction networks (PPINs). For this, we propose three approaches: (1) Studying measures and methods used in social and complex networks. The methods, measures, and properties of social networks allow us to gain an understanding of PPINs via the comparison of different types of network families. We studied models that describe social networks to see which models are useful in describing biological networks. We investigate the similarities and differences in terms of the network community profile and centrality measures. (2) Studying PPINs and their role in evolution. We are interested in the relationship of PPINs and the evolutionary changes between species. We investigate whether the centrality measures are correlated with the variability and similarity in orthologous proteins. (3) Studying protein features that are important to evaluate, classify, and predict interactions. Interactions can be classified according to different characteristics. One characteristic is the energy (that is the attraction or repulsion of the molecules) that occurs in interacting proteins. We identify which type of energy values contributes better to predicting PPIs. We argue that the number of energetic features and their contribution to the interactions can be a key factor in predicting transient and permanent interactions. Contributions of this thesis include: (1) We identified the best community sizes in PPINs. This finding will help to identify important groups of interacting proteins in order to better understand their particular interactions. We furthermore find that the generative model describing biological networks is very different from the model describing social networks A generative model is a model for randomly generating observable data. We showed that the best community size for PPINs is around ten, different from the best community size for social and complex network (around 100). We revealed differences in terms of the network community profile and correlations of centrality measures; (2) We outline a method to test correlation of centrality measures with the percentage of sequence similarity and evolutionary rate for orthologous proteins. We conjecture that a strong correlation exists. While not obtaining positive results for our data. Therefore, (3) we investigate a method to discriminate energetic features of protein interactions that in turn will improve the PPIN data. The use of multiple data sets makes possible to identify the energy values that are useful to classify interactions. For each data set, we performed Random Forest and Support Vector Machine with linear, polynomial, radial, and sigmoid kernels. The accuracy obtained in this analysis reinforces the idea that energetic features in the protein interface help to discriminate between transient and permanent interactions. / Graduate / 0984

Protein interaction data is being generated at unprecedented rates thanks to advancements made in high throughput techniques such as mass spectrometry and DNA microarrays. Biomedical researchers, operating under budgetary constraints, have found it difficult to scale their efforts to keep up with the ever-increasing amount of available data. They often lack the resources and manpower required to analyze the data using existing methodologies. These research deficiencies impede our ability to understand diseases, delay the advancement of clinical therapeutics, and ultimately costs lives. One of the most commonly used techniques to analyze protein interaction data is the construction and visualization of protein interaction networks. This research investigated the effectiveness and efficiency of novel domain-specific algorithms for visualizing protein interaction networks. The existing domain-agnostic algorithms were compared to the novel algorithms using several performance, aesthetic, and biological relevance metrics. The graph drawing algorithms proposed here introduced novel domain-specific forces to the existing force-directed graph drawing algorithms. The innovations include an attractive force and graph coarsening policy based on semantic similarity, and a novel graph refinement algorithm. These experiments have demonstrated that the novel graph drawing algorithms consistently produce more biologically meaningful layouts than the existing methods. Aggregated over the 480 tests performed, and quantified using the Biological Evaluation Percentage metric defined in the Methodology chapter, the novel graph drawing algorithms created layouts that are 237 percent more biologically meaningful than the next best algorithm. This improvement came at the cost of additional edge crossings and smaller minimum angles between adjacent edges, both of which are undesirable aesthetics. The aesthetic and performance tradeoffs are experimentally quantified in this study, and dozens of algorithmically generated graph drawings are presented to visually illustrate the benefits of the novel algorithms. The graph drawing algorithms proposed in this study will help biomedical researchers to more efficiently produce high quality interactive protein interaction network drawings for improved discovery and communication.

Genome-wide studies of diseases and chronic conditions frequently fail to uncover marked or consistent differences in RNA or protein concentrations. However, the developing field of kinetic proteomics has made promising discoveries in differences in the turnover rate of these same proteins, even when concentrations were not necessarily different. The situation can theoretically be modeled mathematically using bifurcation equations, but uncovering the proper form of these is difficult. To this end, we developed TWIG, a method for characterizing bifurcations that leverages information geometry to identify drivers of complex systems. Using this, we characterized the bifurcation and stability properties of all 132 possible 3- and 22,662 possible 4-node subgraphs (motifs) of protein-protein interaction networks. Analyzing millions of real world protein networks indicates that natural selection has little preference for motifs that are stable per se, but a great preference for motifs who have parameter regions that are exclusively stable, rather than poorly constrained mixtures of stability and instability. We apply this knowledge to mice on calorie restricted (CR) diets, demonstrating that changes in their protein turnover rates do indeed make their protein networks more stable, explaining why CR is the most robust way known to extend lifespan.