Alexandru Topirceanu

Network Fidelity: A Metric to Quantify the Similarity and Realism of Complex Networks

The analysis of complex networks revolves around the common fundamental properties found in most natural and synthetic networks that surround us. Each such network is commonly described through a standard set of graph metrics, yet there is no unified and efficient method of quantitatively compare networks to each other. This paper introduces the network fidelity metric delta (δ) which is aimed at offering the possibility to compare networks to each other based on individual metric measurements. Depending on the nature of the comparison, it can offer insight on network model similarity or synthetic model realism compared to a real world network. We apply the metric in a social network modeling scenario and also compare it to another metric - the fractal dimension - highlighting the superior analytic power of our metric.

Clustering drug-drug interaction networks with energy model layouts: community analysis and drug repurposing.

Analyzing drug-drug interactions may unravel previously unknown drug action patterns, leading to the development of new drug discovery tools. We present a new approach to analyzing drug-drug interaction networks, based on clustering and topological community detection techniques that are specific to complex network science. Our methodology uncovers functional drug categories along with the intricate relationships between them. Using modularity-based and energy-model layout community detection algorithms, we link the network clusters to 9 relevant pharmacological properties. Out of the 1141 drugs from the DrugBank 4.1 database, our extensive literature survey and cross-checking with other databases such as Drugs.com, RxList, and DrugBank 4.3 confirm the predicted properties for 85% of the drugs. As such, we argue that network analysis offers a high-level grasp on a wide area of pharmacological aspects, indicating possible unaccounted interactions and missing pharmacological properties that can lead to drug repositioning for the 15% drugs which seem to be inconsistent with the predicted property. Also, by using network centralities, we can rank drugs according to their interaction potential for both simple and complex multi-pathology therapies. Moreover, our clustering approach can be extended for applications such as analyzing drug-target interactions or phenotyping patients in personalized medicine applications.

Uncovering the fingerprint of online social networks using a network motif based approach

Large-scale computational generation and motif distribution analysis for the synthetic topology classes. We obtain a distinct motif pattern for each such class.Comprehensive motif analysis of online social networks (Facebook, Twitter, Google Plus) from which we obtain three quantifiable characteristic motif fingerprints.Mapping and similarity assessment of empirical networks onto topology classes, and defining a general methodology for such an approach.Correlation and discussion of the individual motifs that occur in each fingerprint, and an outlining of the functional properties behind the three online social platforms. Complex networks facilitate the understanding of natural and man-made processes and are classified based on the concepts they model: biological, technological, social or semantic. The relevant subgraphs in these networks, called network motifs, are demonstrated to show core aspects of network functionality and can be used to analyze complex networks based on their topological fingerprint. We propose a novel approach of classifying social networks based on their topological aspects using motifs. As such, we define the classifiers for regular, random, small-world and scale-free topologies, and then apply this classification on empirical networks. We then show how our study brings a new perspective on differentiating between online social networks like Facebook, Twitter and Google Plus based on the distribution of network motifs over the fundamental topology classes. Characteristic patterns of motifs are obtained for each of the analyzed online networks and are used to better explain the functional properties behind how people interact online and to define classifiers capable of mapping any online network to a set of topological-communicational properties.

Tolerance-based interaction: a new model targeting opinion formation and diffusion in social networks

One of the main motivations behind social network analysis is the quest for understanding opinion formation and diffusion. Previous models have limitations, as they typically assume opinion interaction mechanisms based on thresholds which are either fixed or evolve according to a random process that is external to the social agent. Indeed, our empirical analysis on large real-world datasets such as Twitter, Meme Tracker, and Yelp, uncovers previously unaccounted for dynamic phenomena at population-level, namely the existence of distinct opinion formation phases and social balancing. We also reveal that a phase transition from an erratic behavior to social balancing can be triggered by network topology and by the ratio of opinion sources. Consequently, in order to build a model that properly accounts for these phenomena, we propose a new (individual-level) opinion interaction model based on tolerance. As opposed to the existing opinion interaction models, the new tolerance model assumes that individual’s inner willingness to accept new opinions evolves over time according to basic human traits. Finally, by employing discrete event simulation on diverse social network topologies, we validate our opinion interaction model and show that, although the network size and opinion source ratio are important, the phase transition to social balancing is mainly fostered by the democratic structure of the small-world topology. Subjects Network Science and Online Social Networks, Scientific Computing and Simulation, Social Computing

MuSeNet: Collaboration in the Music Artists Industry

Motivated by the constantly growing interest and real-world applicability shown in social networks, we model and analyze the network formed by music artists all around the world, which we call MuseNet. Inspired by similar approaches, we compare our analytic results with generic online friendship models and with the collaboration networks of actors. We are the first to fully create such a network, and by using centrality measures, we discover the most influential nodes in MuseNet. In light of current advances in social networks, we also highlight the importance of music producers in terms of meritocracy versus topological positioning, and discuss the differentiation between collaboration networks using a network fidelity approach. Finally, we show that MuseNet has a characteristic sociability -- a measure which is introduced in this paper - in comparison with other empirical networks.

A network motif based approach for classifying online social networks

Complex networks facilitate the understanding of natural and man-made processes and are classified based on the concepts they model: biological, technological, social or semantic. The relevant subgraphs in these networks, called network motifs, are demonstrated to show core aspects of network functionality. They are used to classify complex networks based on that functionality. We propose a novel approach of classifying complex networks based on their topological aspects using motifs. We define the classifiers for regular, random, small-world and scale-free topologies, as well as apply this classification on empirical networks. The study brings a new perspective on how we can classify and differentiate online social networks like Facebook, Twitter and Google Plus based on the distribution of network motifs over the fundamental network topology classes.

Statistical fidelity: a tool to quantify the similarity between multi-variable entities with application in complex networks

ABSTRACT Complex networks are often characterized by their underlying graph metrics, yet there is no unified computational method for comparing networks to each other. Given that complex networks are entities characterized by a set of known properties, our problem is reduced to quantifying the similarity between the multi-variable entities. To address this issue, we introduce the new statistical fidelity metric, which can compare any types of entities, characterized by specific individual metrics, in order to gauge the similarity of the entities under the form of a single number between 0 and 1. To test the efficiency of statistical fidelity, we apply our composite metric in the field of complex networks, by assessing topological similarity and realism of social networks and urban road networks. Pinned against other statistical methods, such as the cosine similarity, Pearson correlation, Mahalanobis distance and fractal dimension, we highlight the superior analytic power of statistical fidelity.

Genetically Optimized Realistic Social Network Topology Inspired by Facebook

Social network analysis is receiving an increased interest from multiple fields of science since more and more natural and synthetic networks are found to share similar features which help us understand their underlying topological properties. One desire is to create a model of the human society, however, the complexity of such a model is increased by the nature of human interaction, and present studies fail to create a fully realistic model of the societies we live in. Our approach is inspired from studies of online social networking and the ability of genetic algorithms (GA) to optimize topological data in a natural manner. We combine the properties of the small-world and scale-free models to create a community-based social network, which is then rearranged using empirically obtained data from Facebook friendship networks, and optimized using GAs. As a result, our synthetically generated social network topologies are more realistic, with a proposed realism fidelity metric that is with 63 % closer to the observed real-world parameters than the best existing model.

Social cities: Quality assessment of road infrastructures using a network motif approach

Motivated by the constantly growing interest and real-world applicability shown in complex networks, we model and analyze the network formed by road networks in cities from an innovative perspective. Inspired by similar approaches of comparing networks, we create a methodology that proposes the assessment of city road networks based on their motif distributions. To the best of our knowledge, we are the first to fully interpret the roads infrastructure by using network motifs. Based on the similarity of the motif distributions, we choose six diverse cities, create a similarity graph, and discuss the urban influences one has on each other. Through our analysis, we coin the title of Social City to any city which meets particular criteria in terms of optimal roads distribution.

Network science meets respiratory medicine for OSAS phenotyping and severity prediction

Obstructive sleep apnea syndrome (OSAS) is a common clinical condition. The way that OSAS risk factors associate and converge is not a random process. As such, defining OSAS phenotypes fosters personalized patient management and population screening. In this paper, we present a network-based observational, retrospective study on a cohort of 1,371 consecutive OSAS patients and 611 non-OSAS control patients in order to explore the risk factor associations and their correlation with OSAS comorbidities. To this end, we construct the Apnea Patients Network (APN) using patient compatibility relationships according to six objective parameters: age, gender, body mass index (BMI), blood pressure (BP), neck circumference (NC) and the Epworth sleepiness score (ESS). By running targeted network clustering algorithms, we identify eight patient phenotypes and corroborate them with the co-morbidity types. Also, by employing machine learning on the uncovered phenotypes, we derive a classification tree and introduce a computational framework which render the Sleep Apnea Syndrome Score (SASScore); our OSAS score is implemented as an easy-to-use, web-based computer program which requires less than one minute for processing one individual. Our evaluation, performed on a distinct validation database with 231 consecutive patients, reveals that OSAS prediction with SASScore has a significant specificity improvement (an increase of 234%) for only 8.2% sensitivity decrease in comparison with the state-of-the-art score STOP-BANG. The fact that SASScore has bigger specificity makes it appropriate for OSAS screening and risk prediction in big, general populations.