Projects / Programmes
Unravelling Biological Networks
| Code |
Science |
Field |
Subfield |
| 1.07.01 |
Natural sciences and mathematics |
Computer intensive methods and applications |
Algorithms |
| Code |
Science |
Field |
| B110 |
Biomedical sciences |
Bioinformatics, medical informatics, biomathematics biometrics |
| Code |
Science |
Field |
| 1.01 |
Natural Sciences |
Mathematics |
bioinformatics, computational biology, graph theory, complex networks, biological networks, geometric graphs
Researchers (9)
Organisations (1)
Abstract
Recent technological advances in experimental biology allowed for a massive quantity of new data on gene and protein interactions to be obtained. Making sense of these large datasets is among the key problems in modern biology.
Complex systems of genes or proteins can be conveniently represented as networks, whose links model their interaction. Critical for understanding the functioning of a bionetwork is to investigate and quantify its topology.
My current research focuses on development of novel methods for analyzing the topologies of bionetworks. My approach is based on the idea of graphlet, which I define to be a small, connected subgraph of the original network. Graphlets allow for the local topology around a node to be quantified, generalizing the traditional node degree. Calculating the graphlet appearance frequency gives a highly sensitive statistical characterization of network's local structure, and allows for measuring the difference between networks.
Over the past decades, the analysis of genetic sequences have enabled a revolution in our understanding of biological systems. However, my preliminary results indicate that by investigating bionetworks' topology using my graphlet-based methods, we can extract new information that cannot be extracted from sequencing. In particular, I demonstrated the relation between local topology around a protein in a protein interaction network, with this protein's biological function and involvement in disease.
Additionally, my recent findings include an algorithm for embedding of a general network into a low-dimensional Euclidean space. I as showed, globally modeling bionetworks via this geometric graph model, fits the empirical data much better than traditionally used “scale-free” model.
These encouraging results indicate that advances in network analysis and graph-theoretic modeling could substantially contribute towards our understanding of biological systems, and ultimately help improve therapeutics. This project aims at substantially extending the results of my preliminary work, in the direction of developing and testing new computational tools for network topology analysis. Specifically, the goals of this project are the following:
design of new graph-theoretic algorithms for analyzing network topology, both locally and globally. This will include methods for predicting new network links;
analysis of new biological datasets. This specifically refers to bio-technology company DiaGenomi, which is my partner in this project;
construction and investigation of 'disease network', whose nodes represent diseases, and links indicate common cellular cause;
application of my tools to non-biological real-world networks, in particular social networks;
publication of the designed computational tools in form of opensource software.
This project comes at the unique point in scientific history, where development of theoretical network analysis methods can directly and easily contribute towards understanding of biological systems, which in turn can improve public health. The results of this work are likely to make an impact beyond biology and computer science – precisely quantifying topology or any real-world network is crucial for understanding its function and devising strategies for its control.
Significance for science
Network science developed by combining the graph theory and social network analysis with recent empirical findings of universalities in real systems. Complex networks paradigm offers an elegant way to study natural, social or technological systems that are composed of many interacting units. While the twentieth century being the age of reductionism, we are now witnessing the increasing appreciation of the “emergent” phenomena. Stephen Hawking’s opinion that “the twenty-first century would be the century of complexity” comes precisely in this context. This trend suggests that complex networks are likely to remain a dynamic and expanding scientific field. This project has contributed concrete shifts in both bioinformatics and general network science. Below we list the core methodological and thematic contributions to science: METHODOLOGICAL CONTRIBUTIONS TO SCIENCE We systematically evaluated the performances of the available alignment-free network comparison methods. We have done this by measuring accuracy of each method (in a systematic precision-recall framework) in terms of how well a method can group (cluster) topologically similar networks. By testing this on both synthetic and real-world networks from different domains, we showed that GCD remains the most accurate, noise-tolerant and computationally efficient alignment-free method, while being less computationally expenisve than most contenders. We constructed a heterogeneous network that represents the integrated large-scale biological data from different repositories. In these networks the nodes were bio-entities such as disease, genes, pathways or GO terms. Associations between a pairs of these bio-entities were represented as edges. We used this framework to gain further understanding of how diseases associate between each other. We developed a statistical framework that allows for improved understanding of topology-function relationships and of their conservation among species. Using the graphlet degrees to represent the wiring around proteins in PINs and Gene Ontology annotations to describe their functions, our framework characterizes statistically significant topology-function relationships in a given species and uncovers the functions that have conserved topology in PINs of different species, which we term topologically orthologous functions. TOPICAL CONTRIBUTIONS TO SCIENCE We analysed the topology of the ever growing collection of human interactome data. By virtue of improved experimental techniques in biology and medicine, the human interactome is getting denser and more complete. This growth of the available data is accompanied by increasing patterns found in them. We performed a network analysis of the human PPI networks published in the past decade in order to study its topological evolution. The results confirm that the PPI network is becoming more compact and that the presence of patterns in it is growing. We analysed disease-gene association and large-scale molecular networks to find new insight into disease relationships. We used the data from four publicly available disease-gene association datasets and systematically evaluated disease associations. Our work relied on different correlation measures between diseases that can be found in common functions or in topological similarity between the molecular networks.
Significance for the country
The main importance of this project for Slovenia in the development of modern (computational) methods for the analysis of data behind complex diseases, which leads to improvement of methods for their clinical treatment. In Slovenia, for certain diseases predictive models based on specific mutations and clinical data set were already implemented, but unfortunately, have not proven to be the most optimal. Using mathematical and computational approaches resulting from research project, we have discovered new potential biomarkers, which are now in the testing phase. In the coming years we expect a further development in this direction. More specifically, methods that have been developed help to discover new biomarkers for a specific disease or a medical condition. For many diseases specific mutations of this disease are already known and some of them may predict the patient’s response to treatment (pharmacogenomics). With the ability integrate very different databases and specific already known biomarkers, we can for a specific disease discover new potential biomarkers. Only those we can increase the predictive power of existing approaches, or even lead to discovery of new biological routes that play a role in the development of certain diseases. Consequently, the discovery of such new routes leads to the development of new treatments. On the other hand, study of complex networks is automatically a study of real world - in fact, society is nothing but a network of individuals. It is therefore only natural to expect that our scientific efforts will be connected to current social and economics needs. In particular, outputs of our work might benefit several economic sectors, including high-tech industries such as biotechnology, pharmacy or communications, where the Slovenian industry has a long and successful history. Computational social science results might be of interest to policy makers and wider public, as they will grasp intricate social phenomena that help us understand our society’s dynamics. Except on purely scientific front, our long-term results will benefit Slovenia in several other ways. By pursuing interdisciplinary work and organizing scientific events, we intend to strengthen collaborations worldwide, which at present include Oxford University, Harvard University, UCL, Boston University, University of California Santa Barbara, National University of Singapore, East Chine University of Science and Tokyo University. Exchange of young scientists will be a key part our work, not only enhancing scientific thought in our country, but also improving the circulation of ideas in and out of Slovenia. This can create initiatives for new EU project and bring funding into the country.
Most important scientific results
Annual report
2013,
2014,
2015,
final report
Most important socioeconomically and culturally relevant results
Annual report
2013,
2014,
2015,
final report
