Guest post by Teresa Przytycka, PhD, Senior Investigator, Computational and Systems Biology section of the Computational Biology Branch at the National Library of Medicine’s National Center for Biotechnology Information, National Institutes of Health.
The functioning of any complex system involves interactions between elements of that system. This is true at the cellular level, the macro level, and every point in between.
For example, within a cell, diverse molecules coordinate their activities and work together to carry out specific cellular functions. Cells then interact with each other to shape an organism’s development, tissue-level organization, and immune response. In turn, the organisms themselves interact to form various types of connections. In the case of people, those connections form the backbone of social systems.
These many interactions, from the microscopic to the macroscopic, can be described as networks, which comprise nodes connected via links that designate the relationships between nodes.
But how can we discover which nodes are connected? And how can we learn about the nature of those connections?
My research group and I try to answer these questions using network analysis, that is, working at the network level to uncover insights about the underlying system.
We can trace the beginnings of network analysis (also known as graph theory in mathematical circles) to Gottfried Leibniz’s geometria situs (“geometry of position”), a mathematical discipline focused on the relationship between positions and objects. The first recorded application of this new way of thinking was the famous solution to the problem known as “The Bridges of Königsberg,” published by Leonhard Euler in 1736.
Königsberg (now Kaliningrad) straddles the Pregel River. In Euler’s time, seven bridges connected the various parts of the city, including two islands in the middle of the river. The question asked was whether one could chart a walk through the city that required crossing each bridge only once and return to the start.
Euler tackled this question using what we today call network theory. If the regions of Königsberg are nodes and the bridges are links, Euler showed that, for such a walk to be possible, each node must have an even number of links. Why an even number? Because if we cannot cross the same bridge twice, then for every way into each region of the city there must be a new way out. Because that property did not hold for Königsberg, Euler concluded that no such walk could be devised.
As Euler’s argument shows, representing a complex system as a network of nodes and links can help uncover properties of the system that might have otherwise been obscured.
In biology and medicine, such network-centric approaches coincided with the emergence of high-throughput experimental techniques and advanced methods for collecting and storing diverse biomedical data. The protein interaction network for yeast became one of the first large biological networks obtained from high-throughput experiments. Analyzing that network revealed that a small fraction of proteins interacted with a disproportionately large number of other proteins. Additional research showed that these “hub” proteins are essential to the cell’s survival.
This intriguing relationship between a network property and a biological property begged for an explanation.
Our 2008 paper demonstrated that the majority of protein hubs are essential because of their involvement in complex, densely connected modules that carry out functions essential for cell survival. These results illustrate that, in addition to reporting binary relationships between individual nodes, interaction networks encode hidden higher-level organization.
In many networks, including biological and social ones, groups of nodes that interact with each other more tightly than with the rest of the network can be identified. We call these groups “modules.” In the context of biological networks, modules are often associated with groups of genes that work together to perform a specific biological function. At the same time, we’re beginning to see that complex diseases, such as cancer, are more likely caused by the dysregulation of a specific functional module than a dysfunction of an individual gene. That’s why, in recent years, cancer research has turned its attention to identifying dysregulated modules.
This effort includes several methods developed by our research group—Module Cover, Mutual Exclusivity Module Cover, and BeWith. These methods combine information from the human-interaction network with disease-specific information, such as abnormally expressed or mutated genes, to identify disease-associated modules. These modules can shed light on the mechanism of the disease, suggesting areas for further study and possible means of intervention.
We also use networks to discover how information flows between individual nodes. For example, by applying the principles of current flow within an electric circuit, we have been able to identify the causal (altered) disease genes and the pathways they dysregulate, providing another way to discover groups of genes involved in a disease.
Unfortunately, most currently available interaction networks are static depictions of a dynamically changing system. They typically do not account for tissue types, developmental stages, disease status, and other factors. As a result, these networks cannot tell us the full story.
To consider these and other factors, we need a context-specific network. But to build one, we must use context-specific data. How can we do that, given the myriad conditions we must consider? Our new method, NetREX, moves us in that direction.
Network biology has facilitated progress in many areas of biomedical science. This simple, yet powerful, concept allows us to abstract the essence of relations between genes and proteins, predict interactions between drugs, study disease comorbidity, and discover important associations. Of course, discovering an association is just the first step in uncovering a mechanism, but it is often a crucial step.
Teresa M. Przytycka, PhD, leads the Algorithmic Methods in Computational and Systems Biology section at the National Center for Biotechnology Information. Dr. Przytycka is particularly interested in the dynamical properties of biological systems, including spatial, temporal and contextual variations, and exploring how such variations impact gene expression, the functioning of biological pathways, and the phenotype of the organism.