The newly established scientific discipline of complexity science is concerned with studying huge sets of relationships among entities, to make sense of the complex interrelationships within a particular structure or phenomenon. Several new scenarios from sciences or everyday life benefit from formulating a relationship-network between entities as a graph. A typical application scenario for complexity science is network structures in biology. This is because the life functions are organized in a relationship of interacting elements and chemical compounds, forming super complex networks of reactions carried out all over the entire life form. Biologist often semantically organize those networks by segmenting them into sub-networks, known as ‘pathways’, through corresponding types of functionality. Pathways thus form sets of graphs containing dozens of chemical elements representing a particular element of life. For example, ‘Glycolysis’ is a metabolic pathway that decomposes glucose and releases free energy by generating molecules ATP, which is known as a molecular unit of currency for intracellular energy transfer. Unfortunately, this molecular unit is involved in a large number of biological reactions, which leads to the challenge of network visualizations in high quality.
The main challenge in BioNetIllustration is because, in living systems, one molecule is commonly involved in several distinct physiological functions. Initially, biochemists can study pathways primarily for one scenario; they only marginally considered their modification in the context of other roles of that molecule in physiology. The roles of molecules are summarized in pathway diagrams, which are abstract, hierarchically nested and complex across multiple scales, and thus pose significant challenges to comprehend the fundamental reactions. Recently, multiple roles of one molecule are further examined, and the corresponding simulation methods are substantially proposed due to the increasing size of the networks and the development of analysis techniques. Automatic layout algorithms thus become indispensable in the sense that manually creating diagrams of large networks is a very time-consuming or an impractical task, due to the fast data change and the corresponding layout updates. For example, glucose is traditionally considered as a supply of energy, while it is nowadays demonstrated that it also affects cancer metabolism. The pathway designers would need to deliberately move the glucose to a new position in the diagram based on its changed functionalities. Those changes may disturb setting up and revising the mental models of scientists and the public since it induces a significant update of the existing knowledge a person owns.
The primary goal of the BioNetIllustration project in visualization is to intuitively support the comprehensive understanding of relationships among biological networks using interactively computed illustrations. The goal includes (1) on algorithmic layout customization of complex networks, (2) on automatic generation of hand-drawn like illustrations, (3) on smooth transitions between (1) and (2) across multiple levels of detail, and (4) on formulating them to be computationally efficient. The outcome of this research will be a set of new visualization techniques for effective and expressive network illustrations to convey the process and its spatial context in an intuitive form.
After the investigation, the BioNetIllustration project successfully formulated the human-preferred design criteria into algorithmic constraints and thus provided more intuitive understanding visualization on the biological datasets. This is appreciated with the collaborations with several domain experts, including biologists, illustrators, pathway database providers, as well as system users. We, therefore, investigate our research questions and assumptions, and demonstrate the feasibility of the developed techniques through the project.
