Speaker: Steffi Prast
Metabolic networks represent interconnected reactions of chemical entities which take place within cells.
These networks are used in the Life sciences for knowledge exchange and are drawn in domain-specific notations.
Since they can contain thousands of nodes, automatic layouts are required that conserve the meaning of these networks.
There are many graph drawing algorithms including hierarchical, topology-shape-metric, force-directed, and constraint-based approaches.
They typically consider only a subset of the requirements needed to faithfully visualize metabolic networks and rarely support domain-specific notations.
In this work, we present a holistic approach to visualize metabolic networks compliant with the SBGN.
Our approach starts with loading a metabolic network and mapping it to a clustered graph structure to model the hierarchy of subcellular locations.
The nodes are then arranged through vectorized stress majorization using domain-specific constraints in a multilevel setup.
This leads to a SBGN-compliant layout.
To distinguish certain reactions at subcellular locations, we developed a visualization technique that produces distinct shapes in analogy to an elastic band.
To explore large networks, we provide an expand and collapse interaction in combination with motif simplification.
We determine the degree of the layout's compliance with the SBGN by proposing domain-specific quality metrics.
Our results demonstrate that the formulation of SBGN-specific constraints in the framework of vectorized stress majorization is feasible.
Finally, our evaluation corroborates that our layout approach can faithfully represent metabolic networks.