Natural science is where the objective and the subjective meet. When we build data collections, we want them to reflect the objective relations inherent in nature, but also classify these collections, adding subjective relations between data objects. Sometimes these two approaches conflict. We present two examples of blending objective and subjective relations in a single data collection, and discuss some possible generalizations. First example deals with the visualization of metabolic networks in three dimensions. We show that the network topology has a substantial subjective component in this case and discuss possible ways to minimize human inference. In the second example we present a collection of metabolic networks as a (subjective) hierarchical structure with (objective) binary relations between objects, and show how the data visualization may be made more logical via the procedure of "hierarchization of relations".

MOTIVATION: Currently a substantial research effort is devoted to automated representation of metabolic and gene networks. Automatic visualization plays a significant role in such efforts, and becomes an important problem on its own. Graphical visualization of metabolic pathways has to be information dense and not 'overloaded', recognizable and unified, close to traditional and algebraically consistent. The use of three-dimensional 'virtual reality' visualizations may help to understand better the intricate topology of metabolic and regulatory networks. RESULTS: A system of visualizing metabolic networks as graphs in three-dimensional space by means of Virtual Reality Modeling Language (VRML) is presented. The system is based on an XML-compliant MNV ('Metabolic Network Visualizer') language, and comprises MNV language standard and parser, MNV to VRML translator, and interactive pathway constructor, all unified by the HTML graphic user interface. AVAILABILITY: The MNV can be accessed in viewer mode at http://www.patronov.net/sciencevr/mnv/indexview.html or in constructor mode at http://www.patronov.net/sciencevr/mnv/indexmake.html SUPPLEMENTARY INFORMATION: The figures for the paper as well as the Appendices may be found at http://www.patronov.net/sciencevr/mnv/screenshots.html

A successful data collection visualization should satisfy conflicting requirements: unification of diverse data formats, support for serendipity research, support of hierarchical structures, algorithmizability, vast information density, Internet-readiness, etc. The traditional 2D representations of metabolic pathways lack compactness and information density. Virtual reality (VR) has made significant progress in engineering, architectural design, entertainment and communication. Evident advantages of using VR for scientific visualization are subconscious orientation, compromise between overview and detail, memorizing search results, and customization. We experiment with the possibility of using immersive abstract 3D visualizations of metabolic networks. We present the trial Metabolic Network Visualizer (MNV) software, which produces graphical representation of a metabolic network as a VRML (Virtual Reality Modeling Language) world. The entities in the metabolic network are identified as "compartment", "species", "reaction" and "transporter". The properties of the above objects are described using a simple scripting language. The demonstration of the MNV software is available at http://130.239.139.41/sciencevr/mnv/indexview.html.

Proper information visualization has always been viewed as an important tool for research. Traditionally, however, it was regarded as only a tool, with information content and visualization being analogous to an object and its representation, respectively. These days, visualization is increasingly seen as an integral part of information content, without which data collections may lose significant value. This paper is devoted to applying a new visualization scheme-SequenceWorld-to a genetics database. We view a collection of genetic sequences as a browsable virtual world. Sequences cluster according to their homologies. Their relative position, colour and shape codings offer a direct visual perception of a significant amount of information, which would otherwise be available only via extensive analysis of numerical and textual data. We discuss the usability of SequenceWorld for completing traditional and novel genetic database search tasks, particularly since a holistic view of a complete genome has recently become possible.