Figure 2 shows what the user sees as he moves down following a pathway, in this case a step in the G1 phase of the mitotic cell cycle. The navigation bar on the left shows an expandable outline of the current pathway, keeping track of the context of the current step in the context of the larger pathway. The detail panel on the right features a human-readable summary of the step, an optional diagram, and a summary of all complexes, protein products, and other macromolecules that participate in this step. The detail panel also lists orthologous reactions that occur or are predicted to occur in other model organisms/species, and literature citations directly relevant to the step.(The citations and orthologous reactions are scrolled out of view in Fig. 2.) All gene product names currently link out to SwissProt. Links to EnsEMBL, Ref-Seq, and model organism databases will be phased in during the autumn of 2003/winter of 2004. Naturally, pathways do not stand alone but are interconnected. When a user is viewing the details of a molecular reaction that participates in multiple pathways, the navigation bar at the left expands to show him each of the pathways that the reaction participates in. The bioinformaticists’ view of GK, using an experimental “pathfinder” viewer, is shown in Figure 3. Here we have just executed a query to find the shortest path between D-fructose and pyruvate. GK finds a path through the reactions that define intermediate metabolism and displays them in an interactive format. Although this representation is easiest to understand in the context of classical biochemical pathways, it can be used to explore any biological process; for example, to find the steps that connect a primary RNA transcript located in the nucleus to a capped, spliced, polyadenylated messenger RNA located in the cytoplasm.

From the bioinformaticist’s point of view, GK is a database that organizes proteins and other macromolecules into reactions and organizes these in turn into pathways. This information is invaluable when searching for the biological signatures hidden in large-scale data sets such as microarray expression studies and proteininteraction screens.