Databases and Systems

influenzae lacks the upper portion of the cycle shown in Figure 3. Interestingly, H.
pylori is complementary, having only the upper portion and lacking the lower
portion. Furthermore, by examining the KEGG ortholog group table, these two
portions turn out to be coded in different operons whenever the operon structure is
observed. These observations suggest that the TCA cycle is actually formed by two
sets of pathways that are under different regulatory control mechanisms.
Network comparison
The four types of networks (Table 2) can be compared by using the search
capabilities of KEGG (Table 4). For example, the genome-pathway comparison is
done as follows. Starting from the genome map of a given organism, the user displays
the area of interest in the enlarged window and asks where in the known biochemical
pathways the genes in the window function. The query can be done by simply
clicking on the button marked PATHWAY. A typical result would be a gene cluster
in the genome forming a functional unit in the biochemical pathway; namely, the
genes in the window code for a set of proteins in successive steps of, say, amino acid
biosynthesis.
Another example of network comparison invlolves a hierarchy versus biochemical
pathways. For example, in the KEGG table of contents page, select the hierarchical
classification (molecular catalog) of enzymes by SCOP 3D folds. By opening the
third-level data for beta/alpha (TIM)-barrel in the hierarchy, the user can search all
occurrences of TIM-barrel proteins against the known metabolic pathways. This is
done by clicking on "Pathway Search by EC" and choosing "Search against 3D
structures in PDB" to limit the search for only those enzymes with known structures.
One of the results of this query is Phenylalanine, tyrosine and tryptophan
biosynthesis, where the last steps of tryptophan biosynthesis are populated by TIM
barrel proteins, which suggests possible gene duplications in the evolution of
pathway formation [ 1].
Pathway reconstruction with reference
In the traditional similarity search of individual genes (or proteins) against
repositories of all known sequences, it is always problematic to determine an
appropriate level of sequence similarity that can be extended to functional similarity.
The prediction tools in KEGG incorporate an additional feature that is used for
interpretation of sequence similarity; namely, the requirement for reconstructing a
complete pathway or a complete functional unit from a set of genes or proteins. The
reference for reconstruction is the set of pathway diagrams, and the refined data set of
ortholog group tables for a limited, but increasing, number of functional units. For
example, by searching sequence similarities against the KEGG ortholog group table
for ABC transporters using a set of consecutive genes in the genome as a query (an
ABC transporter is often coded in an operon), a transporter can be reconstructed with
prediction of substrate specificity according to the subgrouping of the ortholog group
table.
73