KEGG Virus is an interface to virus data that are part of the GENES, KO, GENOME, BRITE, PATHWAY, NETWORK, DISEASE and DRUG databases in KEGG. It also contains a virus-specific dataset of virus ortholog clusters.

The virus category of the GENES and GENOME databases is generated from the bimonthly release of the NCBI RefSeq database. RefSeq GeneID's are used as gene identifiers and the organism code "vg" (and T40000 identifier) is used for the entire set of viral genes. Each virus is distinguished by the NCBI taxonomy identifier, which is used as the "vtax" identifier in the GENOME database.

• Viruses in taxonomy

Based on experimental evidence an increasing number of virus specific KOs are being defined as summarized in the following BRITE hierarchy file.

• Viral proteins

Most virus KOs are defined at the family or genus level, but sometimes at the species level as well. Furthermore, some KOs are defined for the "vp" category of the GENES database, which contains mature peptides cleaved from polyproteins for selected viruses. Such distictions are made to better represent functional orthologs in the pathway maps and other protein interaction networks.

Due to the lack of experimental evidence, defining and assigning KOs for virus genes will be very limited. The KO assignment rate is less than 5% in viruses, while it is over 50% in cellular organisms. In order to supplement KOs, a new attempt has been initiated to computationally generate virus ortholog clusters using the same annotation resource of GFIT tables (see KO assignment tools). Currently, virus ortholog clusters are generated by a simple procedure shown below.

1. The result of vg-vg comparison is stored in the the paralog GFIT table, a variant form of which may be viewed from the "Paralog" button of each vg entry page.
2. The measure of similarity is defined by a modified identity with weighting of min(1, overlap*2/(aalen1+aalen2)) for the identity of the overlap (aligned) region.
3. For each gene its GFIT table is used to collect similar genes above a given threshold of modified identity.
4. GFIT tables of similar genes are then used to collect additional similar genes and this process is repeated until no addition is made.
5. This can be viewed as single-linkage clustering of truncated GFIT tables, which are processed in the order of the decreasing table size.

Virus ortholog cluster (Voc) data are shown in the Voc page linked from the "Voc" button of each vg entry page such as the following:

• Voc page example for vg:944565

Note that the Voc number identifiers starting with 3, 5 and 7 for the threshold values of 30%, 50% and 70%, respectively, are not stable identifiers and will change when the original RefSeq data are updated.

Voc has also been integrated in the keyword search tool shown above, the taxonomy mapping tool linked from the Voc page and the KEGG Genome Browser for searching conserved gene orders in viral genomes.

KEGG Genome Browser is now available for viral genomes and is linked from the "Genome browser" button in the Position field of each vg entry page such as the following:

• KEGG Genome Browser example for vg:43740568

Here a viral genome is defined by the NCBI taxonomy ID (or the vtax identifier of the GENOME database). When multiple RefSeq (NC number) entries are associated to the same taxonomy ID, each can be selected from the Chr menu.

Selected viruses with relevance to human diseases or plant diseases are given T4 identifiers and are part of the GENOME database.

• KEGG selected viruses

Note that each T4 entry is a collection of virus genomes with the same taxonomy ID. Note also that the specific identifier T40000 is not part of KEGG GENOME; it simply represents the virus category of KEGG GENES.

All the viruses present in KEGG GENES are classified according to the NCBI taxonomy, which is based on the ICTV (International Committee on Taxonomy of Viruses) classification, supplemented by KEGG with the traditional Baltimore classification (see br08620). The correspondence between the ICTV realm, kingdom, phylum, class, order, family classification and the seven types of Baltimore classification is shown below.

Virus specific Brite hierarchy files and Brite table files are being developed.

The pathway maps for viral infections are interaction networks of both human proteins (colered in green and linked to human gene entries) and viral proteins (colored in blue and linked to virus KOs).