KEGG databases

KEGG is an integrated database resource consisting of fifteen manually curated databases and a computationally generated database in four categories as shown in Table ​Table1.1. The databases in the systems information category are PATHWAY, BRITE and MODULE, which constitute the reference knowledge base for understanding higher-level systemic functions of the cell and the organism, including metabolism, other cellular processes, organismal functions and human diseases. The KO database in the genomic information category is another unique database, in which knowledge of molecular-level functions is organized with the concept of functional orthologs. This category also contains GENOME and GENES, which are derived from RefSeq (3), Genbank (4) and NCBI Taxonomy (5) databases and given KEGG original annotations. SSDB is an auxiliary, computationally generated database used for KO-based annotation of the GENES database.

The databases in the chemical information category are COMPOUND, GLYCAN, REACTION, RCLASS and ENZYME, which are collectively called KEGG LIGAND. ENZYME is taken from the enzyme nomenclature database ExplorEnz (6) and given additional annotation of sequence data links (1). The health information category consists of DISEASE, DRUG, DGROUP and ENVIRON shown in Table ​Table11 and two outside databases for drug labels: Japanese drug labels obtained from the JAPIC database and FDA drug labels obtained from the DailyMed database. KEGG MEDICUS represents the health information category integrated with these drug labels.

As of 1 October 2016 the following changes were made to KEGG in order to simplify its overall architecture without the loss of content. First, the RPAIR database for reactant pairs (7) was discontinued, but its main content for defining reaction classes (8) remains to be available as pairs of COMPOUND entries. Second, the DGENES database (7) for eukaryotic draft genomes was merged into the GENES database. Third, the plasmids category (1) of the GENES database was discontinued because plasmid genes could be, whenever necessary, incorporated in the addendum category.

Improvement of GENES and KO databases

In protein sequence databases, such as UniProt (9), functional annotation is included in individual sequence entries. In contrast, the KEGG GENES database is annotated by simply assigning KO identifiers (K numbers) and making links to the KO database. The definition field of each GENES entry contains the data source name in parentheses, such as (RefSeq) and (GenBank), indicating that the definition was given by the original database as shown in Figure ​Figure1A.1A. KEGG original annotation is given in the following KO subfield, K19188 in this case. As shown in Figure ​Figure2,2, the KO entry K19188 is based on the experimentally characterized protein sequence CAD47941, which is stored in the addendum category of the GENES database (Figure ​(Figure1B).1B). Among over 20 000 KO entries, about 80% contain references (PubMed links) and about 60% contain sequence data links (in October 2016). Sequence data links are given to the references when sequences submitted by the authors to GenBank/ENA/DDBJ are identified or when sequences used in the experiments can be retrieved by cited references or sequence accessions.

Open in a separate window

Figure 1.

Definition and KO fields of GENES entries. (A) GenBank derived entry nca:Noca_0613 and (B) manually created entry ag:CAD47941 in the addendum category.

Open in a separate window

Figure 2.

KO entry K19188 defined from an experimentally characterized protein sequence, ag:CAD47941 for EC:3.7.1.19.

The addendum category has made it possible to define KOs independently from KEGG organisms (complete genomes) and has filled the gaps of missing sequence links in the KEGG pathway maps. It is now used to associate sequence data to Enzyme Nomenclature and to create various sequence data collections, including antimicrobial resistance genes (1) and cytochrome P450s. The EC number list of Enzyme Nomenclature contains references reporting experimental characterization of enzymatic reactions. Unfortunately, however, it does not contain any enzyme sequence information, and links between sequences and EC numbers are left to interpretation of individual users and database developers. The KEGG ENZYME database is now annotated, whenever possible, with sequence data links given by the same criteria mentioned above.

Improvement of PATHWAY database

The KEGG PATHWAY database has been and will continue to be the main database in KEGG. It consists of manually drawn reference pathway maps together with organism-specific pathway maps that are computationally generated by matching KO assignments in the genome with reference pathway maps. In regular maps rectangular nodes are linked to KOs and the matching is shown by coloring. For example, the organism-specific glycolysis pathway hsa00010 with green coloring is generated from the hsa (Homo sapiens) genome and the reference pathway map00010. Table ​Table66 shows the list of manually drawn pathway maps by category. Chemical structure transformation maps and drug structure maps are special maps showing relationships between chemical/drug structures and are not subject to genome-based expansion. Global maps (map numbers 01100s) and overview maps (map numbers 01200s) are intended to show global and overall features of metabolism. No rectangles are used and lines connecting circles (chemical compounds) are linked to KOs in the reference pathway maps and colored in the organism-specific pathway maps. An additional feature of these maps is that KEGG modules and reaction modules (12) are embedded in the pathway diagrams.

On average one or two new maps are released every month, and existing maps are constantly updated. As indicated in Table ​Table66 human diseases have become one of the major categories of the KEGG PATHWAY database. Recent additions include the subcategories of drug resistance for both antimicrobial and antineoplastic agents. For example, Figure ​Figure4A4A is a disease pathway map for non-small cell lung cancer (hsa05223), where known genetic alterations are marked in red. Targeted drug names are given on this map: Gefitinib (D01977) and Erlotinib (D04023) for EGFR mutation and Crizotinib (D09731) for ALK fusion gene. Figure ​Figure4B4B is a drug resistance map for EGFR tyrosine kinase inhibitor resistance (hsa01521) caused by a secondary mutation in EGFR or activation of alternative pathways such as involving MET. A newer generation of EGFR inhibitor, Osimertinib (D10766), is for the former type of resistance.

Open in a separate window

Figure 4.

KEGG pathway maps for (A) non-small cell lung cancer (hsa05223) and (B) EGFR tyrosine kinase inhibitor resistance (hsa01521). Some of the well-known molecular target drugs are included in these maps.

Perturbed KEGG molecular networks

The KEGG pathway maps describe, for example, how beta-lactam antibiotics are synthesized as natural products (map00311 for penicillin and cephalosporin, map00332 for carbapenem and map00261 for monobactam). However, the pathway map of peptidoglycan biosynthesis (map00550) does not describe how beta-lactams interact with penicillin binding proteins and affect bacterial cell wall biosynthesis, because this type of interaction is considered as perturbation to the normal pathway. Similarly, drugs as well as genetic and environmental factors of diseases are considered as perturbants to the KEGG molecular networks, and they are not explicitly shown in the KEGG pathway maps except for a small number of disease associated genes marked in red (Figure ​(Figure4).4). Instead, perturbations are described in the DISEASE and DRUG databases.

The DISEASE database is a collection of disease entries, each consisting of a list of disease genes, carcinogens (for cancers), pathogens (for infectious diseases) and other environmental factors. The DRUG database, which is a comprehensive collection of approved drugs, contains drug target information and drug metabolism information, the latter further divided by metabolizing enzymes and transporters, as well as by substrates, inhibitors and inducers. Thus, the DISEASE and DRUG databases may be integrated with PATHWAY and other molecular network databases to understand perturbed molecular networks. Computationally generated pathway maps are available with the special five-letter organism code ‘hsadd’ for disease genes and drug targets shown on human pathway maps with coloring of pink and light blue, respectively.