KEGG Syntax for comparison of organisms, organism groups, and viruses by conserved gene repertoires

Abstract

The KEGG Orthology (KO) system has been used as a basic framework of the KEGG database for linking genes in the genome to pathways and other molecular networks. KOs are manually defined orthologs representing conserved genes with relevance to molecular and higher‐level functions. Here, the KO system is used to define genomic features, including the gene content by the composition of KOs and the gene order on the chromosome by the sequence of KOs, and to compare organisms, organism groups and viruses according to the similarity of KO compositions and KO sequences. The compositions of KEGG modules, manually defined functional sets of KOs, and VOGs, computationally derived ortholog groups for viruses, are also used as genomic features. The similarity analysis of organism groups by KO compositions has revealed a close relationship of Bacteria and Archaea, which differs from the three domains of life in evolution, but which is likely to reflect similarity of metabolic and other cellular functions represented by KOs. A collection of tools is named KEGG Syntax (https://www.kegg.jp/kegg/syntax/) and can be used to analyze the user's data as well as the genomes already registered in KEGG.

1. INTRODUCTION

The KEGG database (https://www.kegg.jp) has been developed as a model of the biological systems (Kanehisa et al., 2025), which enables understanding of higher‐level functions of the cell and the organism from molecular‐level data such as genomic and transcriptomic data. The KEGG model consists of KEGG pathway maps and other molecular networks of interactions, reactions and relations, as well as the KEGG Orthology (KO) system for linking genes and molecules to molecular networks. The KEGG molecular networks are manually created in a generic way using KOs as network nodes, which are orthologs with conserved function defined from experimental evidence in specific organisms and expanded to other organisms by sequence similarity. For a given molecular dataset, the procedure called KEGG mapping (Kanehisa et al., 2022; Kanehisa & Sato, ²⁰²⁰) identifies specific parts of the KEGG molecular networks that can be related to specific higher‐level functions. The KEGG database also contains an increasing number of organisms with completely sequenced genomes, currently over 11,000, which are taken from NCBI RefSeq (Goldfarb et al., 2025) and GenBank databases (Sayers et al., 2025). Protein‐coding genes and RNA genes in each genome are annotated with KOs, and organism‐specific pathways and other molecular networks are generated by KEGG mapping.

Since KOs were originally developed for linking genomes to pathways, the coverage of potential orthologs was not sufficient in the past, in comparison to a number of ortholog databases available (Majidian et al., 2025), some of which have been used to assess the quality of genome assemblies and gene repertoires (Tegenfeldt et al., 2024). As the KO annotation rate increases, currently about 52% on average, it has become possible to define genomic features by KOs, such as the gene content by the composition of KOs and the gene order on the chromosome by the sequence of KOs (Kanehisa et al., 2025). KEGG Syntax is a collection of tools for comparison and analysis of organisms, organism groups and viruses, as well as the user's data, by such conserved gene repertoires. In comparison to KEGG Mapper reported in the previous Protein Science tools issues (Kanehisa et al., 2022; Kanehisa & Sato, ²⁰²⁰), KEGG Syntax utilizes the entire set of conserved genes in the genome. It aims at uncovering shared characteristics of metabolic and other cellular functions among biological organisms as encoded in their genomes, which may be compared with standard taxonomy such as the three domains of life (Woese et al., 1990).

2. DATASETS

2.1. Conserved genes

Conserved genes are termed here as orthologs irrespective of their evolutionary origin. They are represented in KEGG as manually defined KOs (KEGG Orthology groups). Each KO is defined from experimentally characterized sequence data in specific organisms and computationally expanded by sequence similarity to cover other organisms. The level of sequence similarity is variable. Tight groups may be defined to represent specific functions in pathways, while broad functional categories may be defined in Brite hierarchical classifications. A tight group may also be defined by extracting a part from a broad group, in which case the broad KO is the remaining part. Furthermore, one KO may contain multiple sequence similarity groups, which is often the case for legacy KOs originally defined from EC (Enzyme Commission) numbers. Each KO is given the K number identifier consisting of K followed by a five‐digit number. As of March 2026, about 28,000 KOs are defined; 94% of them are associated with references (PubMed records), 90% are linked to experimentally characterized sequence data, and 52% and 83% appear in KEGG pathway maps and Brite hierarchies, respectively. Table 1 shows URLs for using KEGG Syntax and for viewing current statistics.

TABLE 1.

Relevant URLs for using KEGG Syntax.

Content	URL
KEGG top page	https://www.kegg.jp
KEGG Syntax	https://www.kegg.jp/kegg/syntax/
KEGG Orthology (with statistics)	https://www.kegg.jp/kegg/ko.html
KO assignment statistics	https://www.kegg.jp/kegg/annotation/ann_stat.html
KEGG Module (with statistics)	https://www.kegg.jp/kegg/module.html
KEGG Virus (with statistics)	https://www.kegg.jp/kegg/genome/virus.html
BlastKOALA	https://www.kegg.jp/blastkoala/
KEGG Mapper	https://www.kegg.jp/kegg/mapper/
Pathway viewer	https://www.kegg.jp/pathway/kid
Brite viewer	https://www.kegg.jp/brite/kid
Module viewer	https://www.kegg.jp/module/kid
Genome browser	https://www.kegg.jp/genome/kid

Note: kid, KEGG identifier shown in: https://www.kegg.jp/kegg/kegg3.html.

For viruses where the KO annotation rate is low, currently about 9%, computationally generated protein orthologs are also used. They are originally called VOCs (Virus Ortholog Clusters) as reported in another Protein Science tools issue (Jin et al., 2023), but they are now called VOGs (Virus Ortholog Groups) in order to limit the term “clusters” to mean positional correlations, such as gene clusters on the chromosome. VOGs are computed at three similarity thresholds: 30%, 50%, and 70%, and each group is given a six‐digit number starting with 3, 5, and 7, respectively. The VOG dataset is updated whenever the NCBI RefSeq release for viruses is updated and is given a new set of group numbers. All proteins in KEGG organisms are compared against VOGs on a weekly basis and are given group numbers when matches are found.

2.2. Conserved gene sets

There are two types of conserved gene sets: functional units in the pathway and conserved gene orders in the genome. Functional units are represented by manually defined KEGG modules, most of which correspond to conserved subpathways of successive reaction steps in metabolic pathways. Each module is given the M number identifier, M followed by a five‐digit number, and defined by the logical expression of K numbers representing the complexity of relationships between enzyme genes (KOs) and catalyzed reactions. The logical expression is suitable for evaluating whether the module is present or absent in the genome, namely, a specific metabolic capacity is present or absent in the organism.

Conserved gene orders in the genome are termed here as synteny. They are computationally identified by pairwise comparison of genomes, for which chromosomal positions of genes are known (NCBI assembly level of Complete Genome or Chromosome). As reported previously (Kanehisa et al., 2025), gene orders are represented by sequences of KOs, which are then compared by a standard sequence alignment method to comprehensively identify locally similar gene orders as matching KO subsequences. Instead of KOs, VOGs may also be used to identify syntenic regions among viruses and between viruses and cellular organisms.

2.3. Taxonomy and organism groups

Taxonomic classifications of cellular organisms (KEGG organisms) and viruses (KEGG viruses) are based on the NCBI Taxonomy database (Schoch et al., 2020). There are differently simplified versions of NCBI taxonomy, which are all represented as Brite hierarchy or table files, and used for taxonomy mapping (Kanehisa et al., 2022). The default taxonomy files are br08611 for cellular organisms and br08621 for viruses organized in the fixed levels of taxonomic ranks: phylum, class, order, family, genus and species (use Brite viewer in Table 1 to examine the content). The KEGG taxonomy file br08601 is further simplified but still follows the NCBI taxonomy. In addition, KEGG organism and virus groups (Brite table file br08608) are defined to capture overall taxonomic distributions of KOs and modules. The grouping is in three levels with the top level consisting of Animals, Plants, Fungi, Protists, Bacteria, Archaea and Viruses. Figure 1 shows the top level organism groups and the Animals part of the second level and how they are defined from the NCBI taxonomy.

FIGURE 1.

A simple classification scheme of KEGG organism and virus groups (https://www.kegg.jp/kegg/tables/br08608.html) is used to capture overall distributions of conserved genes (KOs and VOGs) and conserved gene sets (modules). Here, the top level and the animals part of the second level are shown, together with the correspondence to the NCBI taxonomy.

3. ANALYSIS TOOLS

Table 2 shows the list of KEGG Syntax tools currently available, all of which involve ortholog‐based analysis using either KOs or VOGs. For analysis of the user's data, such as genome or metagenome sequence data, genes must first be annotated with KOs using, for example, the BlastKOALA annotation server (Kanehisa et al., 2016).

TABLE 2.

KEGG Syntax tools.

Tool	Data type	KEGG data analysis	User data analysis
Genome alignment	Sequence of KOs	Pairwise comparison of gene orders among KEGG organisms and viruses	Conserved gene orders between user data and KEGG organism
	Sequence of VOGs	Pairwise comparison of gene orders among KEGG viruses and organisms
	Precomputed VOG alignments	VOG comparison result against all viruses
Genome similarity	KO composition	KEGG organisms and viruses similar to a given genome	KEGG organisms similar to user data
	Module composition	KEGG organisms similar to a given genome	KEGG organisms similar to user data (after module completeness check)
	VOG composition	KEGG viruses and organisms similar to a given genome
	Combined KO composition	Similarity matrix and dendrogram for organism groups	Organism groups similar to user data
Taxonomy mapping	Organism code and taxonomy ID	Taxonomic distribution of KEGG organisms and viruses associated with KOs, modules and VOGs

3.1. Genome alignment

The term genome alignment is used here for comparison of gene orders, how similar genes (orthologs) are aligned in similar ways in different genomes. The similarity of genes is represented by whether genes have the same K number, meaning that they belong to the same ortholog of KO. Thus, the problem of gene order alignment reduces to the problem of sequence alignment, except that the KO sequence consists of 28,000 different units rather than 4 or 20 letters in nucleotide or amino acid sequences. Nevertheless, a standard sequence alignment method can be applied (Goad & Kanehisa, 1982) to identify comprehensively locally similar gene orders in two genomes (Kanehisa et al., 2025). Furthermore, the method can also be applied to VOG sequences, which consist of 51,000 different units at the 30% similarity threshold level, to identify conserved gene orders among viruses and between viruses and cellular organisms.

Figure 2 shows an example of using the genome alignment tool. The gene orders of Escherichia coli K‐12 MG1655 (KEGG organism code: eco) and Salmonella enterica subsp. enterica serovar Typhimurium LT2 (stm) are compared both as KO sequences and VOG sequences. The best match by VOGs consists of five genes and is found to be a part of the longer match by KOs (Figure 2a). This stretch of VOGs is also found in two Klebsiella phages according to the precomputed VOG alignment dataset, from which conserved gene orders between a given organism and all viruses can be retrieved. KEGG genome browser may then be used to examine gene orders of Escherichia, Salmonella and one of the Klebsiella phage genomes by specifying the starting genes to be aligned (Figure 2b). Color coding of genes by the functional category of KOs helps to visually understand how gene orders are conserved. There is actually a tRNA gene shared by the three genomes and located after the five‐gene stretch. This can be detected by KOs but not by VOGs, since VOGs are only for protein‐coding genes.

FIGURE 2.

An example of using the Genome alignment tool in KEGG Syntax. Gene orders in the Escherichia coli (eco) and Salmonella enterica (stm) genomes are compared both as KO sequences and VOG sequences. (a) A local alignment of five genes is identified by VOG sequences (lower), while a longer alignment containing this five‐gene segment is identified by KO sequences (upper). (b) From the precomputed VOG alignment dataset, the five‐gene segment is also found in Klebsiella phages. KEGG genome browser is used to compare three genomes, including a phage genome, by aligning three genes, eco:B3064, stm:STM3208 and vg:55812000. Color coding of genes (boxes) indicates the functional category of KOs.

The genome alignment tool may be used for aligning gene orders of the user's genome and a KEGG organism by the KO sequence comparison. Use BlastKOALA or other annotation servers to assign KOs to an ordered set of genes along the chromosome. Then, prepare a query file consisting of two columns for gene IDs and assigned KOs (or blanks), which is the default output file format of BlastKOALA. Select a KEGG organism to be compared and run the query. Repeat this selection, if necessary, by finding other similar KEGG organisms.

3.2. Genome similarity

The gene order alignment result may be used to define a genome similarity, for example, by the coverage of the aligned regions that include all locally similar regions caused by genome rearrangements. Here, a much simpler measure of genome similarity is introduced using a genome feature, which is the composition of KOs, modules or VOGs. The composition is taken as a distinct set of KOs in the genome, for example, without considering the number of paralogs assigned to the same KO. Each KEGG organism or virus is considered to be characterized by this composition, and the measure of similarity is defined in three different ways:

$$\text{similarity}=\text{match}/\left(\text{num1}+\text{num2}-\text{match}\right)$$

$$\text{similarity1}=\text{match}/\text{num1}$$

$$\text{similarity2}=\left(\text{match}/\text{num1}+\text{match}/\text{num2}\right)/2$$

where num1 = number of distinct KOs in genome 1, num2 = number of distinct KOs in genome 2, match = number of matching KOs in genomes 1 and 2, and KOs should be replaced by modules or VOGs as appropriate.

As an example, when Mycoplasmoides pneumoniae M129 (mpn) with 375 assigned KOs is compared against Enterococcus faecalis V583 (efa) with 1285 assigned KOs, 345 matching KOs are found. Then, the three types of similarity measures are:

$$\text{similarity}=345/\left(375+1285-345\right)=0.262$$

$$\text{similarity1}=345/375=0.92$$

$$\text{similarity2}=\left(345/375+345/1285\right)/2=0.594$$

The first measure is suitable for identifying closely related genomes with similar sizes. Internally this similarity measure for KOs and modules is used to check the quality of KEGG organisms by comparing against related organisms in the same genus or species. The second one would detect local similarity of a shorter genome embedded in a longer genome. This example indicates that most KOs in a small Mycoplasmoides genome are shared in a much larger Enterococcus genome. The third type is introduced to reduce the influence of size differences in the first type as described below.

3.3. Organism group similarity

The KEGG organism group is a simplified taxonomic group (Figure 1) in three hierarchy levels. Here each group in the top and the second levels is characterized by the combined and distinct set of KOs. The third type of similarity measure was most appropriate for comparison of organism groups because it was necessary to normalize the size differences of KO datasets, num1 and num2. Figure 3 shows the result of applying this similarity measure, creating similarity matrices and drawing dendrograms for the six top‐level groups (animals, plants, fungi, protists, bacteria and archaea) and 105 second‐level groups. Since the dendrogram for the second level was consistent with the top level, only the eukaryotes part of 34 groups is shown here. Color‐coding represents the six top‐level groups and indicates the correspondence between the top and second levels, except for protists, which are more like eukaryotes other than animals, plants and fungi.

FIGURE 3.

An example of using the Genome similarity tool in KEGG Syntax. In addition to comparing individual organisms, this tool allows comparison of organism groups by the combined KO composition and displays resulting similarity dendrograms. The dendrograms shown here are for the similarity relationships of six top‐level groups and 34 second‐level groups of eukaryotes. Coloring is based on the top‐level groups. These dendrograms indicate the distinction of eukaryotes and prokaryotes (or the similarity of bacteria and archaea) and the diversity of eukaryotes other than animals, plants, and fungi, collectively named protists.

According to Figure 3 the similarity of bacteria and archaea, namely, the dissimilarity of prokaryotes and eukaryotes is apparent. Instead of using universally conserved 16S rRNA genes, KEGG Syntax compares all known conserved genes represented by KOs. The majority of KOs are limited to specific taxonomic groups and only 2% are shared among all the six top‐level groups (see current statistics in the KEGG Orthology page listed in Table 1). The conserved gene repertoire represents whether specific KOs are present in the genome, which can be related to whether specific cellular functions are present in the organism. Thus, the dendrograms in Figure 3 may be considered as similarity relationships of cellular life, irrespective of its evolution.

4. DISCUSSION

Since 90% of virus genes are associated with VOGs leaving 10% as singletons, the VOG sequence alignment is expected to capture all conserved gene orders in viruses. Although VOGs are not associated with any functional meanings, they are still useful to uncover gene exchanges between viruses and cellular organisms. In contrast, KOs are defined with functional meanings, but due to the lower KO assignment rate of 52%, the KO sequence alignment cannot detect genes without KOs unless they appear inside the aligned region. This disadvantage of KO sequence alignment will be alleviated as the KO system is expanded and improved.

The KEGG database is daily updated. The KEGG Syntax pages, as well as the entire KEGG website, display the most up‐to‐date statistics for various types of data. The similarity dendrograms (Figure 3) are dynamically generated from such the most updated datasets of KOs and KEGG organisms. It remains to be seen how the dendrograms will change as more organisms are added and KO annotations are improved.

AUTHOR CONTRIBUTIONS

Minoru Kanehisa: Conceptualization; software; project administration; resources; writing – original draft. Yoko Sato: Software; resources.

FUNDING INFORMATION

Japan Science and Technology Agency, JPMJND2203.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Kanehisa M, Sato Y. KEGG Syntax for comparison of organisms, organism groups, and viruses by conserved gene repertoires. Protein Science. 2026;35(6):e70652. 10.1002/pro.70652

DATA AVAILABILITY STATEMENT

All tools and data in KEGG Syntax are made freely available at https://www.kegg.jp/kegg/syntax/.