KEGG

Sixteen databases wearing one name. KEGG - the Kyoto Encyclopedia of Genes and Genomes, maintained by Kanehisa Laboratories since 1995 - is the reference resource that turns molecular parts lists into statements about what biological systems actually do, and Datadory delivers its contents as typed rows rather than a browse interface.

The sixteen fall into four groups. Systems information: PATHWAY's 587 reference maps, BRITE functional hierarchies and MODULE functional units. Genomic information: the KO orthology table at 28,430 entries, with GENES holding more than 67.8 million genes across 11,949 organisms beneath per-organism GENOME records. Chemical information: COMPOUND's 19,623 metabolites, GLYCAN structures, 12,473 reactions and ENZYME nomenclature at 6,972 entries. Health information: NETWORK interaction maps, 3,098 DISEASE entries, 12,896 DRUG entries and MEDICUS.

That gene-to-function linkage is why KEGG remains the default citation for systems biology, metabolomics interpretation and biotech target mapping - the consensus layer that pure volume repositories lack. Within Datadory's catalog of 1,744 datasets across 159 viable industries, this record scores 8/10 against a 7.81 catalog mean, one of twenty-four biotechnology records. Get a sample of this dataset cut to your pathways and organisms.

One structured flat-file block per entry, plus the two-column list view and the statistics header that sizes each database:

# one structured entry per object - here, a human pathway map
ENTRY            hsa04010                    Pathway
NAME             MAPK signaling pathway - Homo sapiens (human)
CLASS            Environmental Information Processing; Signal transduction
PATHWAY_MAP      hsa04010  MAPK signaling pathway

# the identifier-list view - two columns, one row per entry
map01100         Metabolic pathways
map01110         Biosynthesis of secondary metabolites

# the statistics header that sizes each database
database=ko      entries=28,430     stamp=2026/08/20

# shape of the slice
grain=entry      types=pathway|ortholog|gene|genome|compound|reaction|enzyme|drug|disease

Read what the rows prove. The pathway block carries identity (ENTRY), a source-qualified name, a classification path (Environmental Information Processing; Signal transduction reads as category, then subcategory) and the map pointer that hangs drawing-level detail off the same identifier. The list view underneath is the shape most join keys take - one row, one identifier, one preferred name - and the statistics header is why scale claims stay checkable: the ko database states 28,430 entries against its own 2026/08/20 stamp rather than asking you to trust a brochure. Rows shown illustrate the verified record shape; a requested sample pulls live entries scoped to the pathways, organisms or molecules you name.

Fifteen documented fields ride on the entry structure, covering identity, classification, genomic placement, chemistry and literature provenance. Definitions below were verified against live records during the August 2026 research pass. Where a field's definition is documented but no worked example cell was captured in that pass, the example column stays empty rather than being decorated with an invented figure - and anything adjacent confirmed during sample preparation folds under additional fields on request rather than being promised blind.

Geography - the subject is life itself rather than any jurisdiction: reference content spans all domains of life plus viruses, and organism - not country - is the filter that matters. A kinase conserved from yeast to human carries one functional meaning across every organism code that shares it.

Temporal - continuously maintained since 1995, which makes KEGG one of the longest-lived curated resources in bioinformatics. The record dates itself: current release notes are dated July 1, 2026, and individual databases carry their own update stamps - the orthology table's most recent read 2026/08/20 - so any extract quotes a checkable currency figure instead of a vibe. Retaining successive deliveries lets you diff release over release rather than re-profiling anything yourself.

Granularity - one structured entry per object: pathway map, ortholog group, gene, genome, compound, reaction, enzyme, drug or disease. Scale check: 587 reference maps, 28,430 orthology groups, more than 67.8 million genes across 11,949 organisms, 19,623 metabolites, 12,473 reactions, 6,972 enzymes and 12,896 drug entries. The orthology layer is the quiet asset - attaching one functional meaning to genes from thousands of organisms took thirty years of curation, which is exactly what no amount of raw sequencing replicates.

API, files, or your warehouse. Daily, weekly, or hourly.

Pick the channel your stack already speaks. The rows arrive identical either way - entries parsed out of the flat files, identifiers resolved, cross-references kept attached - so nothing in your pipeline re-parses reference prose by hand.

Every delivery ships with the field dictionary above plus the sample rows for validation, so your first join attempt happens against evidence rather than hope.

• Functional annotation of gene lists - a differential-expression hit list becomes a pathway story when every gene resolves onto its KO ortholog group and from there onto the maps it participates in; worked flows continue on our ML model training page.
• Metabolomics interpretation - measured peaks resolve against 19,623 COMPOUND entries and their reaction links, so an annotated spectrum ends at named biochemistry instead of an unnamed mass.
• Target and mechanism mapping - pathway membership plus enzyme classification puts candidate targets inside their mechanism context before a program commits; see market sizing for the downstream sizing workflow.
• Knowledge-graph construction - gene-to-pathway-to-compound-to-drug edges arrive typed and keyed, ready to extend outward to sequence, structure and interaction resources on identifiers both sides already share.
• Cross-species comparability - KO ortholog groups are organism-neutral by design, so a mechanism demonstrated in mouse maps onto human without fuzzy symbol matching.
• Competitive and diligence work - which mechanisms have reference maps, and how densely drugs cluster on them, is readable straight from the health-information group.

Data scientists and ML engineers get a functional label for every gene - orthology membership, pathway placement, enzyme class - which turns identifier soup into supervised features; the fuller workflow lives on data scientists working in biotechnology. Bioinformaticians and computational biologists get thirty years of curated consensus behind their enrichment analyses, so a pathway claim survives review. Market researchers and consultants ground technology landscapes in accepted biology instead of vendor paraphrase. Developers building data products get a stable identifier space and a self-describing entry structure that loads once and joins outward cleanly; see developers building in biotechnology. Investors and diligence teams treat the citation footprint as platform moat. Journalists, academics and students get citable answers to 'what does this gene actually do'.

Notes that pair well with this page:

• Biotechnology data hub - the pooled industry view this record sits inside, alongside sequence, structure and study corpora.
• KEGG DRUG Database - the health-information sibling: 12,896 approved-drug entries with targets, metabolism and classification codes.
• The KEGG source profile - publisher scope, resource layout and dating behavior of the underlying estate.
• UniProt protein sequence data - the sequence-and-function layer these records cross-reference on shared identifiers.
• Human Protein Atlas - tissue-level protein evidence to test whether a mapped function shows up where it matters.
• STRING protein interaction data - put each mapped gene back inside its interaction context.