KEGGconverter

KEGGconverter is a web-based tool that transforms metabolic pathway data from the KEGG Pathways database into simulation-ready computational models. It was published in 2009 (BMC Bioinformatics) by Konstantinos Moutselos, Ioannis Kanaris, Aristotelis Chatziioannou, Ilias Maglogiannis, and Fragiskos N. Kolisis, working in the Greek systems-biology and biomedical-engineering community. Its purpose is narrow but pointed: take one or more KEGG pathway descriptions in the KEGG KGML format (KGML) and emit a merged, biochemically coherent model in the standard SBML exchange format (SBML is today's revision; KEGGconverter targets SBML Level 2 Version 1), the lingua franca consumed by quantitative simulators such as COPASI and cytoscape-adjacent CellDesigner.

The conversion runs as a three-stage pipeline. First it merges multiple KGML files, concatenating their entries, relations and reaction elements into one file while rewriting ID attributes to track provenance and stripping redundant cross-pathway links. Second it converts KGML to SBML using XSL stylesheet transformations followed by DOM processing in Java: XSL maps KGML <entry> elements to SBML <species> and <reaction> nodes, and the Java layer handles deduplication and nomenclature. Cryptic KEGG identifiers are decoded against local tab-delimited flat files (compounds, enzyme names, glycans, enzyme-organism), producing readable labels such as "D-Glucono-1,5-lactone 6-phosphate [C01236]". Third, it inserts default kinetics as MathML rate laws — hyperbolic modifier kinetics for simple 1-reactant/1-product/1-modifier reactions and mass-action laws otherwise — so the output is immediately runnable rather than a static diagram.

The tool's real contribution is fixing problems that plain KEGG-to-SBML export inherits from KGML's visualization orientation. It eliminates duplicate species and reactions across merged maps, optionally removes orphan nodes, and converts KEGG map-link relations into explicit SBML reactions so metabolic cross-talk between adjacent pathways can actually be simulated. This makes it a specialized sibling of general KGML converters like KEGGgraph, KEGGtranslator and kgml2sif, which target graph or network formats rather than executable kinetic models.

Limitations are honestly documented. The default rate laws are placeholders requiring manual tuning for quantitative accuracy; the output carries no MIRIAM or SBO annotations; and when several genes encode one enzyme, it uses only the first ortholog match, oversimplifying isozyme structure. As a decade-plus-old academic web service its long-term availability is uncertain, but its merging and default-kinetics workflow remains a clear reference design for turning KEGG cartography into SBML models.