Skip to content

Changes from RAVEN 2

raven-toolbox is not a direct port of RAVEN 2.0 — it is a fresh Python implementation that makes deliberate architectural choices where RAVEN 2.0 showed its age or where the Python ecosystem offers better alternatives. This page documents the most significant departures and the reasoning behind them.

What is not in raven-toolbox

MetaCyc / BioCyc

RAVEN 2.0 includes getMetaCycModelForOrganism for building draft models from the MetaCyc / BioCyc databases. raven-toolbox does not include MetaCyc support. Reasons:

  • The MetaCyc API and data formats have changed substantially since the original implementation, making the RAVEN 2.0 connector unreliable with current database releases.
  • Maintaining an up-to-date connector requires ongoing manual upkeep tied to MetaCyc licensing and format changes.
  • In practice, homology-based reconstruction from a curated template model or KEGG produces better-quality drafts for most organisms; MetaCyc is more useful as an annotation cross-reference than as a primary reconstruction source.

If you rely on MetaCyc reconstruction, continue using RAVEN 2.0 or extract the relevant reactions manually and import them via SBML or Excel.

Architectural changes

Model data structure

RAVEN 2.0 represents models as MATLAB structs with fields .rxns, .mets, .S, .lb, .ub, etc. raven-toolbox uses cobrapy's cobra.Model as its native data structure. This means:

  • All cobrapy-compatible tools work directly with raven-toolbox models.
  • Solver calls go through cobrapy's unified interface.
  • The MATLAB struct format is not used internally; models are loaded directly into cobra.Model objects.

Python packaging

raven-toolbox is distributed via PyPI (pip install raven-toolbox), enabling standard reproducible environments and straightforward CI integration.

Type annotations

All public functions in raven-toolbox carry full Python type annotations and pass static analysis with mypy. This was not possible in the MATLAB toolbox.

What stayed the same

  • Core algorithms (homology search, gap-filling, KEGG reconstruction) use the same logic as RAVEN 2.0 with equivalent default parameters.
  • SBML (.xml) and RAVEN Excel (.xlsx) model formats are fully interoperable between RAVEN 2.0 and raven-toolbox.
  • Function naming follows the same conceptual structure, converted to snake_case (see the function mapping table).