RetroRules

Download

Get RetroRules datasets and pre-parsed rule/template archives for reproducible research and workflows.

Current release: 3.1.0 • Released 2026-09-06

Licenses & provenance

Unless stated otherwise, data are available under CC BY 4.0 .

RetroRules integrates identifiers, chemical structures and annotations from the public databases MNXref, Rhea, USPTO, ChEBI and UniProt. Those databases may have different licensing terms. Check licenses of each referenced database before redistribution, especially for commercial use.

Current release — v3.1.0

Release v3.1.0 includes three datasets derived from MetaNetX, Rhea, and the USPTO. Each provides reaction templates at radii 0-10 (implicit hydrogens), matching the current web app. Downloads are available in JSON, CSV and TSV formats.

What's new in v3.1.0

  • Updated Rhea, MetaNetX and USPTO RetroRules datasets.
  • Improved detection of cofactors with ill-defined structures.
  • Additional biochemical reactions modeled.

The underlying Rhea, MetaNetX and USPTO source versions are unchanged from v3.0.0.

Datasets

Format

TEMPLATE_ID: str
Template identifier (i.e., RR:…).
TEMPLATE: str
Reaction SMARTS pattern (implicit H).
REACTIONS: list[str]
Source reaction identifiers (e.g., MNXR), semicolon-separated.
REACTIONS_COUNT: int
Number of linked source reactions.
ECS: list[str]
Associated EC numbers, comma-separated.
ECS_COUNT: int
Number of distinct EC numbers.
RADIUS_MIN: int
Minimum modeled radius level.
RADIUS_MAX: int
Maximum modeled radius level.
RADII: list[int]
Radius levels modeled (integers 0-10), comma-separated.
SCORE: float
Enzyme-likeness score in [0, 1].
VALID: bool
Template SMARTS validation flag (1/0).
DATASETS: list[str]
Provenance tags (metanetx, rhea, uspto), comma-separated.

Archives

Previous RetroRules releases.

v3.0.0

Previous multi-dataset release for MetaNetX, Rhea and USPTO.

▸

Released 2025-10-20.

rr02

Pre-parsed rule archives for RetroPath tools.

▸
rr01

Initial RetroRules release

▸