RetroRules 2026: an expanded database combining biochemical and organic reaction templates for pathway discovery

Area of Science:

• Biochemistry
• Computational Chemistry
• Metabolic Engineering

Background:

• RetroRules is a vital resource for reaction templates used in metabolic pathway discovery, reaction prediction, and enzyme engineering.
• Existing versions primarily focused on biochemical reactions.

Purpose of the Study:

• To update and expand the RetroRules database with new biochemical and organic chemistry reaction data.
• To improve the efficiency, specificity, and usability of reaction template representation and access.

Main Methods:

• Integrated updated biochemical sources (MetaNetX, Rhea) and organic chemistry reactions (USPTO).
• Simplified template encoding using implicit hydrogens and minimal atomic descriptors.
• Implemented transformer-based RXNMapper for improved reaction mapping accuracy.
• Redesigned website with an updated Online Template Generator and OpenAPI-defined API.

Main Results:

• RetroRules 2026 contains over 1.17 million templates from nearly 93,000 reactions, covering 5,796 fourth-level EC numbers.
• Expanded scope includes organic chemistry reactions, mass-imbalanced reactions, and a refined radius range (0-10) for specificity control.
• Enhanced data exploration, visualization, and export capabilities through the redesigned interface and API.

Conclusions:

• RetroRules 2026 serves as a comprehensive, cross-domain resource bridging biochemistry and organic chemistry.
• Updates provide broader coverage, controllable specificity, and enhanced usability for high-throughput pathway design, reaction prediction, and enzyme engineering.