Reactant pairs and reaction organization patterns produced by a new rule-based approach.

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Accurate computational detection of reaction similarity is essential for advancing bioinformatics applications.
• Enzyme identification, classification, inhibitor mining, and pathway prediction rely on precise reaction similarity analysis.

Purpose of the Study:

• To provide a validated set of substrate-product pairs clustered by reaction similarity.
• To compare computational predictions with manually curated datasets for accuracy assessment.

Main Methods:

• A new method was employed to split reactions into compound pairs and individual compounds, termed 'architectures'.
• Analysis of 7491 curated reactions from the KEGG-Ligand dataset.
• Reactions were represented in string and tree formats, with clusters of similar splitting patterns identified.

Main Results:

• The study presents substrate-product pairs clustered by shared chemical transformation patterns.
• Comparison with KEGG-RCLASS dataset provided match precision values.
• The developed method facilitates the selection of accurate substrate-product pairs.

Conclusions:

• The provided dataset and methodology offer a reliable resource for computational reaction similarity.
• This work supports the development of more accurate bioinformatics tools for enzyme-related research.