SimCAL

SimCAL computes reaction similarity by integrating molecular fingerprints and physicochemical features to support enzyme identification, enzyme classification, and inhibitor mining.

Key Features:

• Dual-Level Similarity Assessment: Computes similarity at the reaction level across all substrates and products and at the transformation level focused on the transformation center with variable molecular neighborhood.
• Novel Physicochemical Features: Incorporates stereochemistry, mass, volume, and charge into reaction fingerprint computation.
• Customizable Fingerprint Types and Similarity Measures: Supports four different fingerprint types and nine molecular similarity measures for tailored analysis.
• Comparative Assessment Capability: Enables comparative evaluation of different molecular properties to determine their impact on similarity scores.
• Performance and Validation: Evaluated on 3,688,122 reaction pairs with Enzyme Commission (EC) numbers from MetaCyc, achieving area under the curve (AUC) greater than 0.9 and showing strong correlation with EC-BLAST and molecular signature-based approaches.

Scientific Applications:

• Enzyme Identification: Uses reaction similarity to assist in linking reactions to candidate enzymes.
• Enzyme Classification: Supports classification of enzymes based on reaction similarity and EC number relationships.
• Inhibitor Mining: Facilitates identification of potential inhibitors by comparing reaction transformations and molecular features.
• Metabolic Pathway and Mechanism Analysis: Supports detailed analyses of metabolic pathways and reaction mechanisms via comparative similarity assessments.

Methodology:

Computes similarity at reaction and transformation levels; generates reaction fingerprints incorporating stereochemistry, mass, volume, and charge; applies four fingerprint types and nine molecular similarity measures; and evaluates performance on 3,688,122 MetaCyc reaction pairs with EC numbers, reporting AUC > 0.9 and comparisons to EC-BLAST and molecular signature-based approaches.