Large-scale similarity search profiling of ChEMBL compound data sets

Area of Science:

• Computational chemistry
• Cheminformatics
• Drug discovery

Background:

• Large-scale similarity searching is crucial for identifying novel drug candidates.
• Two-dimensional (2D) fingerprints are widely used for molecular similarity calculations.
• Evaluating the performance of different 2D fingerprints and search strategies is essential.

Purpose of the Study:

• To conduct a large-scale investigation of similarity search performance using various 2D fingerprints and search strategies.
• To generate a comprehensive similarity search profile of the ChEMBL database.
• To categorize compound activity classes based on search performance and provide a benchmark.

Main Methods:

• Analysis of 266 compound activity classes from the ChEMBL database.
• Application of two distinct 2D fingerprints: MACCS structural keys and Extended Connectivity Fingerprint with bond diameter four (ECFP4).
• Implementation of three nearest neighbor search strategies for each fingerprint.

Main Results:

• A high success rate (approximately 76%) was achieved, with 203 out of 266 cases showing at least 50% compound recovery.
• Early enrichment of active compounds in database search results was a common trend.
• A compendium categorizing activity classes by similarity search performance was developed.

Conclusions:

• State-of-the-art 2D fingerprinting demonstrates significant effectiveness for compound data sets across diverse pharmaceutical targets.
• The study delineates the performance range of current 2D fingerprinting techniques.
• The generated similarity search profile and performance categorization aid in optimizing virtual screening strategies.