A fast clustering algorithm for analyzing highly similar compounds of very large libraries

Area of Science:

• Computational chemistry
• Cheminformatics
• Drug discovery

Background:

• High-throughput screening (HTS) has led to a rapid increase in available screening compounds.
• Chemical compound libraries are often highly redundant, posing challenges for efficient analysis.
• Existing clustering methods are often too slow for very large compound datasets.

Purpose of the Study:

• To develop a fast and efficient clustering method for analyzing large chemical compound libraries.
• To address the redundancy issue in vast collections of screening compounds.
• To provide a tool suitable for handling millions of compounds on standard hardware.

Main Methods:

• An incremental clustering algorithm was employed.
• Two-dimensional (2D) molecular fingerprints were used to represent compounds.
• The developed method, implemented as cd-hit-fp, processes large datasets rapidly.

Main Results:

• The cd-hit-fp method demonstrates high speed, clustering millions of compounds within hours on a single computer.
• The approach effectively groups similar compounds into families, addressing library redundancy.
• The method is suitable for large-scale chemical data analysis.

Conclusions:

• A fast and scalable clustering method for large compound libraries has been developed.
• cd-hit-fp offers an efficient solution for managing and analyzing redundant chemical screening data.
• The tool is publicly available for researchers in drug discovery and cheminformatics.