Database clustering with a combination of fingerprint and maximum common substructure methods

Area of Science:

• Cheminformatics
• Computational Chemistry
• Drug Discovery

Background:

• Clustering large chemical databases is crucial for scaffold identification and library prioritization.
• Existing methods may lack efficiency or precision in handling vast datasets.

Purpose of the Study:

• To develop an efficient, stepwise method for clustering large chemical databases.
• To identify conserved substructures and distinct chemical entities within databases.

Main Methods:

• Stepwise clustering using an extended exclusion sphere algorithm with Tanimoto coefficients and Daylight fingerprints.
• Iterative extraction of maximum common substructures from clusters.
• Merging of clusters based on common substructures and similarity to singletons.

Main Results:

• The method successfully identifies tight clusters with conserved substructures.
• It generates singletons only for truly distinct chemical structures.
• Demonstrated utility in identifying frequent scaffolds, selecting analogues, and prioritizing commercial libraries.

Conclusions:

• The developed method provides an efficient and accurate approach for chemical database clustering.
• It aids in understanding chemical space, facilitating drug discovery and library management.