Structuring Chemical Space: Similarity-Based Characterization of the PubChem Database

Area of Science:

• Computational chemistry and cheminformatics.
• Data mining and machine learning applications in drug discovery.

Background:

• The vastness of chemical space presents challenges for comprehensive analysis.
• Understanding molecular relationships is crucial for identifying novel compounds with desired properties.

Purpose of the Study:

• To develop and apply a hierarchical Affinity Propagation (AP) clustering algorithm for analyzing large molecular databases.
• To uncover the intrinsic structural organization of chemical databases based on molecular similarity.

Main Methods:

• Utilized a hierarchical version of the Affinity Propagation (AP) clustering algorithm.
• Applied the algorithm to a LINGO-based similarity matrix of a 500,000-molecule subset from the PubChem database.
• Employed numerical diagonalization of the similarity matrix through hierarchical clustering.

Main Results:

• Successfully analyzed a large-scale molecular database (PubChem subset) efficiently and without bias.
• The hierarchical AP clustering revealed the intrinsic, target-independent structure of the chemical database.
• Mapped molecules with similar physical properties or experimentally verified binding to the same biological target.

Conclusions:

• The combination of hierarchical AP clustering and LINGO similarity is a powerful, unbiased method for large database analysis.
• This approach effectively reveals inherent molecular relationships and structures within chemical databases.
• Facilitates the discovery of compounds with similar properties or biological activities.