Structural Key Bit Occurrence Frequencies and Dependencies in PubChem and Their Effect on Similarity Searches

Area of Science:

• Cheminformatics
• Computational Chemistry
• Drug Discovery

Background:

• PubChem utilizes 881-bit structural keys for compound categorization and comparison.
• Limited literature exists on the characterization and impact of these structural keys.

Purpose of the Study:

• To characterize PubChem's 881-bit structural keys by analyzing their frequencies and dependencies.
• To evaluate the effect of these structural keys and their dependencies on compound similarity searching.

Main Methods:

• Frequency analysis of 881-bit structural keys within the PubChem compound database.
• Determination of bit dependencies (co-occurrence patterns).
• Comparison of similarity searches using full versus reduced (dependency-removed) structural keys with FDA-approved drugs.

Main Results:

• The majority of PubChem's structural bits occur infrequently.
• Significant dependencies exist among the structural bits.
• Bit dependencies impact similarity scores and the compounds retrieved in similarity searches.

Conclusions:

• Structural key dependencies influence PubChem compound similarity assessments.
• Careful selection of bits is crucial for effective differentiation of related compounds in drug discovery.