Database searching for compounds with similar biological activity using short binary bit string representations of...

Area of Science:

• Computational chemistry
• Cheminformatics
• Drug discovery

Background:

• Identifying biologically active molecules is crucial for drug discovery.
• Similarity searching in compound databases relies on molecular representations.
• Existing molecular fingerprints can be computationally intensive.

Purpose of the Study:

• To develop and evaluate novel, efficient molecular representations called minifingerprints (MFPs).
• To assess the performance of MFPs in similarity searching for biologically active compounds.
• To compare MFP performance against traditional molecular fingerprints.

Main Methods:

• Designed short binary bit strings (MFPs) encoding structural fragments, aromaticity, flexibility, and hydrogen-bonding capacity.
• Analyzed molecular descriptor and fragment distributions in large compound collections for MFP design.
• Conducted systematic "one-against-all" similarity searches on a 364-compound database.
• Determined optimal similarity cutoff values for different fingerprints.

Main Results:

• An MFP with 32 structural fragments achieved <2% false positives and identified ~40% of active compounds.
• Incorporating three 2D molecular descriptors improved MFP performance by 15%.
• The optimized MFP outperformed a complex 2D fingerprint, which had zero false positives but identified <10% of active compounds.

Conclusions:

• Minifingerprints offer an efficient and effective approach for similarity searching in drug discovery.
• Optimized MFPs balance the trade-off between false positives and true positive identification.
• MFPs represent a promising alternative to complex fingerprints for large-scale virtual screening.