Related Experiment Videos
Do structurally similar molecules have similar biological activity?
Yvonne C Martin1, James L Kofron, Linda M Traphagen
1Global Pharmaceutical Research and Development, Abbott Laboratories, Abbott Park, IL 60064-6100, USA. yvonne.c.martin@abbott.com
Journal of Medicinal Chemistry
|September 6, 2002
Summary
Computational chemists often discard similar compounds in drug discovery. However, this study reveals only a 30% chance that a similar compound is active, highlighting the probabilistic nature of library design.
Area of Science:
- Computational chemistry
- Medicinal chemistry
- Drug discovery
Background:
- Combinatorial library design and compound selection for screening collections often involve rejecting compounds with high similarity (> or =0.85) to existing ones.
- Similarity calculations are frequently based on Daylight fingerprints and Tanimoto similarity metrics.
Purpose of the Study:
- To evaluate the effectiveness of similarity-based compound rejection in enhancing the proportion of active compounds in drug discovery efforts.
- To investigate the relationship between chemical similarity and biological activity.
Main Methods:
- Analysis of IC(50) values from 115 high-throughput screening assays.
- Utilizing Daylight fingerprints and Tanimoto similarity calculations to assess compound relatedness.
- Comparing the activity enrichment of similarity-rejected compounds against random screening and docking methods.
Main Results:
- Compounds with a Tanimoto similarity of > or =0.85 to an active compound showed only a 30% probability of being active themselves.
- This enrichment, while better than random screening, indicates a significant proportion of inactive compounds within highly similar sets.
- Limitations in Daylight fingerprints and Tanimoto calculations, as well as variable target-macromolecule interactions, contribute to this finding.
Conclusions:
- Similarity-based compound selection in library design is a probabilistic process with inherent limitations.
- Perfect enrichment of active compounds cannot be guaranteed solely through high chemical similarity.
- Drug discovery strategies should account for the statistical nature of chemical space and activity prediction.