Related Experiment Videos
Further development of reduced graphs for identifying bioactive compounds.
Edward J Barker1, Eleanor J Gardiner, Valerie J Gillet
1Department of Information Studies and Krebs Institute for Biomolecular Research, University of Sheffield, Western Bank, Sheffield S10 2TN, United Kingdom.
Journal of Chemical Information and Computer Sciences
|March 26, 2003
Summary
Reduced graphs offer summary chemical structure representations, performing comparably to Daylight fingerprints in similarity searches. Combining reduced graphs with Daylight fingerprints enhances retrieval performance across diverse datasets.
Area of Science:
- Cheminformatics
- Computational Chemistry
- Drug Discovery
Background:
- Reduced graphs offer simplified representations of chemical structures.
- These representations are valuable for computational analysis and similarity searching.
- Existing methods like Daylight fingerprints are widely used but have limitations.
Purpose of the Study:
- To compare the performance of various reduced graph types in chemical similarity searches.
- To evaluate the efficacy of combining reduced graphs with Daylight fingerprints for improved retrieval.
- To explore the utility of reduced graphs in building Structure-Activity Relationship (SAR) models.
Main Methods:
- Comparison of different reduced graph representations in similarity searches.
- Data fusion experiments combining reduced graphs and Daylight fingerprints.
- Development of SAR models using recursive partitioning on reduced graph data.
Main Results:
- Reduced graphs demonstrated comparable performance to Daylight fingerprints in retrieving active compounds.
- No single reduced graph type consistently outperformed others across all datasets.
- Data fusion of reduced graphs and Daylight fingerprints improved performance in 10 out of 11 datasets.
- A predictive SAR model was successfully built using recursive partitioning with minimal data splits.
Conclusions:
- Reduced graphs are effective tools for chemical structure representation and similarity searching.
- Combining diverse reduced graph types with established fingerprints offers synergistic benefits.
- Reduced graphs show significant potential for efficient SAR model development in drug discovery.