Related Experiment Videos
Predicting protein-ligand binding affinities: a low scoring game?
Philip M Marsden1, Dushyanthan Puvanendrampillai, John B O Mitchell
1Unilever Centre for Molecular Science Informatics, Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, UK.
Organic & Biomolecular Chemistry
|November 10, 2004
Summary
This study evaluated five scoring functions for predicting protein-ligand binding affinities. Overall performance was disappointing, but consensus scoring functions showed slight improvements.
Area of Science:
- Computational chemistry
- Structural biology
- Drug discovery
Background:
- Accurate prediction of protein-ligand binding affinities is crucial for drug discovery.
- Existing scoring functions vary in their performance across different protein-ligand complex types.
Purpose of the Study:
- To assess the performance of five established scoring functions on a diverse set of 205 protein-ligand complexes.
- To evaluate the effectiveness of consensus scoring functions derived from individual methods.
Main Methods:
- Benchmarking five scoring functions against experimental binding constants for 205 protein-ligand complexes.
- Analyzing performance on both a diverse dataset and specific subsets (proteinases, aspartic proteinases).
- Developing and testing two consensus scoring algorithms: linear combination and rank averaging.
Main Results:
- No individual scoring function achieved high predictive accuracy (r(2) > 0.32) on the diverse dataset.
- Performance varied across subsets; functions performed better on proteinases than aspartic proteinases.
- Consensus scoring functions demonstrated a modest improvement over the best individual function.
Conclusions:
- Current scoring functions have limitations in predicting binding affinities for diverse protein-ligand interactions.
- Consensus approaches offer a potential strategy to enhance predictive accuracy.
- Further development of scoring functions is needed for robust application in drug design.