Related Experiment Video
Updated: Jun 28, 2026
16:41
A Protocol for Computer-Based Protein Structure and Function Prediction
Published on: November 3, 2011
Virtual screening for R-groups, including predicted pIC50 contributions, within large structural databases, using
Richard D Cramer1, Phillip Cruz, Gunther Stahl
1Tripos International, 1699 South Hanley Road, St. Louis, Missouri 63144, USA. cramer@tripos.com
Journal of Chemical Information and Modeling
|October 30, 2008
Summary
This study introduces an objective method for R-group searching in drug discovery, improving pIC50 predictions and enabling efficient virtual screening of chemical databases for novel drug candidates.
Area of Science:
- Medicinal Chemistry
- Computational Chemistry
- Drug Discovery
Background:
- R-group searching in large structural databases is desirable for drug discovery but currently impractical.
- Existing 3D-QSAR (quantitative structure-activity relationship) approaches for pIC50 prediction are subjective and difficult.
- Objective methods are needed to facilitate R-group analysis and enhance virtual screening.
Purpose of the Study:
- To develop an objective and practical method for R-group searching and pIC50 contribution forecasting.
- To evaluate the performance of topomer-based 3D-QSAR models in predicting R-group activity.
- To assess the utility of virtual screening using these models for identifying potent R-groups.
Main Methods:
- Utilized objectively generated topomer poses to replace manual alignments in 3D-QSAR modeling.
- Developed a novel leave-one-R-group-out (LOORG) protocol for rigorous prediction accuracy testing.
- Employed Receiver Operating Curve (ROC) analysis to evaluate virtual screening performance in identifying highly active R-groups.
- Applied topomer-CoMFA models to screen a large chemical database (ZINC) for R-group candidates.
Main Results:
- Topomer-based 3D-QSAR models achieved statistical quality comparable to published manual alignment models with minimal effort.
- The LOORG protocol yielded an average pIC50 prediction error of 0.805 and an average predictive r(2) of 0.495.
- ROC analysis showed an average area of 0.729, indicating a 3-to-1 odds of identifying highly active R-groups.
- Virtual screening identified an average of 5705 R-groups per search, with top candidates showing significantly higher predicted pIC50 values.
Conclusions:
- Objective topomer poses provide a practical and effective alternative to manual alignments for 3D-QSAR modeling.
- Topomer-CoMFA enables accurate pIC50 prediction and successful virtual screening for potent R-groups.
- This approach significantly enhances the efficiency and scope of R-group searching in drug discovery.
