Related Experiment Videos
Evaluation of comparative protein modeling by MODELLER
1Rockefeller University, New York, NY 10021, USA.
Proteins
|November 1, 1995
Summary
3D protein models generated by MODELLER show good accuracy, especially when template proteins share high sequence identity. Errors are mainly in exposed loops and insertions, comparable to NMR vs. X-ray crystallography differences.
Area of Science:
- Structural biology
- Computational biology
- Biophysics
Background:
- Comparative protein modeling is crucial for understanding protein function.
- MODELLER is a widely used software for homology modeling.
- Assessing the accuracy of predicted protein structures is essential.
Purpose of the Study:
- To evaluate the accuracy of 3D protein models generated by MODELLER.
- To identify regions prone to errors in comparative protein modeling.
- To establish benchmarks for model accuracy based on template similarity.
Main Methods:
- Comparative protein modeling using MODELLER.
- Analysis of stereochemistry and root-mean-square deviation (RMSD) against crystallographic structures.
- Evaluation of model accuracy in relation to template sequence identity and structural features.
Main Results:
- MODELLER-generated models exhibit good stereochemistry and accuracy comparable to template structures.
- Largest errors are observed in exposed loops, insertions, and non-conserved regions.
- Models with >40% sequence identity to templates achieve ~1 Å RMSD for 90% of mainchain atoms.
Conclusions:
- MODELLER provides reliable 3D protein models, particularly with high-sequence-identity templates.
- Understanding error-prone regions aids in refining protein modeling strategies.
- The accuracy of MODELLER predictions is comparable to experimental structure determination methods.