Performance of 3D-database molecular docking studies into homology models

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Protein active site modeling is crucial for structure-based drug design.
• Homology modeling is a common method for constructing protein models.
• The accuracy of homology models for docking performance is not fully established.

Purpose of the Study:

• To evaluate the performance of docking studies using homology-modeled protein active sites.
• To compare docking results from homology models with those from crystal structures.
• To determine the impact of sequence identity on docking accuracy.

Main Methods:

• Investigated docking performance using CDK2 and factor VIIa screening data.
• Employed homology modeling to construct protein active site models.
• Analyzed the correlation between sequence identity and the identification of active compounds.

Main Results:

• Docking into homology models with >50% sequence identity identified approximately 5 times more active compounds than random.
• The performance of docking to homology models was comparable to docking to crystal structures.
• Higher sequence identity in homology models significantly improved docking accuracy.

Conclusions:

• Homology modeling is a viable strategy for constructing protein active sites for docking studies.
• Accurate homology models, particularly those with high sequence identity, can yield reliable drug screening results.
• This approach offers a cost-effective alternative to using experimental structures for certain drug discovery applications.