Protein kinases: docking and homology modeling reliability.

Area of Science:

• Biochemistry
• Structural Biology
• Computational Chemistry

Background:

• Molecular docking is crucial for drug discovery, particularly for kinases, a major drug target class.
• Assessing the reliability of docking procedures is essential for accurate structure-based drug design.
• Existing docking methods often struggle with predicting correct ligand poses in kinase active sites.

Purpose of the Study:

• To evaluate the reliability of 17 different molecular docking procedures across six software packages using kinase structures.
• To develop an automated method to enhance the accuracy and statistical reliability of kinase-ligand docking.
• To explore the application of ligand similarity in homology modeling for predicting unknown kinase structures.

Main Methods:

• Performed self- and cross-docking studies on a database of approximately 700 high-resolution kinase structures.
• Analyzed over 80,000 docking calculations to determine pose prediction accuracy.
• Developed and validated an automated procedure incorporating ligand similarity for improved docking and model building.

Main Results:

• Standard docking procedures showed a low success rate (30-37%) for correctly predicting ligand poses in kinase active sites.
• The proposed automated procedure significantly improved docking accuracy by leveraging ligand similarity.
• Ligand similarity-based modeling offered a competitive alternative to sequence homology-based methods for kinase structure prediction.

Conclusions:

• Molecular docking of unknown ligands into kinase targets has limited accuracy with current standard methods.
• An automated, ligand similarity-driven approach enhances docking reliability and can improve kinase structure modeling.
• This strategy provides a valuable alternative for developing kinase models when experimental structures are unavailable.