Improving molecular docking through eHiTS' tunable scoring function.

Area of Science:

• Computational chemistry
• Drug discovery
• Molecular modeling

Background:

• Scoring functions in molecular docking are system-dependent and lack universal accuracy.
• Existing scoring functions often struggle with precise binding energy and bio-activity predictions.

Purpose of the Study:

• To present three score-tuning approaches to enhance molecular docking performance.
• To address the challenge of non-universal scoring function accuracy in molecular docking.

Main Methods:

• Utilizing experimental data to customize scoring functions for specific docking scenarios.
• Adjusting relative weights of energy terms in the eHiTS software for improved pose and energy predictions.
• Rescaling docking scores using ligand-based similarity for enhanced virtual screening enrichment.

Main Results:

• Demonstrated improved score-RMSD and score-affinity correlations through score tuning.
• Achieved better enrichment factors in virtual screening by rescaling docking scores.
• Validated the methods on diverse targets like CDK2, BACE1, and neuraminidase, and benchmarks like PDBBind and Directory of Useful Decoys.

Conclusions:

• Score tuning offers a robust solution to the limitations of universal scoring functions in molecular docking.
• The presented automated utility in eHiTS enhances the reliability of docking predictions.
• Customized scoring functions significantly improve accuracy in pose prediction, virtual screening, and affinity assessment.