DyScore: A Boosting Scoring Method with Dynamic Properties for Identifying True Binders and Nonbinders in...

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Accurate protein-ligand binding affinity prediction is crucial for computer-aided drug discovery.
• Current scoring functions often lack accuracy due to simplified or omitted binding factors, particularly protein-ligand dynamics.

Purpose of the Study:

• To develop novel features characterizing dynamic properties of protein-ligand binding from static structures.
• To create an improved classification model for distinguishing binders from non-binders.

Main Methods:

• Proposed two novel features: geometry-shape matching and dynamic stability.
• Developed DyScore, a binary classification model using Extreme Gradient Boosting.
• Integrated novel features with classical scoring functions.

Main Results:

• DyScore achieved state-of-the-art performance in distinguishing active ligands from decoys on benchmark datasets.
• Both novel features significantly contributed to the improved performance of DyScore.
• DyScore demonstrated superior performance in early recognition for virtual screening.

Conclusions:

• The proposed dynamic features are valuable complements to existing scoring functions.
• DyScore offers a more accurate and efficient approach for protein-ligand binding affinity prediction.
• DyScore advances virtual screening and de novo drug design by improving candidate identification.