Related Experiment Video
Updated: Jul 13, 2026
11:13
Identification of Kinase-substrate Pairs Using High Throughput Screening
Published on: August 29, 2015
8.6K
AlphaFold Kinase Optimizer: Enhancing Virtual Screening Performance Through Automated Refinement of AlphaFold-Based
Sergei Evteev1, Yan Ivanenkov1, Andrew Aiginin2
1Insilico Medicine Hong Kong Ltd., Hong Kong, Pak Shek Kok, Hong Kong.
Proteins
|September 16, 2025
Summary
AF Optimizer refines protein structures for drug design. This deep learning tool improves binding site geometry, leading to more accurate molecular docking and successful drug candidate identification in virtual screening studies.
Area of Science:
- Computational Biology
- Structural Biology
- Drug Discovery
Background:
- AlphaFold generates protein 3D structures but has limitations for structure-based drug design.
- Refining protein models is crucial for accurate prediction of drug-target interactions.
Purpose of the Study:
- Introduce AF Optimizer, a deep learning approach to refine protein geometry for drug design.
- Enhance the accuracy of AlphaFold-generated models for virtual screening and molecular docking.
Main Methods:
- Developed AF Optimizer, integrating neural network scores and binding free energy calculations.
- Applied AF Optimizer to refine the TTK protein structure.
- Performed virtual screening using the optimized AlphaFold model.
Main Results:
- Reduced steric clashes between ligands and the refined protein model.
- Achieved more precise molecular docking and virtual screening outcomes.
- Demonstrated hit enrichment in a prospective in vitro study, validating the approach.
Conclusions:
- AF Optimizer significantly enhances the utility of AlphaFold models in drug discovery.
- The refined protein models improve the efficiency and accuracy of virtual screening pipelines.
- This deep learning-assisted method holds promise for accelerating structure-based drug design.
