Related Experiment Video
Updated: Aug 7, 2025
08:49
Incorporating Target Protein Structure Flexibility and Dynamics in Computational Drug Discovery Using Ensemble-Based Docking Analysis
Published on: June 20, 2025
437
Are Deep Learning Structural Models Sufficiently Accurate for Virtual Screening? Application of Docking Algorithms to
Anna M Díaz-Rovira1, Helena Martín2, Thijs Beuming3
1Barcelona Supercomputing Center, Jordi Girona 29, E-08034 Barcelona, Spain.
Journal of Chemical Information and Modeling
|March 9, 2023
Summary
Machine learning models like AlphaFold2 show promise for drug discovery. However, using their predicted protein structures directly in virtual screening requires post-processing for accurate hit-finding.
Area of Science:
- Structural biology
- Computational drug discovery
- Protein structure prediction
Background:
- Machine learning models, including AlphaFold2, have revolutionized protein structure prediction.
- Their application in drug discovery, particularly virtual screening, is under active investigation.
- Few studies have assessed virtual screening performance using models with limited template information.
Purpose of the Study:
- To evaluate the utility of AlphaFold2-predicted protein structures, generated with minimal template data, for hit-finding in virtual screening.
- To investigate the impact of post-processing on the accuracy of docking studies using these models.
Main Methods:
- Developed a modified AlphaFold2 version excluding templates with >30% sequence identity.
- Performed rigid receptor-ligand docking studies using the generated protein structures.
- Compared results with and without post-processing modeling of the binding site.
Main Results:
- Out-of-the-box AlphaFold2 models, even with low template identity, are suboptimal for direct virtual screening.
- Post-processing modeling is crucial to refine the predicted binding site into a more realistic holo model for improved docking accuracy.
- Previous work demonstrated quantitative accuracy with free energy perturbation methods.
Conclusions:
- AlphaFold2 holds potential for drug discovery, but its direct application in virtual screening needs refinement.
- Implementing post-processing steps to generate realistic holo models is essential for successful hit-finding campaigns.
- Further research is warranted to optimize the use of AI-predicted structures in drug discovery pipelines.
Related Concept Videos
01:24Protein Organization
6.7K
Proteins are polymers of amino acid residues. They are versatile and responsible for different cellular functions, including DNA replication, molecular transport, catalysis, and structural support. Proteins have a hierarchical structure comprising at least three levels of organization: primary, secondary, and tertiary structure. Some large proteins have a quaternary structure where individual protein subunits are linked together.
The primary structure of a protein is its amino acid sequence....
The primary structure of a protein is its amino acid sequence....
6.7K
01:49Conserved Binding Sites
4.3K
Many proteins’ biological role depends on their interactions with their ligands, small molecules that bind to specific locations on the protein known as ligand-binding sites. Ligand-binding sites are often conserved among homologous proteins as these sites are critical for protein function.
Binding sites are often located in large pockets, and if their location on a protein’s surface is unknown, it can be predicted using various approaches. The energetic method computationally...
Binding sites are often located in large pockets, and if their location on a protein’s surface is unknown, it can be predicted using various approaches. The energetic method computationally...
4.3K