tinyIFD: A High-Throughput Binding Pose Refinement Workflow Through Induced-Fit Ligand Docking.

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Accurate prediction of ligand binding poses is crucial for structure-based drug discovery.
• Protein side chain flexibility complicates accurate pose prediction using traditional docking methods.
• Expensive refinement steps are often required, increasing the time and cost of drug development.

Purpose of the Study:

• To develop a high-throughput and flexible ligand pose refinement workflow named tinyIFD.
• To improve the accuracy and efficiency of predicting ligand conformations within therapeutic targets.
• To address the limitations of current screening methods in handling protein side chain dynamics.

Main Methods:

• Utilized specialized high-throughput, small-system molecular dynamics (MD) simulation code (mdgx.cuda).
• Implemented an actively learning model zoo approach for enhanced prediction.
• Developed a flexible ligand pose refinement workflow (tinyIFD).

Main Results:

• Achieved 66% and 76% success rates in finding crystal-like poses within the top-2 and top-5 predictions, respectively, on a diverse set of protein targets.
• Demonstrated the workflow's efficacy on SARS-CoV-2 main protease (Mpro) inhibitors.
• Showcased the benefits of the active learning component in the tinyIFD workflow.

Conclusions:

• The tinyIFD workflow offers a significant advancement in high-throughput ligand pose refinement.
• This method improves the accuracy of predicting ligand conformations, reducing the need for costly post-docking refinements.
• The active learning aspect enhances the workflow's performance, particularly for specific targets like SARS-CoV-2 Mpro.