How many crystal structures do you need to trust your docking results?

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Structure-based drug discovery relies on accurate prediction of protein-ligand complex poses.
• Experimental structures are ideal but costly; thus, leveraging existing data is crucial for efficient design extrapolation.

Purpose of the Study:

• To assess popular strategies for generating docked poses in structure-enabled drug discovery.
• To explore the trade-off between crystal structure acquisition cost and pose prediction accuracy.
• To identify robust methods for predicting poses of newly designed molecules.

Main Methods:

• Utilized data from the open science COVID Moonshot project, including crystallographic screening data.
• Employed retrospective and prospective analyses to evaluate docking pose generation strategies.
• Applied molecular similarity to identify relevant structures for template-guided docking.

Main Results:

• A straightforward strategy using molecular similarity for template-guided docking successfully predicted poses for SARS-CoV-2 main viral protease.
• Template-based docking of scaffold series proved robust, even with limited available structural data.
• The study developed an open-source pipeline and curated datasets for automated pose modeling.

Conclusions:

• Leveraging existing structural data through template-guided docking is an efficient and robust strategy for structure-based drug discovery.
• The developed pipeline and datasets facilitate downstream tasks like pose scoring, machine learning, and binding free energy calculations.
• This approach enhances the prioritization of compounds for synthesis in drug discovery programs.