Shape-Based Descriptors for Efficient Structure-Based Fragment Growing.

Area of Science:

• Computational chemistry
• Drug discovery
• Structural biology

Background:

• Fragment-based drug design (FBDD) is crucial for identifying high-affinity binders.
• Current computational FBDD workflows are often slow and non-interactive.
• Shape-based descriptors offer speed advantages for virtual screening.

Purpose of the Study:

• To evaluate shape-based directional descriptors, termed ray volume matrices (RVMs), for interactive fragment growing.
• To assess the performance of RVMs in both self-growing and cross-growing scenarios.
• To analyze the runtime and robustness of RVM-based screenings.

Main Methods:

• Development and application of ray volume matrices (RVMs) as shape-based descriptors.
• Evaluation of RVM performance on two datasets of protein-ligand complexes.
• Analysis of screening runtime and 3D perturbation robustness.

Main Results:

• RVMs effectively prefilter fragment candidates for drug design.
• Achieved high accuracy ( < 2 Å RMSD) in up to 84% of self-growing and 66% of cross-growing cases.
• Demonstrated fast query speeds of ~30,000 conformations per second.

Conclusions:

• RVMs facilitate interactive fragment growing workflows in drug discovery.
• The speed and accuracy of RVMs enable rapid exploration of fragment libraries.
• This approach enhances the efficiency of structure-based drug design.