Dynamic clustering threshold reduces conformer ensemble size while maintaining a biologically relevant ensemble.

Area of Science:

• Computational chemistry
• Drug discovery
• Structural biology

Background:

• Virtual screening requires accurate 3D ligand structures, but multi-conformer ensembles are computationally demanding.
• Reducing conformer numbers is crucial for efficient screening without losing bioactive conformation accuracy.

Purpose of the Study:

• To develop and evaluate novel methods for selecting computer-generated conformers to reduce computational cost in virtual screening.
• To introduce NMRCLUST, a cut-off independent clustering algorithm for optimizing conformer databases.

Main Methods:

• Tested four conformer selection approaches: OMEGA, NMRCLUST, RMS filtering, and averaged-RMS filtering.
• Utilized 65 ligands from the Protein Data Bank, including those with multiple bound conformations.
• NMRCLUST was used to filter OMEGA-generated conformers.

Main Results:

• NMRCLUST reduced conformer numbers by an average of 10-fold compared to OMEGA, with comparable performance in identifying bioactive conformations.
• Both NMRCLUST and OMEGA outperformed RMS and averaged-RMS filtering.
• Proposed specific OMEGA root-mean square filtering thresholds based on the number of rotatable bonds.

Conclusions:

• NMRCLUST effectively filters conformer ensembles, maintaining biological relevance and significantly reducing computational burden.
• The proposed protocol offers a generalizable strategy for optimizing multi-conformer compound collections for virtual screening.
• This approach alleviates complexity in downstream data processing for virtual screening experiments.