4D Flexible Atom-Pairs: An efficient probabilistic conformational space comparison for ligand-based virtual...

Area of Science:

• Computational chemistry
• Cheminformatics
• Drug discovery

Background:

• 3D virtual screening methods are limited by compound conformations, reducing robustness compared to 2D approaches.
• Utilizing multiple conformers for 3D methods improves results but significantly increases computational cost.
• A novel conformational space encoding using Gaussian mixture models was developed to address these limitations.

Purpose of the Study:

• To develop an efficient and robust virtual screening similarity function.
• To overcome the conformational dependency of 3D virtual screening methods.
• To enable feasible real-world applications of advanced virtual screening.

Main Methods:

• Developed a unique conformational space encoding technique utilizing Gaussian mixture models.
• Created a novel similarity function operating on these Gaussian mixture models.
• Implemented a 4D flexible atom-pair approach for virtual screening.

Main Results:

• The 4D flexible atom-pair approach demonstrated superior performance compared to 15 state-of-the-art 2D and 3D methods on 40 benchmark datasets.
• Achieved an average Area Under the Curve (AUC) of 0.78, outperforming the best 2D (0.74) and 3D (0.72) methods.
• The method achieved an average rank of 1.25 across multiple evaluation metrics, indicating robust performance.

Conclusions:

• The 4D method provides a robust performance for pharmaceutically relevant targets.
• Conformational space encoding allows efficient comparisons, circumventing the weaknesses of single-conformation 3D approaches.
• The 4D flexible atom-pair approach is computationally feasible for real-world applications, with over 100,000 similarity calculations possible on a single CPU.