DSX: a knowledge-based scoring function for the assessment of protein-ligand complexes.

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Scoring functions are crucial for molecular docking accuracy.
• Existing functions like DrugScore have limitations with specialized atom types.
• Robust and efficient scoring is needed for modern drug discovery pipelines.

Purpose of the Study:

• Introduce DSX, a novel knowledge-based scoring function.
• Enhance molecular docking performance through improved potentials.
• Provide a robust and flexible tool for the scientific community.

Main Methods:

• Developed distance-dependent pair potentials with specialized atom types.
• Incorporated novel torsion angle and solvent accessible surface-dependent potentials.
• Presented a method for robust pair potential derivation for arbitrary atom types.

Main Results:

• DSX demonstrated superior docking and ranking power compared to other scoring functions.
• Achieved improved performance with reduced runtime requirements.
• Validated the beneficial combination of DSX with torsion angle and desolvation potentials.

Conclusions:

• DSX offers a robust, flexible, and high-performing scoring function for molecular docking.
• The new potentials and derivation method overcome limitations of previous approaches.
• DSX is compatible with popular docking engines and freely available.