Extending quantum-mechanical benchmark accuracy to biological ligand-pocket interactions.

Area of Science:

• Computational chemistry
• Drug discovery
• Quantum mechanics

Background:

• Accurate prediction of ligand-protein binding affinity is essential for drug design.
• Existing quantum-mechanical benchmarks for ligand-pocket systems are scarce and suffer from method disagreements.
• Ligand-pocket flexibility involves complex electronic interactions requiring robust computational models.

Purpose of the Study:

• Introduce the QUantum Interacting Dimer (QUID) benchmark framework for ligand-pocket systems.
• Provide highly accurate interaction energies for diverse non-covalent binding motifs.
• Evaluate the performance of various computational methods for predicting non-covalent interactions.

Main Methods:

• Developed the QUID benchmark with 170 non-covalent systems.
• Utilized complementary Coupled Cluster (CC) and Quantum Monte Carlo (QMC) methods for robust binding energies.
• Employed symmetry-adapted perturbation theory to analyze non-covalent binding motifs.

Main Results:

• QUID covers a broad range of non-covalent binding motifs and energetic contributions.
• CC and QMC methods achieved excellent agreement (0.5 kcal/mol) for binding energies.
• Dispersion-inclusive density functional approximations showed accurate energy predictions but varied van der Waals forces.
• Semiempirical methods and empirical force fields need improvement for out-of-equilibrium geometries.

Conclusions:

• The QUID benchmark provides highly accurate interaction energies beyond current "gold standard" QM benchmarks.
• QUID facilitates the development and validation of more reliable computational methods for drug design.
• Accurate modeling of non-covalent interactions is critical for predicting binding affinity in ligand-protein systems.