Numerical errors in minimization based binding energy calculations.

Area of Science:

• Computational chemistry
• Molecular modeling
• Drug discovery

Background:

• Molecular mechanics generalized Born surface area (MM-GBSA) methods are widely used for predicting ligand-binding energies.
• Accurate binding energy prediction is crucial for efficient drug discovery and development.

Purpose of the Study:

• To investigate the impact of small input perturbations on binding energy calculations using MM-GBSA methods.
• To identify the sources of numerical instability in these simulations.

Main Methods:

• Applied small perturbations to ligand coordinates (translations and atom permutations) in binding site.
• Utilized Prime MM-GBSA and MOE MM-GB/VI computational approaches.
• Incorporated protein flexibility into MM-GBSA calculations.

Main Results:

• Minor input changes (e.g., 0.1 Å translation) led to significant binding energy variations (up to 17 kcal/mol).
• Binding energies clustered around discrete values corresponding to specific ligand poses.
• Increased sensitivity and numerical instability were observed when including protein flexibility, particularly due to protein strain terms.

Conclusions:

• MM-GBSA calculations are highly sensitive to small input perturbations, potentially yielding unreliable binding energy predictions.
• Ligand pose multiplicity and protein flexibility significantly contribute to computational instability.
• Limiting the flexible region in protein-flexible MM-GBSA calculations to within 6 Å is advisable to mitigate instability.