Reproducibility of QM/MM Calculations for the SARS-CoV-2 Main Protease

Area of Science:

• Computational Chemistry
• Biochemistry
• Enzyme Catalysis

Background:

• Combined quantum mechanics and molecular mechanics (QM/MM) is widely used for enzyme reaction mechanism studies.
• Recent studies have raised concerns about the reproducibility of QM/MM calculations across different software packages.
• The SARS-CoV-2 main protease and carmofur inhibitor interaction serves as a critical test case for evaluating computational methods.

Purpose of the Study:

• To extend the investigation of QM/MM reproducibility by including additional software (ComQum, ORCA, AMBER) alongside NWChem and Q-Chem.
• To identify the primary sources of variation in QM/MM calculated energies for enzyme reaction mechanisms.
• To assess the impact of QM region size, protein setup, and solvent treatment on calculation reproducibility.

Main Methods:

• Performed QM/MM calculations using five different software packages (NWChem, Q-Chem, ComQum, ORCA, AMBER).
• Utilized a standardized test case: the covalent attachment of carmofur to SARS-CoV-2 main protease Cys-145.
• Investigated the effect of varying QM region size (up to ~1400 atoms), protein relaxation, and solvent models (continuum solvent with dielectric constant 80).

Main Results:

• Significant variations in reaction (ΔE) and activation (ΔE‡) energies were observed across different QM/MM software.
• The primary source of variation was identified as the treatment of charge redistribution at the QM/MM boundary.
• Increasing the QM region size to include atoms within 8 Å (~1400 atoms) led to reasonably converged energies.
• Differences in protein setup and water molecule placement introduced substantial energy variations (up to 114 kJ/mol and 18-57 kJ/mol, respectively).
• QM/MM calculations using minimized structures for solvent-exposed active sites showed poor reproducibility.

Conclusions:

• The choice of QM/MM software and specific implementation details significantly impacts the reproducibility of enzyme reaction mechanism calculations.
• Careful consideration of the QM/MM boundary treatment, QM region size, and protein/solvent environment is crucial for reliable results.
• For solvent-exposed active sites, QM/MM methods incorporating dynamic effects and free energy calculations are recommended over static minimized structures.