2 Background, Tasks, Modeling Methods …
25
On the other hand, semi-empirical (SE) methods have been created that adopt
experimental values or empirical parameters as expedient replacement for certain
complicated time-consuming integral items of ab initio QM methods. Therefore,
the SE methods are much faster than the ab initio QM methods, but less accurate
especially when the molecule or system has non-typical conformations (such as those
of transition states). Well-known SE methods include MNDO (modified neglect of
diatomic overlap) [62], AM1 (Austin model 1) [63] etc.
Besides, DFT is a promising method based on slightly different assumptions from
that of QM theories, which can simulate larger systems, with accuracy comparable
to ab initio QM methods. Therefore, DFT has become a widely applied method for
computational toxicology. For example, the DFT method has been employed to probe
the hydrolysis pathways of antibiotics [64] and sites of metabolism of brominated
flame retardants by reactive oxo-heme of P450 enzymes [65].
For biomacromolecules such as proteins with thousands of atoms or inorganic
systems with explicit solvent molecules in condensed state, the DFT method is generally not feasible. For these systems, the theory must be further simplified. An
empirical force field (FF) that describes the interaction between atoms from a classical mechanics perspective provides a feasible route for the large-scale simulations.
In some widely employed academic force fields such as CHARMM (Chemistry at
HARvard Molecular Mechanics) [66] and AMBER (Assisted Model Building with
Energy Refinement) [67] force fields, potential energy functions are used to describe
bonding lengths, angles, dihedrals, electrostatic/Coulomb interactions, and Van der
Waals interactions. The parameters of FF-based methods/software are more complex
than those of QM or DFT methods/software. Therefore, it is also relatively complicated to establish model systems with FF-based methods. CHARMM and AMBER
force fields have provided compatible topologies and high-quality FF parameters for
common biomolecules such as proteins, DNAs, RNAs, lipids, and carbohydrates,
making it very convenient to simulate these biological systems. Besides, CHARMM
and AMBER also provide CGenFF (CHARMM General Force Field) [68] and
GAFF (General AMBER Force Field) [69] for small molecules, which are useful
for toxicological systems involving xenobiotic molecules. FF-based methods typically include molecular mechanics (MM) that minimize energy of a conformation to
obtain its best geometry, and molecular dynamics (MD) or Monte Carlo simulations
that sample the ensemble space of the simulated systems and generate trajectories
for real-time/post-treatments to obtain useful physical quantities [28].
Unlike QM or DFT methods, typical FF-based methods do not consider the forming and breaking of covalent bonds. However, it is of interests to toxicologists to
simulate chemical reactions taking place in either an inorganic environment or in an
enzymatic environment [70]. Thus, schemes of interfaces between QM and MM
methods have been developed to simulate these special cases, with the reactive
site handled by QM methods and the rest of the system handled by MM methods.
QM/MM method is now mostly utilized for enzymatic systems and for explanatory
purposes.
25
On the other hand, semi-empirical (SE) methods have been created that adopt
experimental values or empirical parameters as expedient replacement for certain
complicated time-consuming integral items of ab initio QM methods. Therefore,
the SE methods are much faster than the ab initio QM methods, but less accurate
especially when the molecule or system has non-typical conformations (such as those
of transition states). Well-known SE methods include MNDO (modified neglect of
diatomic overlap) [62], AM1 (Austin model 1) [63] etc.
Besides, DFT is a promising method based on slightly different assumptions from
that of QM theories, which can simulate larger systems, with accuracy comparable
to ab initio QM methods. Therefore, DFT has become a widely applied method for
computational toxicology. For example, the DFT method has been employed to probe
the hydrolysis pathways of antibiotics [64] and sites of metabolism of brominated
flame retardants by reactive oxo-heme of P450 enzymes [65].
For biomacromolecules such as proteins with thousands of atoms or inorganic
systems with explicit solvent molecules in condensed state, the DFT method is generally not feasible. For these systems, the theory must be further simplified. An
empirical force field (FF) that describes the interaction between atoms from a classical mechanics perspective provides a feasible route for the large-scale simulations.
In some widely employed academic force fields such as CHARMM (Chemistry at
HARvard Molecular Mechanics) [66] and AMBER (Assisted Model Building with
Energy Refinement) [67] force fields, potential energy functions are used to describe
bonding lengths, angles, dihedrals, electrostatic/Coulomb interactions, and Van der
Waals interactions. The parameters of FF-based methods/software are more complex
than those of QM or DFT methods/software. Therefore, it is also relatively complicated to establish model systems with FF-based methods. CHARMM and AMBER
force fields have provided compatible topologies and high-quality FF parameters for
common biomolecules such as proteins, DNAs, RNAs, lipids, and carbohydrates,
making it very convenient to simulate these biological systems. Besides, CHARMM
and AMBER also provide CGenFF (CHARMM General Force Field) [68] and
GAFF (General AMBER Force Field) [69] for small molecules, which are useful
for toxicological systems involving xenobiotic molecules. FF-based methods typically include molecular mechanics (MM) that minimize energy of a conformation to
obtain its best geometry, and molecular dynamics (MD) or Monte Carlo simulations
that sample the ensemble space of the simulated systems and generate trajectories
for real-time/post-treatments to obtain useful physical quantities [28].
Unlike QM or DFT methods, typical FF-based methods do not consider the forming and breaking of covalent bonds. However, it is of interests to toxicologists to
simulate chemical reactions taking place in either an inorganic environment or in an
enzymatic environment [70]. Thus, schemes of interfaces between QM and MM
methods have been developed to simulate these special cases, with the reactive
site handled by QM methods and the rest of the system handled by MM methods.
QM/MM method is now mostly utilized for enzymatic systems and for explanatory
purposes.