28
Z. Wang and J. Chen
2.4.1 Nuisance in Parameters for In Silico Models
One vital issue for all computational models is that their simulation can never begin
until the in silico objects of the models are completely and properly parameterized. This issue is especially typical for macroscale models, i.e., the fugacity model,
PBTK model, or systems toxicology models [37]. For example, the fugacity model
requires partition coefficients between each pair of environmental compartments and
rate constants of certain types of transforming reactions. Obtaining values of these
parameters for certain models is not a trivial job. Generally speaking, the macroscale
models based on empirical equations are parameter-intensive models in their very
nature. Although experiments can be performed under the OECD or EPA test guidelines to determine a part of the parameters, sometimes wet experiments are timeconsuming and even impossible due to extremely inert reactivity, low solubility, etc.,
of the queried chemicals. It is also not pragmatic to conduct wet experiments for
all the queried chemicals. Hopefully, if adequate experimental data around certain
parameters have been elaborately collected, QSAR tools could be built to quickly predict those parameters. However, QSAR models still suffer from limited application
domains and weak mechanistic interpretation.
Computational chemistry models can directly simulate the microscopic process
of certain partitions or reactions of chemical substances and calculate the properties
that can be used as parameters for the macroscale models. For example, advanced
quantum chemistry modeling techniques have been proven to be able to calculate the
gaseous reaction rate constants of semi-volatile organic compounds (SOCs) with airborne radicals very accurately [82, 83]. Meanwhile, molecular dynamics simulation
with enhanced sampling strategy could re-establish free energy curves of water–air
partition of certain chemicals with properly developed force field parameters [84],
then the hydration free energy can be readily translated into water–air partition coefficients needed for the macroscale models. A warning on these models is that establishment of the modeling systems is sometimes truly sophisticated and simulation
for the modeling systems requires a huge amount of computational resources and
time. Of note, the parameterization of empirical force fields for xenobiotic small
molecules could be a very serious issue when practicing associated simulations.
Note that even with CGenFF or GAFF, the generated FF parameters for complex
xenobiotic molecules may still bear very high penalties indicating that these generated FF parameters are not suitable for the simulation. Closer cooperation between
computational toxicologists and theoretical chemists, thus, should be encouraged in
order to overcome this issue. Nevertheless, computational chemistry models based
on relatively rigorous theories could provide us unprecedented details on molecular mechanisms of atom-level behavior of chemicals, which seems to envision an
ultimate solution for predicting the macroscale parameters.
Z. Wang and J. Chen
2.4.1 Nuisance in Parameters for In Silico Models
One vital issue for all computational models is that their simulation can never begin
until the in silico objects of the models are completely and properly parameterized. This issue is especially typical for macroscale models, i.e., the fugacity model,
PBTK model, or systems toxicology models [37]. For example, the fugacity model
requires partition coefficients between each pair of environmental compartments and
rate constants of certain types of transforming reactions. Obtaining values of these
parameters for certain models is not a trivial job. Generally speaking, the macroscale
models based on empirical equations are parameter-intensive models in their very
nature. Although experiments can be performed under the OECD or EPA test guidelines to determine a part of the parameters, sometimes wet experiments are timeconsuming and even impossible due to extremely inert reactivity, low solubility, etc.,
of the queried chemicals. It is also not pragmatic to conduct wet experiments for
all the queried chemicals. Hopefully, if adequate experimental data around certain
parameters have been elaborately collected, QSAR tools could be built to quickly predict those parameters. However, QSAR models still suffer from limited application
domains and weak mechanistic interpretation.
Computational chemistry models can directly simulate the microscopic process
of certain partitions or reactions of chemical substances and calculate the properties
that can be used as parameters for the macroscale models. For example, advanced
quantum chemistry modeling techniques have been proven to be able to calculate the
gaseous reaction rate constants of semi-volatile organic compounds (SOCs) with airborne radicals very accurately [82, 83]. Meanwhile, molecular dynamics simulation
with enhanced sampling strategy could re-establish free energy curves of water–air
partition of certain chemicals with properly developed force field parameters [84],
then the hydration free energy can be readily translated into water–air partition coefficients needed for the macroscale models. A warning on these models is that establishment of the modeling systems is sometimes truly sophisticated and simulation
for the modeling systems requires a huge amount of computational resources and
time. Of note, the parameterization of empirical force fields for xenobiotic small
molecules could be a very serious issue when practicing associated simulations.
Note that even with CGenFF or GAFF, the generated FF parameters for complex
xenobiotic molecules may still bear very high penalties indicating that these generated FF parameters are not suitable for the simulation. Closer cooperation between
computational toxicologists and theoretical chemists, thus, should be encouraged in
order to overcome this issue. Nevertheless, computational chemistry models based
on relatively rigorous theories could provide us unprecedented details on molecular mechanisms of atom-level behavior of chemicals, which seems to envision an
ultimate solution for predicting the macroscale parameters.