24
Z. Wang and J. Chen
CSBP models are based purely on the topology of signaling networks, not considering the heterogeneous cellular surroundings. If the crowded space of real cells is
needed, then agent/individual-based models (ABMs) might be considered [50–53].
But generally ABMs would have single cells as their agents, which are typically
used to simulate the cellular or tissue-level behavior/effects such as tumorigenesis
or vasculogenesis [54]. The agents or cells would act according to a predefined set
of rules. Specialized software and platform tools such as CompuCell3D [55] and
NetLogo (http://ccl.northwestern.edu/netlogo/) could be employed to perform these
simulations. With a similar strategy, tissues, and organs such as hepatic lobule can
also be simulated if certain functional and/or survival dose–response curve data is
integrated into a model, which enables one straightforward form of so-called virtual
tissues [56, 57].
In conclusion, the Systems toxicology models introduced in this section are at the
cutting edge of computational toxicology, most of which are explanatory, tentative,
and not ready for prediction. Nonetheless, in the future context of toxicology and
chemicals risk assessment, these models would promisingly serve as better alternatives than the current non-testing systems.
2.3.4 Molecular Models
As suggested by AOPs, toxic effects of chemicals originate from MIEs, i.e., the interaction between chemical molecules and biomacromolecules [19, 58]. In a generalized
sense, partition/adsorption and transformation of chemicals in either an inorganic or
biological environment can also originate from molecular behavior/events. As previously discussed, the experimental sector of toxicology generally cannot observe
atom-level behavior of the molecules. With the advent of theoretical and computational chemistry, computational toxicologists are now able to establish so-called
molecular models that include the queried chemical molecules and their situational
objects, i.e., their surrounding molecules to be interacted with. In an MIE, the situational objects are typically functional proteins [58]. Meanwhile, in a simulation for
the gaseous transformation of a chemical, the situational objects might be airborne
reactive species such as hydroxyl radical or chlorine radical.
QM methods from computational chemistry can be adopted to calculate electronic
structures of molecules and elementary steps along a chemical reaction pathway. Ab
initio QM methods, including Hartree–Fock (HF) [59], configuration interaction
(CI) [60], many-body perturbation (MP) theory, and coupled-cluster (CC) theory
[61], etc., are based only on fundamental assumptions, such as Born–Oppenheimer
approximation which assumes that the motion of atomic nuclei and electrons in a
molecule can be separated. Among ab initio QM methods, post-HF methods, e.g., CI,
MP, and CC are proved to be able to give very accurate results that are very consistent
with experimental observations. However, ab initio QM methods, especially the postHF methods require a large amount of computer resources. For larger systems with
up to a hundred atoms, these methods would not be feasible currently.
Z. Wang and J. Chen
CSBP models are based purely on the topology of signaling networks, not considering the heterogeneous cellular surroundings. If the crowded space of real cells is
needed, then agent/individual-based models (ABMs) might be considered [50–53].
But generally ABMs would have single cells as their agents, which are typically
used to simulate the cellular or tissue-level behavior/effects such as tumorigenesis
or vasculogenesis [54]. The agents or cells would act according to a predefined set
of rules. Specialized software and platform tools such as CompuCell3D [55] and
NetLogo (http://ccl.northwestern.edu/netlogo/) could be employed to perform these
simulations. With a similar strategy, tissues, and organs such as hepatic lobule can
also be simulated if certain functional and/or survival dose–response curve data is
integrated into a model, which enables one straightforward form of so-called virtual
tissues [56, 57].
In conclusion, the Systems toxicology models introduced in this section are at the
cutting edge of computational toxicology, most of which are explanatory, tentative,
and not ready for prediction. Nonetheless, in the future context of toxicology and
chemicals risk assessment, these models would promisingly serve as better alternatives than the current non-testing systems.
2.3.4 Molecular Models
As suggested by AOPs, toxic effects of chemicals originate from MIEs, i.e., the interaction between chemical molecules and biomacromolecules [19, 58]. In a generalized
sense, partition/adsorption and transformation of chemicals in either an inorganic or
biological environment can also originate from molecular behavior/events. As previously discussed, the experimental sector of toxicology generally cannot observe
atom-level behavior of the molecules. With the advent of theoretical and computational chemistry, computational toxicologists are now able to establish so-called
molecular models that include the queried chemical molecules and their situational
objects, i.e., their surrounding molecules to be interacted with. In an MIE, the situational objects are typically functional proteins [58]. Meanwhile, in a simulation for
the gaseous transformation of a chemical, the situational objects might be airborne
reactive species such as hydroxyl radical or chlorine radical.
QM methods from computational chemistry can be adopted to calculate electronic
structures of molecules and elementary steps along a chemical reaction pathway. Ab
initio QM methods, including Hartree–Fock (HF) [59], configuration interaction
(CI) [60], many-body perturbation (MP) theory, and coupled-cluster (CC) theory
[61], etc., are based only on fundamental assumptions, such as Born–Oppenheimer
approximation which assumes that the motion of atomic nuclei and electrons in a
molecule can be separated. Among ab initio QM methods, post-HF methods, e.g., CI,
MP, and CC are proved to be able to give very accurate results that are very consistent
with experimental observations. However, ab initio QM methods, especially the postHF methods require a large amount of computer resources. For larger systems with
up to a hundred atoms, these methods would not be feasible currently.