Processes 2018, 6,56
In following, the spatial sum of protein levels was calculated and redistributed within
the endothelial cells and membrane agents. The modeling paradigm closely follows that
of a Model Mechanism, where features reflect those of a biological mechanism (Table 1).
An important observation of the simulations was that, by removing information that could
influence simulated cell biasing, the simulated Dll4-notch lateral inhibition mechanism
could generate an alternating pattern of cell fates characteristic of normal tip cell selection.
The authors inferred from simulation results that abnormal patterning could be attributed to
the dynamics of this particular sub-system, rather than any uncontrolled bias.
5.3.2. VII—A Computational Model Mechanism
This characterization differs from that in VI in one important way. All features of a Model
Mechanism instantiated in software meet the definition of a mechanism during operation and may
include all of the features in Table 1. To do so, five requirements are specified early in the workflow
to guide software engineering, mechanism instantiation and simulation refinements. (1) Evidence is
presented that entities, and activities of the virtual mechanism are biomimetic in prespecified ways.
(2) Features of the Model Mechanism during execution are measurable. (3) Measurement of features of
one or more simulation solutions match or mimic measurements of the target phenomenon within
some tolerance (e.g., see [33,40]). (4) Arguments can be presented that, during execution, the Model
Mechanism will have a biological counterpart (blue arrow in Figure 4b). (5) Biomimetic phenomena
are generated during execution. Here are three examples.
Example VII.1: Enhanced mechanism-based explanations are needed to anticipate, prevent
and reverse the liver injury caused by acetaminophen and other drugs. A characteristic
acetaminophen phenomenon—the target phenomenon for this example—is that hepatic
necrosis begins adjacent to central veins in hepatic lobules and progresses upstream.
The prevailing (mechanism-oriented spatiotemporal) explanation (PE) is that location
dependent differences in reactive metabolite formation within hepatic lobules (called
zonation) are necessary and sufficient requisites to account for the phenomenon. Progress
has been stymied because challenging that hypothesis in mice would require sequential
intracellular measurements at different lobular locations within the same mouse, which
is infeasible. Smith et al. [33] circumvent that impediment by performing experiments on
virtual Mouse Analogs, where each is equipped with an in-silico liver that achieved multiple
validation targets. Components and spaces at all levels of granularity are written in Java,
utilizing the MASON multi-agent simulation toolkit. An accurate causal model of the PE that
exhibits all Table 1 features was instantiated and parameterized so that, upon dosing with
objects representing acetaminophen, metabolism and pharmacokinetic validation targets
were achieved. However, the authors demonstrated that the PE failed to achieve the
target phenomenon. Two parsimoniously more complex variants also failed to achieve
the target phenomenon but a fourth variant met stringent tests of sufficiency. Execution
of that forth Computational Model Mechanism provided a multilevel biomimetic causal
explanation of key temporal features of acetaminophen hepatotoxicity in mice including
the target phenomenon. The authors argue that the causal explanation provided during
execution is strongly analogous to the actual causal mechanism in mice.
Example VII.2: Inflammation is not the result of one cell or molecule acting alone. It is a
multicellular process that can be highly localized and yet also have diffuse actions. One of the
keys to understanding tissue-level morphogenesis and spatially localized or heterogeneous
processes such as inflammation is to explicitly study the spatial component—how the cells are
arranged in the tissue and the influences that they have on each other. Thus, to gain insight
into the pathogenesis of gastrointestinal inflammatory diseases, Cockrell et al. [41] developed
a multi-level, discrete-event Model Mechanism that is used to study scenarios of how
196
In following, the spatial sum of protein levels was calculated and redistributed within
the endothelial cells and membrane agents. The modeling paradigm closely follows that
of a Model Mechanism, where features reflect those of a biological mechanism (Table 1).
An important observation of the simulations was that, by removing information that could
influence simulated cell biasing, the simulated Dll4-notch lateral inhibition mechanism
could generate an alternating pattern of cell fates characteristic of normal tip cell selection.
The authors inferred from simulation results that abnormal patterning could be attributed to
the dynamics of this particular sub-system, rather than any uncontrolled bias.
5.3.2. VII—A Computational Model Mechanism
This characterization differs from that in VI in one important way. All features of a Model
Mechanism instantiated in software meet the definition of a mechanism during operation and may
include all of the features in Table 1. To do so, five requirements are specified early in the workflow
to guide software engineering, mechanism instantiation and simulation refinements. (1) Evidence is
presented that entities, and activities of the virtual mechanism are biomimetic in prespecified ways.
(2) Features of the Model Mechanism during execution are measurable. (3) Measurement of features of
one or more simulation solutions match or mimic measurements of the target phenomenon within
some tolerance (e.g., see [33,40]). (4) Arguments can be presented that, during execution, the Model
Mechanism will have a biological counterpart (blue arrow in Figure 4b). (5) Biomimetic phenomena
are generated during execution. Here are three examples.
Example VII.1: Enhanced mechanism-based explanations are needed to anticipate, prevent
and reverse the liver injury caused by acetaminophen and other drugs. A characteristic
acetaminophen phenomenon—the target phenomenon for this example—is that hepatic
necrosis begins adjacent to central veins in hepatic lobules and progresses upstream.
The prevailing (mechanism-oriented spatiotemporal) explanation (PE) is that location
dependent differences in reactive metabolite formation within hepatic lobules (called
zonation) are necessary and sufficient requisites to account for the phenomenon. Progress
has been stymied because challenging that hypothesis in mice would require sequential
intracellular measurements at different lobular locations within the same mouse, which
is infeasible. Smith et al. [33] circumvent that impediment by performing experiments on
virtual Mouse Analogs, where each is equipped with an in-silico liver that achieved multiple
validation targets. Components and spaces at all levels of granularity are written in Java,
utilizing the MASON multi-agent simulation toolkit. An accurate causal model of the PE that
exhibits all Table 1 features was instantiated and parameterized so that, upon dosing with
objects representing acetaminophen, metabolism and pharmacokinetic validation targets
were achieved. However, the authors demonstrated that the PE failed to achieve the
target phenomenon. Two parsimoniously more complex variants also failed to achieve
the target phenomenon but a fourth variant met stringent tests of sufficiency. Execution
of that forth Computational Model Mechanism provided a multilevel biomimetic causal
explanation of key temporal features of acetaminophen hepatotoxicity in mice including
the target phenomenon. The authors argue that the causal explanation provided during
execution is strongly analogous to the actual causal mechanism in mice.
Example VII.2: Inflammation is not the result of one cell or molecule acting alone. It is a
multicellular process that can be highly localized and yet also have diffuse actions. One of the
keys to understanding tissue-level morphogenesis and spatially localized or heterogeneous
processes such as inflammation is to explicitly study the spatial component—how the cells are
arranged in the tissue and the influences that they have on each other. Thus, to gain insight
into the pathogenesis of gastrointestinal inflammatory diseases, Cockrell et al. [41] developed
a multi-level, discrete-event Model Mechanism that is used to study scenarios of how
196