Processes 2018, 6,56
Technically, the simulation output is a model of solutions to the relational and mathematical
descriptions under particular conditions; and the mathematical descriptions are a model of the
mechanistic explanation in I given particular assumptions and constraints. Consequently, when
“mechanistic model” or “mechanistic simulation model” is used to describe the work product, it can be
difficult for a reader to know which model is being identified. To avoid misinterpretations, this type of
work product can be identified accurately as a Simulation of a Mechanistic Explanation. The following
are two related examples.
Example IV.1: The gamma rhythm is one of several characterized oscillations of activity
in the brain (brain waves). The alpha rhythm of about 8 Hz is powerful enough that it can
be readily detected outside of the head, something discovered in the 1920s by Hans Berger.
In contrast to alpha, gamma oscillations are faster (~40 Hz) and more spatially localized, best
detected by electrodes placed directly on the brain surface or into the brain parenchyma.
A Simulation of a Mechanistic Explanation [22] helped explore how these gamma oscillations
could be generated through inhibitory inputs, which were classically thought of as delaying
or eliminating neural activity. Wang and Buzsaki demonstrated a mechanistic explanation
wherein inhibitory inputs could in some cases paradoxically facilitate activity [23]. The dual
roles of inhibition and facilitation allow it to entrain cell activity to a signal originating in
inhibitory cells.
A relatively fine-grained, multi-formalism model is required to represent an entrainment
mechanism by a simulated cell’s inputs, at one scale, and the synchronization of multiple cells to
plausibly generate gamma waves at a network scale. These simulations comprised local systems of
ODEs, combined with a coarse PDE approximation to represent the single neuron, with event-driven
techniques to connect cells into networks. To illustrate where this example fits into the spectrum of
types (Figure 2a), it is useful to focus on the way the authors modeled ion channels, as systems of ODEs.
Two cross-model alternatives were used, a coarse 3-channel and a fine 11-channel representation,
both ultimately derived from the underlying Hodgkin-Huxley framework. Practically, using these
alternatives\helped allow for cross-model validation in the face of the greater computational
complexity of the 11-channel simulations. However, from a model of explanation perspective, it
is important to note that the 11-channel parameterization maps more closely to ion channel biophysics.
So, while both alternatives are simulations of mechanistic models, in that they are numerical solutions
to systems of ODEs, the finer grained 11-channel representation is further to the right on the Figure 2a
spectrum, toward an Analogous-mechanism Model and, ultimately, a Model Mechanism. Hence,
this example exhibits different locations along the spectrum of types. It also demonstrates the use of
methods for moving back and forth along that spectrum.
Example IV.2: More recent mechanistic explorations of gamma oscillations have focused on
their possible role in the genesis of schizophrenia, where abnormalities in gamma oscillations
have been demonstrated. Other clues to the biological explanation of schizophrenia have
come from analogies with psychotomimetic drugs, such as ketamine. More recently,
possible roles of particular molecular abnormalities have been suggested by a genome-wide
association study. These many scales of causality were assessed by Neymotin et al. [24]
using multiscale simulations of a mechanistic explanations to explore how alterations
in one of the neural receptors at molecular scale might produce alterations in gamma
oscillations in neuronal receptors at the molecular scale. By using both dynamical and
information theoretic measures, simulation suggested how anomalies in neuronal activity
might produce disturbances in function—disturbances in information flow. Thus, the
model illustrates several levels of mechanistic explanation, connecting molecular anomalies
with cellular anomalies, network anomalies and information transmission disturbance.
Neurons were modeled with piecewise integrated difference equations, including inputs
on the soma and dendrites, representing transmitted as well as background molecules
191
Technically, the simulation output is a model of solutions to the relational and mathematical
descriptions under particular conditions; and the mathematical descriptions are a model of the
mechanistic explanation in I given particular assumptions and constraints. Consequently, when
“mechanistic model” or “mechanistic simulation model” is used to describe the work product, it can be
difficult for a reader to know which model is being identified. To avoid misinterpretations, this type of
work product can be identified accurately as a Simulation of a Mechanistic Explanation. The following
are two related examples.
Example IV.1: The gamma rhythm is one of several characterized oscillations of activity
in the brain (brain waves). The alpha rhythm of about 8 Hz is powerful enough that it can
be readily detected outside of the head, something discovered in the 1920s by Hans Berger.
In contrast to alpha, gamma oscillations are faster (~40 Hz) and more spatially localized, best
detected by electrodes placed directly on the brain surface or into the brain parenchyma.
A Simulation of a Mechanistic Explanation [22] helped explore how these gamma oscillations
could be generated through inhibitory inputs, which were classically thought of as delaying
or eliminating neural activity. Wang and Buzsaki demonstrated a mechanistic explanation
wherein inhibitory inputs could in some cases paradoxically facilitate activity [23]. The dual
roles of inhibition and facilitation allow it to entrain cell activity to a signal originating in
inhibitory cells.
A relatively fine-grained, multi-formalism model is required to represent an entrainment
mechanism by a simulated cell’s inputs, at one scale, and the synchronization of multiple cells to
plausibly generate gamma waves at a network scale. These simulations comprised local systems of
ODEs, combined with a coarse PDE approximation to represent the single neuron, with event-driven
techniques to connect cells into networks. To illustrate where this example fits into the spectrum of
types (Figure 2a), it is useful to focus on the way the authors modeled ion channels, as systems of ODEs.
Two cross-model alternatives were used, a coarse 3-channel and a fine 11-channel representation,
both ultimately derived from the underlying Hodgkin-Huxley framework. Practically, using these
alternatives\helped allow for cross-model validation in the face of the greater computational
complexity of the 11-channel simulations. However, from a model of explanation perspective, it
is important to note that the 11-channel parameterization maps more closely to ion channel biophysics.
So, while both alternatives are simulations of mechanistic models, in that they are numerical solutions
to systems of ODEs, the finer grained 11-channel representation is further to the right on the Figure 2a
spectrum, toward an Analogous-mechanism Model and, ultimately, a Model Mechanism. Hence,
this example exhibits different locations along the spectrum of types. It also demonstrates the use of
methods for moving back and forth along that spectrum.
Example IV.2: More recent mechanistic explorations of gamma oscillations have focused on
their possible role in the genesis of schizophrenia, where abnormalities in gamma oscillations
have been demonstrated. Other clues to the biological explanation of schizophrenia have
come from analogies with psychotomimetic drugs, such as ketamine. More recently,
possible roles of particular molecular abnormalities have been suggested by a genome-wide
association study. These many scales of causality were assessed by Neymotin et al. [24]
using multiscale simulations of a mechanistic explanations to explore how alterations
in one of the neural receptors at molecular scale might produce alterations in gamma
oscillations in neuronal receptors at the molecular scale. By using both dynamical and
information theoretic measures, simulation suggested how anomalies in neuronal activity
might produce disturbances in function—disturbances in information flow. Thus, the
model illustrates several levels of mechanistic explanation, connecting molecular anomalies
with cellular anomalies, network anomalies and information transmission disturbance.
Neurons were modeled with piecewise integrated difference equations, including inputs
on the soma and dendrites, representing transmitted as well as background molecules
191