Processes 2018, 6,56
and their receptors. Networks of simulated neurons were composed according to a fixed
relationship between three different neuron types. Simulated current injections were
used to drive the network to a baseline activity and then tuned to generate baseline
theta, gamma and theta-modulated gamma oscillations in a Local Field Potential (LFP)
spanning the simulated pyramidal neurons. The LFP oscillations provide the distinguishing
phenomena. The simulated intervention mechanism consisted of turning on and off the
NMDA (N-methyl-D-aspartate) inputs across 16 different cellular locations. Because the
interventions are below the network scale, instantiated by the underlying software and
mapped to the derived properties of the LFP oscillations, this model provides an excellent
example of Simulation of a Mechanistic Explanation. Further, each neuron is, itself, an
example of IV, in that it is a collection of sections (soma and dendrites), each of which is a
system of difference equations propagating the inputs. However, the neuronal network is
designed using random connectivity, since there are no data on actual cell-to-cell connectivity.
Therefore, at this level, the model is only structurally evocative of the referent and thus
approaches a Simulation of an Analogous-mechanism model (V). By using the information
theoretic measures to relate the external inputs to spike outputs, the authors were able to
demonstrate an inverse relation between gamma activity and the ability of the network to
transmit information, to demonstrate how gamma oscillation might underlie information
processing and how gamma oscillation anomalies could underlie the abnormal information
processing in schizophrenia.
5.2.2. V—Simulation of an Analogous-mechanism Model
When starting with a description of an Analogous-mechanism Model (II), the simulation research
goal is often to translate the knowledge contained within its description into simulation output that is
qualitatively and quantitatively similar to measurements of the target phenomenon. When successful,
an accurate descriptor of the work product is Simulation of an Analogous-mechanism Model.
An increasing fraction of computational explanations of phenomena reported in the literature,
including some “mechanistic models” described as being “multiscale” [25], fit reasonably well under
that descriptor (e.g., see [26–31]).
Figure 3b is a snapshot of the process of building upon descriptions in II during two workflow
activities that differ from those for IV in important ways. (1) The scientist creates mathematical
descriptions of the Analogous-mechanism Model in operation. Continuum equations are adapted
from descriptions of engineering, physical, mechanical, chemical, and/or electronic mechanisms.
An important subset of those mathematical descriptions, for example, finite element analysis, goes
beyond continuum mathematical descriptions because they also require numerical analysis techniques.
(2) The mathematics is instantiated in software; features to support users are added; and solvers
are selected. Computational solutions involve solving equations subject to boundary conditions
and/or initial conditions and the implementation undergoes verification. (3) Authors undertake the
iterative process of achieving qualitative and quantitative similarity between simulation output and
measurements of the target phenomenon within some tolerance. The product of that process is output
from selected parameterizations of a Simulation of an Analogous-mechanism Model. The following
are examples.
Example V.1: Based on epidemiological studies, high-density lipoprotein (HDL) is believed
to play an important role in lowering the risk of cardiovascular disease by mediating reverse
cholesterol transport. Therapies that raise HDL-cholesterol, however, have been unable to
confirm this hypothesis and demand a re-examination of the proposed mechanism. It is
known that lipid-poor ApoA-I plays a role in initiating reverse cholesterol transport and that
the drug RG7232 increases HDL-cholesterol. However, the influence of RG7232 on lipid-poor
ApoA-I and reverse cholesterol transport is unclear because their direct measurement during
dosing intervals is problematic. Lu et al. [27] developed an Analogous-mechanism Model
192
and their receptors. Networks of simulated neurons were composed according to a fixed
relationship between three different neuron types. Simulated current injections were
used to drive the network to a baseline activity and then tuned to generate baseline
theta, gamma and theta-modulated gamma oscillations in a Local Field Potential (LFP)
spanning the simulated pyramidal neurons. The LFP oscillations provide the distinguishing
phenomena. The simulated intervention mechanism consisted of turning on and off the
NMDA (N-methyl-D-aspartate) inputs across 16 different cellular locations. Because the
interventions are below the network scale, instantiated by the underlying software and
mapped to the derived properties of the LFP oscillations, this model provides an excellent
example of Simulation of a Mechanistic Explanation. Further, each neuron is, itself, an
example of IV, in that it is a collection of sections (soma and dendrites), each of which is a
system of difference equations propagating the inputs. However, the neuronal network is
designed using random connectivity, since there are no data on actual cell-to-cell connectivity.
Therefore, at this level, the model is only structurally evocative of the referent and thus
approaches a Simulation of an Analogous-mechanism model (V). By using the information
theoretic measures to relate the external inputs to spike outputs, the authors were able to
demonstrate an inverse relation between gamma activity and the ability of the network to
transmit information, to demonstrate how gamma oscillation might underlie information
processing and how gamma oscillation anomalies could underlie the abnormal information
processing in schizophrenia.
5.2.2. V—Simulation of an Analogous-mechanism Model
When starting with a description of an Analogous-mechanism Model (II), the simulation research
goal is often to translate the knowledge contained within its description into simulation output that is
qualitatively and quantitatively similar to measurements of the target phenomenon. When successful,
an accurate descriptor of the work product is Simulation of an Analogous-mechanism Model.
An increasing fraction of computational explanations of phenomena reported in the literature,
including some “mechanistic models” described as being “multiscale” [25], fit reasonably well under
that descriptor (e.g., see [26–31]).
Figure 3b is a snapshot of the process of building upon descriptions in II during two workflow
activities that differ from those for IV in important ways. (1) The scientist creates mathematical
descriptions of the Analogous-mechanism Model in operation. Continuum equations are adapted
from descriptions of engineering, physical, mechanical, chemical, and/or electronic mechanisms.
An important subset of those mathematical descriptions, for example, finite element analysis, goes
beyond continuum mathematical descriptions because they also require numerical analysis techniques.
(2) The mathematics is instantiated in software; features to support users are added; and solvers
are selected. Computational solutions involve solving equations subject to boundary conditions
and/or initial conditions and the implementation undergoes verification. (3) Authors undertake the
iterative process of achieving qualitative and quantitative similarity between simulation output and
measurements of the target phenomenon within some tolerance. The product of that process is output
from selected parameterizations of a Simulation of an Analogous-mechanism Model. The following
are examples.
Example V.1: Based on epidemiological studies, high-density lipoprotein (HDL) is believed
to play an important role in lowering the risk of cardiovascular disease by mediating reverse
cholesterol transport. Therapies that raise HDL-cholesterol, however, have been unable to
confirm this hypothesis and demand a re-examination of the proposed mechanism. It is
known that lipid-poor ApoA-I plays a role in initiating reverse cholesterol transport and that
the drug RG7232 increases HDL-cholesterol. However, the influence of RG7232 on lipid-poor
ApoA-I and reverse cholesterol transport is unclear because their direct measurement during
dosing intervals is problematic. Lu et al. [27] developed an Analogous-mechanism Model
192