Processes 2018, 6,56
A phenomenon that is explained using an Analogous-mechanism Model will be to the
right of I in Figure 1b. As explanatory insight improves and the research workflow advances,
one encounters research reports in which an earlier Mechanistic Explanation is replaced by an
Analogous-mechanism Model. At that stage, authors typically assign names to some or all of
the components of their model that are identical to real components and features of the referent
biological system, that is, they draw directly from vocabularies of anatomical, biochemical, and
physiological ontologies. While conceptually useful, such labeling may encourage conflating model
explanation features with reality, which reduces both clarity and scientific credibility.
5.1.3. III—Model Mechanism
As explanatory knowledge about a phenomenon increases (moving further right on the Figure 1b
spectrum), researchers begin conceptualizing and describing (hypothesizing about) a particular
mechanism-based explanation of the phenomenon (Figure 2c) that is biomimetic; it is not an analogy of
something else. Researchers strive to specify and characterize some or all of the explanatory features
in Table 1. Model Mechanism is an accurate descriptor of a product of that process. Model Mechanisms
are less cartoonish than II and more structured. An early stage model of explanation of this type
would likely be assigned a central location on the Figure 1a spectrum. As the description matures, its
location on all three Figure 1 spectra shifts rightward. Mappings exist between the Model Mechanism’s
discrete entities and activities and biological counterparts. The expectation is that, measurements
of a phenomenon generated during simulation of a Model Mechanism would adequately match
measurements of the actual target phenomenon qualitatively and quantitatively.
Example III.1: An illustrative example is the two-dimensional model mechanism developed
by Norton et al. [19] to facilitate achieving two related goals: (1) improve explanatory insight
into the generation of the four distinguishable morphologies of ductal carcinoma in situ
of the breast. (2) Disentangle the mechanisms involved in tumor progression. Additional
examples are included with those provided below under Group C.
5.2. Group B: Using Simulation to Support and Enhance I and II
5.2.1. IV—Simulation of a Mechanistic Explanation
A frequent simulation research goal is to translate a Mechanistic Explanation (I) into simulation
output that is (or is expected to be) qualitatively or quantitatively similar to reported measurements
of the target phenomenon. An additional goal may be providing predictions and/or further
improving insight into how the target phenomenon (and possibly other phenomena) may be generated.
A Simulation of a Mechanistic Explanation (Figure 3a) builds upon I during three workflow activities.
(1) Relational and continuum mathematical descriptions are developed of the salient explanatory
information within the Mechanistic Explanation. (2) Those descriptions are instantiated in software;
features to facilitate exploratory simulations are added; solvers are selected and the implementation
undergoes verification. (3) An iterative workflow process achieves the desired qualitative and
quantitative similarity between simulation output and measurements of the target phenomenon.
During that process, the model and mathematical descriptions may be revised. To enable another
modeler to independently reproduce reported simulation results, details of those workflow decisions
should be made available when results are published [20]. For the second and third activities, it is
increasingly common for researchers to rely on mathematical modeling tools, such as Matlab (The
MathWorks, Inc., Natick, MA, USA), and/or proprietary or open source systems, including, for
example, physiologically based simulation or emulation packages (e.g., see [21]). Use of standardized
software increases credibility, reliability, and reproducibility by providing some assurance that
the underlying numerical techniques are handled correctly. Use of open source software further
improves reproducibility by making the simulation widely available while also opening the underlying
techniques to later examination for correctness.
190
A phenomenon that is explained using an Analogous-mechanism Model will be to the
right of I in Figure 1b. As explanatory insight improves and the research workflow advances,
one encounters research reports in which an earlier Mechanistic Explanation is replaced by an
Analogous-mechanism Model. At that stage, authors typically assign names to some or all of
the components of their model that are identical to real components and features of the referent
biological system, that is, they draw directly from vocabularies of anatomical, biochemical, and
physiological ontologies. While conceptually useful, such labeling may encourage conflating model
explanation features with reality, which reduces both clarity and scientific credibility.
5.1.3. III—Model Mechanism
As explanatory knowledge about a phenomenon increases (moving further right on the Figure 1b
spectrum), researchers begin conceptualizing and describing (hypothesizing about) a particular
mechanism-based explanation of the phenomenon (Figure 2c) that is biomimetic; it is not an analogy of
something else. Researchers strive to specify and characterize some or all of the explanatory features
in Table 1. Model Mechanism is an accurate descriptor of a product of that process. Model Mechanisms
are less cartoonish than II and more structured. An early stage model of explanation of this type
would likely be assigned a central location on the Figure 1a spectrum. As the description matures, its
location on all three Figure 1 spectra shifts rightward. Mappings exist between the Model Mechanism’s
discrete entities and activities and biological counterparts. The expectation is that, measurements
of a phenomenon generated during simulation of a Model Mechanism would adequately match
measurements of the actual target phenomenon qualitatively and quantitatively.
Example III.1: An illustrative example is the two-dimensional model mechanism developed
by Norton et al. [19] to facilitate achieving two related goals: (1) improve explanatory insight
into the generation of the four distinguishable morphologies of ductal carcinoma in situ
of the breast. (2) Disentangle the mechanisms involved in tumor progression. Additional
examples are included with those provided below under Group C.
5.2. Group B: Using Simulation to Support and Enhance I and II
5.2.1. IV—Simulation of a Mechanistic Explanation
A frequent simulation research goal is to translate a Mechanistic Explanation (I) into simulation
output that is (or is expected to be) qualitatively or quantitatively similar to reported measurements
of the target phenomenon. An additional goal may be providing predictions and/or further
improving insight into how the target phenomenon (and possibly other phenomena) may be generated.
A Simulation of a Mechanistic Explanation (Figure 3a) builds upon I during three workflow activities.
(1) Relational and continuum mathematical descriptions are developed of the salient explanatory
information within the Mechanistic Explanation. (2) Those descriptions are instantiated in software;
features to facilitate exploratory simulations are added; solvers are selected and the implementation
undergoes verification. (3) An iterative workflow process achieves the desired qualitative and
quantitative similarity between simulation output and measurements of the target phenomenon.
During that process, the model and mathematical descriptions may be revised. To enable another
modeler to independently reproduce reported simulation results, details of those workflow decisions
should be made available when results are published [20]. For the second and third activities, it is
increasingly common for researchers to rely on mathematical modeling tools, such as Matlab (The
MathWorks, Inc., Natick, MA, USA), and/or proprietary or open source systems, including, for
example, physiologically based simulation or emulation packages (e.g., see [21]). Use of standardized
software increases credibility, reliability, and reproducibility by providing some assurance that
the underlying numerical techniques are handled correctly. Use of open source software further
improves reproducibility by making the simulation widely available while also opening the underlying
techniques to later examination for correctness.
190