Rule 1 – Define context clearly.

Plan and develop the M&S activity with clear definition of the intended purpose or context accommodating end-users needs.

• The goals of our models are clearly defined by the Specific aims of the project, as follows. Specific Aim 1 is to develop predictive multiscale models for blood flow and oxygen transport in the mouse cerebral cortex, and validate these models using experimental data derived from multimodal imaging of the cortex microvasculature. Specific Aim 2 is to develop multiscale models for blood flow autoregulation and neurovascular coupling in the mouse cerebral cortex, and to test and refine these models using experimental data derived from multimodal imaging of the cortical microvasculature. The computational tools are being developed in such a way that they can be used by others for studies in any tissue for which detailed information is available about microvascular structure.

Rule 2 – Use appropriate data.

Use data relevant to the M&S activity, which can ideally be traced back to the source.

• The project is carried out in close collaboration between an experimental group and a theoretical group, facilitating frequent comparisons and “reality checks.” In the first two years of the work, we compared predicted flow velocities in vascular networks of the mouse cerebral cortex with experimental observations, and found that the predictions and experimental results showed major discrepancies. These discrepancies were traced to systematic problems with the experimental procedures. In the past 12 months, our collaborators have significantly refined their methods for velocity measurements and have recently provided us with new experimental data that we are using to test our modeling methods.
• Specific Aim 1 is structured to allow model testing with experimental data other than those used in the initial model development. The model for flow and oxygen transport is parameterized using data in the resting state. Then flow rates and oxygen levels in main inflow and outflow vessels are measured in conditions of reduced blood pressure or blood oxygen levels, and used as boundary conditions for model simulations to predict flow rates and tissue oxygen fields throughout the observed region. These predictions are compared with independently measured values.

Rule 3 – Evaluate within context.

Evaluate the M&S activity through verification & validation, uncertainty quantification, and sensitivity analysis faithful to the context/purpose/scope of the M&S efforts, with clear and a-priori definition of evaluation metrics and including test cases.

• We test the code under conditions for which the correct behavior is known or can be calculated independently. For example, our Greens function method for oxygen transport was tested by comparing its solution with corresponding solutions using the Krogh cylinder model.
• We continuously generate graphical output during program execution, to check for Inconsistent or unexpected behavior. Graphics files showing network structure, hemodynamic variables, oxygen fields on slices through 3D domains, histograms of relevant variables, etc., are generated and monitored.
• In Specific Aim 2, the models are used to test hypotheses regarding the mechanisms of flow regulation in the brain, by generating multiple models in which specific mechanisms are turned on or off. We anticipate that many of these models will be unable to predict behavior consistent with observations, regardless of assumed parameter values. These “failures” will guide the choice of mechanisms to be included in the eventual model. Comparisons with observed responses to several types of experimental conditions will aid in establishing the credibility of these models.
• We carry out sensitivity analyses of model results to key unknown parameters. These analyses are used to assess model robustness, to obtain estimates of uncertainty of model predictions where key parameters are not precisely known, and to predict the effects of parameter changes that occur in various physiological and pathological conditions. For example, we will examine the dependence of tissue oxygen distribution and hypoxic fraction on oxygen consumption rate and on perfusion.

Rule 4 – List limitations explicitly.

Provide an explicit disclaimer on the limitations of the M&S to indicate under what conditions or applications the M&S may or may not be relied on.

• Simplifying assumptions and limitations of our methods are explicitly described in publications describing our theoretical models.

Rule 5 – Use version control.

Implement a version control system to trace the time history of the M&S activities, including delineation of contributors' efforts.

• We have established repositories on GitHub for two key pieces of computer software: FlowEstimateV1, Estimation of blood flows in microvascular networks, version 1; and GreensV4, Greens function method for simulation of solute transport, version 4. They can be found at https://github.com/secomb. Within the past year, we have placed several more computer codes on GitHub and performed numerous upgrades. The following repositories are now available:
  • GreensTD19_GPU: Greens function method for time-dependent solute transport by microvessel networks. Updated on Sep 2, 2019.
  • NetFlowV2. Simulation of microvascular network hemodynamics. Updated on Aug 5, 2019.
  • GreensV4. Greens function method for simulation of solute transport, version 4. Updated on Jul 11, 2019.
  • GreensV4BC. Greens function method with estimation of oxygen boundary conditions. Updated on Jul 11, 2019.
  • GreensV4_GPU. GPU version of GreensV4 - simulation of solute transport in vascular networks. Updated on Jul 1, 2019.
  • FlowEstimateV1. Estimation of blood flows in microvascular networks. Updated on Jun 21, 2019.
  • AddNetworkBdySegs. Program to restore boundary segments to microvessel networks derived from image stacks. Updated on Jun 15, 2019.
  • StackEnhanceV1. Preprocessing of image stacks from optical microscopy of microvessel networks. Updated on Jun 13, 2019.
  • GreensV4BC_GPU. Greens function method with estimation of oxygen boundary conditions - GPU version. Updated on May 24, 2019.
  • AnalyzeAmiraFile. Postprocessing of spatial graphs generated by Amira from image stacks of microvascular networks. Updated on May 3, 2019.
  • CondenseNetwork. Combine short segments in Network.dat files describing networks of microvessels. Updated on May 3, 2019.

Rule 6 – Document adequately.

Document all M&S activities, including simulation code, model markup, scope and intended use of M&S activities, users' and developers' guides.

• While developing model code, we simultaneously write a draft of a paper describing the model. In this way, we keep the code in close correspondence with the methods and assumptions of the model.
• Code is commented with explanatory text at the head of main modules and brief comments throughout, particularly where code is not self-explanatory. While not providing a comprehensive explanation of the code, these comments will help other users who wish to understand and/or modify the programs that we have developed.

Rule 7 – Disseminate broadly.

Disseminate appropriate components of M&S activities, including simulation software, models, simulation scenarios and results.

• We make experimental data sets, model code with sample input data files, and sample output files publicly available via the internet, both on our website https://physiology.arizona.edu/people/secomb and on GitHub https://github.com/secomb.

Rule 8 – Get independent reviews.

Have the M&S activity reviewed by independent third-party users and developers, essentially by any interested member of the community.

• Our consultants (Drs. Roy and Devor) contribute to the assessment of model credibility. We have met or teleconferenced with both consultants within the past 12 months.
• The assessment of our work through peer review of the resulting manuscripts remains an important component of establishing model credibility. Two manuscripts have been accepted for publication in Journal of Cerebral Blood Flow and Metabolism, a premier journal in this field.
• Several other groups are using pieces of our software and we are highly responsive to requests for assistance and fixes to problem or limitations that they encounter.

Rule 9 – Test competing implementations.

Use competition of multiple implementations to check the conclusions of different implementations of the M&S processes against each other.

• As already mentioned, we test the code for simple geometric configurations where the correct behavior is known or can be calculated independently.
• We are not aware of any competing simulation system that would allow simulations for the complex network structures that we are using, with thousands of vessel segments, with a comparable level of spatial resolution.

Rule 10 – Conform to standards.

Adopt and promote generally applicable and discipline specific operating procedures, guidelines, and standards accepted as best practices.

• We follow generally accepted practices in terms of software testing, documentation, dissemination, etc., as outlined under the preceding Rules.
• We are not aware of any existing standards that are specific to simulation of vascular network flow and mass transport, and therefore we are developing our own standard, i.e. the file format that we use for describing microvascular network structure, geometry and flow distributions, and promoting its use.

Critical issues and concerns: Due to the complexity of biological systems and the difficulty of characterizing all aspects of their behavior, biological validity is not absolute, and its assessment is subjective. Theoretical models can and often do fail to represent biological reality in significant ways. In our experience, such failures are often very informative outcomes and drive further conceptual and model development. These processes are continuous throughout all stages of modeling.