1. What Does Model Scope Mean?
Model scope defines the scientific and biological territory that a quantitative systems pharmacology (QSP) model is intended to represent. It specifies the biological processes, compartments, populations, pathways, mechanisms, and outcomes that are relevant to the question being investigated.
Scope is broader than simply listing model variables. Before deciding whether a cytokine, receptor, cell population, signaling pathway, or organ should appear as a state variable, the modeler should first determine whether that component is relevant to the scientific purpose.
For example, a QSP model designed to investigate a monoclonal antibody's effect on tumor growth may need to represent target binding, immune-cell activity, tumor-cell populations, and downstream signaling. It may not need to represent every cytokine, every immune-cell subtype, or every metabolic pathway in the organism.
QSP model development starts with a scientific question and uses that question to define the biological scope required to generate useful predictions.
2. Start With the Scientific Question
The most important step in defining a QSP model boundary is to state the question the model is supposed to answer.
Questions such as “How does the immune system work?” are too broad to provide a practical modeling boundary. A more useful question might be:
This question immediately suggests several requirements: the model needs a drug exposure component, the relevant target or pathway, tumor-cell dynamics, and mechanisms capable of producing altered response or resistance.
It does not automatically require a complete representation of every component of the immune system, liver metabolism, renal physiology, or intracellular signaling network.
| Question type | Possible QSP scope | Potential outputs |
|---|---|---|
| Mechanism of action | Target, binding, signaling, downstream biology | Target engagement, pathway activity, biomarker response |
| Dose-response behavior | PK, target engagement, pharmacology | Exposure-response relationships |
| Combination therapy | Mechanisms affected by both treatments | Combination response and interaction behavior |
| Resistance | Relevant adaptive or alternative mechanisms | Resistance trajectories and treatment-response scenarios |
| Translational prediction | Mechanisms shared across species or populations | Human-relevant predictions under specified assumptions |
3. What Is a Model Boundary?
The system boundary separates processes represented explicitly by the model from processes treated as external inputs, fixed quantities, assumptions, or omitted mechanisms.
Consider a simplified immune-oncology model. The model might explicitly represent tumor cells, activated immune cells, drug concentration, target occupancy, and an inflammatory mediator. Other biological factors may enter as parameters or external inputs.
A model boundary defines which processes are represented dynamically and which processes are supplied from outside the model.
There is no universally correct boundary. The appropriate boundary depends on the scientific question, available evidence, intended predictions, and acceptable model complexity.
4. Decide What Belongs Inside the Model
A practical scope exercise is to classify biological components into four categories:
- Explicit states: variables whose dynamics are represented by differential or algebraic equations.
- Parameters: quantities that affect model behavior but are not themselves dynamically simulated.
- External inputs: quantities supplied to the model, such as dosing schedules or measured baseline values.
- Excluded processes: mechanisms intentionally omitted because they are not required for the intended question.
| Component | Possible representation | Example rationale |
|---|---|---|
| Drug concentration | Dynamic state | Needed to drive target engagement over time |
| Target abundance | Dynamic state or parameter | Dynamic if regulation of target abundance is scientifically important |
| Binding affinity | Parameter | Controls target engagement without requiring its own dynamic state |
| Dose | External input | Determined by the treatment scenario |
| Patient age | Covariate or parameter modifier | Included only if it materially affects the modeled biology |
| Unrelated metabolic pathway | Excluded | Not required for the intended prediction |
5. Choosing the Right Biological Granularity
After deciding what belongs inside the model, the next question is how much detail should be used to represent each component.
For example, an immune-cell population could be represented as a single aggregate compartment:
Alternatively, the model could distinguish multiple populations:
The second representation may be scientifically useful if activation and effector function have different roles in the question being studied. If they do not, the additional states may add complexity without adding useful predictive information.
| Granularity choice | Potential benefit | Potential cost |
|---|---|---|
| Aggregate cell population | Simple, interpretable dynamics | May obscure important subpopulation differences |
| Multiple cell populations | Represents differentiation or distinct mechanisms | More parameters and states |
| Single pathway activity | Efficient representation of downstream biology | May hide mechanistic detail |
| Detailed signaling network | Can represent specific mechanistic hypotheses | Greater complexity and data requirements |
The appropriate level of detail is therefore a scientific decision, not simply a computational one.
6. Define the Relevant Time and Spatial Scales
Scope also includes the time scale and, when relevant, the spatial scale of the model.
A model intended to study receptor phosphorylation over minutes may require signaling dynamics that are unnecessary in a model intended to predict tumor growth over months.
| Time scale | Potential processes |
|---|---|
| Seconds to minutes | Binding, phosphorylation, rapid signaling |
| Hours | Gene expression, protein turnover, biomarker response |
| Days | Cell proliferation, immune activation, treatment response |
| Weeks to months | Tumor growth, disease progression, resistance, chronic adaptation |
Spatial scope can be similarly important. A model might treat an organ as a single well-mixed compartment, while a more detailed model could distinguish plasma, interstitial space, intracellular space, and tissue regions.
7. Different Types of Model Boundaries
QSP boundaries are not limited to biology. Several different boundaries should be considered during model development.
Biological boundary
Defines which biological pathways, tissues, cell populations, and mechanisms are explicitly represented.
Temporal boundary
Defines the time horizon over which the model is expected to describe the system.
Spatial boundary
Defines which tissues, organs, compartments, or spatial regions are represented.
Population boundary
Defines the populations to which the model is intended to apply, such as healthy subjects, patients with a particular disease state, or a specific disease subtype.
Intervention boundary
Defines which treatments, doses, combinations, or perturbations are represented.
Prediction boundary
Defines which quantities the model is intended to predict and under what conditions those predictions are considered relevant.
| Boundary | Key question |
|---|---|
| Biological | Which mechanisms must be represented? |
| Temporal | Over what time interval should the model operate? |
| Spatial | Which tissues or compartments matter? |
| Population | For whom is the model intended? |
| Intervention | Which treatment scenarios should be simulated? |
| Prediction | What outputs should the model support? |
8. Make Boundary Assumptions Explicit
Every model boundary introduces assumptions. The danger is not making assumptions; the danger is allowing important assumptions to remain implicit.
For example, a model may assume that a biomarker concentration is proportional to the activity of a signaling pathway. Another model may assume that a tissue compartment is well mixed. Another may treat an upstream pathway as constant because its dynamics are outside the intended scope.
| Assumption | Modeling consequence | Question to ask |
|---|---|---|
| Well-mixed compartment | No explicit spatial gradients | Could spatial heterogeneity alter the prediction? |
| Constant baseline | No dynamic baseline process | Could baseline adaptation matter? |
| Fast equilibrium | Binding or reaction may be algebraic rather than dynamic | Is the equilibrium approximation appropriate over the time scale? |
| Fixed parameter | No individual or temporal variability | Could variability materially affect the prediction? |
| Excluded pathway | No feedback through that mechanism | Could the omitted pathway change the intervention response? |
9. Check Whether the Boundary Cuts an Important Feedback Loop
One of the most important QSP-specific boundary problems occurs when an apparently minor component participates in a feedback loop.
Suppose drug inhibition reduces pathway activity, pathway activity controls production of a biomarker, and the biomarker feeds back to regulate the target. Removing the biomarker from the model may break a mechanistic loop and change the predicted behavior.
Boundary decisions should be checked for broken feedback loops, because seemingly secondary variables may control the behavior of the larger system.
This does not mean that every feedback mechanism must be included. It means that feedback should be considered explicitly when deciding whether a process can safely remain outside the model.
10. Let the Intended Evidence Inform the Boundary
QSP models are mechanistic, but mechanistic ambition should still be connected to available evidence.
A component may be biologically plausible but poorly characterized. If it has a major influence on the prediction while its parameters are essentially unconstrained, the model may become highly dependent on assumptions.
Useful evidence for defining scope can include:
- Preclinical pharmacology studies.
- Clinical pharmacokinetic and pharmacodynamic data.
- Biomarker measurements.
- In vitro binding and functional assays.
- Cell-based experiments.
- Gene-expression or proteomic measurements.
- Published mechanistic studies.
- Prior validated QSP or systems-biology models.
- Physiological reference information.
11. Scope, Complexity, and Identifiability
A major reason to define model boundaries carefully is identifiability. Adding biological mechanisms generally introduces additional states, parameters, and interactions. The available data may not contain enough information to distinguish all of them.
Consider a simple interaction:
If only the final effect is observed, multiple combinations of intermediate parameters may produce essentially the same output.
Adding more mechanisms can therefore increase biological realism while simultaneously making individual parameters less identifiable.
| Scope decision | Potential consequence |
|---|---|
| Add another pathway | More mechanistic detail, more parameters |
| Split one population into two | More biological resolution, more state variables |
| Add feedback | Potentially more realistic dynamics, additional interactions |
| Add multiple tissue compartments | More spatial realism, more transport parameters |
| Remove weakly supported mechanism | Lower complexity, potentially less mechanistic coverage |
The objective is not to minimize the number of parameters at all costs. Rather, model complexity should be commensurate with the scientific question and the evidence available to support the added structure.
12. The Boundary Test
A useful way to evaluate a proposed model boundary is to ask a sequence of questions for every major component.
- Does this component affect the scientific question?
- Does removing it change an important predicted outcome?
- Does it participate in a feedback or interaction that matters?
- Do we have evidence to characterize it?
- Does including it improve the prediction enough to justify the added complexity?
- Can the resulting parameters and mechanisms be meaningfully evaluated?
This creates a practical decision framework:
This is not a formal statistical decision rule. It is a modeling principle for disciplined scope definition.
13. Worked Example: Defining a QSP Model for an Oncology Therapy
Suppose a development team wants to understand how an investigational targeted therapy may affect tumor growth and whether baseline pathway activity could influence response.
Step 1: Define the question
The initial question is:
Step 2: Identify the minimum biological chain
A reasonable initial mechanism is:
Step 3: Identify candidate components
| Component | Initial decision | Rationale |
|---|---|---|
| Drug exposure | Include | Required to connect dose to pharmacology |
| Target concentration | Include | Required for target engagement if target abundance varies |
| Target binding | Include | Direct mechanism of drug action |
| Downstream pathway | Include | Connects target inhibition to tumor biology |
| Tumor cells | Include | Primary predicted outcome |
| Every immune-cell subtype | Initially exclude | Not required by the initial question |
| Unrelated metabolic pathways | Exclude | Outside intended mechanism |
Step 4: Test the boundary
Suppose later evidence indicates that an immune-cell population is strongly affected by pathway inhibition and contributes substantially to tumor response. The boundary can then be expanded:
Step 5: State the resulting scope
The model might therefore be documented as follows:
This statement is considerably more useful than simply saying that the model is an “oncology QSP model.”
14. When Should the Model Boundary Change?
Model scope should not necessarily remain fixed throughout development. New evidence, new questions, or new intended applications can justify changing the boundary.
Reasons to expand the boundary
- A newly identified mechanism materially affects the intended prediction.
- A feedback loop appears to explain an observed response pattern.
- A previously omitted tissue or cell population becomes scientifically important.
- The model needs to answer a new question that requires additional mechanisms.
- Existing predictions fail systematically in a way that suggests a missing mechanism.
Reasons to simplify the boundary
- A component does not materially affect the predictions of interest.
- A mechanism is poorly supported and contributes mostly speculative complexity.
- Two components behave similarly enough that an aggregate representation is adequate.
- The intended prediction does not require the added biological resolution.
- Additional parameters create substantial uncertainty without improving predictive performance.
15. Use Modules to Manage Model Boundaries
Complex QSP models are often easier to develop when biological processes are organized into modules.
A hypothetical model might contain:
- PK module: dose, distribution, clearance, and drug concentration.
- Target-engagement module: binding between drug and molecular target.
- Signaling module: intracellular or extracellular pathway activity.
- Cellular-response module: proliferation, death, activation, or differentiation.
- Disease module: tumor burden or disease progression.
- Biomarker module: measurable pharmacodynamic biomarkers.
Modularization makes the boundary easier to communicate. It also makes it possible to ask whether a module is necessary for a specific modeling objective.
A modular structure can make a complex QSP model easier to develop, communicate, test, and extend.
16. How to Document the Model Boundary
A model boundary should be documented before the equations become so detailed that the original scientific purpose becomes difficult to recognize.
A practical scope document can contain the following elements:
| Element | What to document |
|---|---|
| Scientific objective | The precise question the model is intended to address |
| Population | Species, disease state, patient population, or experimental system |
| Time horizon | Relevant simulation and prediction interval |
| Spatial scope | Tissues, organs, compartments, or spatial regions |
| Included mechanisms | Processes represented explicitly |
| Excluded mechanisms | Important processes deliberately outside the boundary |
| External inputs | Doses, baseline values, covariates, or other supplied quantities |
| Outputs | Variables and endpoints the model is intended to predict |
| Key assumptions | Important simplifications and approximations |
| Known limitations | Situations in which the model should not be interpreted as predictive |
This documentation creates a traceable connection between the scientific question and the final mathematical implementation.
17. Common Scope and Boundary Mistakes
1. Starting with biology instead of the question
Listing every known pathway can create an enormous model before anyone has established what the model needs to predict.
2. Assuming more detail is always better
Additional detail can introduce parameters and interactions that are poorly supported by available evidence.
3. Treating the model boundary as permanent
New evidence may reveal that a previously excluded mechanism is necessary. A useful model boundary can evolve.
4. Ignoring feedback
An apparently minor component can become essential if it closes an important feedback loop.
5. Mixing inputs and states without explanation
A quantity that is externally imposed should be distinguished from a quantity whose dynamics are generated by the model.
6. Failing to document exclusions
Readers need to understand not only what the model contains, but also what it intentionally does not contain.
7. Defining scope after model construction
When scope is not defined early, model structure can become driven by available equations or datasets rather than by the scientific objective.
18. A Practical QSP Scope-Definition Workflow
- Write the scientific question. State exactly what the model needs to explain or predict.
- Define the intended population. Specify species, disease context, treatment context, and other relevant population boundaries.
- Define the prediction horizon. Identify the time scale over which the model must operate.
- Map the causal mechanism. Sketch the minimum biological chain from intervention to outcome.
- Identify required mechanisms. Determine which processes are necessary to preserve the causal relationships relevant to the question.
- Choose biological granularity. Decide whether populations, pathways, tissues, or compartments should be aggregated or explicitly represented.
- Identify feedback loops. Check whether excluded components participate in important interactions.
- Classify components. Assign each important quantity to an explicit state, parameter, external input, or exclusion.
- Check evidence. Determine whether the available data and literature can support the proposed structure.
- Evaluate complexity. Ask whether each added mechanism materially improves the intended scientific prediction.
- Document assumptions and exclusions. Record why important processes are inside or outside the boundary.
- Revisit the boundary during development. Expand or simplify the model when evidence or scientific objectives justify the change.
19. A Reusable QSP Scope Statement
A concise scope statement can be structured using the following template:
Included biology: The model explicitly represents [mechanisms, pathways, cell populations, tissues, or compartments] because they are required to address the scientific objective.
Time scale: The model describes the system over [time horizon].
External inputs: [Dose, exposure, baseline values, covariates, or other inputs] are supplied externally.
Excluded biology: [Processes] are outside the model boundary because they are not required for the intended predictions or are insufficiently supported for explicit mechanistic representation.
Primary outputs: The model is intended to predict [biomarkers, target engagement, disease measures, treatment response, or other endpoints].
Key assumptions: The model assumes [major simplifications or approximations].
Intended use: Predictions are intended for [specified scenarios] and should not automatically be extrapolated to [out-of-scope scenarios].
This format creates a clear bridge between the scientific objective, biological model, mathematical implementation, and intended use.
20. Key Takeaways
- QSP model scope defines the biological territory required to answer a specific scientific question.
- A model boundary separates mechanisms represented explicitly from processes treated as parameters, inputs, assumptions, or exclusions.
- The scientific question should determine the boundary—not the availability of equations, software components, or biological information alone.
- Model scope includes biological, temporal, spatial, population, intervention, and prediction boundaries.
- Biological granularity should be chosen according to the prediction being made and the evidence available to support the added detail.
- Feedback loops require particular attention because apparently minor components can become important to system-level behavior.
- Adding mechanisms can improve biological representation but can also increase parameter uncertainty and identifiability challenges.
- Important exclusions should be documented with explicit scientific rationale.
- QSP model scope can evolve as new evidence reveals mechanisms that are necessary—or unnecessary—for the intended predictions.
- Modular model architecture can help manage complex biological boundaries while preserving the interactions required by the scientific question.
- A useful QSP model is not necessarily the model containing the most biology; it is the model whose scope is appropriate for its intended scientific purpose.
Where to Go Next
Once the QSP model boundary has been defined, the next step is to translate the biological map into a mechanistic model structure: identify state variables, parameters, governing equations, reaction relationships, feedback mechanisms, and model inputs.
A natural progression is to study building a QSP model from biological knowledge, followed by model assumptions, mechanistic equations, parameterization, calibration, sensitivity analysis, uncertainty analysis, and model qualification.