1. What Is QSP Modeling in Metabolic Disease?
Quantitative systems pharmacology (QSP) uses mathematical models to represent how biological mechanisms interact with one another and how pharmacologic interventions perturb those mechanisms. In metabolic disease, this can involve glucose production and utilization, insulin signaling, pancreatic beta-cell function, adipose tissue, lipid metabolism, hepatic metabolism, inflammation, body weight, and disease progression.
Unlike a narrowly focused exposure-response model, a QSP model attempts to represent a network of interacting biological processes. The purpose is not simply to reproduce a concentration-time curve or fit one biomarker. Instead, the model provides a mechanistic framework for connecting drug exposure to molecular targets, pathways, physiological processes, biomarkers, and clinical outcomes.
A metabolic QSP model can connect drug exposure and molecular mechanisms to physiological biomarkers and clinical outcomes through a network of quantitative relationships.
2. What Questions Can QSP Help Answer?
Metabolic disease is particularly suited to systems modeling because many physiological processes are tightly interconnected. Changes in insulin sensitivity can alter glucose concentrations; glucose can influence insulin secretion; energy intake and expenditure influence body weight; adipose tissue can affect lipid and inflammatory pathways; and hepatic metabolism can influence both glucose and lipid homeostasis.
A QSP model can therefore address questions that span multiple biological levels.
| Question | QSP component | What it helps describe |
|---|---|---|
| How does a drug alter glucose regulation? | Glucose-insulin physiology | Interactions among glucose production, glucose utilization, insulin secretion, and insulin action |
| How does target engagement propagate to physiology? | Mechanistic pathway model | The sequence from molecular target modulation to downstream biological effects |
| Why does treatment change body weight? | Energy-balance model | Relationships among food intake, energy expenditure, adipose stores, and body mass |
| How might treatment alter lipids? | Lipoprotein and lipid metabolism | Production, transport, conversion, and clearance of lipid-related species |
| How could disease progression affect treatment response? | Disease progression component | Time-dependent changes in physiological state and treatment sensitivity |
| What happens when multiple pathways are perturbed? | Integrated QSP network | Potential interactions among mechanisms and combination therapies |
The important distinction is that QSP does not simply ask whether a biomarker changes. It attempts to represent why that biomarker changes and how the change relates to other components of the biological system.
3. The Metabolic System as a Dynamic Network
Metabolic physiology can be represented as a network of compartments, state variables, fluxes, feedback loops, and regulatory processes. A simplified model might contain plasma glucose, insulin, hepatic glucose production, peripheral glucose uptake, pancreatic insulin secretion, and glycogen stores.
A more comprehensive model could additionally represent adipose tissue, free fatty acids, triglycerides, cholesterol, lipoproteins, inflammatory mediators, body composition, energy intake, energy expenditure, and disease progression.
Metabolic physiology contains interconnected feedback loops rather than independent pathways. QSP models attempt to preserve the relationships that are important for the scientific question.
The level of detail should be driven by the intended use of the model. A model designed to study glucose lowering may not require a detailed representation of every lipid species. Conversely, a model intended to investigate cardiometabolic effects may need to represent interactions among glucose, insulin, lipids, body weight, and inflammation.
4. State Variables, Fluxes, and Feedback
A QSP model typically represents biological quantities as state variables. These may correspond to concentrations, amounts, physiological pools, receptor states, cell populations, or other quantities that change over time.
The evolution of each state variable can be represented with a differential equation. In a generic form:
For example, a simplified glucose model might describe the change in plasma glucose as the balance between glucose entering the circulation and glucose being removed from it:
The rates themselves may depend on other model variables. Hepatic glucose production may depend on insulin, while peripheral glucose uptake may depend on both glucose and insulin sensitivity. This creates feedback within the system.
5. Modeling Glucose-Insulin Regulation
Glucose-insulin regulation is a natural starting point for metabolic QSP modeling. In simplified terms, glucose concentration reflects the balance between glucose appearance and glucose disposal, while insulin regulates several processes that influence this balance.
A conceptual model might include:
- hepatic glucose production;
- glucose uptake by skeletal muscle;
- glucose uptake by adipose tissue;
- insulin secretion from pancreatic beta cells;
- insulin-mediated suppression of hepatic glucose production;
- insulin-mediated stimulation of peripheral glucose uptake;
- glucose-dependent insulin secretion; and
- feedback between glucose and insulin.
A simplified insulin-sensitive uptake relationship might be written as:
where \(S_I\) represents an insulin-sensitivity parameter, \(I\) represents insulin, and \(G\) represents glucose. The equation is intentionally simplified: real QSP models may use saturable, nonlinear, delayed, or compartment-specific relationships.
Likewise, insulin secretion can be modeled as a function of glucose:
The function \(f(G)\) may incorporate thresholds, nonlinear stimulation, secretion capacity, and disease-dependent changes in beta-cell function.
6. Representing Insulin Resistance
Insulin resistance can be represented in a QSP model as a change in the relationship between insulin signaling and downstream metabolic processes. Rather than treating insulin resistance as simply a label attached to a patient, the model can represent it through altered parameters or mechanisms.
For example, if insulin stimulates glucose uptake through an efficacy parameter \(S_I\), progressive insulin resistance might be represented by a reduction in effective insulin sensitivity:
where \(D_{\mathrm{IR}}(t)\) represents the degree of insulin-resistance-related impairment.
The impairment could itself be linked to other biological processes. For example, adipose tissue expansion, circulating free fatty acids, inflammation, or ectopic lipid accumulation could be represented as contributors to reduced insulin sensitivity.
7. Modeling Pancreatic Beta-Cell Dysfunction
Type 2 diabetes and related metabolic disorders can involve changes in pancreatic beta-cell function. A QSP model can represent beta-cell capacity as a dynamic state rather than assuming that insulin secretion remains constant.
A simplified beta-cell state \(B(t)\) might follow:
Insulin secretion could then depend on both beta-cell capacity and glucose:
where \(S(G)\) represents glucose-stimulated secretion.
This structure makes it possible to distinguish two mechanisms that may produce similar glucose concentrations at one point in time: reduced insulin sensitivity and reduced beta-cell capacity. Their longer-term consequences can differ substantially.
8. Modeling Hepatic Metabolism
The liver is central to metabolic homeostasis because it participates in glucose production, glycogen storage, fatty-acid metabolism, triglyceride synthesis, lipoprotein production, and other processes.
A metabolic QSP model may therefore represent hepatic glucose production as a regulated flux:
where \(f_I\), \(f_G\), and \(f_D\) represent the effects of insulin, glucose, and disease state, respectively.
The exact mathematical form depends on the biological hypothesis and available data. A simple inhibitory relationship might use a Hill-type function:
This type of relationship can represent diminishing effects as insulin concentration increases. More detailed models may distinguish glycogenolysis and gluconeogenesis or incorporate hormonal and substrate regulation.
9. Modeling Adipose Tissue and Energy Storage
Adipose tissue is more than an energy reservoir. It can influence circulating free fatty acids, inflammatory signaling, insulin sensitivity, and lipid metabolism. QSP models can therefore use adipose tissue as an important mechanistic connection between energy balance and metabolic disease.
A simple energy-storage relationship can be expressed as:
Body mass can then be related to stored energy through an appropriate conversion:
where \(W\) represents body weight and \(\rho_E\) represents an effective energy density parameter.
This is a deliberately simplified representation. Human body weight regulation involves adaptive changes in energy expenditure, appetite, body composition, fluid balance, and other processes. QSP models can add these mechanisms when they are relevant to the scientific question.
10. Modeling Lipid and Lipoprotein Metabolism
Metabolic disease frequently involves abnormalities in triglycerides, cholesterol, free fatty acids, and lipoprotein particles. These variables can be incorporated into QSP models as interacting pools rather than isolated biomarkers.
For example, a simplified triglyceride pool could be described as:
Production and clearance may depend on other variables. Hepatic lipid synthesis could depend on substrate availability and insulin signaling, while clearance could depend on lipoprotein lipase activity and other processes.
| Biological component | Possible QSP representation |
|---|---|
| Free fatty acids | Circulating substrate pool with tissue-specific production and uptake |
| Triglycerides | Production, transport, conversion, and clearance |
| LDL-related cholesterol | Lipoprotein production and receptor-mediated clearance |
| HDL-related processes | Particle formation, remodeling, and clearance |
| Hepatic lipid synthesis | Mechanistically regulated flux |
The appropriate level of lipid detail depends on the purpose of the model. A model intended to describe a drug's effect on triglycerides may require fewer lipid species than a model designed to investigate detailed lipoprotein biology.
11. QSP Models of MASLD and Metabolic Liver Disease
Metabolic dysfunction-associated steatotic liver disease (MASLD) provides an example of a disease where multiple metabolic mechanisms converge on a tissue-level phenotype.
A QSP model can potentially connect:
- energy surplus and adipose tissue expansion;
- free fatty-acid delivery to the liver;
- hepatic de novo lipogenesis;
- hepatic triglyceride accumulation;
- lipid export;
- oxidative and metabolic stress;
- inflammatory signaling;
- cellular injury; and
- progression toward more advanced liver phenotypes.
The conceptual structure might be represented as:
A QSP model does not need to include every molecular detail in this chain. The model should include the mechanisms that are necessary to address the scientific question and make useful predictions.
12. Representing Drug Mechanisms in a Metabolic QSP Model
One of the major strengths of QSP is the ability to connect pharmacologic action to downstream biology. A drug can be represented at several levels depending on the mechanism.
| Level | Example representation |
|---|---|
| Exposure | Plasma or tissue drug concentration |
| Target engagement | Receptor occupancy or target inhibition |
| Signaling | Activation or inhibition of a pathway |
| Cellular response | Altered secretion, uptake, synthesis, or cellular activity |
| Physiology | Changes in glucose, lipids, appetite, energy expenditure, or body weight |
| Clinical outcome | Changes in biomarkers or disease-related endpoints |
A simple target-engagement relationship can be written as:
where \(C\) is drug concentration and \(TE\) is fractional target engagement.
The resulting target effect can then modify one or more biological rates in the QSP model.
This creates a mechanistic chain from exposure to physiological response.
13. Connecting PK to QSP
QSP models frequently use a PK model to provide the drug concentration or exposure driving pharmacologic effects. The PK and QSP components can therefore be connected sequentially.
For example, an oral drug might first be represented by an absorption and disposition model. The resulting concentration-time profile then drives target engagement, which changes a metabolic pathway and ultimately affects glucose or another biomarker.
This separation is useful because PK describes drug exposure, whereas QSP describes how that exposure interacts with the biological system.
14. Linking QSP Mechanisms to Biomarkers
Biomarkers provide important observations for calibrating and evaluating metabolic QSP models. A model may contain unobserved mechanistic states that influence measurable biomarkers.
For example, a model might contain an unobserved insulin-sensitivity state \(S_I\), while the study measures fasting glucose, fasting insulin, HbA1c, triglycerides, or body weight.
An observation model can connect the latent mechanistic state to a measured quantity:
where \(X(t)\) represents the model state, \(h_j\) maps the state to biomarker \(j\), \(\theta\) represents model parameters, and \(\epsilon_j\) represents measurement or residual variability.
15. Modeling Disease Progression
Many metabolic disorders change over time. A QSP model can therefore include disease progression rather than treating the patient's physiological state as fixed.
For example, beta-cell function could decline gradually:
Alternatively, disease progression could be represented through a slowly changing latent disease state \(D(t)\) that modifies several processes simultaneously:
The disease state could influence insulin sensitivity, beta-cell capacity, hepatic metabolism, inflammation, or other components.
This approach allows the model to distinguish between a drug that temporarily changes a biomarker and a treatment that alters the underlying disease trajectory.
16. Why Feedback Loops Matter
Metabolic physiology contains many feedback loops. These loops can produce nonlinear behavior, delays, adaptation, and apparent differences between short-term and long-term treatment effects.
Consider a simplified relationship in which insulin lowers glucose and glucose stimulates insulin secretion:
This is a negative-feedback structure. An increase in glucose stimulates insulin, which increases glucose disposal and tends to reduce glucose.
If a drug changes one component of this loop, the final response can depend on the behavior of the entire system rather than on the direct pharmacologic effect alone.
17. Nonlinear Behavior in Metabolic QSP Models
Metabolic pathways frequently contain nonlinear processes. Examples include receptor binding, enzyme kinetics, saturable transport, hormonal signaling, feedback regulation, and capacity-limited physiological processes.
A generic saturable process can be represented as:
At low substrate concentration, the relationship is approximately linear. At high concentration, the process approaches a maximum rate \(V_{\max}\).
Nonlinearities can have important implications for treatment. A proportional increase in drug exposure does not necessarily produce a proportional increase in target engagement or physiological response.
18. How Are Metabolic QSP Models Calibrated?
QSP model development typically combines biological knowledge with experimental and clinical data. Parameters may come from published literature, in vitro experiments, animal studies, clinical pharmacology studies, biomarker measurements, or clinical trials.
A simplified calibration workflow is:
- Define the biological scope. Decide which pathways and physiological processes must be represented.
- Define the model structure. Specify state variables, fluxes, feedback relationships, and drug mechanisms.
- Compile parameter information. Gather values and plausible ranges from experimental and clinical evidence.
- Identify uncertain parameters. Determine which quantities cannot be directly informed from available data.
- Fit or calibrate the model. Adjust uncertain parameters so the model reproduces relevant observations.
- Evaluate the model. Compare model predictions with data not used directly for calibration where possible.
- Perform sensitivity analysis. Determine which parameters and mechanisms have the greatest influence on outputs.
- Use the model for simulation. Generate predictions under treatment scenarios that may not have been directly observed.
Calibration should not be confused with validation. A model can reproduce the data used for calibration and still fail to predict new observations.
19. Parameter Identifiability in Metabolic QSP
A major challenge in QSP modeling is that complex models can contain many parameters relative to the amount of information available in the data.
Two different parameter combinations may produce very similar model predictions. In such a situation, the individual parameters may not be uniquely identifiable even though the model's overall predictions are useful.
| Issue | Meaning | Potential response |
|---|---|---|
| Structural non-identifiability | The model structure does not permit unique parameter estimation even with ideal data | Reparameterize or simplify the model |
| Practical non-identifiability | Available data are insufficiently informative for precise parameter estimation | Add informative data or constrain parameters using prior knowledge |
| Parameter correlation | Multiple parameters compensate for one another | Examine parameter relationships and model sensitivity |
| Overparameterization | Model complexity exceeds the information available | Reduce unnecessary degrees of freedom |
A mechanistic model is not automatically improved by adding more parameters. Complexity should be justified by the scientific question and the information available to support the model.
20. Sensitivity Analysis
Sensitivity analysis asks how changes in model parameters or assumptions affect model outputs. It is especially important in QSP because the models may contain many interacting parameters.
A local sensitivity measure can be expressed conceptually as:
where \(Y_i\) is an output and \(\theta_j\) is a model parameter.
Sensitivity analysis can help identify:
- parameters that strongly control treatment response;
- biological mechanisms that dominate a particular outcome;
- parameters that require better experimental characterization;
- potentially influential assumptions; and
- opportunities for simplifying the model.
Global sensitivity analysis can additionally examine parameter variation across plausible ranges and can capture nonlinear interactions among parameters.
21. Virtual Patients and Population Variability
Metabolic diseases are heterogeneous. Patients can differ in insulin sensitivity, beta-cell function, body composition, hepatic metabolism, disease severity, and response to treatment.
A QSP model can represent this heterogeneity by allowing model parameters or baseline states to vary across virtual individuals.
For example:
where \(\theta_i\) is an individual parameter, \(\theta_{\mathrm{pop}}\) is a typical population value, and \(\eta_i\) represents between-individual variation.
Virtual populations can then be simulated to investigate how different baseline phenotypes respond to treatment.
22. QSP Modeling of Combination Therapy
Metabolic disease is often treated with combinations of interventions that act through different mechanisms. QSP models can provide a framework for exploring how those mechanisms interact.
Suppose treatment A reduces hepatic glucose production while treatment B increases peripheral glucose uptake. A simplified combined model could contain:
The interaction does not necessarily need to be represented as a simple additive effect. If the mechanisms interact biologically, the model can represent the relevant dependency explicitly.
This is one reason QSP can be useful in combination-development settings: the model can evaluate the consequences of perturbing multiple nodes in the same biological network.
23. Worked Example: A Simplified Glucose-Lowering QSP Model
Consider a hypothetical metabolic QSP model containing three state variables:
- \(G(t)\): plasma glucose;
- \(I(t)\): plasma insulin; and
- \(B(t)\): effective beta-cell functional capacity.
Assume, for illustration, that glucose changes according to:
and that peripheral glucose uptake depends on insulin sensitivity:
Suppose the baseline parameter values are:
| Parameter | Hypothetical value |
|---|---|
| Baseline hepatic glucose input | 10 units/h |
| Insulin concentration | 1 unit |
| Insulin sensitivity \(S_I\) | 0.020 h−1·unit−1 |
| Glucose concentration | 100 units |
Step 1: Calculate baseline glucose uptake
Step 2: Calculate the net glucose change
The hypothetical values above do not represent a physiologically calibrated human model; they simply illustrate the structure of a dynamic system. If the drug increases effective insulin sensitivity by 50%, then:
Step 3: Recalculate uptake
The direct consequence of the assumed mechanism is therefore an increase in modeled glucose uptake. The downstream glucose response would then depend on how the complete system responds over time, including insulin secretion, hepatic glucose production, and feedback.
24. What Can a Metabolic QSP Model Simulate?
Once calibrated and evaluated, a metabolic QSP model can be used to simulate scenarios that may be difficult or expensive to study experimentally.
- different drug doses;
- different dosing schedules;
- different levels of target engagement;
- changes in baseline metabolic phenotype;
- progression or regression of disease mechanisms;
- combination treatments;
- changes in insulin sensitivity;
- changes in beta-cell function;
- effects on glucose and lipid biomarkers;
- body-weight trajectories; and
- hypothetical perturbations of biological pathways.
The model can also be used to examine counterfactual questions: what might happen if a particular pathway were inhibited more strongly, if a biological parameter differed across patients, or if two mechanisms were targeted simultaneously?
25. Mechanism Versus Biomarker Response
An important advantage of QSP is the ability to separate a measured biomarker response from the underlying mechanism responsible for that response.
For example, two hypothetical drugs might both lower fasting glucose but do so through different pathways:
| Drug | Primary modeled mechanism | Potential downstream effect |
|---|---|---|
| Drug A | Increase peripheral insulin sensitivity | Greater glucose uptake |
| Drug B | Reduce hepatic glucose production | Lower glucose appearance |
If only fasting glucose is considered, the two treatments may appear similar. A mechanistic model can reveal that their effects occur through different parts of the biological system.
This distinction can be important when considering combination therapy, resistance mechanisms, safety liabilities, or differences in response across patient phenotypes.
26. How Should a Metabolic QSP Model Be Evaluated?
Model evaluation should address more than whether one dataset can be reproduced. A useful evaluation strategy examines whether the model captures important biological behavior across multiple datasets, conditions, and outputs.
Useful checks may include:
- Baseline behavior: Does the model reproduce relevant physiological ranges?
- Biomarker trajectories: Does it describe observed time courses?
- Dose-response behavior: Does it reproduce known treatment effects?
- Mechanistic perturbations: Does the model behave plausibly when pathways are perturbed?
- External prediction: Can the model reproduce observations that were not used for calibration?
- Sensitivity: Are important outputs controlled by biologically plausible mechanisms?
- Robustness: Do reasonable changes in uncertain assumptions produce reasonable predictions?
Model evaluation should also consider whether a model's predictions remain credible outside the experimental conditions used during development.
27. What QSP Models Do Not Tell Us Automatically
QSP models can integrate large amounts of biological information, but their complexity does not guarantee biological truth.
- A mechanistic model is still a model. It is a structured representation of biological hypotheses.
- More detail is not automatically better. Additional mechanisms require additional assumptions and information.
- Good calibration does not prove correctness. Multiple mechanisms can sometimes reproduce the same observations.
- Parameter uncertainty matters. Predictions can depend strongly on uncertain parameters.
- Structural uncertainty matters. Alternative model structures may lead to different predictions.
- Virtual patients are not real patients. Their realism depends on the assumptions and calibration of the model.
- Extrapolation requires caution. Predictions far outside the observed data may depend heavily on model assumptions.
28. A Practical Workflow for Metabolic QSP Modeling
- Define the scientific question. Start with the decision or biological hypothesis the model needs to address.
- Define the biological scope. Determine which tissues, pathways, biomarkers, and disease mechanisms must be represented.
- Map the mechanism. Draw the causal relationships among targets, pathways, physiological processes, and outcomes.
- Choose state variables. Identify the biological quantities that need to change dynamically.
- Specify equations. Translate biological relationships into quantitative rate laws, algebraic relationships, and differential equations.
- Connect the PK model. Provide the drug concentration or exposure driving pharmacology.
- Represent drug action. Connect exposure to target engagement and downstream mechanisms.
- Compile evidence. Gather experimental, preclinical, clinical, and literature information for parameterization.
- Calibrate the model. Estimate uncertain parameters while maintaining biologically plausible constraints.
- Evaluate the model. Compare predictions with observations and assess sensitivity and uncertainty.
- Generate virtual populations where appropriate. Represent clinically relevant heterogeneity in baseline state and response.
- Simulate prospective scenarios. Use the model to explore treatment, dosing, combination, or mechanistic scenarios.
- Document assumptions. Clearly distinguish experimentally supported relationships from model assumptions.
29. QSP Versus Traditional PK/PD and Pharmacometrics
QSP and traditional pharmacometric approaches overlap substantially, but they often emphasize different levels of biological representation.
| Approach | Typical emphasis | Example question |
|---|---|---|
| PK | Drug concentration and disposition | What concentration follows a given dose? |
| PK/PD | Exposure-response relationship | How does concentration affect glucose? |
| Population PK/PD | Variability across individuals | Which covariates explain response differences? |
| QSP | Mechanistic biological network | How does target modulation propagate through metabolic physiology? |
These approaches are complementary rather than mutually exclusive. A QSP model can incorporate a population PK model, an exposure-response relationship, or other pharmacometric components.
30. How QSP Can Support Metabolic Drug Development
Metabolic drug development often requires decisions about mechanism, dose, combination therapy, patient selection, biomarkers, and expected clinical response. A QSP framework can integrate evidence across these areas.
Potential applications include:
- evaluating target hypotheses;
- connecting target engagement to physiological response;
- exploring dose-response relationships;
- understanding differences among patient phenotypes;
- investigating combination mechanisms;
- simulating biomarker trajectories;
- evaluating alternative mechanistic hypotheses;
- supporting experimental design; and
- identifying biological measurements that could reduce uncertainty.
The value of QSP is greatest when the model is used as a quantitative framework for integrating evidence rather than as a substitute for experimental or clinical evidence.
31. From Mechanistic Model to Translational Prediction
The ultimate objective of many QSP models is translation: using mechanistic knowledge to make quantitative predictions about humans or clinical scenarios.
A conceptual translational chain is:
Each step introduces uncertainty. The model should therefore communicate not only a predicted value but also the assumptions, uncertainty, and biological evidence supporting that prediction.
This makes QSP particularly useful for hypothesis generation and quantitative integration, while also making transparent model qualification essential.
32. Key Takeaways
- Quantitative systems pharmacology models represent interconnected biological mechanisms and their dynamics.
- Metabolic disease is well suited to QSP because glucose, insulin, lipids, adipose tissue, liver metabolism, inflammation, and body weight are strongly interconnected.
- A metabolic QSP model can connect drug exposure to target engagement, signaling, physiological processes, biomarkers, and clinical outcomes.
- Glucose-insulin regulation provides a useful foundation for metabolic QSP modeling.
- Insulin resistance and beta-cell dysfunction can be represented as mechanistic changes in model parameters or biological states.
- Liver, adipose tissue, and lipid metabolism can be integrated into the same model when they are relevant to the scientific question.
- QSP models can represent disease progression rather than treating the metabolic state as fixed.
- Feedback loops and nonlinearities are important because metabolic physiology is a dynamic system rather than a collection of independent pathways.
- Biomarkers provide observations that can inform and evaluate the underlying mechanistic model, but a biomarker is not necessarily identical to the biological state that generates it.
- Parameter identifiability, structural uncertainty, and model assumptions are critical considerations in complex QSP models.
- Virtual populations can be used to explore heterogeneity in baseline biology and treatment response, but their credibility depends on model qualification.
- QSP can support combination-therapy analysis by representing how multiple mechanisms interact within the same biological network.
- QSP and traditional PK/PD approaches are complementary: PK describes exposure, PK/PD describes exposure-response relationships, and QSP can extend the framework to interconnected biological mechanisms.
- The most useful QSP model is not necessarily the most detailed model. It is the model that is sufficiently mechanistic, identifiable, and validated for the scientific question at hand.
Where to Go Next
A natural progression is to study specific metabolic QSP systems in greater detail. Useful next topics include QSP Models of Glucose-Insulin Regulation, QSP Models of Obesity and Energy Balance, QSP Models of Lipid Metabolism, QSP Models of MASLD, QSP Models of Type 2 Diabetes, and QSP Modeling of Metabolic Drug Combinations.
From there, the framework can be extended to quantitative treatment-response modeling, virtual populations, disease progression, biomarker qualification, and translational prediction.