Tutorials › Pharmacometrics › QSP Modeling in Infectious Disease
Pharmacometrics · Quantitative Systems Pharmacology

QSP Modeling in Infectious Disease

Learn how quantitative systems pharmacology models integrate pathogen dynamics, host biology, drug exposure, immune responses, biomarkers, and treatment mechanisms to simulate infectious disease across multiple biological scales.

Intermediate QSP Modeling Infectious Disease Pharmacometrics
01 · The big picture

1. What Is QSP Modeling in Infectious Disease?

Quantitative systems pharmacology (QSP) uses mechanistic mathematical models to connect drug exposure with biological processes across multiple levels of a disease system. In infectious disease, those levels can include pathogen replication, target cells, innate immunity, adaptive immunity, inflammatory mediators, tissue compartments, biomarkers, and treatment effects.

Unlike a purely empirical exposure-response model, an infectious-disease QSP model attempts to represent the biological mechanisms that generate observed disease trajectories and treatment responses.

Pathogen replication clearance QSP model host response immune dynamics drug mechanisms tissue compartments Clinical outputs pathogen load · biomarkers response · treatment effects Mechanistic connections allow observations from different biological levels to inform one model.

An infectious-disease QSP model connects pathogen dynamics, host biology, and drug mechanisms to clinically observable outcomes.

Core idea: QSP modeling is useful when the scientific question depends on interactions among multiple biological mechanisms rather than on a single exposure-response relationship.
02 · Why QSP?

2. Why Use QSP for Infectious Disease?

Infectious diseases are dynamic systems. Pathogen replication changes the host environment, the host mounts immune responses, treatment changes pathogen burden, and changes in pathogen burden can in turn alter the immune response.

These feedback relationships make infectious disease a natural setting for mechanistic modeling.

Scientific questionRelevant QSP componentPotential model output
How does the pathogen expand?Pathogen replication dynamicsPathogen load over time
How does the host control infection?Innate and adaptive immunityImmune-cell or mediator trajectories
How does treatment affect the pathogen?Mechanism-of-action modelReduced replication or increased clearance
Why do patients respond differently?Parameter variability and covariatesIndividual response trajectories
What happens under a new regimen?SimulationPredicted pathogen and biomarker trajectories
How could resistance emerge?Multiple pathogen states or strainsSelection of resistant populations

The central advantage is not that QSP automatically produces more accurate predictions. Rather, the mechanistic structure provides a framework for integrating heterogeneous information and testing how assumptions about biology affect predictions.

03 · The disease system

3. What Does an Infectious-Disease QSP Model Represent?

An infectious-disease QSP model may contain several interacting subsystems. The precise components depend on the pathogen, disease, therapeutic modality, and scientific question.

SubsystemExamples of modeled quantities
PathogenVirions, bacteria, parasites, infected cells, pathogen burden
Target cellsSusceptible, infected, activated, or damaged host cells
Innate immunityInterferons, inflammatory mediators, innate immune cells
Adaptive immunityT cells, B cells, antibodies, memory responses
DrugConcentrations, target engagement, inhibition, immune modulation
ResistanceSusceptible and resistant pathogen populations or genotypes
Tissue compartmentsBlood, lung, lymphoid tissue, gastrointestinal tract, or other relevant sites
Clinical biomarkersPathogen measurements, inflammatory markers, cell counts, clinical scores
Important modeling principle: not every biologically plausible component belongs in the model. A QSP model should contain mechanisms that are relevant to the scientific question and sufficiently informed by available data.
04 · Pathogen dynamics

4. Modeling Pathogen Replication and Clearance

A natural starting point is a dynamic model for pathogen burden. Let \(P(t)\) denote the amount or concentration of pathogen at time \(t\).

A simple replication-clearance model can be written as:

\[ \frac{dP}{dt}=rP-cP \]

where \(r\) represents an effective pathogen replication rate and \(c\) represents an effective clearance rate.

Equivalently:

\[ \frac{dP}{dt}=(r-c)P \]

If replication exceeds clearance, pathogen burden increases. If clearance exceeds replication, pathogen burden declines.

Real infectious diseases are usually more complicated. Replication can depend on target-cell availability, immune pressure, tissue environment, nutrient availability, or other biological constraints.

05 · Host cells

5. Adding Target Cells to the Model

Many infectious agents depend on host cells for replication. A classic mechanistic framework therefore distinguishes susceptible target cells from infected cells.

Let \(T(t)\) denote susceptible target cells and \(I(t)\) infected cells. A simplified target-cell model is:

\[ \frac{dT}{dt}=s-d_TT-\beta PT \]
\[ \frac{dI}{dt}=\beta PT-\delta I \]

Here:

  • \(s\) is the source rate of susceptible cells.
  • \(d_T\) is the loss rate of susceptible cells.
  • \(\beta\) describes infection of susceptible cells by pathogen.
  • \(\delta\) describes loss of infected cells.

The model now contains an important feedback loop: pathogen infects target cells, infected cells generate additional pathogen, and the resulting infection changes the number of available target cells.

Target cells T(t) infection Infected cells I(t) production Pathogen P(t) pathogen drives new infection of target cells

A simple target-cell model creates a mechanistic feedback loop between susceptible cells, infected cells, and pathogen burden.

06 · Host response

6. Modeling the Immune Response

Infectious disease cannot generally be understood from pathogen dynamics alone. The host response can determine both pathogen clearance and the clinical manifestations of infection.

A QSP model may therefore introduce variables representing immune mediators or immune-cell populations.

For example, let \(E(t)\) represent an effective immune response that increases pathogen clearance:

\[ \frac{dP}{dt}=rP-c_0P-k_EE P \]

The term \(k_EEP\) represents additional pathogen loss associated with the immune response.

The immune response itself could depend on pathogen burden:

\[ \frac{dE}{dt}=k_{\text{act}}\frac{P}{K_P+P}-d_EE \]

This creates a feedback system: pathogen stimulates the immune response, and the immune response reduces pathogen burden.

Mechanistic insight: the same observed decline in pathogen burden can arise from different combinations of intrinsic pathogen clearance and immune-mediated clearance. QSP models can explicitly represent these competing mechanisms.
07 · Drug mechanisms

7. Connecting Drug Exposure to Pathogen Dynamics

The pharmacology of an anti-infective can be incorporated by linking drug exposure to a specific biological mechanism.

Suppose \(C(t)\) is the drug concentration and the drug reduces pathogen replication. A simple inhibitory relationship might be:

\[ r_{\text{eff}}(C)=r_0\left(1-\frac{C}{IC_{50}+C}\right) \]

The pathogen equation then becomes:

\[ \frac{dP}{dt}=r_{\text{eff}}(C)P-cP \]

This structure connects pharmacokinetics to pathogen dynamics:

\[ \text{Dose}\rightarrow PK\rightarrow C(t)\rightarrow \text{drug mechanism}\rightarrow P(t) \]

A QSP model can go further by representing target engagement, intracellular drug concentrations, viral or bacterial replication stages, or downstream biological effects.

08 · Mechanism of action

8. Why Mechanism of Action Matters

Two drugs can produce similar reductions in pathogen burden while acting through very different biological mechanisms. A QSP model can represent those mechanisms separately.

Drug mechanismPossible QSP representationPotential consequence
Replication inhibitionReduce pathogen production or replication rateSlower pathogen expansion
Entry inhibitionReduce infection rate \(\beta\)Fewer new infected cells
Pathogen neutralizationIncrease effective clearanceFaster pathogen removal
Immune stimulationIncrease immune activation or persistenceGreater host-mediated clearance
Inflammation modulationModify mediator production or signalingChanged host-response dynamics
Combination therapyMultiple simultaneous mechanismsPotentially complementary or interacting effects

This mechanistic distinction becomes particularly valuable when simulating doses, combinations, resistance, or treatment sequences that have not yet been extensively studied clinically.

09 · Pharmacokinetics

9. Integrating Pharmacokinetics Into an Infectious-Disease QSP Model

Drug concentration is often a key input into the disease model. The PK component may range from a simple compartment model to a physiologically based representation of tissue distribution.

For example, a one-compartment model with first-order elimination can be represented by:

\[ \frac{dC}{dt}=-kC \]

For repeated dosing, the resulting concentration profile can then drive a pharmacologic effect term in the pathogen model.

The overall structure becomes:

\[ \text{Dose}\rightarrow C(t)\rightarrow \text{target engagement}\rightarrow \text{pathogen/host response} \]

This is one of the defining features of pharmacometric QSP: drug exposure is not treated as an isolated predictor. It becomes a mechanistic input to the biological system.

10 · Biomarkers

10. Linking the Model to Biomarkers

QSP models often contain biological variables that are not directly observed. Biomarkers provide a bridge between model states and experimental measurements.

For example, suppose \(E(t)\) represents an underlying immune-response state, while \(B(t)\) represents an observed biomarker. A simple observation relationship might be:

\[ B(t)=B_0+\alpha E(t) \]

More realistic observation models can include baseline values, nonlinear relationships, delays, measurement error, and other factors.

This distinction between model state and observed biomarker is important. A measured biomarker does not necessarily correspond directly to a single biological process.

Modeling principle: distinguish the biological state represented by the model from the measurement used to observe that state. The two are related, but they are not necessarily identical.
11 · Spatial biology

11. Tissue and Compartment Models

Many infectious diseases involve substantial differences between pathogen and drug behavior in blood and tissues. A QSP model can therefore divide the system into biologically meaningful compartments.

For example, a simplified model might contain plasma and tissue concentrations:

\[ \frac{dC_p}{dt}=-k_{pt}C_p+k_{tp}C_t-k_eC_p \]
\[ \frac{dC_t}{dt}=k_{pt}C_p-k_{tp}C_t \]

where \(C_p\) is plasma concentration and \(C_t\) is tissue concentration.

Pathogen dynamics can then be localized to the tissue compartment where infection occurs. This can be particularly important when systemic exposure does not accurately represent exposure at the site of infection.

12 · Resistance

12. Modeling Antimicrobial or Antiviral Resistance

Resistance is a natural application of mechanistic infectious-disease modeling because treatment can change the relative abundance of pathogen populations with different sensitivities.

Suppose \(P_S\) represents a drug-sensitive population and \(P_R\) a resistant population:

\[ \frac{dP_S}{dt}=r_S(C)P_S-c_SP_S \]
\[ \frac{dP_R}{dt}=r_R(C)P_R-c_RP_R \]

The two populations may differ in drug sensitivity, replication fitness, or immune susceptibility.

A treatment that strongly suppresses the sensitive population can alter the competitive environment experienced by the resistant population. This can produce changes in pathogen composition even when total pathogen burden initially declines.

Why this matters: resistance modeling shifts the question from simply asking whether a treatment reduces pathogen burden to asking how treatment changes the evolutionary and competitive dynamics of pathogen populations.
13 · Combination therapy

13. QSP Models of Combination Therapy

Combination therapy is another setting where mechanistic models can be especially useful. Different agents may act at different points in the pathogen life cycle or host-response network.

For two drugs with concentrations \(C_1(t)\) and \(C_2(t)\), a simplified model might represent separate inhibitory mechanisms:

\[ r_{\text{eff}}=r_0 \left(1-I_1(C_1)\right) \left(1-I_2(C_2)\right) \]

where \(I_1\) and \(I_2\) are fractional inhibitory effects.

The exact interaction model should reflect the biological hypothesis. Additive, independent, synergistic, antagonistic, or mechanistically distinct effects should not be assumed to be equivalent.

14 · Feedback systems

14. Why Infectious-Disease QSP Models Are Dynamic

One of the defining characteristics of infectious disease is feedback.

Pathogen burden Immune response Drug exposure stimulates treatment modifies system dynamics

The biological system is coupled: pathogen burden influences immunity, immunity affects pathogen clearance, and treatment modifies one or more components of the system.

These feedback loops mean that changing one component can have consequences elsewhere in the system. That is precisely the type of problem for which mechanistic dynamic models can be useful.

15 · Worked example

15. Worked Example: Simulating Drug-Mediated Pathogen Suppression

Consider a simplified hypothetical infection with pathogen burden \(P(t)\), intrinsic pathogen replication rate \(r=0.30\) h\(^{-1}\), and baseline clearance rate \(c=0.10\) h\(^{-1}\).

Without treatment:

\[ \frac{dP}{dt}=(0.30-0.10)P=0.20P \]

The net growth rate is therefore \(0.20\) h\(^{-1}\).

Step 1: Introduce an antiviral effect

Suppose treatment reduces the replication rate by 80%, so the effective replication rate becomes:

\[ r_{\text{eff}}=0.30(1-0.80)=0.06\text{ h}^{-1} \]

Step 2: Calculate the treated net rate

\[ r_{\text{eff}}-c=0.06-0.10=-0.04\text{ h}^{-1} \]

The sign has changed. Instead of increasing, pathogen burden is now predicted to decline.

Step 3: Calculate the relative change after 24 hours

For the simplified constant-rate model:

\[ P(t)=P_0e^{(r_{\text{eff}}-c)t} \]

After 24 hours:

\[ \frac{P(24)}{P_0}=e^{-0.04(24)}\approx0.383 \]

Thus, the simplified model predicts approximately 38.3% of the starting pathogen burden after 24 hours, assuming the rates remain constant.

Step 4: Interpret the result

The important insight is not the particular numerical prediction. The model shows how a treatment can change the sign and magnitude of the net pathogen growth rate. In a full QSP model, drug concentration would generally vary with time, immune responses would change dynamically, and pathogen replication and clearance would depend on additional biological states.

16 · Calibration

16. How Are Infectious-Disease QSP Models Calibrated?

QSP models often contain parameters that cannot be measured directly in every patient or experiment. Calibration therefore involves integrating information from multiple sources.

  1. Define the biological structure. Specify the states, processes, and interactions that the model is intended to represent.
  2. Gather experimental information. Use in vitro, animal, translational, biomarker, PK, and clinical data where appropriate.
  3. Estimate or fix parameters. Parameters may be estimated from data or constrained using prior experimental knowledge.
  4. Link observations to model states. Specify how measured pathogen loads, biomarkers, and clinical endpoints correspond to model quantities.
  5. Calibrate the model. Estimate uncertain parameters using an appropriate fitting or Bayesian framework.
  6. Evaluate model behavior. Examine whether the model reproduces relevant observations and biological relationships.
  7. Test predictions. Use independent data or prospective experiments when possible.

Because QSP models can be highly parameterized, calibration should be accompanied by careful attention to parameter identifiability and uncertainty.

17 · Identifiability

17. Identifiability and Parameter Uncertainty

A model may contain more parameters than the available data can uniquely determine. This creates an important distinction between a model being mathematically defined and its parameters being empirically identifiable.

For example, if only total pathogen burden is observed, it may be difficult to separately identify several hidden biological processes that all influence that same measurement.

IssueMeaningPotential consequence
Structural identifiabilityWhether parameters can theoretically be distinguished given ideal observationsSome parameters may be inseparable even with perfect data
Practical identifiabilityWhether the available data contain enough information to estimate parameters preciselyWide confidence or credible intervals
Parameter correlationMultiple parameters can compensate for one anotherUncertain individual parameters despite good model fit
Model uncertaintyAlternative biological structures may explain the observationsPredictions may depend on structural assumptions
Key principle: a model can reproduce observed data while still having substantial uncertainty about individual mechanisms. Good calibration does not eliminate structural or parameter uncertainty.
18 · Simulation

18. What Can an Infectious-Disease QSP Model Simulate?

Once calibrated and evaluated, a QSP model can be used to explore scenarios that may be difficult, expensive, or impossible to study exhaustively in experiments.

  • Different dosing regimens.
  • Changes in treatment timing.
  • Combination therapies.
  • Different levels of drug exposure.
  • Changes in pathogen replication or clearance.
  • Differences in immune response.
  • Potential resistance trajectories.
  • Changes in tissue exposure.
  • Biomarker trajectories.
  • Hypothetical patient or disease phenotypes.

Simulation is especially valuable for exploring what-if scenarios. However, simulated outcomes remain conditional on the model structure, parameter values, assumptions, and uncertainty.

19 · Translation

19. Translating Between Experimental and Clinical Systems

One of the major goals of QSP is to connect observations made at different stages of drug development.

In vitro mechanism · potency QSP model integrated biology PK · PD · pathogen host response Animal translation Clinical patients QSP can provide a common mechanistic framework for integrating evidence across development stages.

A QSP framework can help connect experimental mechanism data with translational and clinical observations, provided the biological assumptions remain appropriate.

The goal is not to assume that one experimental system perfectly represents another. Instead, the model provides an explicit framework for representing differences, assumptions, and uncertainty during translation.

20 · Model development

20. A Practical Infectious-Disease QSP Workflow

  1. Define the scientific question. Decide what the model needs to explain or predict.
  2. Map the biology. Identify pathogen, host, immune, drug, and tissue processes relevant to the question.
  3. Define model states. Translate important biological quantities into mathematical state variables.
  4. Write the interactions. Represent production, loss, activation, inhibition, infection, and feedback processes.
  5. Connect PK to pharmacology. Translate drug exposure into target engagement or biological effect.
  6. Connect hidden states to observations. Define measurement models for pathogen loads, biomarkers, and clinical endpoints.
  7. Assign or estimate parameters. Integrate experimental and clinical evidence.
  8. Perform sensitivity analysis. Determine which parameters and mechanisms have the greatest influence on model outputs.
  9. Calibrate and evaluate. Compare predictions with appropriate data and assess model adequacy.
  10. Validate predictions where possible. Use independent observations or prospective experiments.
  11. Simulate scenarios. Explore alternative treatments, doses, mechanisms, or biological conditions.
  12. Communicate uncertainty. Distinguish model-based predictions from directly observed evidence.
21 · Interpretation

21. What Infectious-Disease QSP Models Do Not Tell Us Automatically

Mechanistic detail does not eliminate uncertainty. A sophisticated QSP model can still produce misleading conclusions if its structure, parameters, observations, or assumptions are inappropriate.

  • A detailed model is not automatically a correct model.
  • Biological plausibility does not establish parameter identifiability.
  • A good fit does not prove that every mechanism is correct.
  • Different mechanisms can sometimes produce similar observable trajectories.
  • Predictions outside the calibration range can be strongly assumption-dependent.
  • Unobserved biological states may remain uncertain even when observed biomarkers are well reproduced.
  • Population heterogeneity may require explicit variability rather than a single representative parameter set.
  • Clinical endpoints may depend on mechanisms that are not represented in the model.
Modeling principle: QSP should make biological assumptions explicit and quantitatively testable; it should not be used to disguise uncertainty behind a highly detailed mathematical structure.
22 · Putting it together

22. From Pathogen to Clinical Outcome

A useful conceptual hierarchy for infectious-disease QSP is:

\[ \text{Dose} \rightarrow \text{PK} \rightarrow \text{Target engagement} \rightarrow \text{Pathogen dynamics} \rightarrow \text{Host response} \rightarrow \text{Biomarkers} \rightarrow \text{Clinical outcome} \]

Each arrow represents a mechanistic hypothesis that can be encoded mathematically.

For example, dose determines drug exposure through PK; exposure determines target engagement; target engagement changes pathogen replication; pathogen burden influences immune activation; immune activity changes both pathogen clearance and potentially clinical biomarkers; and the resulting biological state contributes to clinical outcomes.

This structure allows QSP models to connect measurements that would otherwise be analyzed separately.

23 · Advanced applications

23. Where Infectious-Disease QSP Can Go Next

Once the basic framework is established, substantially richer models can be developed.

Advanced topicWhat can be represented
Immune-pathogen co-dynamicsDetailed innate and adaptive immune interactions
Tissue QSPSite-specific drug, pathogen, and immune dynamics
ResistanceMultiple pathogen populations and selection under treatment
Combination therapyMultiple mechanisms and pharmacologic interactions
Virtual populationsHeterogeneous biological and PK characteristics
Biomarker modelsMechanistic links between latent biological states and measurements
Transmission modelsConnections between within-host dynamics and population-level spread
Vaccination modelsImmune priming, memory, protection, and breakthrough infection
Longitudinal disease modelsAcute infection, recovery, persistence, or relapse
Model-informed drug developmentIntegration of mechanistic simulations into development decisions

24. Key Takeaways

  • Quantitative systems pharmacology uses mechanistic mathematical models to connect drug exposure with biological systems.
  • Infectious-disease QSP can represent pathogen replication, host cells, immune responses, drug mechanisms, tissue compartments, biomarkers, and clinical outcomes.
  • Pathogen dynamics are inherently coupled to host biology, making feedback and dynamic interactions central to many infectious-disease models.
  • PK provides the time-varying drug exposure that drives pharmacologic effects within the QSP model.
  • Mechanism-of-action models translate drug concentration or target engagement into changes in pathogen or host processes.
  • Resistance can be modeled by representing pathogen populations with different drug sensitivities or biological properties.
  • Combination therapies can be represented through multiple mechanistic pathways rather than relying only on empirical interaction terms.
  • Biomarkers provide observations of model states but should not automatically be interpreted as direct measurements of a single biological mechanism.
  • QSP models can integrate evidence from experimental, translational, and clinical settings within a common mechanistic framework.
  • Parameter identifiability, model uncertainty, and extrapolation remain important limitations even for highly detailed models.
  • Simulation allows researchers to explore treatment scenarios that have not been directly observed, but predictions remain conditional on model assumptions and evidence.
  • The most useful QSP model is not necessarily the most complex one; it is the model whose biological structure is appropriate for the scientific question and supported by the available evidence.
Next step

Where to Go Next

A natural progression is to study QSP models of pathogen dynamics in greater detail, followed by target-cell models, immune-pathogen interactions, PK/PD integration, antimicrobial or antiviral mechanisms, resistance, combination therapy, tissue distribution, and virtual populations.

The next tutorial can build directly on this framework by developing a mechanistic pathogen model step by step, showing how replication, target-cell depletion, immune-mediated clearance, and drug effects can be represented using coupled ordinary differential equations.

← Back to Pharmacokinetics Tutorials