How Systems Models Are Revolutionizing Medical Product Development
The future of medicine lies not just in discovering new drugs, but in predicting their behavior inside the complex system of the human body.
Imagine trying to predict a city's traffic patterns by studying a single car. This simplistic approach mirrors a long-standing challenge in medical research. For decades, the development of new therapies and technologies has relied heavily on sequential testing—first in cells, then animals, and finally humans—a process that is not only time-consuming and expensive but often fails to predict real-world outcomes.
The statistics are sobering: achieving one approved drug in the United States can take over a decade and cost more than $1.8 billion9 .
However, a transformative shift is underway. Systems modeling, an approach that uses computational models to simulate the complex, interconnected systems of human biology, is emerging as a powerful tool to revolutionize how we regulate and develop medical products. By moving beyond isolated events to understand the entire complex system, researchers and regulators are beginning to see the "currents on the route ahead" rather than just reacting to the waves1 .
Traditional drug development can take over 10 years from discovery to approval.
Traditional drug development often operates in silos, focusing on cause-and-effect chains—a change in a drug's formula causes a change in a biomarker. However, the human body is not a simple linear system. It is a complex network of interacting feedback loops, where altering one element can have unforeseen consequences throughout the entire system1 .
This complexity explains why a therapy that shows promise in animal models can fail in human trials, or why a drug approved based on limited clinical trial data might reveal safety issues in a broader, more diverse population.
At their core, systems models are computational simulations that integrate vast amounts of biological, physiological, and pharmacological data. They aim to create a virtual representation of a disease process or organ system. The most prominent type in medical development is Quantitative Systems Pharmacology (QSP).
QSP models balance data-driven (top-down) approaches with fundamental biological knowledge (bottom-up) to quantitatively predict how a drug will behave in a complex biological environment7 .
| Aspect | Traditional Approach | Systems Modeling Approach |
|---|---|---|
| Focus | Isolated components | Integrated systems |
| Prediction Method | Extrapolation from limited data | Simulation of complex interactions |
| Adaptability | Static models | Dynamic, evolving models |
| Regulatory Application | Post-hoc analysis | Proactive prediction |
To understand the practical power of systems modeling, consider the challenge of chemotherapy-induced diarrhea (CID), a serious side effect that affects up to 80% of patients and often leads to treatment interruptions7 . Traditionally, predicting this toxicity has been difficult because the gastrointestinal systems of laboratory animals often do not translate well to humans.
The model represented individual cells in the crypt as independent "agents," each programmed with rules based on real biological behaviors (e.g., proliferation, differentiation, death).
The virtual crypt replicated the geometry, physical forces, and cell zonation of a real human crypt. It incorporated key cell types and clinically relevant signaling mechanisms that govern cell turnover and tissue health.
The effects of a chemotherapeutic drug, observed in experiments on human-derived intestinal organoids (mini-organs grown in a lab), were translated into the model.
The model was then run to simulate how the entire crypt system responds to the drug-induced injury over time, predicting the severity of damage that would lead to clinically observed diarrhea in patients.
The ABM served as a crucial "dictionary," translating the limited data from lab-grown organoids into a prediction of a whole-organ response in a human patient7 .
By replicating the complex, emergent behavior of the intestinal system, the model could predict whether a given drug dose and schedule would likely cause unacceptable levels of diarrhea. This allows drug developers to de-risk clinical trials by selecting safer dosing regimens before a drug is ever administered to a person. It also reduces reliance on animal models that are poor predictors for this specific toxicity.
Comparison of prediction accuracy between traditional and systems modeling approaches for chemotherapy-induced diarrhea.
| Aspect | Traditional Approach (Animal Models) | Systems Modeling Approach (ABM) |
|---|---|---|
| Predictive Accuracy | Low (significant interspecies differences) | High (based on human biology and data) |
| Key Limitation | Poor translation from animal to human | Requires high-quality human biological data |
| Primary Application | Reactive; observing toxicity after it occurs | Proactive; predicting toxicity before human trials |
| Regulatory Utility | Limited by translatability concerns | Provides a human-relevant, mechanistic rationale for dose selection |
The potential of systems models is clear, but for them to be used in high-stakes regulatory decisions—such as drug approval—they must be trusted. Regulatory agencies like the U.S. Food and Drug Administration (FDA) are actively developing frameworks to integrate these tools. A major focus is on model evaluation to understand and quantify the uncertainty in a model's predictions9 .
A proposed framework for building credible models emphasizes9 :
The model's purpose must be clearly defined and agreed upon by all stakeholders from the outset.
The model's scope and complexity must be appropriate for the question it is trying to answer.
The model must undergo rigorous testing, including sensitivity analysis and validation against real-world data.
The FDA's AI/ML-Based Software as a Medical Device (SaMD) Action Plan addresses the unique challenge of "adaptive" algorithms5 . It proposes requiring a "Predetermined Change Control Plan," where manufacturers must detail how an AI will learn and change over time while maintaining safety and efficacy. This shifts the regulatory focus from a one-time approval to continuous monitoring of a product's real-world performance.
| Model Type | Primary Application |
|---|---|
| Quantitative Systems Pharmacology (QSP) | Dosage regimen selection, biomarker identification, predicting efficacy and safety. |
| Agent-Based Model (ABM) | Understanding complex tissue-level responses, like immune system dynamics or toxicology. |
| Physiologically-Based Pharmacokinetic (PBPK) | Predicting drug-drug interactions and pharmacokinetics in specific patient populations. |
Increasing regulatory acceptance of systems modeling approaches over time.
Many cutting-edge systems biology approaches rely on advanced genetic tools to visualize and manipulate protein function in live organisms. The following toolkit highlights key reagents from a recent study that developed the GEARs (Genetically Encoded Affinity Reagents) system, a versatile platform for studying endogenous proteins in vivo8 .
A modular system for tagging endogenous proteins to enable their visualization, manipulation, or degradation in live models.
Short epitope tags & cognate bindersHigh-affinity binders for specific epitope tags; identified as particularly effective for visualizing protein localization.
NanobodiesEngineered for proper folding and function at the physiological temperatures of the model organism.
Genetically encoded nanobodies/scFvsAdapts GEARs to trigger the degradation of a target protein, allowing functional studies by protein removal.
F-box protein fused to a nanobody| Tool Name | Type | Function in Research |
|---|---|---|
| GEARs | Short epitope tags & cognate binders | Modular system for tagging endogenous proteins |
| NbALFA & NbMoon | Nanobodies | High-affinity binders for specific epitope tags |
| Codon-Optimized Binders | Genetically encoded nanobodies/scFvs | Proper folding at physiological temperatures |
| CRISPR/Cas9 with ssODNs | Gene editing system | Efficiently inserting small GEAR epitope tags |
| zGrad Degradation System | F-box protein fused to a nanobody | Trigger degradation of target proteins |
The integration of systems models into medical research and regulation is not just an incremental improvement; it is a paradigm shift. By embracing the inherent complexity of human biology, these models offer a path to more efficient, safer, and more personalized medical product development.
They help move the industry from a reactive stance—"steering the boat in one direction or another reacting to the waves"—to a proactive one, where we can finally see and navigate the currents ahead1 .
As these tools continue to evolve and gain regulatory acceptance, they promise a future where the journey of a new therapy from the lab to the patient is shorter, cheaper, and far more predictable. The silent, virtual world of systems modeling is poised to make a very real and loud impact on human health.
Systems modeling represents the convergence of biology, computational science, and medicine to create a more predictive approach to healthcare.