Virtual Twins: What They Are & Their Impact in Life Sciences

Life sciences leaders have been grappling with the same concerns for years without adequate solutions: Why do we still treat drug discovery, clinical development, and manufacturing as separate worlds? What would it take to accelerate drug discovery and shorten trial timelines? How can we eliminate inefficiencies across manufacturing?

The technologies now exist to finally address these questions—specifically in the oncology and rare disease spaces.

The Stakes Are Different in Oncology and Rare Disease

For both oncology and rare diseases—and especially when we consider cell and gene therapies (CGTs)—the development environment is uniquely unforgiving:

• Small patient populations. In rare diseases, there may only be a few hundred eligible patients globally. Every failed screening visit, protocol amendment, or manufacturing deviation carries outsized cost—both human and financial.
• Extreme biological complexity. With personalized autologous therapies, genomic interventions, tumor heterogeneity, the science doesn't forgive oversimplification.
• Manufacturing volatility. Unlike a small molecule that can be synthesized at scale, a CAR-T cell therapy is manufactured for a single patient; and it carries with it a process that must work every time, under tightly controlled conditions.

The traditional "trial and error" approach to development—run the experiment, observe the failure, and adjust—isn't just slow here, it can be ethically untenable. You can't burn through a rare disease population to optimize a protocol that you could have modeled computationally.

This is the problem that virtual twin technology was built to solve.

What Is a Virtual Twin?

A virtual twin isn't just a 3D model or a dashboard. It's a dynamic, computational representation of a real system, such as a molecule, patient population, study protocol, bioreactor, or supply chain, that can be simulated, stress-tested, and optimized before anything happens in the physical world. It’s a living model that connects to real data, evolves over time, and informs decisions continuously.

And when you chain those twins together from the molecule all the way through to the patient, you get something genuinely transformative: a digital thread that turns fragmented data islands into a unified picture of development.

Three Places Where Virtual Twins Are Changing the Game

1. Discovery and Molecular Design

The earliest stages of drug discovery eliminate enormous numbers of candidate molecules. Many compounds fail not because the underlying biology is wrong, but because key physical, chemical, and physiological properties—such as stability, solubility, delivery, off-target interactions, metabolism, or conformational behavior under physiological conditions—are difficult to predict reliably before you've run the experiment. Historically, the only way to address that uncertainty was extensive cycles of synthesis and testing. The field is increasingly looking for a better answer.

Computational methods, including molecular docking, quantitative structure–activity relationship (QSAR) modeling, and cheminformatics have supported drug discovery for decades. More recently, artificial intelligence and machine learning (AI/ML) have expanded the ability to analyze large biological and chemical datasets and generate predictions about molecular activity and properties. These models are data-driven; they learn statistical relationships from large datasets rather than explicitly modeling the underlying physical processes. While fast and scalable, their predictions can be hard to interpret mechanistically and they tend not to generalize well beyond the data they were trained on. They’re statistical, not mechanistic.

Physics-based models, such as molecular dynamics simulations, force-field calculations, and quantum mechanical methods take the opposite approach. They explicitly simulate molecular behavior using approximations to the laws of physics and chemistry, providing genuine mechanistic insight into how molecules interact and move. The limitation has always been computational cost. These methods are expensive to run and therefore constrained in the size of systems and the breadth of chemical space they can meaningfully explore.

The major direction in modern computational drug discovery is the integration of both, by combining the scale and pattern recognition of machine learning with the mechanistic grounding of physics-based models. That's what's changing, and it's changing fast.

For oncology and rare disease programs, the ability to simulate molecular dynamics is where the virtual twin earns its value. Predicting how a protein target behaves—not just its static structure, but how it moves, flexes, and interacts with other molecules across time—has historically required months of experimental work using techniques like x-ray crystallography or cryo-electron microscopy. AI-driven structure prediction tools, including AlphaFold2 for multimer structure prediction and OpenFold for monomers, have compressed what once took months into hours. 

But structure prediction alone is just the starting point. What platforms like BIOVIA (now powered by NVIDIA infrastructure) add is the ability to take those AI-generated structures and subject them to molecular dynamics simulations. Capabilities such as sampling conformational states, studying protein-protein and protein-peptide interactions, assessing formulation stability, and running virtual screening via docking are all available within a single integrated environment—accessible as a cloud-based SaaS application that doesn't require expensive hardware or specialized computational infrastructure.

For biotherapeutics, including antibodies, biologics, or CGT vectors, this is particularly consequential. Viral vector design for CGT involves molecular architectures of extraordinary complexity. Modeling how a vector will behave under physiological conditions, predicting its stability, and identifying potential immunogenicity risks before synthesis are the capabilities that compress the front end of drug development timelines and reduce the risk carried into manufacturing. 

For rare disease programs where patient populations are small, funding is constrained, and every development decision carries disproportionate weight, this isn't a computational luxury. It's a strategic imperative, and the foundation on which everything downstream depends.

2. Clinical Trial Design and Execution

This is where impact is most immediate and perhaps most underappreciated.

Protocol design is still largely an art form. Inclusion/exclusion criteria get debated in conference rooms. Disease progression assumptions get baked in based on historical precedent that may or may not reflect the specific population you're enrolling. Budget forecasts are built on assumptions that unravel the moment reality diverges from the model. The result? Amendments. Delays. Recruitment failures. And compounding costs.

The virtual twin changes this by introducing computation where intuition has been the default. Specifically in oncology and rare disease, where the margins for error are thinnest, this shift does more than improve the process; it changes what's possible.

Here are just a few examples of virtual twins that can shift the trajectory of a development program:

Financial Scenario Planning

Financial scenario planning is an example of the virtual twin applied to the budget. Clinical trials are expensive, and the financial plans built to support them are often fragile—constructed around assumptions that no one has stress-tested. Scenario and assumption modeling lets sponsors simulate variations in trial parameters and immediately see the downstream financial implications: what happens to the budget if enrollment takes three months longer than projected, or if a site in a key country underperforms? 

That's not just a finance function. For CGT and rare disease programs operating with constrained capital, the ability to optimize funding allocation and forecast outcomes across multiple scenarios before committing to a design can be the difference between a program that gets to the finish line and one that doesn't.

AI-powered Trial Design

AI-powered trial design brings the virtual twin concept directly to the patient. Modeling disease progression across different patient subsets lets sponsors predict which patients are most likely to respond to a treatment or to experience an adverse event. In oncology, where tumor biology is heterogeneous and the relationship between molecular profile and clinical response is complex, this has real implications for trial design such as tighter inclusion criteria, more precise stratification, and adaptive dosing strategies grounded in predicted response rather than historical averages. For CGT programs targeting rare genetic variants, the ability to model genomic intervention responses across a small, diverse patient population before the first patient is screened is particularly valuable.

The Synthetic Control Arm (SCA)

SCAs address one of the most persistent ethical and operational tensions in rare disease and oncology trials: the control arm problem. Traditional randomized designs require a meaningful proportion of patients (often a third to a half) to receive a placebo or inadequate standard of care. In indications where the unmet need is severe and the patient population is small, that design constraint raises ethical questions and creates recruitment barriers. The SCA resolves this by constructing an external control group from historical clinical trial data (for example, Medidata draws on a repository spanning 38,000 trials and 12 million patients). Using propensity-score modeling, the SCA matches real patients in the experimental arm with historical patients whose baseline demographics and disease characteristics closely align, creating a rigorous, regulatory-grade comparator.

The Medicenna phase 3 glioblastoma trial, where the FDA accepted a hybrid external control design in an indication that had traditionally relied on randomized controls, demonstrates how far this approach has matured. The SCA isn't a scientific shortcut; it's a mechanism for running more ethical trials with stronger designs that get therapies to patients faster.

Protocol Optimization

Protocol optimization, which on the Medidata platform is integrated into the Medidata Study Experience, takes scenario modeling into the operational architecture of the trial itself. Different protocol design decisions, such as endpoint selection, visit schedules, or inclusion/exclusion criteria, have downstream consequences that are notoriously difficult to anticipate. Sponsors can now simulate those consequences before they're locked in, quantifying how a given design choice will affect patient burden, site burden, and overall trial costs—and identifying the configurations most likely to support both scientific rigor and operational feasibility. This capability becomes a core risk management tool for rare disease programs where protocol amendments can deplete a limited patient population before the primary endpoint is reached.

Study Feasibility and Performance Analytics

These analytics apply AI-predictive modeling to the operational execution of the trial itself. Enrollment duration scenarios, country and site selection optimization, and early risk identification are the levers that determine whether a trial runs on time or doesn't. The ability to model enrollment trajectories before committing to a site strategy and to course-correct in near real-time during execution can compress timelines in ways that matter enormously for both sponsors and patients.

Taken together, the above capabilities represent a fundamentally different way of designing and running clinical trials—one where the consequences of design decisions are modeled before they're lived. This is the clinical virtual twin in practice.

3. Manufacturing and Process Optimization

This is where the personalized medicine revolution runs headfirst into the realities of biology and logistics.

Autologous cell therapies (where the drug is made from and for a single patient) represent perhaps the most complex manufacturing challenge in the history of pharmaceuticals. You're not running a batch process. You're running thousands of parallel, highly individualized processes simultaneously, each with its own timeline and quality requirements.

Virtual twins can model the bioreactor environment, simulate cleanroom logistics, and predict process variability before the manufacturing process is locked. This means identifying failure modes before they occur in a GMP suite, optimizing cell culture conditions computationally rather than through expensive wet lab iteration, and ensuring that the autologous scale-out process is robust before a single patient cell is processed.

For oncology manufacturers scaling from clinical to commercial, this is also where supply chain modeling becomes critical. Virtual twins of the end-to-end supply chain—from leukapheresis through infusion—can identify bottlenecks, model capacity constraints, and stress-test logistics scenarios that would otherwise only reveal themselves through expensive, real-world failures.

The numbers reinforce the value of what virtual twins deliver in manufacturing. According to Dassault Systèmes, the average time to design and install a drug manufacturing production line in a GMP environment is 10 months, with 20% of that total delivery time consumed by mandatory commissioning activities alone and 60% of commissioning time spent fixing software errors. Virtual commissioning—simulating and validating the entire manufacturing system before it exists physically—changes that calculus significantly: a 40% reduction in commissioning time, a 15% reduction in total time-to-market, an 80% reduction in re-engineering deviations, and a 40% reduction in the risk of test batch loss. For CGT manufacturers, where a failed batch isn't just a financial loss but a patient without a therapy, that risk reduction is not an operational metric, it's a patient outcome metric. 

Virtual twins in manufacturing are available in flexible delivery models. Some organizations want the software and the control that comes with building their own virtual twin. Others want the outcomes without the implementation overhead (preferring to have it delivered as a service). While both paths lead to compressed engineering timelines, less on-site testing, or faster commissioning, virtual twin as a service is a shortcut to value.

The Digital Thread: Why Connection Is Everything

Each of the three pillars above is valuable in isolation, but the real opportunity lies in how they connect.

A fully seamless digital thread, where data flows automatically from molecule to patient, is still an aspiration for most organizations. But there's a meaningful difference between aspiration and foundation. When the tools across discovery, clinical, and manufacturing are designed to interoperate on a shared platform, the foundation for the thread is there. That's a very different starting point than assembling disconnected point solutions and hoping they'll talk to each other.

The Human Impact of Virtual Twin Technology

These innovations are not just abstractions. They're the difference between:

• a child with a rare genetic disease receiving a therapy that works compared to one who doesn't
• a CAR-T patient receiving their cells on time compared to waiting in a hospital bed while manufacturing processes fail
• an oncology trial that generates clean data compared to one that requires three amendments and burns through the patient population before the sponsor realizes the protocol was flawed

And the numbers reflect it. Medidata's platform supported 80% of all FDA novel drug approvals in 2025, including 80% of orphan drug approvals and 85% of breakthrough therapy designations. Across oncology, hematology, immunology, and other high-priority therapeutic areas, the support rate for approvals exceeded 95%. These aren't market share numbers. They're a reflection of where the industry has chosen to place its trust when the stakes are highest.

Virtual twin technology is one of the most powerful tools we have right now to close the gap between what's biologically possible and what we can deliver to patients. That's worth talking about, loudly and often.

Learn more about how virtual twins are transforming life sciences.