Traditional drug development relies on a lengthy, sequential, and often uncertain process of experimentation, beginning with target identification and hit discovery, followed by lead optimization, preclinical studies, and extensive multi-phase clinical trials before market approval.
This "trial-and-error" approach involves significant laboratory and human resources, leading to high costs and failure rates. Advances in computational modeling and simulation (M&S) are transforming this process, enabling researchers to design, test, and optimize new therapies more efficiently and at less cost. These tools allow scientists to integrate biological, chemical, and clinical data into predictive models that can forecast how a drug behaves in the human body and how patients might respond.
By combining mechanistic understanding with quantitative analysis, M&S helps guide critical early-stage decisions from selecting the most promising candidates to designing efficient clinical trials. In doing so, these approaches can increase success rates, optimize dosing, and reduce reliance on animal studies, representing a major shift in how new medicines are developed.
Early drug insights
The attrition rate of drugs in clinical trials is still extremely high, with approximately 90 percent of drugs failing to reach approval. A major reason for failure is lack of efficacy, which has led researchers to increasingly use translational pharmacokinetics/pharmacodynamics (PK/PD) modeling. This is the process of linking what the body does to a drug (PK) with what the drug does to the body (PD).
Understanding these relationships early enables teams to predict the right dose and schedule, choose biomarkers and sampling time points for clinical trials, and set clear go/no-go criteria before committing to expensive trials. These models also greatly increase the likelihood of demonstrating proof-of-mechanism (PoM) — evidence that the drug engages its target and produces the expected biological effect — in early clinical trials.
Daniel Veres, co-founder and Chief Scientific Officer at Turbine, emphasized that “these approaches are most useful in the preclinical stage, where they help design and optimize hypotheses before a patient’s health is affected.” By exploring different dosing schedules, biomarker strategies, and exposure-response relationships early, teams can design clinical studies that are more likely to succeed.
“Preclinical mechanistic modeling has a strong and growing track record,” said Piet van der Graaf, Senior Vice President and Head of Quantitative Systems Pharmacology at Certara. He points to a recent analysis of AstraZeneca’s portfolio showing that “projects with robust PK/PD packages achieved an 85 percent PoM success rate, compared with just 33 percent for projects with basic packages.”
Beyond boosting early PoM outcomes, this study also showed that PK/PD frameworks allow researchers to identify weak candidates sooner, saving time and resources while accelerating the development of more promising drugs.
Dose optimization
Another clear example of the increasing value of M&S tools in drug development is in dose optimization.
Establishing a dosing regimen that maximizes clinical benefit while minimizing toxicity, unscheduled dose interruptions, or premature discontinuation is a critical objective for drug developers. By using in silico evaluations, researchers can identify promising dosing scenarios before patients are exposed, reducing risk and increasing the potential for therapeutic success.
According to Mark Davies, Head of Quantitative Pharmacology and Data Science at Physiomics, “the leading area for model-informed drug development (MIDD) currently is in supporting dose selection in oncology, driven in part by the shift from systemic cytotoxic therapies to targeted agents.” Optimizing the benefit/risk ratio is increasingly important, and regulatory initiatives like the FDA’s Project Optimus actively encourage the use of MIDD to guide dose selection in oncology.
Orr Inbar, CEO of QuantHealth, agreed, “we’ve seen some of the biggest impacts in oncology. It’s a tough space — only about 4 percent of oncology trials make it from Phase 1 to approval — and the need for safe, effective new treatments is still growing. Our technology has been able to simulate oncology trials with 88 percent accuracy, allowing pharma teams to design smarter, more successful trials.” Inbar noted that predictive modeling is also proving valuable in autoimmune, cardiometabolic, and gastroenterology programs, with smaller and mid-sized biotechs especially quick to adopt AI to sharpen and accelerate precision medicine approaches.
Simulating clinical trials
Being able to predict entire clinical trial outcomes before a single patient is enrolled represents a major shift in the approach to drug development. Instead of relying solely on costly and time-consuming trial-and-error in the clinic, researchers can use mechanistic models to anticipate how different doses, schedules, or patient populations will respond.
Tawanda Gumbo, co-founder and CEO of Phase Advance, applied predictive modeling to infectious diseases with great success. “Working with Otsuka and the Gates Foundation, we predicted that a triple-drug regimen for tuberculosis would provide a 4-month 100 percent cure rate at its lowest dose, compared to two other doses. A minimal prospective clinical trial later confirmed our prediction, allowing Otsuka to proceed directly with the lowest dose in a larger clinical trial — saving an estimated $90 million and sparing 700 patients from unnecessary risk.”
Scaling, standardization, and trust
Despite these gains, broader adoption of M&S still faces technical and cultural hurdles. Daniel Veres highlights the main challenges: “Models often lack robustness or scalability, and even when the tools work, adoption is slowed by a trust gap. The main challenges remain technical maturity and building confidence in predictions.”
Mark Davies emphasized education as key. “Successful integration also requires helping non-modellers overcome reticence, moving from viewing MIDD as ‘nice to have’ to making it an integral decision-making tool,” he said.
For Orr Inbar, the sticking points are data access and validation. “In clinical research, even getting access to high-quality real-world data can be a major hurdle — it’s often scattered, incomplete, or siloed. And once you have it, the challenge is turning it into usable knowledge that truly reflects real patients. On top of that, validating models to the highest standard is critical to building trust with sponsors and regulators. Integration into clinical development workflows must also safeguard quality, protect data security, and work within existing processes.”
Meanwhile, regulatory frameworks are gradually catching up, providing structure for industry implementation. The FDA-endorsed American Society of Mechanical Engineers (ASME) Verification and Validation 40 (V&V40) and the International Council for Harmonization (ICH) M15 guidance have established best practices for model development, validation, and submission. Van der Graaf noted that PBPK models are essentially mature, with European Medicines Agency guidance already in place, while quantitative systems pharmacology (QSP) approaches are “a few years behind but catching up quickly.”
The future of M&S in drug development
Looking forward, M&S could begin to replace, rather than just supplement, early-phase testing in select areas. Regulatory pathways are already being developed for M&S applications in dermal and topical drugs, rare disease studies with small patient populations, and early toxicology screening. “In the next two to three years, the fastest growth is expected in toxicology and safety predictions,” said Veres. “Predictive technologies are mature enough to integrate into standard research and development practice.”
Van der Graaf adds that convergence of AI with QSP and physiologically based pharmacokinetic models, along with digital twins and virtual patient technologies, will enable more precise, data-driven predictions of drug behavior and treatment outcomes. He cautions that complete replacement of animal studies will take time, but key areas are already seeing reduced reliance thanks to advanced mechanistic and organ-on-a-chip models.
With robust modeling, AI integration, and growing regulatory acceptance, companies are increasingly using virtual tools to guide preclinical and clinical decisions — saving time, reducing cost, and ultimately improving the probability of bringing safe and effective medicines to patients.
This piece is part of a series examining the growing role of modeling and simulation in both drug development and regulatory decision-making. For more on the regulatory side, read our other article on how the FDA, EMA, and others are embracing modeling and simulation to modernize decision-making.
Frequently asked questions (FAQ):
What role do computational models play in early drug development?
Computational modeling and simulation (M&S) tools integrate biological, chemical, and clinical data into predictive frameworks. These models help scientists forecast drug behavior, anticipate patient responses, and make data-driven decisions earlier in development, reducing costs and failure rates.
Can modeling help optimize drug dosing before human trials?
Yes. In silico evaluations allow researchers to explore dosing regimens that maximize benefit and minimize toxicity before patient exposure. This is particularly impactful in oncology, where regulatory initiatives like FDA’s Project Optimus encourage model-informed dose selection.
Is it possible to simulate entire clinical trials virtually?
Emerging approaches now predict trial outcomes before enrollment. For example, predictive modeling has guided tuberculosis trials, confirming dosing strategies while saving millions of dollars and sparing patients from unnecessary risk. Similar methods are expanding into oncology and other therapeutic areas.
What barriers remain for broader adoption of these tools?
Challenges include access to high-quality data, validation of models, and building trust among sponsors, clinicians, and regulators. While frameworks such as FDA’s ASME V&V40 and ICH M15 guidance are advancing standards, cultural and technical hurdles still slow widespread integration.
