Patient digital twins & in-silico trials

01Overview and value chain

Markers: [EC: Directive 2001/83/EC + EMA qualification opinion procedure | OECD: Biotech-health | Regulator: FDA (USA), EMA (European Union), NMPA (China)]

Patient digital twins and in-silico trials use computational patient models and virtual simulation to reduce, accelerate or de-risk clinical drug development, rather than treating patients directly. Unlearn’s AI-generated digital twins are increasingly used to power synthetic control arms in place of a portion of a trial’s control group, and the company published a roadmap on causal inference and digital twins for the future of clinical trials in npj Digital Medicine in June 2026; it also announced a collaboration with Acumen Pharmaceuticals in July 2026 to explore digital-twin analyses of Alzheimer’s disease clinical programs. Certara’s Simcyp Simulator, a physiologically-based pharmacokinetic (PBPK) platform built over 25 years with a consortium of more than 35 pharmaceutical companies, released Version 25 in early 2026 building on EMA qualification, and Certara reported that Simcyp simulation results replaced ten human trials for the chronic myeloid leukemia therapy asciminib. VeriSIM Life is building what it describes as “full-stack predictive infrastructure” to help pharma companies foresee translation challenges from laboratory discovery to human therapy, formalizing a research collaboration with the FDA’s National Center for Toxicological Research in June 2026 under the agency’s Roadmap to Reducing Animal Testing in Preclinical Safety Studies. In France, Novadiscovery (Nova In Silico) prospectively predicted the LDL-C outcomes of Merck’s Phase 3 CORALreef Lipids trial using its jinkō causal-AI in-silico clinical trial platform, announced in November 2025.

The key directions of patient digital twins & in-silico trials are:

1. AI-generated synthetic control arms (AI-Generated Synthetic Control Arms): digital twins that replace or supplement a portion of a trial’s control group, reducing the number of patients needed to receive placebo — Unlearn.AI’s TwinRCT methodology.
2. Physiologically-based pharmacokinetic modeling (Physiologically-Based Pharmacokinetic Modeling): mechanistic simulation of a drug’s absorption, distribution, metabolism and excretion, regulator-qualified to substitute for some human trials — Certara’s Simcyp Simulator.
3. Preclinical-to-clinical translation AI (Preclinical-to-Clinical Translation AI): predictive infrastructure that forecasts how a molecule will behave moving from lab to human trials, aligned with regulatory efforts to reduce animal testing — VeriSIM Life.
4. Causal-AI virtual patient simulation (Causal-AI Virtual Patient Simulation): virtual patient “twins” simulating real-world physiology, disease progression and treatment response for trial design and prediction — Novadiscovery’s jinkō platform.

Sectoral value chain

[patient/physiology model development] ──> [model validation against real trial data] ──> [regulatory qualification (FDA/EMA/NMPA)]
                                                    │
                                       (trial design/simulation deployment)
                                                    │
                                                    ▼
[expanded regulatory acceptance] <─── [prospective prediction accuracy tracking] <─────┘

Value chain levels

Cross-cutting technologies of the sector:

• Causal AI digital twin generation (AI-Generated Synthetic Control Arms, Causal-AI Virtual Patient Simulation): machine-learning methods that generate an individualized predicted patient trajectory, usable as a synthetic control or trial-design input.
• Mechanistic PBPK simulation (Physiologically-Based Pharmacokinetic Modeling): whole-body physiological modeling of drug absorption, distribution, metabolism and excretion, built and refined over decades with pharma-industry consortium input.
• Cross-species translation modeling (Preclinical-to-Clinical Translation AI): predictive models bridging animal/preclinical data to human clinical outcomes, supporting regulatory efforts to reduce animal testing.

02US

The US hosts the leading concentration of digital-twin and in-silico trial companies, spanning synthetic control arms, PBPK modeling and preclinical-to-clinical translation AI, each with direct regulatory or pharma partnership validation.

Unlearn.AI, Certara, VeriSIM Life, FDA

• Unlearn.AI: its AI-generated digital twins are increasingly used to power synthetic control arms, and the company published a roadmap on causal inference and digital twins for the future of clinical trials in npj Digital Medicine in June 2026; it also announced a collaboration with Acumen Pharmaceuticals in July 2026 to explore digital-twin analyses of Alzheimer’s disease clinical programs.
• Certara: its Simcyp Simulator, a PBPK platform built over 25 years with a consortium of more than 35 pharmaceutical companies, released Version 25 in early 2026 building on EMA qualification, and Certara reported that Simcyp simulation results replaced ten human trials for the chronic myeloid leukemia therapy asciminib.
• VeriSIM Life: is building “full-stack predictive infrastructure” to help pharma companies foresee translation challenges from laboratory discovery to human therapy, formalizing a research collaboration with the FDA’s National Center for Toxicological Research in June 2026 under the agency’s Roadmap to Reducing Animal Testing in Preclinical Safety Studies.
• FDA framework: digital-twin and in-silico trial methods require FDA qualification (e.g., through the agency’s Model-Informed Drug Development or Innovative Science and Technology Approaches for New Drugs pathways) to be used as regulatory-grade evidence, with Certara’s EMA-qualified Simcyp platform illustrating the cross-regulator qualification path.

03CN

No dedicated Chinese digital-twin or in-silico clinical trial company cleared source confirmation as of 2026; Chinese sources covering this period center on the broader digital-twin-in-life-sciences market and research commentary (including domestic coverage of a Nature Biotechnology paper on “virtual immune cells”) rather than a specific commercial originator.

market coverage, NMPA pathway

• Market coverage focus: Chinese industry and trade-press reporting on digital twins in life sciences during this period addresses the broader technology category and market outlook rather than a dedicated in-silico-trial platform company.
• NMPA pathway: any digital-twin or in-silico trial method supporting a Chinese regulatory submission would require NMPA acceptance under China’s model-informed drug development framework, following the same qualification logic as FDA and EMA pathways.

04EU

France anchors Europe’s patient digital twin capacity through a causal-AI platform with a prospectively validated, named pharma-trial prediction.

Novadiscovery (Nova In Silico), EMA

• Novadiscovery (Nova In Silico; Lyon, France): prospectively predicted the LDL-C outcomes of Merck’s Phase 3 CORALreef Lipids trial using its jinkō causal-AI in-silico clinical trial platform, announced in November 2025 — a virtual patient “twin” simulating real-world physiology, disease progression and therapeutic response ahead of the trial’s actual readout.
• EMA framework: in-silico trial and digital-twin methods supporting an EU regulatory submission proceed through EMA’s qualification opinion procedure for novel methodologies, the same regulatory-science pathway underlying Certara’s Simcyp qualification.

05Leading companies and research institutes

06Tech stack and innovations

The stack applies computational patient modeling and simulation to reduce, accelerate or de-risk clinical drug development ahead of or alongside human trials.

1. AI-generated synthetic control arms (AI-Generated Synthetic Control Arms):
   • Unlearn.AI’s TwinRCT methodology generates individualized digital twins usable as synthetic control-arm data, reducing reliance on placebo-arm patients, with its approach now the subject of a peer-reviewed roadmap and a named Alzheimer’s-disease pharma collaboration.
2. Physiologically-based pharmacokinetic modeling (Physiologically-Based Pharmacokinetic Modeling):
   • Certara’s Simcyp Simulator mechanistically models drug absorption, distribution, metabolism and excretion, with EMA-qualified results that have replaced ten human trials for a leukemia therapy.
3. Preclinical-to-clinical translation AI (Preclinical-to-Clinical Translation AI):
   • VeriSIM Life’s predictive infrastructure forecasts translation risk from lab to human therapy, formalized through a direct FDA National Center for Toxicological Research research collaboration.
4. Causal-AI virtual patient simulation (Causal-AI Virtual Patient Simulation):
   • Novadiscovery’s jinkō platform builds virtual patient twins simulating physiology and disease progression, validated by a prospective, named prediction of a Merck Phase 3 trial’s outcome.

07Value chains and production pipelines

Industrial pipeline of a patient digital twin/in-silico trial method (FDA MIDD / EMA qualification opinion / NMPA)

┌───────────────────────────┐      ┌───────────────────────────┐
│ 1. Patient/physiology     │ ───> │ 2. Model validation        │
│    model development         │      │    against real trial data  │
└───────────────────────────┘      └───────────────────────────┘
                                                   │
                                                   ▼
┌───────────────────────────┐      ┌───────────────────────────┐
│ 4. Trial design/           │ <─── │ 3. Regulatory              │
│    simulation deployment     │      │    qualification             │
└───────────────────────────┘      └───────────────────────────┘
              │
              ▼
┌───────────────────────────┐      ┌───────────────────────────┐
│ 5. Prospective prediction │ ───> │ 6. Expanded regulatory    │
│    accuracy tracking         │      │    acceptance                │
└───────────────────────────┘      └───────────────────────────┘

Stage 1: Patient and physiology model development

A computational model of patient physiology, disease progression or drug pharmacokinetics is built from existing clinical and biological data, as with Certara’s Simcyp PBPK platform built over 25 years with pharma-consortium input.

Stage 2: Model validation against real trial data

The model’s predictions are tested against actual completed trial outcomes, the validation step behind Certara’s claim that Simcyp results replaced ten human trials for asciminib.

Stage 3: Regulatory qualification

FDA, EMA or NMPA qualification is obtained for the model or method to support regulatory submissions, as with Certara’s EMA-qualified Simcyp platform and VeriSIM Life’s formal FDA NCTR research collaboration.

Stage 4: Trial design and simulation deployment

The qualified model designs a trial, generates a synthetic control arm, or simulates an outcome prospectively, as with Unlearn.AI’s TwinRCT synthetic control arms and Novadiscovery’s prospective CORALreef Lipids prediction.

Stage 5: Prospective prediction accuracy tracking

The model’s prospective predictions are compared against the trial’s actual real-world outcome once available, the evidence base Novadiscovery is building with its named, prospectively-announced Merck Phase 3 prediction.

Stage 6: Expanded regulatory acceptance

A strong accuracy track record supports broader regulatory acceptance across more indications, the pattern behind Certara’s Simcyp Version 25 update building on prior EMA qualification and Unlearn.AI’s expanding pharma partnership base moving from methodology publication into a named Alzheimer’s collaboration.