Digital twins have become one of the most promising tools in Alzheimer’s research, but the bigger story is what happens when they scale across disease areas. In this episode, Dr Andree Bates interviews Aaron Smith, Founder and Head of Machine Learning at Unlearn AI, about how “digital twin generators” can transform trial design by modelling realistic patient progression and improving statistical power without compromising the fundamentals of randomised controlled trials.

Aaron shares his journey from academic mathematics into computer vision and machine learning, then into biopharma, where Unlearn began by building generative models that learn the joint distribution of clinical variables. In practice, that means the model can take baseline patient measurements and generate likely future progressions that are as indistinguishable from real clinical records as possible.

The conversation dives into a key misconception: digital twins are not only about replacing control arms. Aaron explains a regulatory friendly approach where you keep standard trial structure, but add counterfactual information for every patient into the analysis. Unlearn’s best known method, ProCOVA (prognostic covariate adjustment), summarises a predicted control outcome per patient and uses it for covariate adjustment, creating more efficient treatment effect estimates. The headline result is simple: you can increase power, or reduce recruitment burden while maintaining power, potentially speeding time to results.

Finally, Aaron explains why scaling across diseases is genuinely hard. Data structures differ wildly by indication, missingness can block transfer learning, and areas like oncology require modelling complex treatment histories. He also highlights that combining sources is not just “more data”, it demands careful harmonisation and context modelling to avoid biased predictions, especially when bringing in real world evidence.

Topics Covered

What “digital twin generators” are in clinical trials
Generative modelling of clinical records and disease progression
Counterfactual prediction under standard of care
Why replacing control arms is not the only use case
ProCOVA and prognostic covariate adjustment
Getting more statistical power and reducing trial size
FDA openness to digital twins in trials and what it enables
Why scaling across disease areas is not just parameter tuning
Missing data, confounding context, and data harmonisation
CNS versus oncology modelling challenges
Real world evidence and how to validate digital twin models

About the Podcast

AI For Pharma Growth is the podcast from pioneering Pharma Artificial Intelligence entrepreneur Dr Andree Bates, created to help pharma, biotech and healthcare organisations understand how AI-based technologies can save time, grow brands, and improve company results.

This show blends deep sector experience with practical conversations that demystify AI for biopharma leaders, from start-up biotech right through to Big Pharma. Each episode features experts building AI-powered tools that are driving real-world results across discovery, R&D, clinical trials, medical affairs, market access, regulatory, insights, sales, marketing, and more.

Dr. Andree Bates LinkedIn | Facebook | X

E210: Beyond Alzheimer’s: Scaling Digital Twins Across Disease Areas

AI For Pharma Growth

E210: Beyond Alzheimer’s: Scaling Digital Twins Across Disease Areas