Episode Description
Summary
In Part 1 of this two-part conversation, Anastassia and Dr. Andrée Bates take the concept of digital twins from its industrial roots — NASA rockets and GE power plants — all the way into the human body. Andrée unpacks what a true clinical-grade digital twin actually requires (individuation, credibility evidence, uncertainty quantification, and regulator-aligned analytical roles), and why many things called “digital twins” in healthcare today are really just well-marketed predictive models. The conversation travels through clinical trials, rare disease drug development, AI-assisted drug repurposing, and lands in genuinely mind-expanding territory: brain cells powering server farms, a non-invasive headband restoring speech to paralysed patients, and the bold thesis that AI alone is not enough — that medicine needs physics embedded into its models.
Key Takeaways
- A real digital twin has three parts: a physical reference (the human), a virtual representation, and a live data link that continuously updates — without all three, it’s just a predictive model
- Synthetic control arms are already FDA- and EMA-accepted in clinical trials, especially for rare diseases where putting patients in a placebo arm would be unethical
- Clinical-grade digital twins require four properties: individuation, formal verification/validation for regulators, calibrated uncertainty quantification (not point estimates), and a regulator-aligned statistical analysis plan
- The FDA approved digital twins for clinical trials in late 2022
- AI alone is insufficient for drug development — despite ~$20 billion invested, no AI-discovered drug has reached market yet; physics-based modeling (“world models”) is the missing layer
- AI excels at drug repurposing, demonstrated powerfully during COVID with baricitinib and atazanavir identified from existing approved drugs
- 8,000 rare diseases exist, but only ~100 have treatments — AI-driven matching of existing drugs to rare disease profiles is a massively under-leveraged opportunity
- Full-body digital twins remain a decade+ away due to the complexity of organ-system interaction and computational cost — individual organ twins are mature, but integration is the hard problem
Guest Bio
Dr. Andrée Bates is the Chairwoman, Founder, and CEO of Eularis, AI consultancy for the pharmaceutical and life sciences industry. She hosts her own podcast with over 220 episodes on AI in pharma.
Chapters
- 00:00 The Emergence of Digital Twins in Medicine
- 03:03 Understanding Digital Twins: Definition and Applications
- 10:09 Digital Twins in Clinical Trials: A New Paradigm
- 10:17 Dynamic Systems and AI in Drug Development
- 39:53 Leveraging AI for Drug Repurposing
- 41:38 Regulatory Landscape for AI and Digital Twins
- 42:45 Exploring the Digital Twin Concept
- 43:51 Regulatory Landscape and AI in Medicine