
Powering AI-Driven Drug Discovery with President of Isomorphic Labs Max Jaderberg
Hosted by Sarah Guo
- 39m
- Transcript
Summary
Written by a model from the transcript; not yet reviewed.
This episode covers the application of advanced artificial intelligence and machine learning to biopharma, drug discovery, and human health at Isomorphic Labs. The guest explains how the traditional biopharmaceutical industry relies on an empirical, trial-and-error process that has grown increasingly expensive and stagnant, while the number of scientific papers and potential molecular modalities has exploded. By building deep learning simulation and world models like AlphaFold, Isomorphic Labs aims to move drug design and disease biology onto computers to achieve experimental accuracy before entering the real world.
The discussion explores how scaling predictive models and generating proprietary datasets can bypass traditional animal models, which are often poor predictors of human clinical toxicity. The conversation also touches on the organization's focus on oncology, immunology, and inflammation, the challenges of targeting complex mechanisms like RAS mutations, and the balance between patient safety regulations and rapid innovation in drug development.
- Achieving experimental accuracy on computer simulation models reduces the need for expensive, time-consuming real-world laboratory iterations. 15:58
- Animal models are poor predictors of human toxicity, leading to high failure rates when drugs reach clinical trials. 21:37
- Developing more precise molecules will eventually enable safer combination drug treatments and complex multi-target therapies. 30:12
- Targeting root node mechanisms in cancer and inflammation can positively impact many downstream indications and patient populations. 27:56
Chapters
- 0:00Introduction and AI in Drug Discovery
Hosts introduce the episode and explain the potential of AI to transform drug discovery.
- 0:45Meet Max Yodberg and Isomorphic Labs
The hosts welcome Max Yodberg, discuss his background, and outline Isomorphic Labs’ mission to solve all disease.
- 1:56From DeepMind to Isomorphic: Founding the Company
Max recounts his transition from DeepMind to founding Isomorphic Labs and the motivations behind it.
- 7:00Challenges of the Traditional Drug Discovery Pipeline
The conversation highlights the empirical, costly, and slow nature of current drug development processes.
- 12:41Building a Cross‑Disciplinary Team
Discussion of hiring strategies, the balance of AI and life‑science talent, and fostering interdisciplinary collaboration.
- 15:50AI’s Role in Accelerating Discovery
Max explains how AI models can predict biology and reduce reliance on expensive real‑world experiments.
- 19:34Data Generation and Experimental Automation
Exploration of generating proprietary data, using automated labs, and integrating organoid technologies to improve predictive accuracy.
- 26:35Partnerships and Internal Development
Max describes how Isomorphic develops drugs internally and partners with pharma to tackle hard problems.
- 27:23Focus Areas: Oncology, Immunology, and Longevity
The team outlines their primary therapeutic targets and discusses the broader impact on healthspan and lifespan.
- 34:26Future Outlook and Clinical Trials
Closing remarks on upcoming clinical work, predictions for cancer treatability, and the broader vision for AI‑driven medicine.
Highlights
- 0:19
Moving Out of Real Worldstrategy
The guest explains the core goal of bringing drug discovery and biology onto computers to bypass the time and cost bottlenecks of synthesizing real atoms and testing real assays.
- 4:07
The guest describes transitioning from an engineering background and a PhD in computer vision pre-AlexNet to developing winning networks in ImageNet and eventually joining DeepMind.
- 5:38
The host notes that despite an exponential increase in scientific papers, new modalities, and costs reaching 1.5 billion dollars per drug, the rate of new drug approvals per year remains flat.
- 21:31
Poor Animal Predictabilitydata
The guest points out that rat models are only about thirty percent predictive of human clinical toxicity, highlighting the need for better human-centric predictive endpoints.
- 27:23
Internal Pipeline Focusstrategy
The guest shares that Isomorphic Labs focuses its internal pipeline on oncology, immunology, and inflammation while keeping their generalist AI drug design engine disease-agnostic.
- 30:12
Barriers to Drug Cocktailsinsight
The guest explains that the inability to safely combine multiple drugs stems from poor molecule engineering that causes liver toxicity and adverse drug-drug interactions.
- 37:50
Future Health Span Goalsinsight
The guest and host discuss the distinction between lifespan and health span, exploring how targeting the root causes of aging could prevent physical and mental degradation in older age.
Quotes
“like, there's toxicity that we didn't account for.”
Unnamed speaker · 0:12 “It went from a hundred, a few hundred million dollars per drug to one point five billion.”
Unnamed speaker · 6:04 “Things got a bit more sophisticated. It's a bit more empirical trial and error.”
Unnamed speaker · 7:10 “synthesizing real atoms and then testing them out in real assays and, you know, experiments.”
Unnamed speaker · 15:27 “you know, choose your study, but it's about thirty percent predictive of human clinical tox, like in a rat.”
Unnamed speaker · 21:27
Sponsored segments
- No Priors 38:58–39:13
About this episode
Drug discovery and development has long relied on expensive, time-consuming empirical trial and error. By shifting the discovery process onto computers, Isomorphic Labs wants to fundamentally change medicine from a process of discovery to one of precision engineering. Elad Gil sits down with Isomorphic’s president Max Jaderberg to discuss his vision for using predictive AI models to achieve experimental accuracy. Max details how the success of AlphaFold provided the first proof of concept that a deep learning model could generalize in the field of biology, and how an AI model’s compounding of learnings allows it to achieve generalizability with scale, thus greatly increasing the probability of success as well as speed in real world clinical trials. He also talks about evolving past the need for animal testing through in-silico models, Isomorphic’s focus on developing drug design engines that are agnostic of disease area, target class, and modality, why accurate molecule engineering unlocks the potential for multidrug cocktails that span complex biological pathways, and how he believes AI molecule optimization will improve both human lifespan and healthspan. Sign up for new podcasts every week. Email feedback to [email protected] Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @MaxJaderberg | @IsomorphicLabs Chapters: 00:00 – Trailer 00:45 – Max Jaderberg Introduction 01:24 – Isomorphic Origins 05:23 – Drug Discovery Stagnation 11:49 – Hiring at Isomorphic 14:35 – Bringing Drug Discovery Onto Computers 19:31 – Isomorphic’s Datasets 20:30 – Rethinking Endpoints 22:50 – External vs. Internal Dependences 23:57 – Automating Biologic Labs 26:22 – Internal Development vs. Partnerships 27:19 – Areas of Human Health Focus 29:21 – Multidrug Approaches 32:17 – Risk Appetite Variation 34:26 – Disease Treatment Outlook 35:42 – Human Longevity and Drug Development 39:16 – Conclusion
Transcript
7,712 words · 3 speakers
Speaker names marked are inferred from speaking time and the show’s known hosts — they are not voice-verified.
Topics
- AI
- DeepMind
- Deep learning
- AlphaFold
- Isomorphic Labs
- Computer vision
- Alphabet
- Reinforcement learning
- Regeneron
- Moderna
Page updated · transcript and summary are machine-generated