Interview with Jay Bradner, M.D., Amgen
James “Jay” Bradner, M.D., Executive Vice President, Research and Development, Artificial Intelligence and Data, Amgen
Interview
Drug Discovery & Chemical Biology
AI & Data in R&D
1
Reaching the Undruggable Proteome: Molecular glues, degraders and what remains out of reach
Bradner describes how chemical biology, molecular glues and induced proximity are opening historically intractable targets, particularly in gene control. But intrinsically disordered proteins and the intracellular delivery of large biomolecules remain major challenges.
Q1
Much of the proteome was once considered undruggable. Molecular glues, degraders, and induced proximity have changed that premise. What can we reach today that was out of reach ten years ago, and what still feels out of reach?
A:
The chemical biology mindset has contributed meaningful in-roads to the discovery of new molecules working in new ways, in many cases on historically intractable targets. Molecular glues, which remain a focus of my research, leverage large protein interfaces adjacent to binding sites to complete molecular recognition events, allowing the engagement of protein targets that would otherwise not be “druggable”. This has been particularly exciting in the field of gene control, which for years was the focus of my scholarship as an investigator. Gene control proteins, in particular transcription factors, frequently possess intrinsically disordered regions that rare adopt coherent secondary structures. High entropic penalties and absent binding sites continue to challenge the drug hunter. Beyond small molecules, approaches to deliver large biomolecules into cells remains a grand challenge, which we and others are working on to this day.
2
Where AI Is Changing Drug Development: From protein design to agentic workstreams
Bradner points to concrete AI applications across discovery, regulatory work and clinical development. His strongest near-term excitement is around agentic workstreams, while emphasizing that AI works best in the hands of disciplinary experts rather than as a substitute for them.
Q2
Your current role brings research, development, AI and data under one remit. Where is AI genuinely changing the probability of developing a successful medicine, and where is the field still mistaking hype for progress?
A:
Artificial Intelligence is proving to be a powerful catalyst for biomedical discovery research and therapeutic development, with a growing number of high-impact applications. Protein structure prediction accelerates both de novo discovery of active large molecules, as well as their optimization for drug-like properties. Document authoring decreases the time to regulatory filing. Pattern recognition identifies new target pathways in complex genomic and phenotypic datasets. Forecasting and clustering optimize clinical trial enrollment. Together, these and other tools cut through complexity, improve decision-making and ultimately improve the probability of success. Most exciting to me at this moment is the systematic application of agentic workstreams to critical phases of pharmaceutical science, and business. AI is an instrument best played by a disciplinary expert, who has accepted the invitation to approach their work in a new way. While there remains some distracting hype around AI across sectors, there is no question that AI is emerging as a powerful adjacency for biomedicine – which we call convergent innovation.
3
What JQ1 Taught Us: Open chemical probes can accelerate an entire field
The JQ1 experience became a case study in how an openly shared chemical probe can accelerate biology far beyond its original hypothesis. Bradner traces its impact from BRD4 biology to protein degradation, oncogenic transcription factors and therapeutics now in clinical development.
Q3
JQ1 became one of the most widely shared chemical probes in modern biology, yet BET inhibitors have proven difficult to convert into approved medicines. What did that teach you about the distance between a beautiful hypothesis and a drug — and about how openly discovery should be shared?
A:
In this post-genomic era, biomedicine is best practiced by hypothesis-testing. Where available, incisive prototype drugs paired with predictive pre-clinical models can expedite therapeutic translation. The experience of creating, characterizing, studying and sharing the BET bromodomain inhibitor, JQ1, provided a powerful learning. Open access to chemical probes opens up a field of study and accelerates medical progress. In the few years following the publication of JQ1, hundreds of laboratories worked with the molecule to learn about the biology of BRD4 in cancer, heart disease, fibrosis, inflammation, fertility, and most importantly the fundamental role of BRD4 in gene control (transcription elongation). In this way, JQ1 powerfully accelerated scientific discovery, bringing BRD4 biology into view in many disparate fields of study. Years later, scientists have used JQ1 as a work-horse of protein degradation, since our publication of the chemical solution to protein degradation and the molecule dBET1. Others have used JQ1 to recruit BRD4 to oncogenic transcription factors, such as the Androgen Receptor (at Halda, now J&J). JQ1-derived therapeutics are presently in advanced stages of clinical investigation as cancer therapies and for cardiovascular disease. I am hopeful that patients will benefit from the learnings of this remarkable molecule.
