科学发现与人工智能科学到患者的旅程：来自该领域领导者的加速器和障碍（英）

The science-to-patient journey: accelerants and obstacles fromleaders across the field Foreword Every decision made today about how AI integrates into the science-to-patient value chain will shape the care patients receive for a generation. The question we put to all of the participants: as AI accelerates everystage of the science-to-patient journey, what does it actually take to What emerged was remarkable — a discovery scientist, a clinical triallead, and a medical publisher, working through the same value chainfrom completely different points, yet arriving at the same conclusions. A new peer-reviewed paper ispublished roughly every ten seconds.A practicing specialist would need toread for twenty-six hours a day just to With AI now touching every stage, the technology is moving fasterthan ever. Yet, the incentives, the trust infrastructure, and thewillingness to act together are not moving at the same speed. That DEFININGTHE OPPORTUNITIES AHEAD What follows is an honest account of that shared diagnosis. Thisreport captures where the opportunities are real, where the barriersare structural, and where the system is genuinely stuck. It’s where the None of this is fixable by working harder, hiring more people, orwriting better review articles. We’re grateful to every leader who participated with the openness andcandor that made this possible. The insights here belong to them, and There’s a gap between what science knows and what patients receive.AI has the potential to close that gap, but only if the whole system We hope what follows is genuinely useful for your thinking, for yourorganization’s decisions, and for the conversations you’ll take back to ASSEMBLING THE PARTICIPANTS The choices made now will determine whether AI transformsscience in ways patients actually feel and benefit from. Wiley and IQVIA convened 25 senior leaders from across thescience-to-patient value chain: physicians, researchers, publishers,technologists, health system leaders and pharmaceutical manufacturerswho rarely share a room, let alone a diagnosis. Wiley sits at the pointwhere knowledge is created and validated; IQVIA at the point where this We hope this sharpens the urgency. Rob KotchiePresident Real World Evidence& Clinical Technology Solutions, IQVIA Armughan RafatChief AI & Data Services Officer, Wiley Gathering input Table of contents Twenty-five senior leaders across the science-to-patient value chain came together to tackle onequestion:How do we responsibly apply the rapidgrowth of AI in all aspects of healthcare and lifesciences to accelerate the translation of rigorousscience into real-world patient impact?What Inside the value chain Five tensions and early thinking11 The transformational shift in AI forscience is already happening — and mostorganizations are treating it like an ITproblem. No single sector has the answer,and the science-to-patient chain is beingrewritten whether we act now or not.The only question is whether the people Cross-cutting themes and unresolved questions Closing — Ethan Mollick, Wharton professor, NYTbestselling author of Co-Intelligence and theforthcoming Co-Existence (October 2026),shared this with the group May 2026 For decades, scientific discovery has operated like a relay race. A discoverymoves from academic lab to clinical trial to peer review to clinical practice,passed baton by baton, each stage completing its leg before handing off to The relay raceof scientific discovery:from sequential to The model was never elegant — handoffs were slow, information was lostin translation, and the runners lacked consistent standards and a sharedlexicon, rarely knowing what was happening in the other legs of the race. AI has broken that assumption. The speed of discovery is now outrunning thesystem designed to validate, disseminate, and apply it. Academic research isgenerating hypotheses faster than clinical trials can test them. Clinical trialsare producing data faster than peer review can evaluate it. AI-generatedmedical information — pulled from everything between peer-reviewed Clinical development &evidence generationB The relay race model doesn’t fail catastrophically under these conditions.It stalls. Each stage, optimized in isolation, becomes a bottleneck for everystage downstream. The result: more output, less accuracy, and a system that Validation, publication, &dissemination of clinicaltrial results What replaces it isn’t a faster relay race. It’s a fundamentally different model:one where discovery, validation, and adoption operate as a continuous,interconnected loop rather than a sequential hand-off. It’s more of a hub-and-spoke model. Where Real-World Evidence (RWE) feeds back into trial design.Where failed experiments are legible to the next researcher. Where patients Real-world application &patient impactD The question these conversations surfaced — and left deliberately open — iswhether the institutions, incentives, and trust structures of science