Key Takeaways
- Amgen is embedding artificial intelligence (AI) throughout the entire drug‑development pipeline—from molecule design to regulatory submission.
- The company’s internal machine‑learning system ATOMIC streamlines site selection, patient recruitment, and registration for clinical trials.
- AI has cut the time needed to prepare regulatory documents from six months to just six weeks.
- “Go/No‑Go” decisions at each clinical stage are now aided by AI, helping to prioritize the most promising candidates before costly late‑stage trials.
- Amgen’s next frontier is the digital twin, a patient‑specific model that predicts treatment trajectories and optimal switch points.
- The AI effort is powered by massive biological datasets from Amgen’s acquisition of Decode Genetics and the NVIDIA super‑computer Freyja in Iceland.
- CMO Paul Burton stresses that AI complements—but does not replace—clinical trials, projecting a ~25 % reduction in total drug‑development time by the early 2030s.
Introduction and Interview Setting
Paul Burton, Amgen’s senior vice president and chief medical officer (CMO), recently sat down with Maeil Business Newspaper at the Shilla Hotel in Seoul to discuss how the global biotechnology firm is harnessing artificial intelligence (AI) across its drug‑development value chain. “Using AI, we can design better molecules and treatments faster and connect patients to appropriate clinical trials,” Burton told the reporter, underscoring the strategic shift toward data‑driven innovation. The interview highlighted Amgen’s conviction that AI is not a peripheral tool but a core engine for accelerating every phase of bringing new therapies to market.
AI Across Drug Development Stages
Amgen has instituted AI “at all stages of new drug development,” according to Burton. Early‑stage medicinal chemistry benefits from generative models that propose novel chemical structures with desirable potency, selectivity, and safety profiles. In preclinical research, machine‑learning algorithms sift through vast omics datasets to pinpoint disease‑relevant targets and predict off‑liability risks. By integrating AI into these upstream steps, the company aims to increase the hit‑to‑lead conversion rate while shrinking the timelines traditionally consumed by trial‑and‑error chemistry.
Accelerating Regulatory Documentation
One of the most tangible efficiencies AI has delivered is in the preparation of regulatory submissions. Burton noted that, historically, compiling the dossiers required after completing clinical trials took six months. With AI‑assisted document generation, automation of data extraction, and natural‑language processing for report writing, the same task now consumes only six weeks. “It used to take six months to prepare related documents after completing clinical trials, but now it only takes six weeks,” he said, illustrating how AI can lift a significant administrative bottleneck and free regulatory teams to focus on strategy and compliance nuances.
Machine Learning Platform ATOMIC for Clinical Trials
To further tighten the clinical trial workflow, Amgen deploys its proprietary machine‑learning system, ATOMIC. The platform identifies trial sites with the highest likelihood of enrolling eligible patients, matches patient profiles to study inclusion criteria, and even automates the registration process. By predicting which investigators will achieve rapid recruitment, ATOMIC reduces site‑selection cycles and minimizes costly delays. Burton emphasized that the system “will find a clinical trial institution that can connect suitable patients to clinical trials and even proceed with patient registration,” showcasing a closed‑loop approach where AI feeds directly into operational execution.
Go/No‑Go Decisions and Candidate Selection
As drug candidates advance, the financial and scientific stakes rise sharply, especially in Phase II and III trials. Amgen employs AI‑driven Go/No‑Go analytics to evaluate interim data, safety signals, and efficacy trends at each decision gate. These models weigh probabilistic outcomes against resource allocation, enabling the company to halt underperforming programs early and double‑down on those with a higher likelihood of success. Burton remarked that “It is important to select promising candidates as the cost and time that need to be put into late-stage clinical trials increase,” highlighting how AI serves as a risk‑mitigation tool that preserves capital while preserving scientific rigor.
Future Focus on Digital Twins
Looking beyond traditional AI applications, Burton expressed particular excitement about the concept of a digital twin—a virtual replica of an individual patient’s physiological state that simulates treatment journeys over time. “The area I’m personally very looking forward to is the digital twin,” he said. “I think it’s a great opportunity to innovate clinical decision-making methods not only in the development of new drugs but also in medicine as a whole.” By continuously updating the twin with real‑world data (e.g., wearable sensor readings, lab results, imaging), clinicians could forecast when a therapy is likely to lose efficacy, anticipate adverse events, and pre‑emptively switch to alternative regimens. Amgen envisions deploying these models first in oncology and immunology, where therapeutic pathways are highly dynamic and patient‑specific.
Data Foundations: Decode Genetics and NVIDIA Freyja
The potency of Amgen’s AI initiatives rests on the depth and quality of its underlying data. The company’s 2012 acquisition of Decode Genetics, an Icelandic firm specializing in population‑scale genomic and phenotypic analysis, granted Amgen access to petabytes (PB) of human disease‑related data. This treasure trove includes whole‑genome sequences, detailed health records, and longitudinal outcomes from a genetically homogeneous yet deeply phenotyped populace. To crunch these massive datasets, Amgen relies on the NVIDIA‑powered super‑computer Freyja, located in Iceland, which delivers the computational horsepower needed for training complex deep‑learning models and running simulations at scale. Burton affirmed, “AI’s performance ultimately depends on what data is put into it,” adding that ongoing model refinement focuses on “how to design new molecules and how to find optimal targets.”
Limitations of AI and Role of Clinical Trials
Despite the optimism, Burton was clear about AI’s boundaries. He stressed that AI cannot replace the clinical trial itself, but rather serves to reduce and accelerate the time required between clinical stages. “Personally, AI will reduce the time required to develop the entire drug by about 25 % in the early 2030s,” he projected. This tempered view acknowledges that while AI can optimize candidate selection, trial design, and data analysis, the ultimate proof of safety and efficacy still demands rigorous human testing under regulated conditions. The commentary reassures stakeholders that AI is positioned as an augmentative force—enhancing human expertise, not supplanting it.
Conclusion and Outlook
Amgen’s AI‑first strategy, articulated by CMO Paul Burton, illustrates a broad transformation underway in the biopharma industry. By embedding intelligent algorithms from molecule conception through regulatory filing, leveraging platforms like ATOMIC for smarter trial execution, and investing in next‑generation concepts such as digital twins, the company aims to shave years off the traditional drug‑development timeline. The foundation of this effort—massive, high‑quality data from Decode Genetics and the computational might of NVIDIA’s Freyja—ensures that AI models are trained on realistic biological complexity. As Burton anticipates, the convergence of these advances could yield a ~25 % acceleration in bringing new medicines to patients by the early 2030s, marking a pivotal shift toward faster, more precise, and data‑driven healthcare innovation.
https://www.mk.co.kr/en/it/12158399