In a groundbreaking advancement bridging artificial intelligence and personalized oncology, researchers from the Yong Loo Lin School of Medicine at the National University of Singapore (NUS Medicine) have successfully demonstrated an AI-driven platform capable of optimizing chemotherapy dosing for patients with advanced solid tumors. This pioneering clinical study marks a significant departure from traditional population-based cancer treatment paradigms, ushering in a new era where drug dosages can be dynamically tailored to the intricate, evolving biological responses of each individual patient.
Until now, much of artificial intelligence’s contributions to healthcare have largely been confined to retrospective analyses or theoretical models, leaving a vast potential unfulfilled in direct clinical application. However, led by Professor Dean Ho, Director of the Institute for Digital Medicine (WisDM) at NUS Medicine, the research team has deployed the CURATE.AI platform in a real-world clinical setting—specifically at the National University Cancer Institute, Singapore (NCIS). Their system employed continuous monitoring of two hallmark cancer biomarkers, carcinoembryonic antigen (CEA) and cancer antigen 125 (CA125), across a cohort of 10 patients diagnosed with advanced solid tumors to develop personalized digital twins. These digital twins serve as intimate virtual replicas of individual patients’ tumor biology and therapeutic responses, enabling precise real-time calibration of chemotherapy doses.
By meticulously analyzing the dynamic biomarker fluctuations in response to varying chemotherapy doses, the CURATE.AI platform guided clinicians to adjust treatment regimens with unprecedented precision. Remarkably, over a treatment period spanning from August 2020 to September 2022, 97.2% of the AI-recommended dose modifications were adopted by clinicians. The adjustments led, on average, to approximately 20% lower drug doses in some patients, spotlighting the promising potential to not only maintain therapeutic efficacy but also reduce chemotherapy-induced toxicity and associated healthcare costs.
Traditional oncology often relies on standardized dosing protocols derived from population averages, largely overlooking the considerable heterogeneity in patient responses and tumor evolution during the course of treatment. This limitation presents a pressing challenge as tumor physiology and drug sensitivity are far from static, varying significantly over time within each patient. CURATE.AI circumvents this challenge by harnessing patient-specific, longitudinal clinical data—integrating drug type, administered dose, and objective biomarker responses—to construct an evolving digital pharmacodynamic model. This model empowers the selection of an optimal chemotherapy dose tailored to the patient’s unique, contemporary tumor landscape.
Professor Dean Ho emphasized the innovative nature of this approach, highlighting that many extant AI systems are constrained by reliance on population-level static datasets or retrospective analyses. By contrast, CURATE.AI dynamically responds to individual patient data in real time, effectively capturing intra-patient variability and the continuous metabolic interplay between chemotherapeutic agents and tumor cells. This represents a crucial paradigm shift, enabling iterative, adaptive treatment optimization and heralding the onset of truly precision-guided oncology.
The clinical lead, Associate Professor Raghav Sundar, underscored the translational importance of this study. He reflected on the historical challenge faced by oncologists striving for personalized chemotherapy dosing due to the lack of suitable tools to objectively and dynamically tailor drug regimens. The CURATE.AI trial’s promising findings lay important groundwork for future expansive randomized controlled trials, poised to rigorously evaluate the platform’s efficacy and safety relative to standard-of-care protocols. The clinical implications extend beyond dosing precision, promising to mitigate adverse drug reactions and enhance patient quality of life.
At the core of CURATE.AI’s success lies its sophisticated algorithmic architecture that synergizes Bayesian optimization with mechanistic understanding of cancer biomarker kinetics. Such integration facilitates high-fidelity forecasting of dose-response curves unique to each patient. Furthermore, by repeatedly recalibrating dose selections based on biomarker feedback, the AI system adapts seamlessly to tumor evolution and drug resistance mechanisms that often undermine long-term chemotherapeutic success.
Beyond the study’s immediate oncology focus, the researchers are optimistic about the wider applicability of the CURATE.AI platform across diverse therapeutic domains. Preliminary adaptations are underway to extend its functionalities into immunotherapy regimens, hypertensive medication titration, and interventions designed to enhance healthspan within the longevity medicine landscape. This versatility underscores CURATE.AI’s foundational potential to revolutionize personalized dosing strategies well beyond its initial cancer cohort.
A vital insight from this work, highlighted by co-author Nigel Foo, is the recognition that therapeutic data efficacy is contingent not merely on volume, but on strategic, context-sensitive acquisition. By synchronizing incremental drug dose changes with concomitant biomarker trajectories, CURATE.AI capitalizes on temporal data richness, exposing nuanced pharmacodynamic interactions that are otherwise obscured in traditional clinical datasets. The concept of digital twins crystallizes this insight, enabling a feedback loop of data-driven, patient-specific care planning.
This research represents one of the first tangible illustrations of an AI-driven platform being integrated into everyday clinical treatment decisions, moving beyond the laboratory or theoretical sphere into the practical domain where patients benefit directly. The feasibility trial lays a robust foundation for subsequent multi-center trials with larger sample sizes designed to scrutinize the platform’s reproducibility and impact on long-term clinical outcomes such as progression-free survival and overall survival.
Published recently in the distinguished journal npj Precision Oncology, the findings position CURATE.AI at the frontier of next-generation oncology therapeutics. While conventional cancer care largely depends on pre-defined dosing schemas resistant to mid-course alterations, CURATE.AI epitomizes an adaptive, continuously learning system. Such agility aligns with the emerging understanding of cancer as a highly heterogeneous and time-variant disease, necessitating equally dynamic treatment strategies.
Ultimately, the success of this AI-enabled personalized dosing platform holds profound implications for healthcare economics. By potentially lowering drug dosages without compromising efficacy, CURATE.AI could alleviate the financial burden on healthcare systems and patients alike while limiting exposure-related toxicities that diminish patients’ quality of life. This dual advantage represents a compelling incentive for accelerating regulatory approval processes and clinical adoption worldwide.
As the oncology community grapples with escalating complexity in cancer management and burgeoning molecular data streams, CURATE.AI exemplifies the transformative convergence of digital health technologies with precision medicine. Its ability to deliver individualized, evidence-based treatment adjustments in real time crystallizes the promise of AI not merely as an analytical tool, but as a direct driver of improved patient outcomes in routine clinical care.
Subject of Research: Personalized dose selection platform for patients with advanced solid tumors using AI-driven digital twins.
Article Title: Personalized dose selection platform for patients with solid tumors in the PRECISE CURATE.AI feasibility trial.
News Publication Date: 21-Feb-2025
Web References:
https://www.nature.com/articles/s41698-025-00835-7
http://dx.doi.org/10.1038/s41698-025-00835-7
Image Credits: NUS Medicine
Keywords: Cancer research, Digital data, Drug therapy, Artificial intelligence, Cancer patients, Cancer medication, Chemotherapy, Drug studies, Chemical analysis
