Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Jun 26.
Published in final edited form as: Mayo Clin Proc. 2025 Oct 1;100(11):1965–1975. doi: 10.1016/j.mayocp.2025.07.028

Individualized Medicine in the Era of Artificial Intelligence

Konstantinos N Lazaridis 1,2, Eric W Klee 3,4,5, Timothy B Curry 6,7, Victor E Ortega 8, William V Bobo 9, Arjun P Athreya 10, Rebekah M Samsonraj 11
PMCID: PMC13294813  NIHMSID: NIHMS2165183  PMID: 41037050

Abstract

The goal of individualized medicine is to provide a precise approach for the prevention, diagnosis, and treatment of disease, utilizing information about an individual’s genomic, environmental, and lifestyle patterns. The rapid advancements in artificial intelligence (AI) have led to the application of sophisticated approaches including machine learning, generative AI, and large language models that could lead to discoveries and practice implementations aimed at improving clinical decision making and health care delivery through diverse applications. Nevertheless, the exact role of AI in supporting individualized medicine continues to evolve. In this position paper, we present our principles of applying AI to bolster the vision of the Mayo Clinic Center for Individualized Medicine. Leveraging the Center for Individualized Medicine’s strategy on innovation and transformation as well as existing state-of-the-art operational architecture and capacity in handling large corpora of data (ie, multiomics, digital, multimodal data), we envision an AI-powered integration of such information toward an enterprise-wide adoption to serve the needs of individualized clinical practice of the future.

Individualized medicine aims to deliver precise strategies for the prevention, diagnosis, and treatment of human disease. The rapid progress in artificial intelligence (AI) has introduced sophisticated approaches that can lead to significant discoveries in medicine with clincial impact. These innovations promise to improve the clinical decision-making and health care delivery across various areas of practice benefiting patient care.

OVERVIEW OF INDIVIDUALIZED MEDICINE

Definition and Impact of Individualized Medicine

Individualized medicine, also known as personalized medicine or precision medicine, is a revolutionizing approach to health care that tailors prevention, diagnosis, and medical treatment to the unique characteristics of each patient.1 Taking into consideration an individual’s genomic information, environmental exposures, and lifestyle to ascertain health status, and develop prevention strategies as well as targeted and effective treatments, individualized medicine has a broad and farreaching impact.2 With a plan to transform the health care landscape through improved accessibility and affordability, individualized medicine empowers health care professionals and patients to make health care decisions and promotes a patient-centric approach. To advance the research of and facilitate the transformation of the clinical practice into individualized medicine, the Mayo Clinic Board of Trustees established the Center for Individualized Medicine (CIM) in 2010. The aim of this article is to describe how the CIM is leveraging advances in AI to bolster and promote individualized medicine from the perspective of discovery and translational sciences.

Evolution of Health Care Toward Individualized Medicine

Driven by advancements in science and our ever-increasing understanding of human biology, there is an evolution of health care from traditional cohort-based medicine toward a more individualized approach that is transforming the way we prevent, diagnose, and treat diseases. With the completion of the Human Genome Project in 2003 alongside innovations in sequencing and high-content screening platforms, it is now possible to obtain comprehensive genetic variant profiles of patients that allow for integrated and precise clinical decision making.1,3 Increasing availability of large-scale complex biological data sets and expanding capabilities in data analytics including bioinformatics have accelerated the understanding of disease mechanisms, development of predictive models, and individualized treatment strategies.4,5 Innovative digital health technologies for continuous monitoring and real-time data collection of vital signs, activity levels, and other health metrics using wearable devices, mobile apps, and electronic health records (EHRs) have facilitated individualized interventions and early detection of diseases.6,7 Integration of multidisciplinary advancements in genomics, bioinformatics, data science, and digital health have together propelled the evolution of health care toward an individualized clinical practice offering tremendous potential for feasible and effective health care outcomes.7 These innovations are also now acknowledged by the US Food and Drug Administration (FDA) in a white paper on the use of AI and machine learning strategies for the development of drugs and biological products.8

Artificial Intelligence

Artificial intelligence is a field of computer science and engineering that applies advanced algorithmic formulations embedded in intelligent systems that can compute data and aid in decision making (either as automation or human augmentation).9,10 In the context of health care, under the umbrella of AI, there are currently 4 broad classes of models that have a wide range of applications (Figure 1). With the overarching principle of using AI for extracting knowledge from data and translating it to the bedside, AI can enable predictions, subtyping, and generation of exemplary data. In particular, generative AI and LLMs could be used for generating text that embodies notes of clinicians’ visits with patients. While traditional AI analyzes user behavior data, generative AI utilizes such data to create personalized content in response to user prompts. Generative AI techniques use complex algorithms and models to generate, on demand, new and original content such as images, text, or based on patterns and structures learned from large data sets.11,12

FIGURE 1.

FIGURE 1.

Open in a new tab

Models of artificial intelligence (AI) health care.

Demonstrating high processing abilities for text generation, summarization, translation, and personalized query response, LLMs are driving major promises by executing cognitive tasks with near-human—level performance that could transform discovery and clinical sciences in health care.

Role of AI in Transforming Medical Research and Practice

Generative AI creates new opportunities to improve the quality of patient care through diverse and multimodal (ie, images, text, audio) data generation for improving the quality of care of patients. For example, rendered by generative AI and LLMs, improvements in core medical competencies such as factual knowledge and health care professional—patient interactions facilitate better communication and patient outcomes.13 Through simplified and streamlined summarization, AI can automate the review of medical documentation, imaging, and laboratory tests, bringing years of clinical data to the physician’s fingertips within seconds.14 Artificial intelligence—powered analysis of large volumes of medical literature can help clinicians and researchers stay up-to-date on new discoveries and emerging trends and have access to the most current, evidence-based medicine. Artificial intelligence—driven high-resolution analysis and screening of biomolecular structures and/or targets have accelerated drug discovery and design, alongside predicting efficacy and functional outcomes. Generative AI has broad applications across several health care segments including health care professionals, payers, biomedical technology, pharmaceutics, services and operations, and public health agencies.15–17 Moreover, generative AI overcomes some of the previous hurdles to AI adoption in health care. Requiring less data, generative AI is more adaptable to unfamiliar situations and can interface better with humans.18,19

Although generative AI technology is promising, the current level of technology is not infallible, and some near-term caution is warranted. Prior to broad adoption and implementation in health care, several inherent risks must be addressed, which include biased outputs, false readouts (ie, hallucinations), patient privacy, misuse, erroneous or biased source attribution (ie, citation of relevant literature), and overdependence. With careful integration and human oversight, it is expected that emerging generative AI and LLM tools can be responsibly and effectively implemented.

Potential of AI Implementation in Individualized Medicine

With powerful computational systems analyzing patient data, LLMs permit accurate disease detection and prediction. Beyond diagnosis, generative AI could accelerate and improve clinical trials and precision medicine therapies through patient-driven research and precision engineering of drugs. For example, digital twinning of clinical trials—including synthetic control groups—has recently been validated.20 Similarly, a tool developed by Synthesized Ltd has the potential to help researchers expand existing drugs beyond their initial use to treat other diseases, which could make medicines more accessible. Exscientia, the first company to automate drug design and to have an AI-designed molecule enter clinical trials, is using generative AI to analyze patient tissue and employ functional precision oncology to improve patient outcomes.21 The use of generative AI at the preclinical and clinical stages could expedite access to therapeutics, even for rare conditions wherein larger dimensions of data can be more efficiently analyzed to identify biomarkers of disease pathophysiology or drug mechanisms.

MAYO CLINIC CIM EFFORTS IN AI

The Mayo Clinic CIM is a visionary leader in omics-informed medical care with the overarching goal to discover, translate, and apply innovative individualized medicine products and services that bring hope and healing to every patient. In the past decade, fueled by integrated clinical and research data, expertise, and cutting-edge technology, the CIM has rapidly advanced preventive medicine, diagnostics, and therapeutics by applying genomics methodologies from rare to common conditions and from cancer to nonmalignant chronic disease. With a synergistic team of scientific, technical, and operational experts in genomics and other omics, the CIM’s efforts are focused to transform discovery science and lead implementation to predict, prevent, diagnose, and treat disease based on an individual’s omics profile.

The CIM’s vision is to transform the health care of every Mayo Clinic patient through full implementation of an omics-driven practice leveraging AI and powered by research discovery for novel diagnostics and therapeutics. By 2030, we expect that (1) genomes will be ubiquitous in practice, reimbursed and accompanied by clinical decision support, (2) novel diagnostics and therapeutics will emerge from analysis of the genome-exposome interaction, (3) many rare diseases (RDs) will be cured with gene-based and/or gene-editing therapies, and (4) cancer will be intercepted or detected early while still curable.

Overview of Mayo Clinic CIM’s Strategic Pillars

To meet the vision of an omics-driven clinical practice, the CIM’s focus is directed at initiatives that are innovative, transformational, and aligned with Mayo Clinic’s “Bold. Forward. Unbound.” strategic initiative, resulting in the establishment of 4 strategic pillars (Figure 2). The CIM has been successful in creating teams, building infrastructure, and realizing early wins in clinical translation of individualized medicine. While leveraging a robust infrastructure of experts in bioethics, bioinformatics, business development, communications, education, and information technology, we align resources and optimize capabilities to provide end-to-end support for our 4 strategic initiatives.

FIGURE 2.

FIGURE 2.

Open in a new tab

Strategic pillars and associated programs of Mayo Clinic Center for Individualized Medicine. AI, artificial intelligence.

Rare Disease Omics.

Rare disease omics aims to scale and simplify genomics for broad clinical transformation of those uncommon diseases, enabling utilization in never-before-considered indications, creating value, impact, and demand via genomics- and other omics-derived diagnoses and treatments for patients with rare and complex diseases across all specialties. Rare diseases—defined as conditions affecting fewer than 200,000 people; altogether, there are more than 7000 known RDs.

Population Omics.

Population omics is focused on establishing scalable approaches for broad genomic and other omics screening to improve the ability to predict/prevent disease; execute decentralized studies to identify new patient populations that may require tertiary care and integrate omics data including the biological effects of exposures into clinical care.

Functional Omics.

Functional omics seeks to innovate through discovery in the biological expression and function of disease by developing reusable models to speed the pace of scientific discovery on how genes work for quicker integration and impact on disease management and therapy.

Digital Omics.

Digital omics is creating a world-class, cloud-based, digital omics infrastructure to comprehensively capture, standardize, store, protect, govern, and deliver discrete omics data. Thereby, allowing these data to be easily queried by scientists and physicians alike, facilitating AI-based algorithm developments, propelling research, enabling a rich platform of genomic and other omics data, and transforming clinical care as the engine of an omics-driven practice.

Leveraging AI Toward Realizing CIM’s Vision for Individualized Health Care

Utilizing AI models for predicting disease diagnosis, outcomes, and personalized treatment recommendations form the core of the CIM’s operational strategy. Leveraging these AI technological innovations and automation, current efforts within each CIM pillar are directed to provide patient-centric outcomes with improved prognostic and diagnostic accuracy and therapeutic efficiency. Below, we provide specific paradigms wherein each pillar is designed to incorporate AI and automation.

Rare Disease Omics.

Impacting a relatively minor percentage of the population individually, RD collectively affects about 300 million lives worldwide. Artificial intelligence is transforming how RDs are diagnosed and treated.22 The CIM has created and is validating AI algorithms to identify Mayo Clinic patients within general practice who might benefit from genetic testing, with the goal of improved patient navigation to reduce the time to diagnosis and creating opportunities for individualized treatments. The CIM has also started to develop and analyze large omics data sets from selected RDs to identify patterns and markers characteristic of those diseases to allow for better understanding of the disease mechanisms. Creation of LLMs for RD applications requires a significant number of comprehensive training data sets to achieve reliable performance, which can be limited by the prevalence of RDs and the scarcity of comprehensive data. However, with advancements in accessibility to EHRs and multimodal databases, generative AI can create synthetic data sets, helping researchers overcome the dearth of real-world data for RDs. By creating LLMs, we expect to standardize and share clinical/medical terminology, which can maximize diagnosis accuracy and efficiency to identify patterns and associations that would otherwise be challenging for clinicians. In addition, AI-powered platforms are shifting the paradigm in finding therapies for RDs. It is now possible to rapidly analyze vast data sets of drugs to predict efficacy, allowing for repurposing of medications because novel drug development is unlikely due to the rarity of the diseases. Also, generative AI has the potential to unravel molecular signatures and pathways associated with RDs, which can be used in the design of precise targeted therapies, such a gene therapy, antisense oligonucleotides, and other small-molecule therapies. Finally, using a digital twin approach, clinical trial similars built using genotypic and phenotypic profiles combined with historical trial data can enable optimization of individualized trial design for RDs for which patient recruitment is generally challenging alongside excessive trial costs. Together, harnessing the potential of AI to transform the practice and research of RDs, the CIM is making continuous enhancements toward establishing new frontiers in individualized health care. Our strengths include Mayo Clinic’s long history of evaluating patients with RDs and our integrated multidisciplinary practice.

Population Omics.

Given that human health and disease data are intrinsically multimodal, generative AI holds the potential to accelerate individualized medicine through self-supervised learning and intelligible structuring of health/disease information while also providing scaled population-level real-world analytics. To this end, we plan to build generative AI workflows to integrate multimodal data such whole-genome sequencing and other omics data (ie, epigenomics, transcriptomics, metabolomics, proteomics, exposomics), behavioral, imaging, and clinical data, to provide an interpretable and holistic view of biological systems useful for individualized medicine. For example, polygenic risk scores that capture the genetic propensity to complex traits for each patient have proved useful for risk stratification and individualized treatment, as well as targeted recruitment for clinical research and trials. For example, the Trans-Omics for Precision Medicine (TOPMed) program and the All of Us Research Program are large national initiatives to build deep multimodal data resources from diverse participants. However, a major bottleneck lies in the insufficiency and/or inaccessibility of biomedical data in shared open data resources for multimodal pretraining. At Mayo Clinic, we have successfully created and implemented a large, decentralized whole-exome sequencing study, coined Tapestry, that has enabled detection of clinically actionable germline variants (for hereditary breast and ovarian cancer, familial hypercholesterolemia, and Lynch syndrome) from a diverse group of nearly 100,000 participants, and the study is currently being used for both clinical practice and research discovery. By simultaneous recruitment of diverse cohorts, the Tapestry study provides opportunities to examine genomic associations with diseases, assess disease risk in underrepresented minorities, and include presumed healthy individuals as control participants in research studies. Of note, Tapestry is characterized by substantial longitudinal assessments, unlike other measurements that are typically one-off snapshots of health status. Through the linking of genomic data to Mayo Clinic’s extensive clinical records with deep phenotypes, this is the first large-scale study, to our knowledge, that combines clinically interpreted and research whole-exome sequencing data on a diverse group of participants within a comprehensive, integrated, and accessible EHR system in a tertiary medical center. Key strengths for the CIM in this space hinge on our active plans to launch Tapestry 2.0, a multiomics, multispecimen prospective study of more than 10,000 participants.

Functional Omics.

With remarkable outcomes in sequencing in recent years, the vast amount of high-resolution biological data obtained from the genome, proteome, transcriptome, epigenome, metabolome, and microbiome can be analyzed at the single cell level, permitting improved insight into individual patients’ conditions and driving informed clinical decision making for patient care. Key elements of this pillar include the creation of functional laboratories for cultivating cell lines and advancing technologies in epigenetics and single-cell RNA sequencing. These technologies are important for studying how genes are activated and deactivated, as well as analyzing the functions of individual cells. With improved accessibility and sharing of genomic and other omics data, genome-wide association studies and characterization of complex human diseases and traits have now become realizable. We propose the creation of LLMs for integration of this diverse set of data types, allowing new insights into disease processes and creating a powerful tool for individualized medicine. Given the high-performance abilities of generative AI to rapidly synthesize patient data from diverse sources, such as omics, EHRs, and medical literature, functional omics when fully automated will become an indispensable tool for clinicians. Artificial intelligence—driven methods for repurposing existing FDA-approved drugs to match the genomic cause of disease will be enhanced by the creation of precision models of an individual patient’s disease, enabling these new drug applications to be tested for efficacy. Integrating other modalities of data such as imaging and histopathology with multiomics profiles through graph neural networks and dimensionality reduction approaches is expected to inform clinical trajectories and allow for precise and individualized prevention, diagnosis, and treatment. Key strengths for the CIM in this space lie in our establishment of the Translational Omics Program, providing access to a functional omics resource to generate multiomics data and execute functional studies on patient-derived cell lines, heterologous cell-based systems, and preclinical animal models. This pillar plans to intentionally take real-world clinical phenotypes (diagnoses, responses to treatment) from real-world patient populations, traced backward translationally to the bench, to generate novel biological markers and mechanistic hypotheses underlying the identified phenotypes, thus providing a molecular “dissection” of complicated phenotypes.

Digital Omics.

Our aim is to maintain a cloud-based, digital omics infrastructure to comprehensively capture, standardize, store, protect, and deliver discrete omics data, allowing it to be easily queried by scientists and physicians alike. Key to this goal is enabling automated access to a wide spectrum of data types, allowing novel AI models to be developed, tested, and implemented, and driving innovation in the care of our patients. Artificial intelligence—powered natural language processing has potential to organize and capture voice and text records of patient information, reducing the administrative load on health care professionals and creating the structured data necessary for these discoveries. We see opportunities for digitizing clinical trials, wearable sensor—based realtime monitoring of vital functions including heart rate, sleep, physical exercise, oxygen saturation, biomarkers and metabolites, glucose levels, and environmental exposure to pollutants and infectious agents in high-risk/disease outbreak areas, as well as the availability of digital AI-assisted virtual health coaches to provide timely advice based on the current needs of the patient in a personalized manner. For example, LLMs are being tested for increased ease in validating data included in this platform and furthering the usability of the systems developed by allowing clinicians to ask plain-text questions of this vast data resource. Integration of these efforts with Mayo Clinic Platform operations will further the discovery, validation, and deployment of health care solutions by creating access to an even greater amount of multi-institutional data spanning the globe. Collaborating with the Mayo Clinic Platform on integrating clinical and research data, the CIM is well positioned to provide personalized, predictive, and innovative care. Strengths for the CIM include our accessibility to large and diverse cohorts of research and clinical data and the availability of a data science—informed informatics team that can enable development of approaches to pretrain models toward the establishment of realistic and comprehensive benchmarks representative of the inherent diversity in health/disease data.

Merged patients’ omics data and clinical outcomes, as derived from the EHRs augmented by the CIM’s efforts in the generative AI space, are streamline to facilitate (1) development of structured and quarriable data sets that are comprehensive for institution-wide (or nation-wide?) adoption; (2) improved access and sharing of valuable biological data for informed clinical decision making and recommendations, fostering discovery science and medical education; and (3) “smart patient” education and empowerment to participate in their own health management, thus enriching the overall ecosystem of AI-driven individualized medicine.

CHALLENGES AND OPPORTUNITIES FOR AI IN ADVANCING INDIVIDUALIZED MEDICINE

Developments in AI models within the digital health space have shown potential yet pose concerns about bias, which necessitates cautious treading on unchartered waters and being vigilant of technical challenges and privacy issues.

Technical and Analytic Challenges

The inherently multimodal nature of human health data can be a boon and bane, bringing unique challenges in data collection, integration, and annotation. The complexities of health-related data sets are dependent on factors such as participant cohort sample size, diversity and heterogeneity, extent of phenotypic analyses, time points for follow-up, data collection standards, and breadth/depth of integration between data sources.9,23 To effectively train AI models yielding meaningful outcomes, we are building (ie, the Tapestry study) the diverse and inclusive large biomedical data sets of significant sample sizes with extensive phenotypic characterization and substantial longitudinal data that are required. However, unless automated, the acquisition of such large volumes of data will be technically challenging and financially demanding.24 In the realm of individualized medicine, algorithmic bias is prevalent with a vast majority of data derived from individuals of a specific racial/ethnic ancestry. Underrepresented populations often have fewer data points, leading to disparities in model performances and resulting in data-driven discrimination.25 This bias is also amplified when excluding participants with missing data, leading to introduction and perpetuation of inaccurate information.26 Artificial intelligence implementations can be hindered by inadequate information about the data used to make decisions and recommendations. Strategies to mitigate these biases include training LLMs on domain-specific biomedical data sets and creating models from bottom-up to establish clinical foundation models, which may have better efficiency and predictability. In addition, increased clinician and/or biomedical researcher involvement in model conception, construction, and deployment can eliminate some of the unintended errors otherwise overlooked by data science professionals who may lack the domain expertise associated with the field of medicine. Furthermore, training end-users (eg, health care professionals, scientists, medical students) to input the right queries within generative AI platforms by applying prompt engineering processes can reduce confirmation biases because LLMs can yield inconsistent and inaccurate information with the assumption that all input prompts are veritable. Retraining models on a regular basis with continuously evolving medical data will be key to overcoming the limitations of this valuable technology. Increased emphasis on explainable models that remove the black-box effect from traditional AI methods will also be critical in the adoption of expert decision support tools and will eliminate the potential for error introduced by expert decision-making algorithms. Finally, mandating periodic auditing processes and sensitivity measurements to evaluate algorithmic performance will be critical to improving model performances and ensuring fairness and accuracy.27

Safeguarding Patient Data in the Era of AI

Striking a balance between innovation and privacy is essential while adopting AI models in health care. Data privacy and ethical concerns pervade AI models, requiring meticulous attention to protecting sensitive patient/participant information and strict adherence to regulatory frameworks, such as HIPAA (Health Insurance Portability and Accountability Act). Even with deidentification, privacy breaches are still possible. Implementing safeguards against leaks of confidential information by AI models or protecting against other cybersecurity threats is critical. Adhering to data privacy regulations including strict guidelines for data sharing while adopting AI models should mitigate risks of malicious behavior. To overcome some of these primary concerns, one possibility is to train the models on synthetic data, which can help avoid potential pitfalls. New synthetic data can be artificially synthesized from real-world raw data through a statistical approach. However, the extent to which such data anonymization can benefit individualized medicine remains to be examined. With LLMs, there is limited or no established accountability for consequences of model outputs. Additionally, the extent to which AI should augment human intelligence or automate processes in medicine is an active area of research. It is also imperative to provide transparency about the data sets used for initial training of the AI tools. Home-grown AI solutions (eg, Mayo Clinic’s AI-assisted screening tool for advanced review of an electrocardiogram) are generally designed to have better oversight over patient information while also effectively addressing the specific needs of hospitals and patients in a controlled environment. As with any software and operating systems, which undergo updates as part of maintenance, routine inspection and evaluation of LLMs can help address such concerns. Establishment of governance authorities as well as collaborative consensus initiatives involving both patients and practitioners may serve to mitigate these challenges.27

Regulatory Frameworks and Ethical Guidelines

Equity, safety, and regulation are paramount when delegating human tasks to machines. When AI software and models are imported from an external commercial source into a hospital or health system, little is known about the data used to train the model, and there is uncertainty about ownership and accountability of resulting real-patient data upon implementation. Governance structures and reporting systems will be required to establish accountability standards in every setting in which AI models are a contributor to research findings or clinical recommendations/decisions.28 Recent efforts to ensure trustworthy and high-quality AI have resulted in the establishment of a nation-wide, nonprofit initiative—the Coalition for Health AI (CHAI) to provide a framework for the responsible use of AI in health care (https://www.chai.org). CHAI is steered by leading health care institutions, including Mayo Clinic, Johns Hopkins University, Stanford University, and Duke University, in partnerships with technology giants such as Microsoft and Google, as well as government agencies including the FDA and the Office of the National Coordinator for Health Information Technology, to establish AI assurance laboratories with the goal to bring structured, disciplined, standardized and evidence-based approaches to using AI. This nation-wide network of laboratories is designed to test, certify, and register AI models prior to enterprise-wide adoption. Independent of the framework, appropriate institution-specific guidelines should be put in place for data management while ensuring security and privacy. It remains unclear how regulations will be enforced in today’s world; it will be up to the end users to harness AI with an ethical conscience.

CONCLUSION

Artificial intelligence models and applications have shown promise in the field of individualized medicine. With the use of advanced algorithmic models trained on large multiomics data sets, LLM-based decision support for individualized medicine is now becoming a reality, with rapid developments expected in the future. From predicting patient outcomes to designing tailored therapies, AI models have the potential to revolutionize individualized medicine. As the field continues to evolve and transform the landscape of individualized medicine, it is crucial to address emerging issues surrounding the accountability of AI-driven clinical decision making, data privacy, regulatory considerations, and quality control. Creating platforms, initiatives, and partnerships that bring together researchers, clinicians, and AI experts will catalyze innovation. Finally, promoting education and training to bridge the gap between medicine and AI, while fostering an environment of trust and reliability is a key responsibility shared by our CIM as we will continue to deliver omics-based individualized care to our patients.

ACKNOWLEDGMENTS

The content of this work is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Grant Support:

This work was supported by the Mayo Clinic Center for Individualized Medicine; and in part by award #2041339 from the National Science Foundation (A.P.A.); grants or contracts from the Myocarditis Foundation, the Blue Gator Foundation, the Watzinger Foundation, Mayo Foundation for Medical Education and Research, the National Science Foundation, and the National Institutes of Health (W.V.B.); and grants from the National Heart, Lung, and Blood Institute (R01HL142992) (V.E.O.).

Abbreviations and Acronyms

AI

artificial intelligence

CIM

Center for Individualized Medicine

EHR

electronic health record

FDA

US Food and Drug Administration

LLM

large language model

RD

rare disease

Footnotes

POTENTIAL COMPETING INTERESTS

Dr Klee has patents planned, issued, or pending for a supervised learning method for classifying methylation disorders (disclosure filed) and for revolutionized genetic test selection and interpretation through generative AI and LLM (disclosure filed). Dr Ortega has participated in a data safety monitoring board or advisory board for Sanofi, AstraZeneca, and Regeneron Pharmaceuticals Inc. Dr Bobo has received royalties or licenses from UpToDate and has patents planned, issued, or pending for an analytics and machine learning framework for actionable intelligence from clinical and omics data. Dr Athreya has patents planned, issued, or pending for an analytics and machine learning framework for actionable intelligence from clinical and omics data and for a probabilistic graphical model-based prediction of outcomes in the treatment of major depressive disorder in adolescents. The other authors report no competing interests.

Contributor Information

Konstantinos N. Lazaridis, Center for Individualized Medicine, College of Medicine and Science, Mayo Clinic, Rochester, MN; Division of Gastroenterology and Hepatology, College of Medicine and Science, Mayo Clinic, Rochester, MN.

Eric W. Klee, Center for Individualized Medicine, College of Medicine and Science, Mayo Clinic, Rochester, MN; Department of Clinical Genomics, College of Medicine and Science, Mayo Clinic, Rochester, MN; Department of Quantitative Health Sciences, College of Medicine and Science, Mayo Clinic, Rochester, MN.

Timothy B. Curry, Center for Individualized Medicine, College of Medicine and Science, Mayo Clinic, Rochester, MN; Department of Anesthesiology and Perioperative Medicine, College of Medicine and Science, Mayo Clinic, Rochester, MN.

Victor E. Ortega, Center for Individualized Medicine, College of Medicine and Science, Mayo Clinic, Scottsdale, AZ.

William V. Bobo, Department of Psychiatry and Psychology, College of Medicine and Science, Mayo Clinic, Jacksonville, FL..

Arjun P. Athreya, Department of Molecular Pharmacology and Experimental Therapeutics, College of Medicine and Science, Mayo Clinic, Rochester, MN.

Rebekah M. Samsonraj, Center for Individualized Medicine, College of Medicine and Science, Mayo Clinic, Rochester, MN.

REFERENCES

  • 1.Green ED, Gunter C, Biesecker LG, et al. Strategic vision for improving human health at The Forefront of Genomics. Nature. 2020;586(7831):683–692. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Delpierre C, Lefèvre T. Precision and personalized medicine: what their current definition says and silences about the model of health they promote; implication for the development of personalized health. Front Sociol. 2023;8:1112159. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Aworunse OS, Adeniji O, Oyesola OL, Isewon I, Oyelade J, Obembe OO. Genomic interventions in medicine. Bioinform Biol Insights. 2018;12:1177932218816100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Sharma A, Lysenko A, Jia S, Boroevich KA, Tsunoda T. Advances in AI and machine learning for predictive medicine. J Hum Genet. 2024;69(10):487–497. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Batko K, Ślêzak A. The use of Big Data Analytics in healthcare. J Big Data. 2022;9(1):3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Li C, Wang J, Wang S, Zhang Y. A review of IoT applications in healthcare. Neurocomputing. 2024;565:127017. [Google Scholar]
  • 7.Walter JR, Xu S, Rogers JA. From lab to life: how wearable devices can improve health equity. Nat Commun. 2024;15(1):123. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Food US and Administration Drug. Using artifical intelligence & machine learning in the development of drug & biological products: discussion paper and request for feedback. Revised February 2025. Accessed 1 February 2025. Published May 2023. https://www.fda.gov/media/167973/download
  • 9.Xu Y, Liu X, Cao X, et al. Artificial intelligence: a powerful paradigm for scientific research. Innovation (Camb). 2021;2(4):100179. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Acosta JN, Falcone GJ, Rajpurkar P, Topol EJ. Multimodal biomedical AI. Nat Med. 2022;28(9):1773–1784. [DOI] [PubMed] [Google Scholar]
  • 11.Hanson SJ, Yadav V, Hanson C. Dense sample deep learning. Neural Comput. 2024;36(6):1228–1244. [DOI] [PubMed] [Google Scholar]
  • 12.Singhal K, Azizi S, Tu T, et al. Large language models encode clinical knowledge. Nature. 2023;620(7972):172–180. Published correction appears in Nature. 2023;620(7973):E19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Clusmann J, Kolbinger FR, Muti HS, et al. The future landscape of large language models in medicine. Commun Med (Lond). 2023;3(1):141. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Bhasker S, Bruce D, Lamb J, Stein G. Tackling healthcare’s biggest burdens with generative AI. McKinsey & Company website. Accessed 1 February 2025. Published July 10, 2023. https://www.mckinsey.com/industries/healthcare/our-insights/tackling-healthcares-biggest-burdens-with-generative-ai
  • 15.Mukherjee J, Sharma R, Dutta P, Bhunia B. Artificial intelligence in healthcare: a mastery. Biotechnol Genet Eng Rev. 2024;40(3):1659–1708. [DOI] [PubMed] [Google Scholar]
  • 16.Haymond S, McCudden C. Rise of the machines: artificial intelligence and the clinical laboratory. J Appl Lab Med. 2021;6(6):1640–1654. [DOI] [PubMed] [Google Scholar]
  • 17.Davenport T, Kalakota R. The potential for artificial intelligence in healthcare. Future Healthc J. 2019;6(2):94–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Cadamuro J Rise of the machines: the inevitable evolution of medicine and medical laboratories intertwining with artificial intelligence—a narrative review. Diagnostics (Basel). 2021;11(8):1399. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Haque A, Milstein A, Fei-Fei L. Illuminating the dark spaces of healthcare with ambient intelligence. Nature. 2020;585(7824):193–202. [DOI] [PubMed] [Google Scholar]
  • 20.Katsoulakis E, Wang Q, Wu H, et al. Digital twins for health: a scoping review. NPJ Digit Med. 2024;7(1):77. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Sousa BB, de Almeida CR, Barahona AF, et al. Selective inhibition of Bruton’s tyrosine kinase by a designed covalent ligand leads to potent therapeutic efficacy in blood cancers relative to clinically used inhibitors. ACS Pharmacol Transl Sci. 2022;5(11):1156–1168. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Wojtara M, Rana E, Rahman T, Khanna P, Singh H. Artificial intelligence in rare disease diagnosis and treatment. Clin Transl Sci. 2023;16(11):2106–2111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Fittipaldi S, Legaz A, Maito M, et al. Heterogeneous factors influence social cognition across diverse settings in brain health and age-related diseases. Nature Mental Health. 2024;2:63–75. [Google Scholar]
  • 24.Thirunavukarasu AJ, Ting DSJ, Elangovan K, Gutierrez L, Tan TF, Ting DSW. Large language models in medicine. Nat Med. 2023;29(8):1930–1940. [DOI] [PubMed] [Google Scholar]
  • 25.Loree JM, Anand S, Dasari A, et al. Disparity of race reporting and representation in clinical trials leading to cancer drug approvals from 2008 to 2018. JAMA Oncol. 2019;5(10):e191870. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Haneuse S, Arterburn D, Daniels MJ. Assessing missing data assumptions in EHR-based studies: a complex and underappreciated task [editorial]. JAMA Netw Open. 2021;4(2):e210184. [DOI] [PubMed] [Google Scholar]
  • 27.Meskó B, Topol EJ. The imperative for regulatory oversight of large language models (or generative AI) in healthcare. NPJ Digit Med. 2023;6(1):120. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Mennella C, Maniscalco U, De Pietro G, Esposito M. Ethical and regulatory challenges of AI technologies in healthcare: a narrative review. Heliyon. 2024;10(4):e26297. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES