Machine learning driven precision medicine in tacrolimus dosing: current research and future perspectives in liver and kidney transplantation.

BACKGROUND: Tacrolimus (TAC) is a core immunosuppressant used to prevent transplant rejection after organ transplantation. However, its clinical use is limited by a narrow therapeutic window and substantial interindividual pharmacokinetic variability. Subtherapeutic TAC concentrations are closely associated with acute or chronic transplant rejection, whereas supratherapeutic concentrations often cause severe drug toxicity. As precision medicine advances, integrating multidimensional patient clinical characteristics with machine learning (ML) provides important support for individualized TAC dosing adjustments after liver and kidney transplantation.

AIM: This review summarizes the application of ML in determining the initial dose of TAC and in recommending early blood drug concentrations in liver and kidney transplant patients, analyzes the core challenges in model extrapolation and clinical translation in existing research, compares their characteristics with traditional models, and looks ahead to the construction direction of long-term immunosuppression monitoring and TAC dynamic adjustment models.

METHODS: This narrative review summarizes the benefits of ML for tacrolimus dosing in liver and kidney transplantation. Its key strength is integrating patients' multidimensional clinical data to enable personalized dosing. We focus on ML applications in initial tacrolimus dose selection, early post-transplant concentration range determination, and dosage form conversion. We also systematically discuss current model limitations and future directions.

CONCLUSION: ML has been widely applied to develop precise models for TAC immunosuppressive drug administration in patients after liver and kidney transplantation, providing a reference for individualized treatment. However, such models have not yet been widely applied in clinical practice and generally lack the ability to dynamically regulate drug concentrations over the long term. How to promote the model's true application in clinical practice remains to be explored in depth, but it holds significant prospects for application and research value in individualized treatment.