PPubMed13 Jan 2026Cited 1×

Mechanism‐informed machine learning for individualized tacrolimus dose adjustment in the early post‐kidney transplant period Yu et al.

Hui Yu, Zihan Qin, Logan S. Smith, Jeong M. Park, Hao‐Jie Zhu

Abstract

4## DISCUSSION This study developed and validated mechanism-informed ML models to predict tacrolimus Ctrough and guide dose adjustments in kidney transplant recipients during the early post-transplant period. Both the GRU and XGBoost models, informed by PK principles, outperformed traditional PopPK modelling, demonstrating superior and robust predictive performance in internal and external validations. Although PopPK modelling is widely used in MIPD, its performance may be compromised in dynamic clinical scenarios involving rapidly changing physiology, such as the early post-kidney transplant

A figure from Mechanism‐informed machine learning for individualized tacrolimus dose adjustment in the early post‐kidney transplant period Yu et al. — fig. from the paper

§ The Valyu brief

The paper, in five minutes

Read from the full text via Valyu Contents

Mechanism-informed machine learning models improve early post‑kidney transplant tacrolimus dosing accuracy and therapeutic target attainment.

What’s new

i.Developed mechanism-informed GRU and XGBoost models that outperformed traditional Bayesian PopPK models in predicting tacrolimus trough levels.
ii.Integrated pharmacokinetic principles into ML models, enhancing interpretability and generalizability without losing accuracy.
iii.Created a publicly accessible tool based on the GRU model for individualized tacrolimus dose adjustment.
iv.Validated the models internally with five‑fold cross‑validation and externally on an independent cohort.

How they did it

The study collected 4,311 tacrolimus trough concentrations from 1,624 adult kidney‑transplant recipients within the first 7 days post‑transplant. Two machine‑learning models (GRU and XGBoost) were built incorporating linear pharmacokinetic equations, then compared to a Bayesian population PK model and purely data‑driven ML models using five‑fold cross‑validation and external validation on a separate cohort.

What they found

In internal validation, the mechanism‑informed GRU model achieved MSE 7.81 ng²/mL² and R² 0.537, outperforming the Bayesian PopPK model (MSE 9.27, R² 0.450).
In external validation, the GRU model showed MSE 6.09 ng²/mL² and R² 0.397 versus the PopPK model (MSE 8.96, R² 0.211).
When administered doses fell within the GRU‑recommended range, 51.3% of trough measurements reached the therapeutic target (8‑12 ng/mL) versus 37.0% overall (p < 0.01).

Caveats

✕Genotype information was excluded due to missing data in the external cohort, potentially limiting personalization.
✕External validation showed reduced predictive bias and lower R², indicating limited generalizability across different patient populations.
✕The study focused only on the early (first 7 days) post‑transplant period and adult patients ≥18 years, restricting broader applicability.

§ The Valyu brief

Reading the full paper and taking notes. This takes a few seconds…

§ Ask this paper

Ask a question about this paper

Valyu reads the full text and answers from what the paper actually says.

§ Related work in other archives

Similar papers in all four archives

The same idea searched in arXiv, PubMed, bioRxiv and chemRxiv at once, so a method can lead you to where it is being applied.

AarXiv4 found

PPubMed3 found

BbioRxiv2 found

CchemRxiv2 found

Searching the other archives…