Applying exposure-response analysis to enhance Mycophenolate Mofetil dosing precision in pediatric patients with immune-mediated renal diseases by machine learning models

• Models assist the off-label precision medication use in pediatric nephropathy • Dose recommendations guided by exposure-response analysis are helpful for treatment • Multi-model modeling with interpretable analysis makes the results more reliable Mycophenolate mofetil (MMF), a cornerstone immunosuppressant for lupus nephritis, is increasingly used off-label in pediatric immune-mediated renal diseases. The aims of this study were to develop and validate pharmacokinetic models for mycophenolic acid (MPA), the active metabolite of MMF, to optimize dosing strategies in Chinese pediatric patients with immune-mediated renal diseases guided by exposure-response analysis. We developed and validated PK models using 513 MPA concentration measurements from 171 Chinese pediatric patients with diverse immune-mediated renal diseases, split into training and testing cohorts (4:1 ratio). A two-compartment A two-compartment Population Pharmacokinetic (PopPK) model with first-order absorption was established. In parallel, seven Machine Learning (ML) models were trained using the selected feature with backward elimination process. Model performance was evaluated by multiple metrics to identify the best model. The Exposure-response analysis was performed in 20 refractory nephrotic syndrome patients to determine an effective exposure threshold for guiding individualized dosing. Among evaluated models, Random Forest demonstrated optimal performance, with SHapley Additive exPlanations (SHAP) analysis identifying plasma albumin, calcium, and hepatic/renal function markers as key predictors of MPA exposure. Exposure-response analysis in refractory nephrotic syndrome revealed a critical therapeutic threshold (AUC 0-12 h >30 mg·h/L), correlating with reduced 24-hour urinary protein levels and lower disease progression risk. Weight- and albumin-adjusted dosing regimens were formulated to optimize individualized therapy. Based on exposure-response analyses and established model simulations, we proposed stratified dosing recommendations tailored to patient characteristics, aiming to optimize MMF therapy in pediatric patients.