Optimizing tacrolimus dosage in post-renal transplantation using DoseOptimal framework: profiling CYP3A5 genetic...

Area of Science:

• Pharmacogenomics
• Transplant Medicine
• Machine Learning in Healthcare

Background:

• Optimal tacrolimus dosing is crucial for balancing efficacy and safety in renal transplant recipients.
• CYP3A5 genetic variants and patient variability necessitate personalized dosing strategies.
• Precision medicine approaches are vital for tailoring tacrolimus therapy.

Purpose of the Study:

• To optimize tacrolimus dosage prediction in renal transplant recipients.
• To integrate CYP3A5 genetic polymorphisms into the DoseOptimal framework for enhanced accuracy.
• To improve the interpretability and reliability of tacrolimus dosing decisions.

Main Methods:

• Developed the DoseOptimal framework by integrating top-performing machine learning models.
• Incorporated SHapley Additive exPlanations (SHAP) and decision tree insights for interpretability.
• Utilized clinical, demographic, and CYP3A5 genetic data from 1045 patients.

Main Results:

• The DoseOptimal framework achieved high performance (R²=0.884 training, R²=0.830 testing).
• Mean Absolute Error (MAE) was low (0.40 mg/day training, 0.41 mg/day testing).
• Accurate dosage prediction was achieved for 87.6% of the test cohort, with minimal under/overestimation.

Conclusions:

• The developed framework offers precision medicine-based tacrolimus dosing solutions.
• Tailoring dosage to individual genetic profiles minimizes errors.
• Enhanced dosing strategies improve patient outcomes in renal transplantation.