Numerical modeling and prediction of late estimated glomerular filtration rate in kidney transplant recipients based on machine learning models and the Monte Carlo simulation method.

OBJECTIVES

This paper presents an experimental numerical method for modeling and analyzing stochastic systems. For this purpose, various machine prediction models are trained using the Monte Carlo simulation method. This method is presented using experimental data of a kidney transplantation with an immunosuppressive protocol based on tacrolimus.

METHODS

A multivariate regression model was constructed by previous authors based on a clinical study in which key independent physiological parameters such as serum creatinine and estimated glomerular filtration rate (eGFR) six months after transplantation, as well as the pharmacokinetics of tacrolimus, including the dose-adjusted trough concentration of tacrolimus (C0/D) and intrastation variability (IPV), and eGFR between 13 and 36 were the dependent variable. Using the Monte Carlo simulation method, this model is further applied to obtain the essential data for the optimization of the prediction models. To determine the optimal prediction model, the DecisionTreeClassifier, Random Forest Classifier, and XGBClassifier were trained and compared.

RESULTS

The results indicate that XGBoost is the most accurate, reliable and generalizable model among the classifiers tested, while Monte Carlo simulation represents a significant methodological advance in the field of kidney transplantation.

CONCLUSIONS

Advanced numerical methods for kidney transplant patients' therapy are step forward in optimization of current immunosuppressive protocols.