Construction method of organ transplantation perioperative period precise drug delivery model, precise drug delivery prediction device and electronic equipment

Abstract

The invention provides a construction method of an organ transplantation perioperative period precise drug delivery model, a precise drug delivery prediction device and electronic equipment, and belongs to the technical field of crossing of precise medicine and pharmacology. A primary absorption and elimination two-atrioventricular model with lagging time is adopted as a basic model, covariables are introduced, model fitting is carried out, and an organ transplantation perioperative period precise drug delivery model is constructed. According to the method, internal verification and external verification are carried out on the basis of a goodness-of-fit diagnostic diagram, a prediction correction visual prediction test, a bootstrap method and a normalized prediction distribution error, finally Monte Carlo simulation is carried out to design an administration scheme, an individualized administration scheme is recommended for a patient in the perioperative period, and the model obtained through the construction method provided by the invention has the advantages of high accuracy and high reliability. The adverse reaction occurrence rate of the mycophenolate mofetil capsule can be greatly reduced, the MPA exposure compliance rate is improved, and the occurrence rate of adverse events after organ transplantation is reduced.

Description

Technical Field

[0001] The present invention relates to the interdisciplinary field of precision medicine and pharmacology, and in particular to a method for constructing a precision medication model for the perioperative period of organ transplantation, a precision medication prediction device, and electronic equipment. Background Art

[0002] Organ transplantation is the most effective means of treating end-stage organ failure. The perioperative period of organ transplantation is a critical stage of recovery after organ transplantation. The rational use of postoperative immunosuppressants is crucial to preventing graft rejection. Mycophenolic acid (MPA) drugs are commonly used immunosuppressants that can reversibly inhibit the key enzyme inosine monophosphate dehydrogenase (IMPDH) in the de novo synthesis pathway of guanine monophosphate (GMP) in lymphocytes, thereby inhibiting the proliferation of lymphocytes and thus inhibiting the immune response. After organ transplantation surgery, MPA drugs are usually combined with other immunosuppressants such as tacrolimus and cyclosporine to achieve the best immunosuppressive effect.

[0003] Mycophenolate Mofetil (MMF) is a commonly used MPA drug. During the use of mycophenolate mofetil, due to its extremely narrow therapeutic window, its efficacy and adverse reactions are closely related to blood drug concentration. Therefore, drug monitoring is required to evaluate the in vivo exposure of mycophenolate mofetil (AUC 0-12h ), the patient's AUC 0-12h Patients with low AUC will face a high risk of transplant rejection. 0-12h If the dosage is high, there will be a high risk of drug toxicity and side effects. Based on this, it is necessary to develop a precise drug delivery model to reduce the complexity and risk of mycophenolate mofetil capsule treatment.

[0004] The pharmacokinetic properties of mycophenolate mofetil show significant individual differences and are affected by multiple factors such as age, weight, liver and kidney function, and genetic polymorphisms. However, current clinical dosing regimens are mostly based on population-average pharmacokinetic parameters and lack precise adjustments for individual patient characteristics, which can easily lead to large fluctuations in blood drug concentrations, unstable efficacy, or increased toxic and side effects. In the existing technology, although traditional pharmacokinetic models (such as non-compartmental models or simple linear models) can preliminarily describe the laws of drug metabolism, it is difficult to accurately quantify inter- and intra-individual variations, which leads to insufficient model prediction accuracy. In addition, existing covariate screening methods often rely on a single statistical standard, which may miss key influencing factors or introduce redundant variables, affecting the robustness of the model.

[0005] To address the above issues, there is an urgent need for a precise drug delivery model that can integrate patients' clinical information, genetic characteristics, and dynamic physiological indicators to improve the level of personalized medication of mycophenolate mofetil capsules in organ transplant patients. Summary of the Invention

[0006] The purpose of the present invention is to provide a precise drug delivery model for the perioperative period of organ transplantation.

[0007] To achieve the above objectives, the present invention provides a method for constructing a perioperative precision drug delivery model for organ transplantation, comprising the following steps: S1: Collect clinical information and blood drug concentration data of patients undergoing liver or kidney transplantation, including covariates; S2: The patient's blood drug concentration data were analyzed using a nonlinear mixed-effects model program, using a first-order absorption and elimination and a two-compartment model with a lag time as the basic model. The basic model included a random-effects model, which included both inter-individual and intra-individual variations. Formula (1) is used to express the inter-individual variation: (1) in, is the pharmacokinetic parameter value of the i-th individual, is the typical value of the group, is the random effect value of the i-th individual, The mean is 0 and the variance is Normal distribution; Formula (2), formula (3), and formula (4) are used to express intra-individual variation: (2) (3) (4) in, is the observed value, is the predicted value, and is the residual variation, the residual variation obeys the mean of 0, and the error is Normal distribution; S3: Introduce the covariates in step S1 into the basic model, examine the effects of different covariates on pharmacokinetic parameters in the basic model, and construct the final model.

[0008] The construction method provided by the present invention is based on population pharmacokinetics (PPK). Through blood drug concentration analysis, a nonlinear mixed-effects model and a two-compartment model with lag time are used to establish a precision dosing model for mycophenolate mofetil capsules suitable for the perioperative period of liver or kidney transplantation. The final model constructed by this construction method is a precision dosing model for organ transplantation perioperatively. The final model can accurately predict MPA exposure after taking mycophenolate mofetil for perioperative liver or kidney transplant recipients, facilitating subsequent drug dosage adjustments and establishing a personalized mycophenolate mofetil dosing regimen. This can effectively reduce the incidence of adverse drug reactions and acute rejection reactions. This model has important guiding significance for the precise use of mycophenolate mofetil in liver or kidney transplant patients, and also has good economic and social benefits.

[0009] Preferably, in step S1, the clinical information includes sex (SEX), age (AGE), height (HT), weight (WT), body mass index (BMI), albumin level (ALB), total bilirubin level (TBIL), aspartate aminotransferase level (AST), alanine aminotransferase level (ALT), alkaline phosphatase level (ALP), gamma-glutamyl transferase level (GGT), serum creatinine level (Cre), creatinine clearance rate (CLCR), uric acid level (UA), hemoglobin level (HGB), hematocrit (HCT), combined proton pump inhibitor status (PPI), delayed graft function (DGF), rejection reaction status (REJ), UGT1A9 C311T gene locus, UGT2B7 G211T gene locus, UGT2B7 C802T gene locus, SLCO1B1 388A>G gene locus, SLCO1B1 521T>C gene locus, SLCO1B1 334T>G gene locus, ABCC2 1249G>A gene locus and ABCC2 24C>T gene locus.

[0010] Further preferably, the calculation formula of body mass index (BMI) is shown in formula (5): (5) Delayed graft function (DGF) is a common early complication after organ transplantation, defined as the need for dialysis within 1 week after surgery or failure of serum creatinine to fall below 400 μmol / L 1 week after surgery.

[0011] Further preferably, the calculation formula of male creatinine clearance rate (CLCR) is shown in formula (6): (6) Further preferably, the calculation formula of female creatinine clearance rate (CLCR) is shown in formula (7): (7) Preferably, in step S2, the inter-individual variation is estimated using an exponential model, and the intra-individual variation is estimated using a proportional model.

[0012] Preferably, in step S2, the basic model is constructed by comparing the objective function value (OFV), Akaike information criterion (AIC), and Bayesian information criterion (BIC) values.

[0013] Preferably, in step S3, the objective function value of the model is evaluated using the forward inclusion method and the backward elimination method to examine the effects of different covariates on the pharmacokinetic parameters.

[0014] Preferably, the method for examining the covariates is as follows: Forward inclusion: All covariates were introduced into the basic model separately to obtain the objective function value corresponding to each covariate. The objective function values ​​corresponding to all covariates were evaluated separately. When the decrease in the objective function value compared with the basic model was ΔOFV ≥ 3.84 and P < 0.05, the corresponding covariate was retained to obtain the full model; Backward elimination: Based on the full model, the covariates included in the full model were eliminated one by one. After eliminating a single covariate, the objective function value of the corresponding model after the full model was eliminated was obtained. When the increase in the objective function value compared with the full model was ΔOFV ≥ 7.78 and P < 0.005, the corresponding covariate was retained, and the final model was obtained using the last retained covariate.

[0015] Preferably, the covariates include continuous covariates and categorical covariates, wherein the categorical covariates are directly introduced into the basic model; the continuous covariates need to undergo correlation analysis before being introduced into the basic model. If there is a correlation between any two continuous covariates, any one of the continuous covariates is selected to be introduced into the basic model.

[0016] As a preference, the Pearson test is used to measure the correlation between any two continuous covariates. , there is correlation between continuous covariates.

[0017] Preferably, continuous covariates include age, height, weight, body mass index, albumin level, total bilirubin, alanine aminotransferase, alkaline phosphatase, γ-glutamyl transferase, creatinine clearance, uric acid, and hemoglobin; categorical covariates include the status of combined proton pump inhibitors, the occurrence of delayed graft function recovery, and the rejection reaction status.

[0018] Preferably, if the patient is a renal transplant recipient, the continuous covariate also includes aspartate aminotransferase.

[0019] Preferably, if the patient is a liver transplant patient, the continuous covariate also includes serum creatinine, and the categorical covariate also includes UGT1A9 C311T gene locus, UGT2B7 G211T gene locus, UGT2B7 C802T gene locus, SLCO1B1 388A>G gene locus, SLCO1B1 521T>C gene locus, SLCO1B1 334T>G gene locus, ABCC2 1249G>A gene locus, and ABCC2 24C>T gene locus.

[0020] Preferably, the drug applicable to the final model is mycophenolate mofetil capsules.

[0021] Preferably, the construction method further comprises S4: evaluating the performance of the final model by using goodness of fit plots, prediction-correction visual prediction tests, bootstrapping, normalized prediction distribution errors, precision and accuracy.

[0022] Preferably, the goodness of fit graph includes observed values ​​and individual predicted values, observed values ​​and group predicted values, conditional weighted residuals and group predicted values, and conditional weighted residuals and time after administration.

[0023] The present invention comprehensively evaluates the accuracy, precision and stability of the model through multi-dimensional verification methods such as goodness-of-fit plot, prediction-correction visual prediction test, bootstrap method, normalized prediction distribution error, precision and accuracy, ensures the universality of the model in different populations, and provides a basis for clinical promotion.

[0024] Furthermore, a second aspect of the present invention provides a precise drug administration prediction device, the prediction device comprising: Data acquisition module: used to obtain clinical information, blood drug concentration data, and medication data of organ transplant recipients and organ transplant donors, wherein the clinical information includes covariates; Basic model building module: used to generate basic models containing random effects of inter-individual variation and intra-individual variation in nonlinear mixed effect model procedures; A precision drug delivery model construction module is used to introduce the covariates in the data acquisition module into a basic model that includes inter-individual variation and intra-individual variation, and generate a precision drug delivery model for organ transplantation perioperative period through forward inclusion and backward elimination methods; Validation module: used to internally validate the organ transplant perioperative precision drug delivery model through goodness-of-fit plots, prediction-correction visualization prediction tests, bootstrapping, and normalized prediction distribution errors; Dose prediction module: used to predict the dosage and dosing regimen based on the precise medication model for perioperative organ transplantation; Data storage module: used to store patients' covariates, medication data, blood drug concentration data, model parameters, verification results, and the dosage and dosing regimen generated by the dosage prediction module.

[0025] Furthermore, a third aspect of the present invention provides an electronic device, comprising: at least one processor, and At least one computer-readable storage medium is communicatively connected to the processor, wherein the computer-readable storage medium stores program instructions that can be executed by the processor, and the program instructions can execute the method for constructing a perioperative precision drug delivery model for organ transplantation as described in the first aspect.

[0026] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention uses a nonlinear mixed-effects model program, combined with a first-order absorption and elimination and a two-compartment model with lag time, to simultaneously quantify inter-individual and intra-individual variability. This overcomes the deficiency of traditional multivariate linear models in accurately describing complex pharmacokinetic characteristics, significantly improves the accuracy of blood drug concentration prediction, and helps reduce the blindness of dosage adjustment. 2. This study integrates clinical information, genetic polymorphisms, and dynamic physiological parameters to construct a multidimensional covariate system. This dual screening strategy of forward inclusion and backward elimination effectively identifies key influencing factors and eliminates redundant variables, enhancing the model's stability and clinical applicability. This approach provides a more precise dosing regimen for mycophenolate mofetil capsules for patients undergoing liver or kidney transplantation. 3. This paper adopts the statistical standard of the ΔOFV threshold and combines it with the Bayesian feedback method to ensure the scientific and efficient screening of covariates, avoid the model overfitting or underfitting problems caused by traditional single methods, shorten the modeling cycle and improve the reliability of the model; 4. The perioperative period of organ transplantation is a critical period for graft function and recipient postoperative recovery. Paying attention to perioperative management is of great clinical significance for improving the success rate of transplantation. The model constructed by the construction method provided by the present invention can reduce the trial-and-error cost of dose adjustment, shorten the perioperative and postoperative recovery period of organ transplantation, reduce the cost of complication treatment, and effectively reduce the economic burden on patients, which has important social value and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a graph showing the correlation analysis results of the demographic data in Example 1; Figure 2 This is a graph showing the correlation analysis results of the liver function indicators in Example 1; Figure 3 Graph showing the correlation analysis results of the renal function indicators in Example 1; Figure 4 This is a graph showing the correlation analysis results of the blood routine indicators in Example 1; Figure 5 GOF diagram of the final model in Example 1; Figure 6 This is the Pc-VPC result diagram of the final model in Example 1; Figure 7 This is the NPDE result diagram of the final model in Example 1; Figure 8 This is a diagram showing the simulation results of the dosing regimen and dosage in Example 1; Figure 9 This is a graph showing the correlation analysis results of the demographic data in Example 2; Figure 10 This is a graph showing the correlation analysis results of the liver function indicators in Example 2; Figure 11 Graph showing the correlation analysis results of the renal function indicators in Example 2; Figure 12 This is a graph showing the correlation analysis results of the blood routine indicators in Example 2; Figure 13 GOF diagram of the final model in Example 2; Figure 14 This is the Pc-VPC result diagram of the final model in Example 2; Figure 15 This is the NPDE result diagram of the final model in Example 2. DETAILED DESCRIPTION

[0028] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, specific embodiments of the present invention are described in detail below. It should be noted that the following embodiments are only intended to illustrate the implementation methods and typical parameters of the present invention, and are not intended to limit the parameter ranges described in the present invention. Reasonable variations derived therefrom are still within the scope of protection of the claims of the present invention.

[0029] It should be noted that the endpoints of the ranges and any values ​​disclosed herein are not limited to the precise ranges or values, and these ranges or values ​​should be understood to include values ​​close to these ranges or values. For numerical ranges, the endpoints of each range, the endpoints of each range and individual point values, and the individual point values ​​can be combined with each other to obtain one or more new numerical ranges, and these numerical ranges should be considered to be specifically disclosed herein.

[0030] Unless otherwise defined, all terms, symbols, and other scientific terms used herein are intended to have the same meanings as those commonly understood by those skilled in the art to which the present invention pertains. In some cases, terms with commonly understood meanings are defined herein for purposes of illustration or ease of reference, and such definitions herein should not be construed as representing a significant difference from the conventional understanding in the art. The technical methods described or referenced herein are generally well understood by those skilled in the art and are employed by conventional methods.

[0031] As described in the background art, MPA drugs have an extremely narrow therapeutic window, and traditional pharmacokinetic models suffer from insufficient accuracy and low stability. Therefore, a specific embodiment of the present invention provides a method for constructing a precision drug delivery model for organ transplantation perioperatively based on the guidance of the PPK model. The model construction method comprises the following steps: S1: Collect clinical information and blood drug concentration data of patients undergoing liver or kidney transplantation, including covariates; S2: The patient's blood drug concentration data were analyzed using a nonlinear mixed-effects model program, using a first-order absorption and elimination and a two-compartment model with a lag time as the basic model. The basic model included a random-effects model, which included both inter-individual and intra-individual variations. Formula (1) is used to express the inter-individual variation: (1) in, is the pharmacokinetic parameter value of the i-th individual, is the typical value of the group, is the random effect value of the i-th individual, The mean is 0 and the variance is Normal distribution; Formula (2), formula (3), and formula (4) are used to express intra-individual variation: (2) (3) (4) in, is the observed value, is the predicted value, and is the residual variation, the residual variation obeys the mean of 0, and the error is Normal distribution; S3: Introduce the covariates in step S1 into the basic model, examine the effects of different covariates on pharmacokinetic parameters in the basic model, and construct the final model.

[0032] By using the model constructed using the above-mentioned embodiment, the blood concentration of the metabolite MPA of mycophenolate mofetil capsules can be combined with the patient's individual physiological and chemical indicators through a small amount of blood sampling, and the individualized pharmacokinetic parameters of patients undergoing solid organ transplantation, such as liver transplantation and kidney transplantation, can be calculated, and personalized medication can be implemented, thereby reducing the adverse reaction rate of mycophenolate mofetil capsules and the incidence of postoperative rejection reactions in organ transplantation.

[0033] In step S1 of the above embodiment, the clinical information includes sex (SEX), age (AGE), height (HT), weight (WT), body mass index (BMI), albumin level (ALB), total bilirubin (TBIL), aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT), serum creatinine (Cre), creatinine clearance rate (CLCR), uric acid (UA), hemoglobin (HGB), hematocrit (HCT), combined proton pump inhibitor status (PPI), delayed graft function (DGF), rejection reaction status (REJ), UGT1A9 C311T gene locus, UGT2B7G211T gene locus, UGT2B7 C802T gene locus, SLCO1B1 388A>G gene locus, SLCO1B1 521T>C gene locus, SLCO1B1 334T>G gene locus, ABCC2 1249G>A gene locus and ABCC2 24C>T gene locus.

[0034] The calculation formula of body mass index (BMI) is shown in formula (5): (5) Delayed graft function (DGF) is the most common early complication after kidney transplantation. It can cause post-transplant oliguria, increase the risk of graft immunogenicity and acute rejection, and is an independent risk factor for long-term graft survival. The incidence of DGF in deceased donor kidney transplantation is approximately 24.3%. For kidney transplantation, DGF is defined as the need for dialysis within one week after surgery, or a daily decrease in serum creatinine of less than 10% of the previous day's level for three consecutive days within the first week after surgery within the same hospital, or failure to reduce serum creatinine to below 400 μmol / L within one week after surgery.

[0035] Specifically, the calculation formula for male creatinine clearance rate (CLCR) is shown in formula (6): (6) Specifically, the calculation formula for female creatinine clearance rate (CLCR) is shown in formula (7): (7) The model provided in the above embodiment is a population pharmacokinetic model constructed using a nonlinear mixed-effect model (NONMEM Version 7.4.3). For mycophenolate mofetil capsules, one- and two-compartment structural models with first-order absorption and first-order elimination, as well as models with and without lag time, were attempted. The most suitable basic model was selected by comparing the objective function value (OFV), Akaike information criterion (AIC), and Bayesian information criterion (BIC) values. The model calculations were performed using first-order conditional estimation (FOCE). An exponential model was used to estimate inter-individual variation, and residual variation was fitted using additive, proportional, or mixed models to establish the basic model.

[0036] In the above embodiment, in the process of constructing the basic model, random effects need to be introduced to explain unknown or unobservable variations and model errors, specifically including inter-individual variation and intra-individual variation.

[0037] In the above embodiment, in order to investigate the effects of various factors on the pharmacokinetic parameters of mycophenolate mofetil capsules and thus more accurately describe the behavioral differences of mycophenolate mofetil capsules in different individuals, a specific embodiment of the present invention introduces covariates.

[0038] In a specific embodiment of the present invention, for mycophenolate mofetil capsules, a stepwise method, ie, forward inclusion method and backward elimination method, is used to evaluate the objective function value of the model to investigate the effects of different covariates on pharmacokinetic parameters.

[0039] More specifically, the covariate analysis method is as follows: Forward inclusion: All covariates were introduced into the basic model separately to obtain the objective function value corresponding to each covariate. The objective function values ​​corresponding to all covariates were evaluated separately. When the decrease in the objective function value compared with the basic model was ΔOFV ≥ 3.84 and P < 0.05, the corresponding covariate was retained to obtain the full model; Backward elimination: Based on the full model, the covariates included in the full model were eliminated one by one. After eliminating a single covariate, the objective function value of the corresponding model after the full model was eliminated was obtained. When the increase in the objective function value compared with the full model was ΔOFV ≥ 7.78 and P < 0.005, the corresponding covariate was retained, and the final model was obtained using the last retained covariate.

[0040] More specifically, covariates include continuous covariates and categorical covariates. For continuous covariates, the correlation between continuous covariates needs to be checked before screening. The specific embodiment of the present invention uses the Pearson test to evaluate the correlation between continuous covariates. The correlation coefficient is That is, it is believed that there is correlation between covariates; for categorical covariates, the basic model is directly introduced for investigation.

[0041] Among them, continuous covariates included age, height, weight, body mass index, albumin level, total bilirubin, alanine aminotransferase, alkaline phosphatase, γ-glutamyl transferase, creatinine clearance, uric acid, and hemoglobin. Categorical covariates included the status of combined proton pump inhibitors, delayed graft function recovery, and rejection reaction status.

[0042] More specifically, if the patient is a kidney transplant recipient, the continuous covariate also includes aspartate aminotransferase.

[0043] More specifically, if the patient is a liver transplant recipient, the continuous covariate also includes serum creatinine, and the categorical covariate also includes the UGT1A9 C311T gene locus, UGT2B7 G211T gene locus, UGT2B7 C802T gene locus, SLCO1B1 388A>G gene locus, SLCO1B1 521T>C gene locus, SLCO1B1 334T>G gene locus, ABCC2 1249G>A gene locus, and ABCC2 24C>T gene locus.

[0044] More specifically, according to the precise drug delivery model provided by the present invention, for renal transplant patients, the calculation model of the clearance rate (CL) and bioavailability ratio of mycophenolate mofetil capsules is shown in formula (8): (8) In formula (8), UA is the patient's uric acid level, and its unit is μmol / L.

[0045] More specifically, according to the precise drug delivery model provided by the present invention, for renal transplant patients with DGF, the calculation model of the peripheral compartment apparent distribution volume V2 / F (patients with DGF) of mycophenolate mofetil capsules is shown in formula (9): (9) More specifically, according to the precise drug delivery model provided by the present invention, for renal transplant patients who do not develop DGF, the calculation model of the peripheral compartment apparent distribution volume V2 / F (patients who do not develop DGF) of mycophenolate mofetil capsules is shown in formula (10): (10) More specifically, according to the precise drug delivery model provided by the present invention, for liver transplant patients, if the donor's ABCC2 24C>T genotype is CC, the calculation model of the recipient's peripheral compartment apparent distribution volume V2 / F(CC) of mycophenolate mofetil capsules is as shown in formula (11): (11) More specifically, according to the precise drug delivery model provided by the present invention, for liver transplant patients, if the donor's ABCC2 24C>T genotype is CT, the calculation model of the recipient's peripheral compartment apparent distribution volume V2 / F(CT) of mycophenolate mofetil capsules is as shown in formula (12): (12) More specifically, according to the precise drug delivery model provided by the present invention, for liver transplant patients, if the donor's ABCC2 24C>T genotype is TT, the calculation model of the recipient's peripheral compartment apparent distribution volume V2 / F(TT) of mycophenolate mofetil capsules is as shown in formula (13): (13) More specifically, in formulas (8), (9), (10), (11), (12), and (13), η1 is a random effect parameter related to clearance (CL / F), and η2 is a random effect parameter related to the central volume of distribution (V2 / F).

[0046] In a specific embodiment of the present invention, the method for constructing a precision drug delivery model for perioperative organ transplantation should further include a model evaluation step, which includes internal validation and external validation.

[0047] In another embodiment of the present invention, a precision medication prediction device based on the aforementioned method for constructing a precision medication model for the perioperative period of organ transplantation is provided, comprising: Data acquisition module: used to obtain clinical information, blood drug concentration data, and medication data of organ transplant recipients and organ transplant donors, wherein the clinical information includes covariates; Basic model building module: used to generate a basic model containing random effects of inter-individual variation and intra-individual variation in the nonlinear mixed effect model program. The basic model is a first-order absorption and elimination and a two-compartment model with lag time; A precision drug delivery model construction module is used to introduce the covariates in the data acquisition module into a basic model that includes inter-individual variation and intra-individual variation, and generate a precision drug delivery model for organ transplantation perioperative period through forward inclusion and backward elimination methods; Validation module: used to internally validate the organ transplant perioperative precision drug delivery model through goodness-of-fit plots, prediction-correction visualization prediction tests, bootstrapping, and normalized prediction distribution errors; Dose prediction module: used to predict the dosage and dosing regimen based on the precise medication model for perioperative organ transplantation; Data storage module: used to store patients' covariates, medication data, blood drug concentration data, model parameters, verification results, and the dosage and dosing regimen generated by the dosage prediction module.

[0048] In another embodiment of the present invention, an electronic device is provided, comprising: at least one processor, and At least one computer-readable storage medium communicatively connected to the processor, wherein the computer-readable storage medium stores program instructions executable by the processor, and the program instructions are capable of executing the method for constructing a perioperative precision drug delivery model for organ transplantation as described in the aforementioned embodiment.

[0049] The technical solution of the present invention is further described below through specific embodiments.

[0050] Example 1 Construction of a precision drug delivery model for perioperative renal transplantation This example involved 120 Chinese renal transplant patients who received mycophenolate mofetil capsules at the Second Affiliated Hospital of Nanchang University between 2022 and 2024. All patients received at least 5 days of a triple immunotherapy regimen based on tacrolimus and mycophenolate mofetil capsules after surgery. The initial treatment doses of mycophenolate mofetil capsules were 0.5g, 0.75g, 1g, and 1.25g, respectively, administered orally every 12 hours. Two (2) mL of whole blood was collected from each patient at 0 hour before dosing and 0.5, 1, 1.5, 2, 4, 6, 8, 10, and 12 hours after dosing to determine MPA plasma concentration, blood count, liver function, and renal function. Patients' sex (SEX), age (AGE), height (HT), weight (WT), body mass index (BMI), concomitant use of proton pump inhibitors (PPIs), delayed recovery of renal function (DGF), and rejection (REJ) were obtained from the electronic medical record.

[0051] Model establishment: Data analysis was performed using the nonlinear mixed effect model program (NONMEM Version 7.4.3).

[0052] The one-compartment and two-compartment structural models of primary absorption and elimination, as well as the structural models with and without lag time, were compared. The residual models of additive, proportional, and mixed types were compared. The comparison results are shown in Table 1.

[0053] Table 1

[0054] As shown in Table 1, the two-compartment model with lag time was the best. Finally, the two-compartment model with primary absorption and elimination and lag time was selected as the final basic model. The residual variation was the exponential model, and the inter-individual variation was the proportional model.

[0055] Correlation Analysis of Continuous Covariates: To avoid collinearity and parameter estimation instability in the model, correlations between covariates were assessed before covariate selection. This example uses the Pearson test to measure correlations between variables; a correlation coefficient of |r| ≥ 0.7 is considered significant. If the test results indicate a correlation between variables, only one variable is retained when constructing the covariate model.

[0056] Based on the data source, the covariates are divided into four categories: Demographic data: age (AGE), height (HT), weight (WT), body mass index (BMI); Liver function indicators: albumin (ALB), total bilirubin (TBIL), aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT); Renal function indicators: serum creatinine (CRE), creatinine clearance rate (CLCR), uric acid (UA); Routine blood test indicators: hemoglobin (HGB), hematocrit (HCT).

[0057] Among them, the correlation analysis of demographic data is as follows: Figure 1 As shown by Figure 1 It can be seen that there is a correlation between BMI and WT (r=0.84), so AGE, HT, and WT were included in the covariate screening.

[0058] Among them, the correlation analysis of liver function indicators is as follows: Figure 2 As shown by Figure 2 It can be seen that there is a correlation between AST and ALT (r=0.72), so ALB, TBIL, AST, ALP, and GGT were included in the covariate screening.

[0059] Among them, the correlation analysis of renal function indicators is as follows: Figure 3 As shown by Figure 3 It can be seen that there is a correlation between CRE and CLCR (r=-0.72), so CLCR and UA were included in the covariate screening.

[0060] Among them, the correlation analysis of blood routine indicators is as follows Figure 4 As shown by Figure 4 It can be seen that there is a correlation between HGB and HCT (r=0.98), so HGB was included in the covariate screening.

[0061] Covariate screening: Based on the results of the correlation analysis of continuous covariates, AGE, HT, WT, ALB, TBIL, AST, ALP, GGT, CLCR, UA, and HGB were finally included in the continuous covariates for covariate investigation. For the categorical covariates combined with proton pump inhibitor status (PPI), delayed graft function (DGF), and rejection reaction (REJ), the forward inclusion and backward elimination methods were used to investigate the covariates. The covariate screening process and results are shown in Table 2.

[0062] Table 2

[0063] Internal verification: The GOF diagram of the final model is as follows Figure 5 As shown, Figure 5 a in the figure is the plot of observed value (DV) and individual predicted value (IPRED); Figure 5 b in the figure is the plot of observed value (DV) and population predicted value (PRED); Figure 5 c in the figure is the conditional weighted residual (CWRES) and group predicted value (PRED) plot; Figure 5 d in is the conditional weighted residual (CWRES) and time after administration (TIME) graph, Figure 5 The black solid line is the y=x reference line, the red solid line is the trend line, and the black dotted line is the position line of CWRES at y=±2. Figure 5 It can be seen that most of the scatter points of observed and predicted values ​​are distributed near the y=x reference line, and the trend line is close to the reference line, indicating that the model prediction values ​​are highly consistent with the actual observation values; the CWRES scatter points are symmetrical on both sides of y=0, mostly within the range of ±2, and the trend line has no obvious trend change, indicating that the distribution of model residuals is reasonable and there is no systematic deviation, indicating that the model prediction is good.

[0064] The Pc-VPC results of the final model are as follows Figure 6 As shown, Figure 6 The blue dots represent the actual observations, the blue solid lines represent the 5% and 95% quantiles of the observations from bottom to top, the red solid line represents the median of the observations, and the blue and red shaded areas represent the 95% confidence intervals of the corresponding percentiles of the simulated 1000 data. Figure 6 As can be seen, the majority of observed values ​​fall within the predicted 95% confidence interval, indicating good agreement between the model predictions and observed values. The model has a high degree of fit and effectively captures the changing patterns of drug concentration. Throughout the entire timeframe, the model's prediction interval covers the majority of observed values, demonstrating the model's strong predictive power and reliability.

[0065] The NPDE results of the final model are as follows Figure 7 As shown by Figure 7The NPDE histogram shows a generally normal distribution, with a mean of -0.0373 and a variance of 1.04. The Wilcoxon rank sum test showed a significant difference between the mean and 0 (P=0.11), while the Fisher variance test showed no significant difference between the variance and 1 (P=0.301). The scatter plots generally fall on a straight line on the QQ plot. These NPDE results indicate that the MPA model fits the individual data well.

[0066] Bootstrap analysis was performed 1000 times, and the results are shown in Table 3.

[0067] Table 3

[0068] As shown in Table 3, the parameter values ​​obtained by the bootstrap method were similar to those of the final model established, and all estimated values ​​fell within the 95% confidence interval of the calculated values. The constructed population pharmacokinetic model of MPA in renal transplant recipients had high reliability and stability.

[0069] Therefore, it can be seen that UA has a significant effect on CL / F and DGF has a significant effect on V2 / F. Therefore, the final model incorporates these two factors, UA and DGF. The final model includes: the CL / F model shown in formula (8), the V2 / F (patients with DGF) shown in formula (9), and the V2 / F (patients without DGF) shown in formula (10): (8) (9) (10) External validation: The final model was externally validated using an additional 29-patient dataset to calculate the final model prediction AUC. 0-12h The value was (32.11±15.49) mg·h / L, which was close to the observed value (34.09±17.54) mg·h / L. There was no statistical difference between the model prediction value and the observed value (P=0.738). The MPE and RMSE of the group prediction value were -3.33 and 1.01, respectively. The results were within the acceptable range, indicating that the model group prediction performance was acceptable.

[0070] Dosage regimen and dosage simulation: The results of the research model showed that the patient's uric acid level will affect the CL / F of MPA. As the uric acid level increases, the CL / F of MPA decreases. Since the AUC of MPA is closely related to CL / F, the uric acid level needs to be fully considered when formulating the dosing regimen. The final simulation scheme is as follows: According to the quartiles of uric acid, the patient's uric acid level is divided into four groups: A, B, C, and D. The specific grouping criteria are: Group A uric acid ≤205μmol / L, Group B uric acid level between 205μmol / L and 265μmol / L, Group C uric acid level between 265μmol / L and 380μmol / L, Group D uric acid>380μmol / L. For patients in different groups, the dosing regimens of taking MMF 0.5g, 0.75g, 1g, 1.25g, 1.5g, and 2g are simulated respectively. The simulation results are shown in Figure 8 The two dotted lines in the figure represent the optimal therapeutic level of MPA 30mg·h / L≤AUC 0-12h ≤60mg·h / L. The results in the figure show that as the uric acid level increases, the exposure level AUC0-12h of MPA decreases, which is consistent with the results of the final model. For patients with UA≤205μmol / L, the recommended dose is 1g; for patients with 205<UA≤265μmol / L, the recommended dose is 1.25g; for patients with 265<UA≤380μmol / L and UA>380μmol / L, the recommended dose is 1.5g.

[0071] Example 2 Construction of a precision drug delivery model for perioperative liver transplantation This example involved 48 Chinese liver transplant patients who received mycophenolate mofetil capsules at the Second Affiliated Hospital of Nanchang University between 2022 and 2024. All patients received at least 5 days of a triple immunotherapy regimen based on tacrolimus and mycophenolate mofetil capsules after surgery. The initial treatment doses of mycophenolate mofetil capsules were 0.25g, 0.5g, 0.75g, and 1g, respectively, administered orally every 12 hours. Two (2) mL of whole blood was collected from each patient at 0 hour before dosing and 0.5, 1, 1.5, 2, 4, 6, 8, 10, and 12 hours after dosing to determine MPA blood concentration, blood count, liver function, and renal function. Patients' sex (SEX), age (AGE), height (HT), weight (WT), body mass index (BMI), genetic profiles, concomitant use of proton pump inhibitors (PPIs), and the occurrence of rejection (REJ) were obtained from the electronic medical record.

[0072] Model establishment: Data analysis was performed using the nonlinear mixed effect model program (NONMEM Version 7.4.3).

[0073] The one-compartment and two-compartment structural models of primary absorption and elimination, as well as the structural models with and without lag time, were compared. The residual models of additive, proportional, and mixed types were compared. The comparison results are shown in Table 4.

[0074] Table 4

[0075] As shown in Table 4, the two-compartment model with lag time was the best. Finally, the two-compartment model with primary absorption and elimination and lag time was selected as the final basic model. The residual variation was the exponential model, and the inter-individual variation was the proportional model.

[0076] Correlation Analysis of Continuous Covariates: To avoid collinearity and parameter estimation instability in the model, correlations between covariates were assessed before covariate selection. This example uses the Pearson test to measure correlations between variables; a correlation coefficient of |r| ≥ 0.7 is considered significant. If the test results indicate a correlation between variables, only one variable is retained when constructing the covariate model.

[0077] Based on the data source, the covariates are divided into five categories: Demographic data: age (AGE), height (HT), weight (WT), body mass index (BMI); Liver function indicators: albumin (ALB), total bilirubin (TBIL), aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT); Renal function indicators: serum creatinine (CRE), creatinine clearance rate (CLCR), uric acid (UA); Routine blood test indicators: hemoglobin (HGB), hematocrit (HCT).

[0078] Genetic indicators: (UGT1A9 C311T, UGT2B7 G211T, UGT2B7 C802T, SLCO1B1 388A>G, SLCO1B1 521T>C, SLCO1B1 334T>G, ABCC2 1249G>A, ABCC2 24C>T). All genetic indicators were included in the categorical covariates.

[0079] Among them, the correlation analysis of demographic data is as follows: Figure 9 As shown by Figure 9It can be seen that there is a correlation between BMI and HT (r = 0.90), so AGE, HT, and WT were included in the covariate screening.

[0080] Among them, the correlation analysis of liver function indicators is as follows: Figure 10 As shown by Figure 10 It can be seen that there is a correlation between AST and ALT (r=0.85), and a correlation between ALP and GGT (r=0.75), so ALB, TBIL, AST, and GGT were included in the covariate screening.

[0081] Among them, the correlation analysis of renal function indicators is as follows: Figure 11 As shown by Figure 11 It can be seen that the r values ​​between the three covariates are all <0.7, so all of them were included in the covariate model screening.

[0082] Among them, the correlation analysis of blood routine indicators is as follows Figure 12 As shown by Figure 12 It can be seen that there is a correlation between HGB and HCT (r=0.76), so HGB was included in the covariate screening.

[0083] Covariate screening: Based on the results of correlation analysis of continuous covariates, AGE, HT, WT, ALB, TBIL, AST, GGT, CLCR, UA, and HGB were finally included in the continuous covariates for covariate investigation. For the categorical covariates, proton pump inhibitor status (PPI), delayed graft function (DGF), rejection status (REJ), UGT1A9 C311T, UGT2B7 G211T, UGT2B7 C802T, SLCO1B1 388A>G, SLCO1B1 521T>C, SLCO1B1 334T>G, ABCC2 1249G>A, and ABCC2 24C>T were combined to investigate the covariates directly using the forward inclusion and backward elimination methods. The covariate screening process and screening results are shown in Table 5.

[0084] Table 5

[0085] Internal verification: The GOF diagram of the final model is as follows Figure 13 As shown, Figure 13 a in the figure is the plot of observed value (DV) and individual predicted value (IPRED); Figure 13 b in the figure is the plot of observed value (DV) and population predicted value (PRED); Figure 13 c in the figure is the conditional weighted residual (CWRES) and group predicted value (PRED) plot; Figure 13 d in is the conditional weighted residual (CWRES) and time after administration (TIME) graph, Figure 13 The black solid line is the y=x reference line, the red solid line is the trend line, and the black dotted line is the position line of CWRES at y=±2. Figure 13 It can be seen that most of the scatter points of observed and predicted values ​​are distributed near the y=x reference line, and the trend line is close to the reference line, indicating that the model prediction values ​​are highly consistent with the actual observation values; the CWRES scatter points are symmetrical on both sides of y=0, mostly within the range of ±2, and the trend line has no obvious trend change, indicating that the distribution of model residuals is reasonable and there is no systematic deviation, indicating that the model prediction is good.

[0086] The Pc-VPC results of the final model are as follows Figure 14 As shown, Figure 14 The blue dots represent the actual observations, the blue solid lines represent the 5% and 95% quantiles of the observations from bottom to top, the red solid line represents the median of the observations, and the blue and red shaded areas represent the 95% confidence intervals of the corresponding percentiles of the simulated 1000 data. Figure 14 As can be seen, the majority of observed values ​​fall within the predicted 95% confidence interval, indicating good agreement between the model predictions and observed values. The model has a high degree of fit and effectively captures the changing patterns of drug concentration. Throughout the entire timeframe, the model's prediction interval covers the majority of observed values, demonstrating the model's strong predictive power and reliability.

[0087] The NPDE results of the final model are as follows Figure 15 As shown by Figure 15 The NPDE histogram shows a generally normal distribution, with a mean of -0.0751 and a variance of 1.08. The Wilcoxon rank sum test showed a significant difference between the mean and 0 (P=0.0671), while the Fisher variance test showed no significant difference between the variance and 1 (P=0.24). The scattered points on the QQ plot generally fall on a straight line. These NPDE results indicate that the MPA model provides a good fit for the individual data.

[0088] Bootstrap analysis was performed 1000 times, and the results are shown in Table 6.

[0089] Table 6

[0090] As shown in Table 6, the parameter values ​​obtained by the bootstrap method were similar to those of the final model established, and all estimated values ​​fell within the 95% confidence interval of the calculated values. The constructed population pharmacokinetic model of MPA in liver transplant recipients had high reliability and stability.

[0091] Therefore, it can be seen that during the covariate screening process, no covariates that affect CL / F were found, and the donor gene ABCC2 24C>T was found to have a significant effect on V2 / F. Finally, the donor gene ABCC2 24C>T was included in the model. The final model formula is expressed as follows: V2 / F (CC) model with the donor ABCC2 24C>T genotype being CC, V2 / F (CT) model with the donor ABCC2 24C>T genotype being CT, and V2 / F (TT) model with the donor ABCC2 24C>T genotype being TT. Among them, the V2 / F (CC) model is shown in formula (11), the V2 / F (CT) model is shown in formula (12), and the V2 / F (TT) model is shown in formula (13): (11) (12) (13) As can be seen from Example 1, the present invention establishes a precise medication model for the perioperative period of renal transplantation, incorporates DGF into the model for the first time, and provides guidance for the personalized medication of DGF patients in the perioperative period of renal transplantation for the first time.

[0092] As shown in Example 2, the present invention establishes a precision medication model for perioperative liver transplantation, incorporating ABCC2 24C>T into the model, highlighting the necessity of preoperative genetic screening to guide personalized medication. This is the first time that a precision model based on population pharmacokinetics and Bayesian feedback has been established for perioperative liver transplant recipients, providing guidance for personalized medication for perioperative liver transplant patients.

[0093] The above embodiments demonstrate that specific embodiments of the present invention, based on population pharmacokinetics (PPK) combined with Bayesian feedback analysis, establish a personalized dosing model for mycophenolate mofetil (MMF) capsules in perioperative solid organ transplant patients. This precise perioperative dosing model for organ transplantation can help clinicians accurately predict MPA exposure after perioperative administration of MMF, facilitating medication dosing and dose adjustment. It also reduces the risk of toxic side effects caused by excessive MPA dosages or rejection caused by insufficient dosages, lowers the incidence of acute rejection reactions, and reduces adverse drug events. This model is of great significance for the precise application of MMF in organ transplant patients, with promising economic and social benefits.

[0094] Although the present disclosure is disclosed as above, the protection scope of the present disclosure is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present disclosure, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A method for constructing a precision drug delivery model for organ transplantation perioperative period, characterized in that: The following steps are involved: S1: Collect clinical information and blood drug concentration data of patients undergoing liver or kidney transplantation, including covariates; S2: The patient's blood drug concentration data were analyzed using a nonlinear mixed-effects model program, using a first-order absorption and elimination and a two-compartment model with a lag time as the basic model. The basic model included a random-effects model, which included both inter-individual and intra-individual variations. Formula (1) is used to express the inter-individual variation: （1） in, is the pharmacokinetic parameter value of the i-th individual, is the typical value of the group, is the random effect value of the i-th individual, The mean is 0 and the variance is Normal distribution; Formula (2), formula (3), and formula (4) are used to express intra-individual variation: （2） （3） （4） in, is the observed value, is the predicted value, and is the residual variation, the residual variation obeys the mean of 0, and the error is Normal distribution; S3: Introduce the covariates in step S1 into the basic model, examine the effects of different covariates on pharmacokinetic parameters in the basic model, and construct the final model. The final model is a precise drug delivery model for the perioperative period of organ transplantation, and the drug applicable to the precise drug delivery model for the perioperative period of organ transplantation is mycophenolate mofetil capsules.

2. The construction method according to claim 1, wherein In step S3, the forward inclusion method and the backward elimination method are used to evaluate the objective function value of the model to examine the effects of different covariates on the pharmacokinetic parameters.

3. The construction method according to claim 2, wherein: The specific method of examining covariates is as follows: Forward inclusion: All covariates were introduced into the basic model separately to obtain the objective function value corresponding to each covariate. The objective function values ​​corresponding to all covariates were evaluated separately. When the decrease in the objective function value compared with the basic model was ΔOFV ≥ 3.84 and P < 0.05, the corresponding covariate was retained to obtain the full model; Backward elimination: Based on the full model, the covariates included in the full model were eliminated one by one. After eliminating a single covariate, the objective function value of the corresponding model after the full model was eliminated was obtained. When the increase in the objective function value compared with the full model was ΔOFV ≥ 7.78 and P < 0.005, the corresponding covariate was retained, and the final model was obtained using the last retained covariate.

4. The construction method according to claim 1, wherein The covariates include continuous covariates and categorical covariates, wherein the categorical covariates are directly introduced into the basic model; the continuous covariates need to undergo correlation analysis before being introduced into the basic model. If there is a correlation between any two continuous covariates, any one of the continuous covariates is selected to be introduced into the basic model.

5. The construction method according to claim 4, wherein: The continuous covariates included age, height, weight, body mass index, albumin level, total bilirubin, alanine aminotransferase, alkaline phosphatase, γ-glutamyl transferase, creatinine clearance, uric acid, and hemoglobin. The categorical covariates included the status of combined proton pump inhibitors, delayed graft function recovery, and rejection reaction status.

6. The construction method according to claim 5, wherein: If the patient was a kidney transplant recipient, continuous covariates also included aspartate aminotransferase.

7. The construction method according to claim 5, wherein: If the patient is a liver transplant recipient, the continuous covariate also includes serum creatinine, and the categorical covariate also includes the UGT1A9 C311T gene locus, UGT2B7 G211T gene locus, UGT2B7 C802T gene locus, SLCO1B1 388A>G gene locus, SLCO1B1 521T>C gene locus, SLCO1B1334T>G gene locus, ABCC2 1249G>A gene locus, and ABCC2 24C>T gene locus.

8. The construction method according to claim 1, wherein: The construction method further includes S4: evaluating the performance of the final model by using a goodness of fit plot, a prediction-correction visual prediction test, a bootstrap method, and a normalized prediction distribution error.

9. A precise drug delivery prediction device, characterized in that: The precise drug administration prediction device comprises: Data acquisition module: used to obtain clinical information, blood drug concentration data, and medication data of organ transplant recipients and organ transplant donors, wherein the clinical information includes covariates; Basic model building module: used to generate basic models containing random effects of inter-individual variation and intra-individual variation in nonlinear mixed effect model procedures; A precision drug delivery model construction module is used to introduce the covariates in the data acquisition module into a basic model that includes inter-individual variation and intra-individual variation, and generate a precision drug delivery model for organ transplantation perioperative period through forward inclusion and backward elimination methods; Validation module: used to internally validate the organ transplant perioperative precision drug delivery model through goodness-of-fit plots, prediction-correction visualization prediction tests, bootstrapping, and normalized prediction distribution errors; Dose prediction module: used to predict the dosage and dosing regimen based on the precise medication model for perioperative organ transplantation; Data storage module: used to store patients' covariates, medication data, blood drug concentration data, model parameters, verification results, and the dosage and dosing regimen generated by the dosage prediction module.

10. An electronic device, characterized in that: The electronic device comprises: at least one processor, and At least one computer-readable storage medium communicatively connected to the processor, the computer-readable storage medium storing program instructions executable by the processor, the program instructions being capable of executing the method for constructing a perioperative precision drug delivery model for organ transplantation as described in any one of claims 1 to 8.

Images