A multi-gene SNP site mediated antipsychotic drug blood concentration prediction model and a construction method and application thereof

Abstract

This invention belongs to the fields of biomedicine and detection analysis technology, specifically relating to a predictive model for the blood concentration of antipsychotic drugs mediated by multiple gene SNP sites, its construction method, and its application. This invention screens genetic and non-genetic factors affecting the blood concentration of commonly used antipsychotic drugs and explores the influence of related factors on the in vivo metabolic process of antipsychotic drugs. Based on a nonlinear mixed-effects model, it integrates genetic and non-genetic factors to construct population pharmacokinetic models for antipsychotic drugs such as risperidone, quetiapine, clozapine, and olanzapine. Simultaneously, this invention also employs the Super Learner prediction algorithm to construct a predictive model for the steady-state trough concentration of antipsychotic drugs, providing a quantitative scientific reference for clinicians to determine the dosage for schizophrenia patients, thus possessing significant practical application value.

Description

Technical Field

[0001] This invention belongs to the fields of biomedicine and detection analysis technology, specifically relating to a prediction model of blood drug concentration of antipsychotic drugs mediated by multiple gene SNP sites, its construction method, and its application. Background Technology

[0002] The information disclosed in this background section is intended only to enhance understanding of the overall background of the invention and is not necessarily to be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art.

[0003] Generally, after antipsychotic drug efficacy prediction models are used to develop medication regimens, clinicians often administer a low initial dose to patients and then adjust the dose based on the observed efficacy and adverse reactions. If the dose remains ineffective even after increasing to an effective level, switching to a drug with a different mechanism of action or considering combination therapy may be considered. This method of adjusting the dose or medication regimen based on efficacy and adverse reactions during treatment may result in patients experiencing a prolonged period before achieving effective treatment, and may also lead to adverse reactions due to excessively high blood drug concentrations. This not only delays treatment but also undoubtedly increases the psychological and financial burden on patients and their families. Therefore, identifying factors affecting the pharmacokinetics of antipsychotic drugs and providing clinicians with quantitative, personalized dosing support tools at the beginning of patient admission is of significant clinical importance.

[0004] Numerous studies have demonstrated significant individual differences in the metabolism of antipsychotic drugs. For example, BIGO et al. found that the clearance rate of olanzapine can vary by 4 to 10 times among different individuals. The factors influencing antipsychotic drug blood concentrations are multifaceted. Non-genetic factors such as age, sex, weight, smoking, liver function, kidney function, and concomitant medications all significantly affect antipsychotic drug blood concentrations. The drug-metabolizing cytochrome P450 superfamily of enzymes are the main metabolic enzymes of antipsychotic drugs, and several studies have reported that gene polymorphisms in genes such as CYP2D6 and CYP1A2 are associated with antipsychotic drug blood concentrations. Furthermore, the membrane-associated transporter encoded by the ABCB1 gene is mainly expressed in the kidneys, liver, and blood-brain barrier, participating in drug absorption, distribution, and elimination. Multiple studies have shown that ABCB1 gene polymorphisms are associated with the blood concentrations of antipsychotic drugs such as olanzapine and clozapine. Currently, clinical practice primarily guides drug dosage based on the metabolic classification of cytochrome P450 metabolic enzymes, failing to provide quantitative estimates of blood drug concentrations. Furthermore, there is a lack of quantitative tools capable of utilizing genetic and non-genetic factors influencing blood drug concentrations for personalized medication guidance. Population pharmacokinetics (PPK) combines classical compartmental pharmacokinetic models with statistical principles to study the overall drug transport patterns within a patient population after administration of clinically appropriate doses of a particular drug, as well as the sources of individual differences in drug concentrations. It is one method that can provide relatively precise quantitative personalized dosing. Several studies have constructed population pharmacokinetics models for antipsychotic drugs such as olanzapine and risperidone; however, these models only incorporate limited non-genetic factors, and there is still no population pharmacokinetics model or blood drug concentration prediction model that comprehensively considers both genetic and non-genetic factors. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a population pharmacokinetic model based on the blood concentration of antipsychotic drugs, its construction method, and its application. This invention screens genetic and non-genetic factors affecting the blood concentration of commonly used antipsychotic drugs and explores the influence of these factors on the in vivo metabolism of antipsychotic drugs. Based on a nonlinear mixed-effects model, it integrates genetic and non-genetic factors to construct population pharmacokinetic models for antipsychotic drugs such as risperidone, quetiapine, clozapine, and olanzapine. Simultaneously, this invention employs the Super Learner prediction algorithm to construct a predictive model for the steady-state trough concentration of antipsychotic drugs, providing a quantitative scientific reference for clinicians to determine the dosage for patients with schizophrenia. Based on the above research results, this invention is thus completed.

[0006] This invention is achieved through the following technical solution:

[0007] The first aspect of this invention provides a method for constructing a population pharmacokinetic model of an antipsychotic drug. The method includes: collecting clinical information and blood samples from subjects to obtain relevant data; analyzing the data using a nonlinear mixed-effects model, specifically employing a one-compartment first-order absorption and elimination pharmacokinetic model as the basic structural model, obtaining simulated blood drug concentration data to acquire corresponding model pharmacokinetic parameters; using an additive random-effects model and introducing covariates, fitting the model, examining the influence of different factors on the pharmacokinetic parameters, and constructing a population pharmacokinetic model of the antipsychotic drug.

[0008] Antipsychotic drugs include, but are not limited to, risperidone, olanzapine, clozapine, and quetiapine.

[0009] The covariates include both genetic and non-genetic factors, which can be quantitatively screened and evaluated using linear models.

[0010] When the drug is risperidone, the covariates include, but are not limited to, weight, smoking, drinking habits, combined use of Lianhua Qingwen granules and silymarin meglumine tablets, decreased glomerular filtration function, and CYP2D6 enzyme activity score.

[0011] The final population typical values ​​for Ka, V, and CL are: Ka = 6.113h -1 V = 25.694 * exp(0.022 * body weight), CL = 3.139 * exp(0.155 * smoking - 0.1158 * drinking - 0.280 * Lianhua Qingwen granules + 0.473 * silymarin meglumine tablets - 0.141 * decreased glomerular filtration function + 0.066 * CYP2D6 enzyme activity score).

[0012] When the drug is olanzapine, the covariates include, but are not limited to, aspartate aminotransferase, decreased glomerular filtration rate, combination with sodium valproate, oxcarbazepine, and SNP sites rs7916649 and rs12768009.

[0013] The final population typical values ​​for Ka, V, and CL are as follows: Ka = 0.144h -1 V = 71.197 L, CL = 6.120 * exp(0.009 * AST - 0.130 * glomerular filtration rate decline + 0.253 * sodium valproate + 0.774 * oxcarbazepine + 0.292 * rs7916649 - 0.155 * rs12768009).

[0014] When the drug is clozapine, the covariates include, but are not limited to, age, sex, combination with propranolol, and SNP sites rs1135840 and rs1135822.

[0015] The population typical values ​​of Ka, V, and CL in the final model are as follows: Ka = 0.558h -1 V = 181.874 * exp(1.371 * male), CL = 42.472 * exp(0.239 * male - 0.009 * age - 0.137 * propranolol - 0.148 * rs1135840 - 0.474 * rs1135822).

[0016] When the drug is quetiapine, the covariates include, but are not limited to, age, sex, combination with atorvastatin calcium tablets, endogenous creatinine clearance rate, and rs2242480.

[0017] The final population typical values ​​for Ka, V, and CL are as follows: Ka = 0.163h -1 V = 274.626L, CL = 66.389*exp(-0.195*female-0.007*age+0.004*endogenous creatinine clearance-0.599*atorvastatin calcium tablets-0.119*rs2242480).

[0018] It is important to note that the key to establishing a population pharmacokinetic model is that the selected indicators can be accurately quantified and correctly reflect their relationship with disease treatment. Furthermore, only indicators with a clear dose-response relationship can establish a good population pharmacokinetic model. It is not acceptable to arbitrarily select any pathological or physiological condition related to the disease or any indicator related to drug efficacy as a covariate. This invention uses a large clinical sample as its data foundation. Through scientific analysis, it selects reasonable indicators from numerous sources to comprehensively quantify and evaluate the overall drug efficacy. A scientifically designed pharmacokinetic system capable of quantitatively evaluating drug participation has been developed, and the scientific accuracy of the indicator system has been verified through a population pharmacokinetic model, effectively reflecting the kinetic changes of the overall pharmacodynamic space over time.

[0019] A second aspect of the present invention provides the application of the population pharmacokinetic model of antipsychotic drugs obtained by the above construction method in any one or more of the following:

[0020] (a) To prepare products that accurately predict the dosage of antipsychotic drugs;

[0021] (b) To prepare products for personalized administration of antipsychotic drugs.

[0022] (c) Used for basic research related to the pharmacokinetics of antipsychotic drugs.

[0023] In (c), the pharmacokinetics of the antipsychotic drug includes the prediction of antipsychotic drug blood concentration.

[0024] A third aspect of the present invention provides a system for predicting and assessing blood concentrations of antipsychotic drugs, the system comprising at least:

[0025] The acquisition unit is configured to acquire subject-related data information;

[0026] The data processing unit is configured to predict the blood concentration of the antipsychotic drug in the subject based on the data information obtained by the acquisition unit and a built-in prediction and evaluation model; the prediction and evaluation model is obtained by training the model with a statistical algorithm using relevant data information of the patient collected in advance.

[0027] The output unit is configured to output the predicted blood concentration of the antipsychotic drug for the subject based on the information from the data processing unit.

[0028] The subject-related data includes, but is not limited to, the subject's general information, disease characteristics, antipsychotic drug use, and SNP sites related to the efficacy of antipsychotic drugs;

[0029] The antipsychotic drugs mentioned include, but are not limited to, risperidone, clozapine, olanzapine, and quetiapine.

[0030] When the antipsychotic drug is risperidone, the relevant data information used in the construction of the predictive evaluation model includes, but is not limited to, medication dosage, weight, BMI, age, endogenous creatinine clearance rate, rs17327442, CYP2D6 enzyme activity score, sodium valproate, propranolol, rs7787082, trihexyphenidyl, rs3789243, rs4244285, rs11528090, rs7779562, Di'ao Xinxuekang soft capsules, and rs7625. 51, rs3743484, rs2235047, smoking, rs1135840, rs28371699, rs1058164, gender, traracetam capsules, drinking alcohol, rs1081003, rs1065852, rs58440431, rs2004511, rs16947, rs75276289, rs1080996, silymarin meglumine tablets, rs116917064, enteric-coated aspirin tablets.

[0031] The prediction model is constructed using a super-powerful learning ensemble algorithm to predict the blood concentration of 9-hydroxyrisperidone.

[0032] The antipsychotic drug mentioned is olanzapine. During the construction of the predictive assessment model, the relevant data information includes, but is not limited to, medication dosage, frequency of use, age, alkaline phosphatase, weight, alanine aminotransferase (ALT), aspartate aminotransferase (AST), sodium valproate, CYP2D6 enzyme activity score, rs1065852, rs762551, rs3743484, aripiprazole, rs11528090, rs35280822, rs16947, rs79331140, rs1080996, and rs7787082. Propranolol, rs7779562, rs75276289, rs4646437, rs6583954, rs3758580, rs4244285, smoking, drinking, gender, rs12768009, rs7916649, rs2470890, rs17884832, rs5030865, rs17879992, rs17885098, rs28371725, decreased glomerular filtration function, phenolphthalein tablets, sulpiride, ziprasidone, and sanguisorba officinalis tablets.

[0033] The prediction model is constructed using a super-powerful learning ensemble algorithm to predict olanzapine blood concentration.

[0034] The antipsychotic drug mentioned is clozapine. During the construction of the predictive evaluation model, the relevant data information includes, but is not limited to, medication dosage, creatinine, age, aspartate aminotransferase (AST), creatinine clearance rate, weight, propranolol, risperidone, rs1080996, rs75276289, rs16947, rs11528090, rs762551, gender, medication frequency, rs12535512, tracerasetan capsules, rs1135840, rs1128503, rs12720464, rs3789243, CYP2D6 enzyme activity score, enteric-coated aspirin tablets, rs3842, rs6979885, atorvastatin calcium tablets, rs4728709, and smoking.

[0035] The prediction model is constructed using a super-powerful learning ensemble algorithm to predict clozapine blood concentration.

[0036] The antipsychotic drug mentioned is quetiapine. During the construction of the predictive evaluation model, the relevant data information includes, but is not limited to, medication dosage, age, BMI, alkaline phosphatase, aspartate aminotransferase, creatinine, alanine aminotransferase, weight, sodium valproate, CYP2D6 enzyme activity score, rs12535512, rs2246709, rs12768009, rs2242480, rs1081003, cefuroxime axetil, amisulpride, rs58440431, rs2004511, Dio Xin Xue Kang soft capsules, gender, enalapril maleate tablets, atorvastatin calcium tablets, rs2470890, smoking, abnormal liver function, rs1135822, oxcarbazepine, buspirone, aripiprazole, and Wenxin granules.

[0037] The prediction model is constructed using a super-powerful learning ensemble algorithm to predict quetiapine blood concentration.

[0038] In a fourth aspect, the present invention provides a computer-readable storage medium having a program stored thereon that, when executed by a processor, performs the functions of the system as described in the third aspect of the present invention.

[0039] In a fifth aspect, the present invention provides an electronic device including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to perform the functions of the system as described in the fourth aspect of the present invention.

[0040] The beneficial technical effects of one or more of the above technical solutions:

[0041] The above-mentioned technical solutions screened the genetic and non-genetic factors affecting the blood concentration of commonly used antipsychotic drugs and explored the influence of related factors on the in vivo metabolism of antipsychotic drugs. Based on a nonlinear mixed-effects model, genetic and non-genetic factors were integrated to construct population pharmacokinetic models for antipsychotic drugs such as risperidone, quetiapine, clozapine, and olanzapine. At the same time, this invention also uses a super-powerful learning ensemble algorithm to construct a predictive model for the steady-state trough concentration of antipsychotic drugs, providing a quantitative scientific reference for clinicians to determine the dosage for patients with schizophrenia. Therefore, it has good practical application value. Attached Figure Description

[0042] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0043] Figure 1 This is a schematic diagram of the sampling time points for risperidone blood concentration in an embodiment of the present invention.

[0044] Figure 2 This is a goodness-of-fit diagram of the final risperidone model in this embodiment of the invention.

[0045] Figure 3 This is a residual plot of the final risperidone model in this embodiment of the invention.

[0046] Figure 4 This is a schematic diagram of the final model population pharmacokinetic curve when the risperidone dosing regimen is 2 mg (bid) in an embodiment of the present invention (from top to bottom, these are the 2.5%, 50%, and 97.5% quantiles of the predicted blood drug concentration; the gray circles represent the actual observed values).

[0047] Figure 5 This is a comparison of RMSE for different numbers of candidate variables for risperidone in this embodiment of the invention.

[0048] Figure 6 This is the importance score of the predictor variable for risperidone based on random forest in an embodiment of the present invention.

[0049] Figure 7 This is a scatter plot of the predicted and actual observed values ​​of the 9-hydroxyrisperidone blood pressure concentration prediction model based on Super Learner in this embodiment of the invention.

[0050] Figure 8 This is a goodness-of-fit graph of the final olanzapine model in this embodiment of the invention.

[0051] Figure 9 This is the residual plot of the final model of olanzapine in this embodiment of the invention.

[0052] Figure 10 This is a schematic diagram of the final model population pharmacokinetic curve when the olanzapine dosing regimen is 5 mg (bid) in an embodiment of the present invention (from top to bottom, these are the 2.5%, 50%, and 97.5% quantiles of the predicted blood drug concentration; the gray circles represent the actual observed values).

[0053] Figure 11 This refers to the RMSE based on the random forest prediction model when olanzapine has different numbers of candidate variables in this embodiment of the invention.

[0054] Figure 12 This is the olanzapine predictor variable importance score based on random forest in an embodiment of the present invention.

[0055] Figure 13 This is a scatter plot of the predicted and actual observed values ​​of the olanzapine blood concentration prediction model based on Super Learner in this embodiment of the invention (A: training set 10-fold cross; B: validation set).

[0056] Figure 14 This is a goodness-of-fit graph of the final olanzapine model in this embodiment of the invention.

[0057] Figure 15This is a residual plot of the final model of olanzapine in this embodiment of the invention.

[0058] Figure 16 This is a schematic diagram of the final model population pharmacokinetic curve when the clozapine dosing regimen is 75 mg (bid) in an embodiment of the present invention (from top to bottom, these are the 2.5%, 50%, and 97.5% quantiles of the predicted blood drug concentration; the gray circles represent the actual observed values).

[0059] Figure 17 This is a comparison of RMSE for different numbers of candidate variables for clozapine in this embodiment of the invention.

[0060] Figure 18 This is the importance score of the predictor variable for clozapine based on random forest in an embodiment of the present invention.

[0061] Figure 19 This is a scatter plot of the predicted and actual observed values ​​of the clozapine blood concentration prediction model based on Super Learner in this embodiment of the invention (A: training set 10-fold cross; B: validation set).

[0062] Figure 20 This is a goodness-of-fit graph of the final quetiapine model in this embodiment of the invention.

[0063] Figure 21 This is a residual plot of the final model of quetiapine in this embodiment of the invention.

[0064] Figure 22 This is a schematic diagram of the final model population pharmacokinetic curve when the quetiapine dosing regimen is 300 mg (bid) in an embodiment of the present invention (from top to bottom, these are the 2.5%, 50%, and 97.5% quantiles of the predicted blood drug concentration; the gray circles represent the actual observed values).

[0065] Figure 23 This is a comparison of RMSE for different numbers of candidate variables for quetiapine in this embodiment of the invention.

[0066] Figure 24 This is the importance score of quetiapine based on random forest in an embodiment of the present invention.

[0067] Figure 25 This is a scatter plot of the predicted and actual observed values ​​of the quetiapine blood concentration prediction model based on Super Learner in this embodiment of the invention (A: training set 10-fold cross; B: validation set). Detailed Implementation

[0068] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0069] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. It should be understood that the scope of protection of the present invention is not limited to the specific embodiments described below; it should also be understood that the terminology used in the embodiments of the present invention is for the purpose of describing specific embodiments and not for limiting the scope of protection of the present invention. Experimental methods in the following specific embodiments, unless specific conditions are specified, generally follow conventional biological methods and conditions within the art, which are fully explained in the literature.

[0070] The following examples further illustrate the present invention, but do not constitute a limitation thereof. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, biomedical methods not described in detail in the examples are conventional methods in the art; for specific operations, please refer to biomedical guidelines or product instructions.

[0071] Example

[0072] I. Materials and Methods

[0073] 1.1 Research Subjects

[0074] This study was approved by the Ethics Committee of the School of Public Health, Shandong University. The research team enrolled schizophrenia patients hospitalized at Pingyi County Mental Hospital in Linyi City, Shandong Province from 2021 to 2022. Specific inclusion and exclusion criteria are as follows:

[0075] Inclusion criteria: (1) Clinical diagnostic criteria for schizophrenia according to the International Statistical Classification of Diseases and Related Health Problems (10th Revision) (ICD-10); (2) Age 18-65 years; (3) Regular use of antipsychotic drugs such as risperidone, quetiapine, olanzapine or clozapine as prescribed during hospitalization; (4) Acceptance of routine blood drug concentration monitoring;

[0076] Exclusion criteria: (1) suffering from chronic enteritis, diarrhea or other gastrointestinal diseases; (2) suffering from other serious diseases, such as severe cardiovascular and cerebrovascular diseases, cancer, etc.; (3) unable to obtain complete medical records; (4) extremely poor adherence and unable to take medication regularly for a long time; (5) needing long-term injections to maintain treatment adherence; (6) recently receiving clozapine treatment regularly due to treatment resistance (excluding patients who take clozapine treatment for other reasons).

[0077] 1.2 Data Collection

[0078] The basic information of schizophrenia patients collected from Pingyi County Mental Hospital in this study included: (1) general demographic characteristics: age, gender, height, weight, education level, marital status; (2) comorbidities: diabetes, hypertension, coronary heart disease and stroke, etc.; (3) lifestyle habits such as smoking and drinking; (4) characteristics of schizophrenia: date of first onset, course of disease (acute, subacute, slow onset), family history of mental illness, number of hospitalizations, number of hospital stays, etc.

[0079] This study obtained information on medication treatment during hospitalization of schizophrenia patients from the inpatient prescriptions section of the electronic medical record system of Shandong Provincial Mental Health Center. The medications included antipsychotics such as risperidone, olanzapine, quetiapine, clozapine, amisulpride, aripiprazole, and ziprasidone; mood stabilizers such as lithium carbonate, sodium valproate, and magnesium valproate; antidepressants such as citalopram, mirtazapine, and duloxetine; and anti-anxiety medications such as buspirone and lorazepam. The collected treatment information included: medication name, start and end dates of administration, frequency of administration, and dosage.

[0080] This study obtained laboratory test data of patients with mental disorders during hospitalization from the Laboratory Information System (LIS) of Pingyi County Mental Hospital, including liver function indicators: aspartate aminotransferase (AST), alanine aminotransferase (AST), and alkaline phosphatase (ALP); kidney function indicators: creatinine (Cr) and blood urea nitrogen / creatinine ratio (BUN / Cr); and blood routine indicators: white blood cell count, neutrophil count, hemoglobin count, and platelet count.

[0081] 1.3 Blood Sample Collection

[0082] 1.3.2 Blood Sample Collection for Drug Concentration

[0083] 1.3.2.1 Routine Blood Drug Concentration Monitoring Blood Sample Collection

[0084] Two milliliters of anticoagulated violet blood were collected from each patient, generally at 6:30 AM. Ideally, the target drug should have reached steady state by the time of blood collection, i.e., after 5-6 half-lives of continuous administration at a fixed dose. The half-lives and therapeutic reference concentration ranges for each drug are shown in Table 1.

[0085] Table 1. Reference range of half-life and therapeutic concentration for four antisperm drugs.

[0086]

[0087]

[0088] 1.3.2.2 Collection of blood samples for population pharmacokinetic analysis

[0089] This study uses risperidone as an example to illustrate the process of determining the sampling time for population pharmacokinetic blood samples. Referring to relevant literature, the population pharmacokinetic model of risperidone was set as a one-compartment model with first-order absorption. The initial values ​​of the population pharmacokinetic parameters were set as follows: Cl = 4.6 L / h, kJ / L... a =3.1h -1 V = 250 L. Following a dosing schedule of once at 8:00 AM and once at 8:00 PM, the recommended sampling times for obtaining blood drug concentrations using the optimized population pharmacokinetic design R-package PopED are 0.35 h, 2 h, and 12 h after administration, once the blood drug concentration reaches steady state. However, in clinical practice, schizophrenic patients at Pingyi Mental Hospital typically take their medication at 10:30 AM and 8:00 PM, making the aforementioned ideal sampling requirements difficult to achieve.

[0090] To align with clinical practice and enhance the feasibility of the research protocol, we combined blood sample collection with other relevant testing methods, establishing the following two sampling schemes: After the blood drug concentration reaches steady state (generally considered to be 6 days of continuous administration of the same dose for risperidone and 9-hydroxyrisperidone), samples are taken at 6:30 AM and 9:30 AM, i.e., 10.5 hours and 13.5 hours after administration at 8:00 PM the previous evening; a second sample is taken at 11:30 AM after administration at 10:30 AM. Considering that multiple samplings on the same day may reduce patient compliance, blood samples can be collected on different days, provided the daily dosage and administration time remain constant. For example, after the patient reaches steady-state blood drug concentration, the first blood sample is taken at 6:30 AM on the first day, the second at 9:30 AM on a subsequent day, and the third at 11:30 AM several days later. A schematic diagram of the sampling points is shown below. Figure 1 .

[0091] Table 2 Sampling Time Points for Blood Concentration of Antipsychotic Drugs

[0092]

[0093] 1.4 Blood drug concentration determination

[0094] High-performance liquid chromatography (HPLC) was used to determine the blood concentration of antipsychotic drugs. The "FLC fully automated two-dimensional liquid chromatography system" and its supporting high-performance liquid chromatograph, mobile phase, chromatographic column, protein removal agent, etc. were all provided by Hunan Demeter Instrument Co., Ltd.

[0095] 1.5 Selection of candidate SNP sites and SNP genotyping

[0096] Based on the pharmacogenetic and genomic pharmacology database (PharmGKB) listing of antipsychotic drug metabolism-related genes and SNP sites, and in conjunction with pharmacological research literature, metabolic enzyme genes such as CYP2D6, CPY3A4, CYP1A2, CYP3A5, and CYP2C19, and the ABCB1 transporter gene were identified as candidate genes, and SNP sites within these genes were selected as candidate SNP sites. All drug metabolism-related SNP sites reported in the PharmGKB or SNP database (SNPedia) were included; other candidate gene sites, if located in linkage disequilibrium (LD) regions (r... 2 If the mean value is greater than 0.8, the SNP with the highest minimum allele frequency (MAF) is included as a candidate marker SNP. Genotyping is performed using the Illumina ASA (Asian Screening Array) microarray.

[0097] 1.6 Data Processing and Analysis

[0098] 1.6.1 Data Processing

[0099] Since the original data came from multiple information systems, for ease of statistical analysis, we merged the basic information of the study subjects, inpatient medical order data, antipsychotic drug blood concentration monitoring data, and genotyping data using the hospital number as the unique identifier. The dosages of antipsychotic drugs such as risperidone, olanzapine, quetiapine, and clozapine, as well as other combined medications, were defined as the dosage required in the prescription at the time of blood concentration monitoring of the target antipsychotic drug. All drug dosage units were converted to mg / day, and the frequency of antipsychotic drug use was recorded. Other variable codes are shown in Table 3. Multiple imputation was used to fill missing variables with a missing rate of less than 5%.

[0100] Table 3 Variable Encoding Table

[0101]

[0102]

[0103] Currently, in clinical practice, clinicians often use the metabolic typing of the CYP2D6 gene to guide medication use. To facilitate clinical application, this study referred to the metabolic typing method based on the CYP2D6 gene proposed by CPIC and Professor Cui Yimin [a1a, metabolic typing method of the CYP2D6 gene related to psychotropic drugs in the Chinese population] and performed metabolic typing on the subjects in this study. This method, based on the genotyping results of SNP loci such as rs3892097, rs5030865, rs28371706, rs28371717, rs769258, rs28371725, rs1135822, rs72549348, rs267608297, rs267608289, rs1065852, rs16947, rs1135840, rs28371699, and rs28371701, marks the alleles at these SNP loci as: a) normally functional alleles CYP2D6*1, *2, *33, *35, *35, *39; b) defunctional alleles CYP2D6*9, *10, *14.

[0104] *17, *29, *41, *49, *54, *59; c) Non-functional alleles CYP2D6*3, *4, *5, *6, *8, *11, *51, *59; d) Alleles of unknown function CYP2D6*22, *28, *30, *65. After determining the number of mutations in the four functional alleles, the enzyme activity score (AS) is calculated using the following formula: AS = a + d + 0.5 * b (where *10 is a coefficient of 0.25).

[0105] 1.6.2 Data Description

[0106] This study provided a statistical description from two aspects: (1) a general description of the demographic characteristics, lifestyle habits, comorbidities, and prevalence of mental illnesses of the study subjects; and (2) the distribution of antipsychotic drug dosages and blood drug concentrations. If continuous variables conform to a normal distribution, they are expressed as mean ± standard deviation; if they do not conform to a normal distribution, continuous variables are expressed as median (interquartiles from P25 to P75); categorical variables are expressed as frequency and proportion.

[0107] 1.6.3 Analysis of Factors Affecting Blood Drug Concentration

[0108] Because the study subjects underwent multiple blood drug concentration monitoring during hospitalization, the data in this study are longitudinal data with repeated measurements. Therefore, this study used a generalized estimating equations (GEE) model to screen for non-genetic factors and SNP loci affecting antipsychotic drug blood concentration. Antipsychotic drug dosage, the interval between previous dosing, and candidate variables were simultaneously included in the GEE model to analyze the impact of candidate variables on antipsychotic drug blood concentration.

[0109] 1.6.4 Population Pharmacokinetic Analysis

[0110] To further analyze the influence of factors affecting antipsychotic drug blood concentration and their impact on antipsychotic drug metabolism, this study employed a nonlinear mixed-effects model for population pharmacokinetic analysis. The basic process for constructing the population pharmacokinetic model is as follows:

[0111] The nonlinear mixed-effects model can be expressed by the following general formula:

[0112]

[0113] Among them, Y ij X is the concentration observation value of the i-th individual in the j-th observation. ij The independent variables for a particular individual (such as observation time or dose) are represented by a pi*t-dimensional regression design matrix, which represents t independent intra-individual variables. It refers to all pharmacokinetic parameters of a particular individual, ε ij :P ij * 1. A normal random error vector. f is a Pi*1 dimensional X ij and The nonlinear function, whose functional form depends on the selected pharmacokinetic structure model;

[0114] (1) Determine the pharmacokinetic structural model:

[0115] Pharmacokinetic structural models describe the average effect of a pharmacokinetic population. Differences in administration routes and drug processes in vivo determine the diversity of pharmacokinetic models. All models can be described by a set of formulas and parameters.

[0116] Taking a single dose as an example, the pharmacokinetic structural model of first-order absorption and elimination in a one-compartment system can be represented in the following form:

[0117]

[0118] At time t j The formula for calculating the cumulative dose is as follows:

[0119] D[i,j]=D[i,j-1]exp(-ka t j -t j-1 ])+d[i,j]

[0120] At time t j The formula for calculating plasma concentration is as follows:

[0121]

[0122] d[i,j] represents the state of the i-th research object at time t. j The dosage of the drug taken; F represents the bioavailability of the drug in the population (the relative amount of drug absorbed into the systemic circulation after administration via an extravascular route). V[i] represents the apparent volume of distribution, K a and K a These are the first-order absorption and elimination rate constants, respectively.

[0123] Pharmacokinetic structural models can be flexibly selected based on the characteristics of the drug and the features of the data. In this study, based on relevant literature reports, the pharmacokinetic structural models of risperidone, olanzapine, quetiapine and clozapine were all set as one-compartment first-order absorption and elimination pharmacokinetic models.

[0124] (2) Fixed-effects model: The fixed-effects model is used to quantitatively examine the influence of fixed effects (such as age, sex, SNP sites, etc.) on pharmacokinetic parameters. The model structure includes linear, multiplicative, saturated, and indicator variable models. Currently, the most commonly used models are linear and multiplicative models.

[0125] 1) Linear model: expressed by the following formula:

[0126]

[0127] In the formula WT is the typical value for the population, and θ1 is the standard value for the population. When fixed effects are not considered, it is equal to the typical value for the population; WT is the patient's weight, and θ1 is the fixed-effects parameter for weight.

[0128] 2) Multiplicative model: If the weight range of the study subjects is very large, and the drug clearance rate is related to weight, the following formula can be used:

[0129]

[0130] Taking the logarithm of both sides of the formula yields a linear model on a logarithmic scale. In the formula... WT is the typical value for the population, and θ1 is the standard value for the population. WT is the patient's weight, and θ1 is the fixed-effects parameter of weight.

[0131] (3) Random Effects Model: The random effects model includes between-individual variation and residual variation. The former describes the differences in pharmacokinetic parameters among individuals in the population; while the latter describes the variation that cannot be explained by the observations. It is generally assumed that the pharmacokinetic parameters are normally or log-normally distributed, and the model with the smallest objective function value is selected according to the additive, proportional, or exponential models. The between-individual variation of pharmacokinetic parameters is represented by the random variable η, with a mean of 0 and a variance of ω. 2 This is expressed as η=N(0,ω) 2 The random variable ε represents the residual variation, expressed as η = N(0, ε). 2 Currently, the addition model or the proportional model are commonly used to represent this:

[0132] 1) Additive model: expressed by the following formula:

[0133]

[0134]

[0135] Among them CL j For a certain individual parameter, These are typical values ​​for the parameter group. These are concentration observation values. This represents the model-predicted concentration. ε ij These are the variations between individuals and random variations.

[0136] 2) Proportional Model: Expressed by the following formula:

[0137]

[0138]

[0139] (3) Covariate screening: This examines the effects of non-genetic factors such as age, sex, weight, and concomitant medications, as well as genetic factors, on pharmacokinetic parameters. Currently, covariate screening mainly employs hypothesis testing and forward incrementing or backward decrementing methods. The basic principles of hypothesis testing are as follows:

[0140] The reliability of a nonlinear mixed-effects model can be estimated using the maximum likelihood approach (ML). The maximum likelihood of a series of observations in the model can be expressed as follows:

[0141]

[0142] Maximum likelihood can be represented by the minimum value of -2log(L). θ represents the pharmacokinetic parameters of the established model, and y... i The observed blood drug concentration is f(θ, x). i) represents the simulated blood drug concentration based on the established pharmacokinetic model, where n is the number of observation points, and σ is the concentration of the drug. 2 This represents the variance of the residuals. The first part of the above equation is a constant, so the minimum value depends on the second part. The second part is also known as the objective function value (OFV) of the extended least squares (ELS) method.

[0143]

[0144] The maximum likelihood test is equivalent to the significance test of OFV changes, and conventional statistical significance testing methods are used. First, a simplified model without any fixed effects is established. Then, candidate covariates are added one by one to the simplified model to build an incremental model. The difference in parameters between the simplified and incremental models can be expressed as the likelihood ratio (L1 / L2). It has been proven that -2log(L1 / L2) follows a chi-square distribution. If the difference in OPV between the two models is greater than 3.84, then at a significance level of α = 0.05, the candidate covariate has a statistically significant effect on the parameters. Candidate fixed effects are examined sequentially to obtain the final model. Forward incremental regression involves adding meaningful fixed effects one by one to the simplified model, using ΔOFV as the standard, to obtain the full regression model (FRM). Then, backward decreasing regression is used to set each fixed effect to "0", and hypothesis testing (ΔOFV) is used to eliminate insignificant fixed effects, thus determining the final regression model.

[0145] The goodness of fit of the final population pharmacokinetic model was evaluated using RMSE and R², and assessed by plotting diagnostic graphs, including individual predicted values ​​versus observed values ​​(IPRED vs DV; the closer the trend line is to y = x, the stronger the model's individual predictive ability), population predicted values ​​versus observed values ​​(PRED vs DV; the closer the trend line is to y = x, the stronger the model's population predictive ability), weighted residuals versus population predicted values ​​(CWRES vs PRED; examining whether the conditional residuals bias the predicted values; the closer the weighted residuals are to 0, the stronger the model's population predictive ability), and weighted residuals versus time scatter plot (CWRES vs TIME; examining whether the conditional weighted residuals bias the blood sample collection time). Individual predicted values ​​refer to the fitting results calculated from the model's PK individual parameters. Data points should be randomly and uniformly distributed on both sides of a straight line with an intercept of 0 and a slope of 1. The closer the trend line is to the diagonal, the higher the precision of the fit. The weighted residuals against population predictions and the weighted residuals against time scatter plots can evaluate the overall goodness of curve fitting; that is, a weighted residual between -6 and 6 indicates a good model fit. Furthermore, this study uses the bootstrap method to validate the stability of the population pharmacokinetic model. By performing 500 bootstrap validations on the model, the distribution of population pharmacokinetic parameters and the model's stability are examined.

[0146] II. Test Results

[0147] 2.1 Study subjects and population pharmacokinetics of risperidone

[0148] 2.1.1 Basic Information of the Research Subjects

[0149] This study included 239 patients with schizophrenia. The general characteristics and laboratory test results of the subjects are shown in Table 4. The mean age of the subjects was 44.3 ± 12.0 years; 132 were male and 107 were female, accounting for 55.2% and 44.8% respectively; the mean height and weight were 163.6 ± 8.7 cm and 68.2 ± 13.7 kg respectively, and the mean BMI was 25.4 ± 4.3 kg / m²; among the subjects, 68 were smokers and 44 were drinkers, accounting for 28.5% and 18.4% respectively; the prevalence of diabetes and hypertension was 7.9% and 6.3% respectively. The mean values ​​of liver and kidney function test results were all within the normal reference range, but 49 subjects had decreased glomerular filtration rate and 40 subjects had abnormal liver function, accounting for 20.5% and 16.7% respectively.

[0150] This study only included schizophrenia patients taking risperidone twice daily, with a mean dose of 2.4 ± 0.6 mg per dose. Under steady-state blood drug concentration conditions, risperidone blood concentrations were collected in 1115 cases. The mean blood drug concentrations at 1.5 h, 5.0 h, 9.5 h, and 12.5 h after administration were 70.22 ± 28.54 ng / ml, 53.23 ± 14.68 ng / ml, 51.01 ± 24.33 ng / ml, and 44.89 ± 18.43 ng / ml, respectively. Based on the upper limit of the therapeutic reference range of 60 ng / ml for risperidone, 244 cases had blood drug concentrations exceeding the upper limit at 9.5 h and 11 cases had blood drug concentrations exceeding the upper limit at 12.5 h, representing 26.5% and 23.4% of the cases measured at those time points, respectively.

[0151] Table 4. General information and laboratory test indicators of the study subjects

[0152]

[0153] * Under steady-state conditions, the blood drug concentration at 1.5h, 5.0h, 9.5h, and 12.5h after the last administration; decreased glomerular filtration function: creatinine clearance <80ml / min; abnormal liver function: alanine aminotransferase >40U / L or aspartate aminotransferase >40U / L;

[0154] 2.1.2 Patient pharmacokinetic studies of risperidone

[0155] To determine the mean, standard deviation, and dispersion of various pharmacokinetic parameters in a schizophrenia patient population and to investigate the effects of different factors on risperidone metabolism, this study conducted a population pharmacokinetic analysis of risperidone. Based on literature reports and the characteristics of the data in this study, the fundamental pharmacokinetic model was determined to be a one-compartment, first-order absorption and elimination model, and the random-effects model was an additive model. A linear model was used to quantitatively examine the effects of fixed effects (such as age, sex, and SNP loci) on pharmacokinetic parameters. The genetic and non-genetic factors related to population pharmacokinetic parameters, screened based on a univariate nonlinear mixed-effects model, are shown in Table 5. The results showed that the apparent volume of distribution gradually increased with age and weight, and both had statistically significant effects on the apparent volume of distribution (P<0.05). Genetic factors that had a statistically significant impact on the primary elimination rate (P<0.05) included: rs28371699 (CYP2D6), rs1135840 (CYP2D6), rs1058164 (CYP2D6), rs75276289 (CYP2D6), rs16947 (CYP2D6), and rs72547513 (CYP1A2); non-genetic factors included: smoking, alcohol consumption, decreased glomerular filtration rate, body weight, Lianhua Qingwen granules, and silymarin meglumine tablets. Furthermore, the CYP2D6 enzyme activity score was positively correlated with the primary elimination rate; as the CYP2D6 enzyme activity score increased, the primary elimination rate also increased (P<0.05).

[0156] Table 5 Genetic and non-genetic factors associated with population pharmacokinetic parameters

[0157]

[0158]

[0159] a Ka: First-order absorption rate constant (h) -1 V: Apparent volume of distribution (L); CL: First-order elimination rate constant (L / H);

[0160] Covariates for the population pharmacokinetic model were screened using a forward incremental and backward elimination method, and the final population pharmacokinetic model was determined by combining relevant professional knowledge. The parameter estimates for the final 9-hydroxyrisperidone model are shown in Table 6. In the final population pharmacokinetic model, body weight had a significant effect on the apparent volume of distribution (P<0.05); smoking, drinking habits, combined use of Lianhua Qingwen granules and silymarin meglumine tablets, decreased glomerular filtration function, and CYP2D6 enzyme activity score had significant effects on the first-order elimination rate (P<0.05). The population typical values ​​of Ka, V, and CL in the final model were: Ka = 6.113 h. -1V = 25.694 * exp(0.022 * body weight), CL = 3.139 * exp(0.155 * smoking - 0.1158 * drinking - 0.280 * Lianhua Qingwen granules + 0.473 * silymarin meglumine tablets - 0.141 * decreased glomerular filtration rate + 0.066 * CYP2D6 enzyme activity score). The results of 500 Bootstrap validations show that the model parameter estimates based on measured data are generally close to those based on sampled data. The estimated values ​​of the main PK parameters obtained from the final model are all within the quartile range of the Bootstrap validation parameters, indicating that the model estimation is stable and has good reliability.

[0161] Table 6. Final model parameter estimation and validation results for 9-hydroxyrisperidone

[0162]

[0163] a CL: 9-hydroxyrisperidone scavenging rate (L / H); Ka: absorption rate (h⁻¹) -1 V: Apparent volume of distribution (V);

[0164] The goodness-of-fit plot of the final model for 9-hydroxyrisperidone is shown in the figure. Figure 2 RMSE and R of 99-hydroxyrisperidone individual prediction versus population prediction 2 The concentrations were 9.90 ng / ml, 0.837 ng / ml, and 20.9 ng / ml, 0.329 ng / ml, respectively. The observed blood drug concentrations were evenly distributed on both sides of the trend line (y = x) and largely coincided with the diagonal. The residual plot of the final model for 9-hydroxyrisperidone is shown below. Figure 3 The results showed that the conditionally weighted residuals were uniformly distributed on both sides of y=0 and showed no significant correlation with time and population predictions. A schematic diagram of the final model population pharmacokinetic curves for a risperidone dosage regimen of 2 mg (bid) is shown below. Figure 4 .

[0165] 2.1.3 Routine monitoring of risperidone to predict blood drug concentration

[0166] 2.1.3.1 Variable Selection Based on Recursive Feature Elimination Algorithm

[0167] We employ a recursive feature elimination algorithm based on random forests to select predictor variables for machine learning modeling. The RMSE of 10-fold cross-validation when modeling with different numbers of candidate variables using random forests is shown in [link to RMSE documentation]. Figure 5 .Depend on Figure 5It can be seen that when including the top 36 predictor variables in terms of importance, the random forest algorithm has the smallest RMSE for 10-fold cross-validation. Therefore, the predictor variables selected by the recursive feature elimination algorithm based on random forest include medication dosage, weight, BMI, age, endogenous creatinine clearance rate, rs17327442, CYP2D6 enzyme activity score, sodium valproate, propranolol, rs7787082, benztropine, rs3789243, rs4244285, rs11528090, rs7779562, and dioscorea opposita. The following are listed as predictive variables: Xinxuekang soft capsules, rs762551, rs3743484, rs2235047, smoking, rs1135840, rs28371699, rs1058164, gender, traracetam capsules, alcohol consumption, rs1081003, rs1065852, rs58440431, rs2004511, rs16947, rs75276289, rs1080996, silymarin meglumine tablets, rs116917064, and aspirin enteric-coated tablets. These predictive variables will be used to construct predictive models based on machine learning methods such as random forests and Bayesian cumulative regression trees.

[0168] 2.1.3.2 Model Fitting and Evaluation

[0169] This study fitted prediction models based on gradient boosting trees, support vector machines, random forests, Bayesian cumulative regression trees, and the XGBoost algorithm. Using NNLS as the loss function, ten-fold cross-validation was employed to determine the weight vector that minimized the cross-validation risk of the combined models. The weights for these models were 0.000, 0.156, 0.347, 0.243, and 0.254, respectively. This resulted in the construction of a Super Learner-based model for predicting 9-hydroxyrisperidone blood concentration. This model is suitable for predicting the steady-state 9-hydroxyrisperidone blood concentration in schizophrenic patients at 6:30 AM (medication times are 10:00 AM and 9:00 PM). The weights and parameter settings for each individual algorithm in Super Learner are shown in Table 7.

[0170] Table 7 shows the weights and parameter settings for each individual algorithm in Super Learner.

[0171]

[0172] Note: Algorithm parameters not listed in Table 7 are set to the default values ​​of the R package.

[0173] In this study, we randomly selected 300 routine blood drug concentration monitoring records as the validation set, and the remaining records as the training set. The 9-kirisperidone blood drug concentration prediction model based on Super Learner was evaluated. In the training set, the RMSEs of Super Learner, Gradient Boosting Tree, Support Vector Machine, Random Forest, Bayesian Cumulative Regression Tree, and XGBoost algorithms were 12.91 ng / ml, 14.25 ng / ml, 14.48 ng / ml, 12.69 ng / ml, 16.22 ng / ml, and 11.31 ng / ml, respectively. In the validation set, the RMSEs of the above algorithms were 16.23 ng / ml, 15.67 ng / ml, 16.98 ng / ml, 16.15 ng / ml, 16.98 ng / ml, and 17.31 ng / ml, respectively. Super Learner achieved higher R-values ​​in both the training and validation sets. 2 The values ​​were 0.66 and 0.50 respectively, both showing good performance. Figure 7 ).

[0174] Table 8 Evaluation of the Super Learner-based 9-kirisperidone plasma concentration prediction model

[0175]

[0176] 2.2 Basic information of olanzapine study subjects and population pharmacokinetic studies

[0177] 2.2.1 Basic Information of the Research Subjects

[0178] This part of the study included 131 patients with schizophrenia treated with olanzapine. The general characteristics and laboratory test results of the subjects are shown in Table 9. The mean age of the subjects was 45.6 ± 14.2 years; 84 were male and 47 were female, accounting for 64.1% and 35.9% respectively; the mean height and weight were 163.4 ± 9.2 cm and 67.4 ± 14.8 kg respectively, and the mean BMI was 25.2 ± 4.6 kg / m²; 47 were smokers and 30 were drinkers, accounting for 35.9% and 22.9% respectively; the prevalence of diabetes and hypertension was 5.3% and 9.9% respectively. The mean values ​​of liver and kidney function test results were all within the normal reference range, but 25 patients had decreased glomerular filtration rate and 22 patients had abnormal liver function, accounting for 19.1% and 16.8% respectively. Of the patients, 73 took olanzapine twice daily and 58 took it once nightly, accounting for 55.7% and 44.3% respectively. The mean dose per dose was 6.8 ± 2.4 mg.

[0179] Under steady-state blood drug concentration conditions, this study collected olanzapine blood concentrations from 467 patients. For patients taking the medication twice daily, the mean blood drug concentrations at 1.5h, 5.0h, 9.5h, and 12.5h were 61.54±18.32, 71.72±22.99, 49.38±18.93, and 46.41±20.12 ng / ml, respectively. For patients taking the medication once nightly, the mean blood drug concentration at 9.5h was 34.61±15.39 ng / ml. Based on the upper limit of the therapeutic reference range of 80 g / ml for olanzapine, 22 cases (5.41% of all cases measured at that time point) showed blood drug concentrations exceeding this upper limit at 9.5h.

[0180] Table 9 General information and laboratory test indicators of the study subjects

[0181]

[0182]

[0183] * Under steady-state conditions, the blood drug concentration at 1.5h, 5.0h, 9.5h, and 12.5h after the last administration; decreased glomerular filtration function: creatinine clearance <80ml / min; abnormal liver function: alanine aminotransferase >40U / L or aspartate aminotransferase >40U / L;

[0184] 2.2.2 Group pharmacokinetic studies of olanzapine

[0185] To determine the mean, standard deviation, and dispersion of various pharmacokinetic parameters in a schizophrenia patient population and to investigate the effects of different factors on clozapine metabolism, this study conducted a population pharmacokinetic analysis of olanzapine. Based on literature reports and the characteristics of the data in this study, the fundamental pharmacokinetic model was determined to be a one-compartment, first-order absorption and elimination model, and the random-effects model was an additive model. A linear model was used to quantitatively examine the effects of fixed effects (such as age, sex, and SNP loci) on pharmacokinetic parameters. Genetic and non-genetic factors related to population pharmacokinetic parameters, screened based on a univariate nonlinear mixed-effects model, are shown in Table 10. The results showed that male sex, aspartate aminotransferase (AST) levels, and concurrent use of sodium valproate, oxcarbazepine, and lithium carbonate were associated with a higher olanzapine elimination rate (P<0.05); while decreased glomerular filtration rate was associated with a lower olanzapine elimination rate (P<0.05). The SNP sites rs4244285 (CYP2C19), rs12768009 (CYP2C19), rs7916649 (CYP2C19), rs3758580 (CYP2C19), rs3743484 (CYP1A2), and rs11528090 (CYP2C19) had a statistically significant impact on the primary elimination rate (P<0.05).

[0186] Table 10 Genetic and non-genetic factors associated with olanzapine population pharmacokinetic parameters

[0187]

[0188] a CL: First-order elimination rate constant (L / H); b Aspartate aminotransferase (AST) was a continuous variable; sex, decreased glomerular filtration rate, sodium valproate, oxcarbazepine, and lithium carbonate were binary variables (0 / 1); rs4244285: 0-A / A, 1-A / G, 2-G / G; rs12768009: 0-A / A, 1-G / A, 2-G / G; rs7916649: 0-A / A, 1-G / A, 2-G / G; rs3743484: 0-C / C, 1-G / C, 2-G / G; rs11528090: 0-G / G, 1-T / G, 2-T / T;

[0189] Covariates for the population pharmacokinetic model were screened using a forward incremental and backward elimination method, and the final population pharmacokinetic model was determined by combining relevant professional knowledge. The parameter estimates for the final population pharmacokinetic model of olanzapine are shown in Table 11. In the final population pharmacokinetic model, the following factors were included in the first-order elimination rate fixed-effects model: aspartate aminotransferase (AST), decreased glomerular filtration rate, combination with sodium valproate, oxcarbazepine, and rs7916649 and rs12768009. The population typical values ​​for Ka, V, and CL in the final model are: Ka = 0.144 h.-1 V = 71.197 L, CL = 6.120 * exp(0.009 * AST - 0.130 * glomerular filtration rate decline + 0.253 * sodium valproate + 0.774 * oxcarbazepine + 0.292 * rs7916649 - 0.155 * rs12768009). The results of 500 Bootstrap validations show that the model parameter estimates based on the measured data are generally close to those based on the sampled data. The estimated values ​​of the main PK parameters obtained from the final model are all within the quartile range of the Bootstrap validation parameters, indicating that the model estimation is stable and has good reliability.

[0190] Table 11. Parameter estimation and validation results of the final population pharmacokinetic model of olanzapine.

[0191]

[0192] a CL: Olanzapine clearance rate (L / H); Ka: Absorption rate (h) -1 V: Apparent volume of distribution (V);

[0193] b rs7916649: 0-A / A, 1-G / A, 2-G / G.

[0194] The goodness-of-fit plot of the final population pharmacokinetic model of olanzapine is shown in [the image]. Figure 8 The results showed that the predicted individual and population concentrations fit well with the measured concentrations, and the RMSE and R0 values ​​were satisfactory. 2 The concentrations were 13.946 ng / ml, 0.469 ng / ml, and 16.9 ng / ml, 0.212 ng / ml, respectively. The observed blood drug concentrations were evenly distributed on both sides of the trend line (y = x) and largely coincided with the diagonal. The residual plot of the final population pharmacokinetic model for olanzapine is shown below. Figure 9 The results showed that the conditionally weighted residuals were uniformly distributed on both sides of y=0 and had no significant correlation with time and population predictions. A schematic diagram of the final model population pharmacokinetic curves for a clozapine dosing regimen of 75 mg (bid) is shown below. Figure 10 .

[0195] 2.2.3 Olanzapine routine monitoring for blood drug concentration prediction

[0196] 2.2.3.1 Variable Selection Based on Recursive Feature Elimination Algorithm

[0197] We employ a recursive feature elimination algorithm based on random forests to select predictor variables for machine learning modeling. The RMSE of 10-fold cross-validation when modeling with different numbers of candidate variables using random forests is shown in [link to RMSE documentation]. Figure 11 .Depend on Figure 11It can be seen that when including the top 42 predictor variables in the importance ranking, the random forest algorithm has the smallest RMSE with 10-fold cross-validation. The importance of random forest variables can be found in [the table / reference needed]. Figure 12 The predictive variables selected using the recursive feature elimination algorithm based on random forest include medication dosage, frequency of medication, age, alkaline phosphatase, weight, alanine aminotransferase (ALT), aspartate aminotransferase (AST), sodium valproate, CYP2D6 enzyme activity score, rs1065852, rs762551, rs3743484, aripiprazole, rs11528090, rs35280822, rs16947, rs79331140, rs1080996, rs7787082, propranolol, and rs77 The following variables were identified: 79562, rs75276289, rs4646437, rs6583954, rs3758580, rs4244285, smoking, drinking, gender, rs12768009, rs7916649, rs2470890, rs17884832, rs5030865, rs17879992, rs17885098, rs28371725, decreased glomerular filtration rate, phenolphthalein tablets, sulpiride, ziprasidone, and sanguisorba officinalis tablets. These predictor variables will be used to construct predictive models based on machine learning methods such as random forests and Bayesian cumulative regression trees.

[0198] 2.2.3.2 Model Fitting and Evaluation

[0199] This study fitted prediction models based on gradient boosting trees, support vector machines, random forests, Bayesian cumulative regression trees, and the XGBoost algorithm. Using NNLS as the loss function, ten-fold cross-validation was employed to determine the weight vector that minimized the cross-validation risk of the combined models. The weights for these models were 0.424, 0.454, 0.000, 0.028, and 0.094, respectively. This led to the construction of a Super Learner-based clozapine blood concentration prediction model, suitable for predicting olanzapine blood concentrations in schizophrenic patients at 6:30 AM in a steady-state state (with medication times at 10:00 AM and / or 9:00 PM). The weights and parameter settings for each individual algorithm in Super Learner are shown in Table 12.

[0200] Table 12 Weights and parameter settings for each individual algorithm in Super Learner

[0201]

[0202]

[0203] Note: Algorithm parameters not listed in Table 12 are set to the default values ​​of the R package.

[0204] In this study, we randomly selected 100 routine blood drug concentration monitoring records as the validation set, and the remaining records as the training set. The evaluation of the olanzapine blood drug concentration prediction model based on Super Learner is shown in Table 13. In the training set, the RMSEs of Super Learner, Gradient Boosting Tree, Support Vector Machine, Random Forest, Bayesian Cumulative Regression Tree, and XGBoost algorithms were 9.18 ng / ml, 9.36 ng / ml, 10.12 ng / ml, 9.03 ng / ml, 11.51 ng / ml, and 7.31 ng / ml, respectively. In the validation set, the RMSEs of the above algorithms were 13.90 ng / ml, 13.74 ng / ml, 14.96 ng / ml, 13.74 ng / ml, 15.07 ng / ml, and 14.17 ng / ml, respectively. Super Learner's R-values ​​in the training and validation sets... 2 The values ​​were 0.75 and 0.53 respectively, both showing good performance. Figure 13 ).

[0205] Table 13 Evaluation of the olanzapine plasma concentration prediction model based on Super Learner

[0206]

[0207] 2.3 Basic information of subjects and population pharmacokinetic studies of clozapine

[0208] 2.3.1 Basic Information of the Research Subjects

[0209] This part of the study included 136 schizophrenia patients treated with clozapine. The general characteristics and laboratory indicators of the subjects are shown in Table 14. The mean age of the subjects was 44.8 ± 9.9 years; there were 70 males and 63 females, accounting for 51.5% and 48.5% respectively; the mean height and weight were 163.7 ± 8.8 cm and 70.6 ± 14.5 kg respectively, and the mean BMI was 26.3 ± 4.7 kg / m². 2 Among the study participants, 35 were smokers and 19 were drinkers, accounting for 25.7% and 14.0% respectively; the prevalence of diabetes and hypertension was 11.0% and 8.1% respectively. The mean values ​​of liver and kidney function test indicators were all within the normal reference range, but 27 participants had decreased glomerular filtration function and 30 had abnormal liver function, accounting for 19.9% ​​and 22.1% respectively.

[0210] This study included schizophrenic patients taking clozapine twice or once daily, with a mean dose of 75.9 ± 35.5 mg per dose. Under steady-state blood drug concentration conditions, blood drug concentrations of clozapine were collected from 665 patients. The mean blood drug concentrations at 1.5 h, 5.0 h, 9.5 h, and 12.5 h after administration were 588.2 ± 235.5 ng / ml, 460.2 ± 227.5 ng / ml, 337.4 ± 233.8 ng / ml, and 249.1 ± 198.9 ng / ml, respectively. Based on the upper limit of the therapeutic concentration reference range of 600 ng / ml for clozapine, the blood drug concentrations exceeded the upper limit of the therapeutic concentration range in 7 cases after 1.5 h and in 66 cases after 9.5 h, representing 53.8% and 10.69% of the cases measured at those time points, respectively.

[0211] Table 14 General information and laboratory test indicators of the study subjects

[0212]

[0213]

[0214] * Under steady-state conditions, the blood drug concentration at 1.5h, 5.0h, 9.5h, and 12.5h after the last administration; decreased glomerular filtration function: creatinine clearance <80ml / min; abnormal liver function: alanine aminotransferase >40U / L or aspartate aminotransferase >40U / L;

[0215] 2.3.2 Population pharmacokinetic studies of clozapine

[0216] To determine the mean, standard deviation, and dispersion of various pharmacokinetic parameters in a schizophrenia patient population and to investigate the effects of different factors on clozapine metabolism, this study conducted a population pharmacokinetic analysis of quetiapine. Based on literature reports and the characteristics of the data in this study, the fundamental pharmacokinetic model was determined to be a one-compartment, first-order absorption and elimination model, and the random-effects model was an additive model. A linear model was used to quantitatively examine the effects of fixed effects (such as age, sex, and SNP loci) on pharmacokinetic parameters. The genetic and non-genetic factors related to population pharmacokinetic parameters, screened based on a univariate nonlinear mixed-effects model, are shown in Table 15. The results showed that the apparent volume of distribution was significantly larger in males than in females (P>0.05); however, the apparent volume of distribution gradually decreased with increasing age (P<0.05). Genetic factors that significantly affected the primary elimination rate (P<0.05) included rs1135840 (CYP2D6), rs1065164 (CYP2D6), rs28371699 (CYP2D6), and rs1135822 (CYP2D6); non-genetic factors included age, sex, smoking, aspartate aminotransferase (AST) level, and concomitant propranolol use. Furthermore, the CYP2D6 enzyme activity score was positively correlated with the primary elimination rate; as the CYP2D6 enzyme activity score increased, the primary elimination rate also increased (P<0.05).

[0217] Table 15 Genetic and non-genetic factors associated with clozapine population pharmacokinetic parameters

[0218]

[0219] V: Apparent volume of distribution (L); CL: First-order elimination rate constant (L / H);

[0220] Covariates for the population pharmacokinetic model were screened using a forward incremental and backward elimination method, and the final population pharmacokinetic model was determined by combining relevant professional knowledge. The parameter estimates for the final population pharmacokinetic model of clozapine are shown in Table 16. In the final population pharmacokinetic model, sex had a significant effect on the apparent volume of distribution (P<0.05); age, sex, combination with propranolol, and SNP sites rs1135840 and rs1135822 had significant effects on the first-order elimination rate (P<0.05). The parameter estimates for the final model are shown in Table 18. The results show that the population typical values ​​of Ka, V, and CL in the final model are: Ka = 0.558h. -1V = 181.874 * exp(1.371 * male), CL = 42.472 * exp(0.239 * male - 0.009 * age - 0.137 * propranolol - 0.148 * rs1135840 - 0.474 * rs1135822). The Bootstrap validation results (500 runs) show that the model parameter estimates based on the measured data are generally close to those based on the sampled data. The estimated values ​​of the main PK parameters obtained from the final model are all within the quartile range of the Bootstrap validation parameters, indicating that the model estimation is stable and has good reliability.

[0221] Table 16. Final model parameter estimation and validation results for clozapine.

[0222]

[0223]

[0224] a CL: Clozapine scavenging rate (L / H); Ka: Absorption rate (h) -1 V: Apparent volume of distribution (V);

[0225] b rs1135840: 0-C / C, 1-C / G, 2-G / G; rs1135822: 0-A / T, 1-T / T;

[0226] The goodness-of-fit plot of the final model is shown below. Figure 14 The results showed that the predicted individual concentrations and population concentrations fit the measured concentrations well, with low RMSE and low R0.05. 2 The observed blood drug concentrations were 98.33 ng / ml and 0.857 ng / ml, and 173.45 ng / ml and 0.56 ng / ml, respectively. The observed blood drug concentrations were evenly distributed on both sides of the trend line (y = x) and largely coincided with the diagonal. The residual plot of the final model is shown below. Figure 15 The results showed that the conditionally weighted residuals were uniformly distributed on both sides of y=0 and had no significant correlation with time and population predictions, indicating a good model fit. A schematic diagram of the final model population pharmacokinetic curves for a clozapine dosing regimen of 75mg (bid) is shown below. Figure 16 .

[0227] 2.3.3 Prediction of Clozapine Blood Drug Concentration through Routine Monitoring

[0228] 2.3.3.1 Variable Selection Based on Recursive Feature Elimination Algorithm

[0229] We employ a recursive feature elimination algorithm based on random forests to select predictor variables for machine learning modeling. The RMSE of 10-fold cross-validation when modeling with different numbers of candidate variables using random forests is shown in [link to RMSE documentation]. Figure 17 .Depend on Figure 17 It can be seen that when including the top 28 predictor variables in terms of importance, the random forest algorithm has the smallest RMSE for 10-fold cross-validation. Therefore, the predictor variables selected by the recursive feature elimination algorithm based on random forest include medication dosage, creatinine, age, aspartate aminotransferase, creatinine clearance rate, weight, propranolol, risperidone, rs1080996, rs75276289, rs16947, rs11528090, rs762551, gender, medication frequency, rs12535512, retracetam capsules, rs1135840, rs1128503, rs12720464, rs3789243, CYP2D6 enzyme activity score, enteric-coated aspirin tablets, rs3842, rs6979885, atorvastatin calcium tablets, rs4728709, and smoking. The aforementioned predictor variables will be used to construct predictive models based on machine learning methods such as random forests and Bayesian cumulative regression trees.

[0230] 2.3.3.2 Model Fitting and Evaluation

[0231] This study fitted prediction models based on gradient boosting trees, support vector machines, random forests, Bayesian cumulative regression trees, and the XGBoost algorithm. Using NNLS as the loss function, ten-fold cross-validation was employed to determine the weight vector that minimized the cross-validation risk of the combined models. The weights for these models were 0.271, 0.342, 0.213, 0.155, and 0.019, respectively. This led to the construction of a Super Learner-based clozapine blood concentration prediction model, suitable for predicting clozapine blood concentrations in schizophrenic patients at 6:30 AM in a steady-state state (with medication administration times at 10:00 AM and / or 9:00 PM). The weights and parameter settings for each individual algorithm in Super Learner are shown in Table 17.

[0232] Table 17 Weights and parameter settings for each individual algorithm in Super Learner

[0233]

[0234] Note: Algorithm parameters not listed in Table 19 are set to the default values ​​of the R package.

[0235] In this study, we randomly selected 150 routine blood drug concentration monitoring records as the validation set, and the remaining records as the training set. The evaluation of the olanzapine blood drug concentration prediction model based on Super Learner is shown in Table 18. In the training set, the RMSE values ​​of Super Learner, Gradient Boosting Tree, Support Vector Machine, Random Forest, Bayesian Cumulative Regression Tree, and XGBoost algorithms were 68.88 ng / ml, 69.26 ng / ml, 72.76 ng / ml, 64.41 ng / ml, 80.73 ng / ml, and 55.89 ng / ml, respectively. In the validation set, the RMSE values ​​of the above algorithms were 88.51 ng / ml, 92.08 ng / ml, 90.45 ng / ml, 90.68 ng / ml, 88.98 ng / ml, and 96.31 ng / ml, respectively. Super Learner's R-values ​​in the training and validation sets... 2 The values ​​were 0.72 and 0.58 respectively, both showing good performance. Figure 19 ).

[0236] Table 18 Evaluation of the olanzapine plasma concentration prediction model based on Super Learner

[0237]

[0238]

[0239] 2.4 Basic information of subjects and population pharmacokinetic studies of quetiapine

[0240] 2.4.1 Basic Information of the Research Subjects

[0241] This part of the study included 177 patients with schizophrenia treated with quetiapine. The general characteristics and laboratory test results of the subjects are shown in Table 19. The mean age of the subjects was 46.3 ± 12.6 years; 104 were male and 73 were female, accounting for 58.8% and 41.2% respectively; the mean height and weight were 163.3 ± 8.6 cm and 71.5 ± 16.5 kg respectively, and the mean BMI was 26.7 ± 5.5 kg / m²; among the subjects, 44 were smokers and 35 were drinkers, accounting for 24.9% and 19.8% respectively; the prevalence of diabetes and hypertension was 11.3% and 13.6% respectively. The mean values ​​of liver and kidney function test results were all within the normal reference range, but 38 subjects had decreased glomerular filtration rate and 21 subjects had abnormal liver function, accounting for 21.5% and 11.9% respectively.

[0242] This study only included schizophrenic patients taking quetiapine twice daily, with a mean dose of 242.9 ± 88.3 mg per dose. Under steady-state blood drug concentration conditions, a total of 551 quetiapine blood concentrations were collected. The mean blood drug concentrations at 1.5 h, 5.0 h, 9.5 h, and 12.5 h after administration were 334.5 ± 218.2, 286.3 ± 141.2, 209.1 ± 144.5, and 141.9 ± 125.5 ng / ml, respectively. Based on the upper limit of the therapeutic concentration reference range of 500 ng / ml for quetiapine, 9 cases showed blood drug concentrations exceeding the upper limit at 1.5 h and 18 cases at 9.5 h, representing 25.7% and 3.9% of the cases measured at those time points, respectively.

[0243] Table 19 General Information and Laboratory Testing Indicators of the Study Subjects

[0244]

[0245] * Under steady-state conditions, the blood drug concentration at 1.5h, 5.0h, 9.5h, and 12.5h after the last administration; decreased glomerular filtration function: creatinine clearance <80ml / min; abnormal liver function: alanine aminotransferase >40U / L or aspartate aminotransferase >40U / L;

[0246] 2.4.2 Population pharmacokinetic studies of quetiapine

[0247] To determine the mean, standard deviation, and dispersion of various pharmacokinetic parameters in a schizophrenia patient population and to investigate the effects of different factors on clozapine metabolism, this study conducted a population pharmacokinetic analysis of quetiapine. Based on literature reports and the characteristics of the data in this study, the fundamental pharmacokinetic model was determined to be a one-compartment, first-order absorption and elimination model, and the random-effects model was an additive model. A linear model was used to quantitatively examine the effects of fixed effects (such as age, sex, and SNP loci) on pharmacokinetic parameters. Genetic and non-genetic factors related to population pharmacokinetic parameters, screened based on a univariate nonlinear mixed-effects model, are shown in Table 20. The results showed that the apparent volume of distribution gradually increased with increasing body weight, and the effect of body weight on the apparent volume of distribution was statistically significant (P < 0.05). Genetic factors that had a statistically significant impact on the primary elimination rate (P<0.05) included: rs5030865 (CYP2D6), rs2069526 (CYP1A2), rs2242480 (CYP3A4), and rs2235047 (ABCB1); non-genetic factors included: age, sex, endogenous creatinine clearance rate, alkaline phosphatase, aspartate aminotransferase, traracetam capsules, aripiprazole, atorvastatin calcium tablets, and serum creatinine levels.

[0248] Table 20 Genetic and non-genetic factors associated with quetiapine population pharmacokinetic parameters

[0249]

[0250] a Ka: First-order absorption rate constant (h) -1 V: Apparent volume of distribution (L); CL: First-order elimination rate constant (L / H);

[0251] Covariates for the population pharmacokinetic model were screened using a forward incremental and backward elimination method, and the final population pharmacokinetic model was determined by combining relevant professional knowledge. The parameter estimates for the final population pharmacokinetic model of quetiapine are shown in Table 21. In the final population pharmacokinetic model, age, sex, concomitant use of atorvastatin calcium tablets, endogenous creatinine clearance, and rs2242480 significantly affected the first-order elimination rate (P<0.05). The population typical values ​​of Ka, V, and CL in the final model were: Ka = 0.163 h. -1 V = 274.626L, CL = 66.389*exp(-0.195*female-0.007*age+0.004*endogenous creatinine clearance-0.599*atorvastatin calcium tablets-0.119*rs2242480). The results of 500 Bootstrap validations show that the model parameter estimates based on measured data are generally close to those based on sampled data. The estimated values ​​of the main PK parameters obtained from the final model are all within the quartile range of the Bootstrap validation parameters, indicating that the model estimation is stable and has good reliability.

[0252] Table 21. Parameter estimation for the final population pharmacokinetic model of quetiapine.

[0253]

[0254] a CL: Quetiapine clearance rate (L / H); Ka: Absorption rate (h) -1 V: Apparent volume of distribution (V);

[0255] b rs2242480: 0, C / C; 1, T / C; 2, T / T;.

[0256] The goodness-of-fit plot of the final model is shown below. Figure 20The results showed that the predicted individual and population concentrations fit the measured concentrations well, with RMSE and R² values ​​of 63.11 ng / ml and 0.857, and 136.13 ng / ml and 0.316, respectively. The observed blood drug concentrations were evenly distributed on both sides of the trend line (y = x) and largely coincided with the diagonal. The residual plot of the final model is shown below. Figure 21 The results showed that the conditionally weighted residuals were uniformly distributed on both sides of y=0 and had no significant correlation with time and population predictions, indicating a good model fit. A schematic diagram of the final model population pharmacokinetic curves for a quetiapine dosing regimen of 300 mg (bid) is shown below. Figure 22 .

[0257] 2.4.3 Routine monitoring of quetiapine to predict blood drug concentration

[0258] 2.4.3.1 Variable Selection Based on Recursive Feature Elimination Algorithm

[0259] We employ a recursive feature elimination algorithm based on random forests to select predictor variables for machine learning modeling. The RMSE of 10-fold cross-validation when modeling with different numbers of candidate variables using random forests is shown in [link to RMSE documentation]. Figure 23 As shown in the figure, when including the top 31 predictor variables in terms of importance, the random forest algorithm achieves the lowest RMSE with 10-fold cross-validation. The importance score for predictor variables based on random forest can be found in [the figure / reference needed]. Figure 24 The predictive variables selected using a recursive feature elimination algorithm based on random forest include medication dosage, age, BMI, alkaline phosphatase, aspartate aminotransferase (AST), creatinine, alanine aminotransferase (ALT), weight, sodium valproate, CYP2D6 enzyme activity score, rs12535512, rs2246709, rs12768009, rs2242480, rs1081003, cefuroxime axetil, amisulpride, rs58440431, rs2004511, Dio Xin Xue Kang soft capsules, gender, enalapril maleate tablets, atorvastatin calcium tablets, rs2470890, smoking, abnormal liver function, rs1135822, oxcarbazepine, buspirone, aripiprazole, and Wenxin granules. These predictive variables will be used to construct predictive models based on machine learning methods such as random forest and Bayesian cumulative regression trees.

[0260] 2.4.3.2 Model Fitting and Evaluation

[0261] This study fitted prediction models based on gradient boosting trees, support vector machines, random forests, Bayesian cumulative regression trees, and the XGBoost algorithm. Using NNLS as the loss function, ten-fold cross-validation was employed to determine the weight vector that minimized the cross-validation risk of the combined models. The weights for these models were 0.000, 0.083, 0.674, 0.243, and 0.000, respectively. This resulted in the construction of a Super Learner-based model for predicting 9-hydroxyrisperidone blood concentration. This model is suitable for predicting the steady-state blood concentration of 9-hydroxyrisperidone in schizophrenic patients at 6:30 AM (with medication administration times at 10:00 AM and 9:00 PM). The weights and parameter settings for each individual algorithm in Super Learner are shown in Table 22.

[0262] Table 22 Weights and parameter settings for individual algorithms in Super Learner

[0263]

[0264] Note: Algorithm parameters not listed in Table 22 are set to the default values ​​of the R package.

[0265] In this study, we randomly selected 100 routine blood drug concentration monitoring records as the validation set, and the remaining records as the training set. The evaluation of the quetiapine blood drug concentration prediction model based on Super Learner is shown in Table 23. In the training set, the RMSE values ​​for Super Learner, Gradient Boosting Tree, Support Vector Machine, Random Forest, Bayesian Cumulative Regression Tree, and XGBoost algorithms were 68.43 ng / ml, 67.61 ng / ml, 75.28 ng / ml, 63.27 ng / ml, 88.60 ng / ml, and 48.66 ng / ml, respectively. In the validation set, the RMSE values ​​for the above algorithms were 85.27 ng / ml, 81.09 ng / ml, 84.18 ng / ml, 86.90 ng / ml, 89.35 ng / ml, and 90.23 ng / ml, respectively. Super Learner's R-values ​​in the training and validation sets... 2 The values ​​were 0.65 and 0.46 respectively, both showing good performance. Figure 25 ).

[0266] Table 23 Evaluation of the Super Learner Quetiapine Blood Concentration Prediction Model

[0267]

[0268] It should be noted that the above examples are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the given examples, those skilled in the art can modify or make equivalent substitutions to the technical solutions of the present invention as needed, without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for constructing a population pharmacokinetic model of an antipsychotic drug, characterized in that, The construction method includes: collecting clinical information and blood samples from subjects to obtain relevant data; using a nonlinear mixed-effects model for analysis, specifically using a one-compartment first-order absorption and elimination pharmacokinetic model as the basic structural model, obtaining simulated blood drug concentration data to obtain corresponding model pharmacokinetic parameters; using an additive random-effects model and introducing covariates to fit the model, examining the influence of different factors on pharmacokinetic parameters, and constructing a population pharmacokinetic model for antipsychotic drugs; Among them, antipsychotic drugs include risperidone, olanzapine, clozapine, and quetiapine; The covariates include genetic and non-genetic factors, and a linear model is used for quantitative screening and evaluation.

2. The construction method as described in claim 1, characterized in that, When the drug is risperidone, the covariates include weight, smoking, drinking habits, combined use of Lianhua Qingwen granules and silymarin meglumine tablets, decreased glomerular filtration function, and CYP2D6 enzyme activity score. The final population typical values ​​for Ka, V, and CL are as follows: Ka = 6.113h -1 V = 25.694 * exp(0.022 * body weight), CL = 3.139 * exp(0.155 * smoking - 0.1158 * drinking - 0.280 * Lianhua Qingwen granules + 0.473 * silymarin meglumine tablets - 0.141 * decreased glomerular filtration function + 0.066 * CYP2D6 enzyme activity score); When the drug is olanzapine, the covariates include aspartate aminotransferase, decreased glomerular filtration function, combination with sodium valproate, oxcarbazepine, and SNP sites rs7916649 and rs12768009. The final population typical values ​​for Ka, V, and CL are as follows: Ka = 0.144h -1 V=71.197L, CL=6.120*exp(0.009*AST-0.130*glomerular filtration rate decline+0.253*sodium valproate+0.774*oxcarbazepine+0.292*rs7916649-0.155*rs12768009); When the drug is clozapine, the covariates include age, sex, combination with propranolol, and SNP sites rs1135840 and rs1135822; The population typical values ​​of Ka, V, and CL in the final model are as follows: Ka = 0.558h -1 V = 181.874 * exp(1.371 * male), CL = 42.472 * exp(0.239 * male - 0.009 * age - 0.137 * propranolol - 0.148 * rs1135840 - 0.474 * rs1135822); When the drug is quetiapine, the covariates include age, sex, combination with atorvastatin calcium tablets, endogenous creatinine clearance rate, and rs2242480. The final population typical values ​​for Ka, V, and CL are as follows: Ka = 0.163h -1 V=274.626L, CL=66.389*exp(-0.195*female-0.007*age+0.004*endogenous creatinine clearance-0.599*atorvastatin calcium tablets-0.119*rs2242480).

3. The application of the population pharmacokinetic model of antipsychotic drugs obtained by the construction method according to claim 1 or 2 in any one or more of the following: (a) To prepare products that accurately predict the dosage of antipsychotic drugs; (b) To prepare products for personalized administration of antipsychotic drugs; (c) Used for basic research related to the pharmacokinetics of antipsychotic drugs; Furthermore, in (c), the pharmacokinetics of the antipsychotic drug includes prediction of antipsychotic drug blood concentration.

4. A blood concentration prediction and assessment system for antipsychotic drugs, using a population pharmacokinetic model of antipsychotic drugs obtained by the construction method described in claim 1 or 2, characterized in that, The system includes at least: The acquisition unit is configured to acquire subject-related data information; The data processing unit is configured to: predict the blood concentration of the antipsychotic drug in the subject based on the data information obtained by the acquisition unit and a built-in prediction and evaluation model; the prediction and evaluation model is obtained by training the model using statistical algorithms on relevant data information of the pre-collected patients. The output unit is configured to output the predicted blood concentration of the antipsychotic drug for the subject based on the information from the data processing unit. The subject-related data includes the subject's general information, disease characteristics, antipsychotic drug use, and SNP sites related to the efficacy of antipsychotic drugs; The antipsychotic drugs mentioned include risperidone, clozapine, olanzapine, and quetiapine.

5. The system as described in claim 4, characterized in that, When the antipsychotic drug is risperidone, during the construction of the prediction and evaluation model, the relevant data information includes, but is not limited to, medication dosage, weight, BMI, age, endogenous creatinine clearance rate, rs17327442, CYP2D6 enzyme activity score, sodium valproate, propranolol, rs7787082, trihexyphenidyl, rs3789243, rs4244285, rs11528090, rs7779562, and so on. Aoxinxuekang soft capsules, rs762551, rs3743484, rs2235047, smoking, rs1135840, rs28371699, rs1058164, gender, traracetam capsules, alcohol consumption, rs1081003, rs1065852, rs58440431, rs2004511, rs16947, rs75276289, rs1080996, silymarin meglumine tablets, rs116917064, aspirin enteric-coated tablets; The prediction model is constructed using a super-powerful learning ensemble algorithm to predict the blood concentration of 9-hydroxyrisperidone.

6. The system as described in claim 4, characterized in that, The antipsychotic drug mentioned is olanzapine. During the construction of the predictive assessment model, the relevant data information includes medication dosage, frequency of use, age, alkaline phosphatase, weight, alanine aminotransferase (ALT), aspartate aminotransferase (AST), sodium valproate, CYP2D6 enzyme activity score, rs1065852, rs762551, rs3743484, aripiprazole, rs11528090, rs35280822, rs16947, rs79331140, rs1080996, rs7787082, and propranolol. Lor, rs7779562, rs75276289, rs4646437, rs6583954, rs3758580, rs4244285, smoking, drinking, gender, rs12768009, rs7916649, rs2470890, rs17884832, rs5030865, rs17879992, rs17885098, rs28371725, decreased glomerular filtration function, phenolphthalein tablets, sulpiride, ziprasidone, and sanguisorba officinalis tablets; The prediction model is constructed using a super-powerful learning ensemble algorithm to predict olanzapine blood concentration.

7. The system as described in claim 4, characterized in that, The antipsychotic drug mentioned is clozapine. During the construction of the predictive evaluation model, the relevant data information includes medication dosage, creatinine, age, aspartate aminotransferase (AST), creatinine clearance rate, weight, propranolol, risperidone, rs1080996, rs75276289, rs16947, rs11528090, rs762551, gender, medication frequency, rs12535512, traracetam capsules, rs1135840, rs1128503, rs12720464, rs3789243, CYP2D6 enzyme activity score, enteric-coated aspirin tablets, rs3842, rs6979885, atorvastatin calcium tablets, rs4728709, and smoking. The prediction model is constructed using a super-powerful learning ensemble algorithm to predict clozapine blood concentration.

8. The system as described in claim 4, characterized in that, The antipsychotic drug mentioned is quetiapine. During the construction of the predictive evaluation model, the relevant data information includes medication dosage, age, BMI, alkaline phosphatase, aspartate aminotransferase, creatinine, alanine aminotransferase, weight, sodium valproate, CYP2D6 enzyme activity score, rs12535512, rs2246709, rs12768009, rs2242480, rs1081003, cefuroxime axetil, amisulpride, rs58440431, rs2004511, Dio Xin Xue Kang soft capsules, gender, enalapril maleate tablets, atorvastatin calcium tablets, rs2470890, smoking, abnormal liver function, rs1135822, oxcarbazepine, buspirone, aripiprazole, and Wenxin granules. The prediction model is constructed using a super-powerful learning ensemble algorithm to predict quetiapine blood concentration.

9. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it performs the functions of the system as described in any one of claims 4-8.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the functions of the system as described in any one of claims 4-8.

Images