Blood concentration prediction model based on artificial intelligence and construction method and application thereof
By combining liquid chromatography-tandem mass spectrometry and high-resolution melting curve technology, a blood drug concentration prediction model based on the extreme gradient boosting algorithm was constructed, which solved the problem of high cost of sirolimus blood drug concentration monitoring, enabling its promotion and real-time dynamic monitoring in primary hospitals, and improving treatment efficacy and safety.
Patent Information
- Application Number
- CN202511547883.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-28
- Publication Date
- 2026-02-06
AI Technical Summary
Current technologies for monitoring sirolimus blood concentration rely on high-performance liquid chromatography (HPLC) and liquid chromatography-tandem mass spectrometry (LC-MS/MS), which are costly and difficult to popularize. Furthermore, there is no blood concentration prediction system based on gene polymorphism, resulting in high treatment costs, numerous side effects, and difficulty in promoting its use in primary hospitals.
The blood concentration of sirolimus was detected by liquid chromatography-tandem mass spectrometry, and gene polymorphism was detected by high-resolution melting curve technology. A blood concentration prediction model based on extreme gradient boosting algorithm was constructed, and real-time dynamic monitoring and early warning were realized through mobile devices.
It enables accurate detection and prediction of sirolimus blood concentration, reduces treatment costs, improves treatment efficacy, reduces adverse reactions, and is suitable for promotion in primary hospitals.
Smart Images
Figure CN121483635A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of model construction, in particular to a blood drug concentration prediction model based on artificial intelligence and a construction method and application thereof BACKGROUND
[0002] Kaposiform hemangioendothelioma (KHE) is a rare and aggressive vascular tumor that mainly occurs in infants and children, with an annual incidence of about 9 / 100,000. About 70% of KHE cases are associated with Kasabach-Merritt phenomenon (KMP), which leads to a very high mortality rate. Without effective treatment, the mortality rate can be as high as 20-30%. Traditionally, the use of mTOR blocker sirolimus in the treatment of KHE has achieved remarkable results, effectively controlling tumor growth and reducing tumor size.
[0003] However, as an immunosuppressant, sirolimus can cause multiple systemic injuries during treatment, including anemia, thrombocytopenia, interstitial lung disease, and other serious side effects. The incidence and severity of adverse reactions are closely related to the dose of the drug.
[0004] Currently, the monitoring of sirolimus blood concentration mainly relies on high-performance liquid chromatography analysis (LC / UV) or liquid chromatography tandem mass spectrometry (LC / MS-MS) technology. These methods are technically difficult and costly, making it difficult to popularize in primary hospitals, resulting in a significant increase in the time and economic costs of patients seeking medical treatment.
[0005] At the same time, the pharmacokinetics of sirolimus is significantly affected by genetic polymorphisms of CYP 3A4 / 5, CYP 2C8, and P-glycoprotein (P-gp), but there is currently no blood concentration prediction system based on these genetic polymorphisms applied in clinical practice.
[0006] Therefore, the development of a low-cost, easy-to-promote, and real-time dynamic monitoring of sirolimus blood concentration prediction model is of great significance for improving the treatment effect of KHE children, reducing treatment costs, and reducing adverse reactions.
[0007] To solve the above problems, the applicant proposes a blood drug concentration prediction model based on artificial intelligence. SUMMARY
[0008] The purpose of the present application is to provide a blood drug concentration prediction model based on artificial intelligence to solve the problems in the prior art.
[0009] To achieve the above-mentioned purpose, the present application provides the following technical solutions: a construction method of a blood drug concentration prediction model based on artificial intelligence, comprising the following steps:
[0010] Collecting whole blood samples of patients after taking sirolimus, using liquid chromatography tandem mass spectrometry to detect sirolimus blood concentration;
[0011] Extracting whole blood DNA, using high-resolution melting curve technology to detect CYP 3A4 / 5, CYP 2C8, P-gp gene polymorphism of patients;
[0012] Building a database containing basic information of patients, sirolimus blood concentration and gene polymorphism data;
[0013] Using extreme gradient boosting algorithm to train the data in the database, and establishing an artificial intelligence prediction model of sirolimus blood concentration;
[0014] Verifying and optimizing the prediction model to ensure its prediction accuracy.
[0015] Optionally, the step of detecting sirolimus blood concentration by liquid chromatography tandem mass spectrometry comprises:
[0016] Preparing standard stock solution and working solution;
[0017] Preparing whole blood standard curve and quality control solution;
[0018] Pretreating whole blood samples, including protein precipitation and organic solvent extraction steps;
[0019] Using LC / MS-MS for detection to obtain sirolimus blood concentration data.
[0020] Optionally, the step of detecting gene polymorphism by high-resolution melting curve technology comprises:
[0021] Extracting whole blood DNA;
[0022] Designing and synthesizing primers for CYP 3A4 / 5, CYP 2C8, P-gp genes;
[0023] Performing PCR amplification to obtain target gene fragments;
[0024] Using HRM technology to analyze the melting curve of PCR products to determine gene polymorphism.
[0025] Optionally, the construction of the database comprises:
[0026] Collecting and organizing the basic information of patients, treatment records, sirolimus blood concentration and gene polymorphism data;
[0027] Importing data into the database management system for data cleaning, desensitization and labeling;
[0028] Establishing a structured database to facilitate subsequent data analysis and model construction.
[0029] Optionally, the step of training the prediction model using the extreme gradient boosting algorithm comprises:
[0030] preprocessing and feature extraction of the data in the database;
[0031] constructing a regression prediction model using the XGBoost algorithm and setting relevant parameters;
[0032] training and validating the model using five-fold cross-validation method;
[0033] adjusting the model parameters and optimizing the model performance according to the validation results.
[0034] An artificial intelligence-based blood concentration prediction system, comprising:
[0035] a data acquisition module for acquiring basic information of patients, sirolimus blood concentration and genetic polymorphism data;
[0036] a data processing module for preprocessing and feature extraction of the acquired data;
[0037] a prediction model module loading a sirolimus blood concentration artificial intelligence prediction model constructed according to the method of any one of claims 1 to 5;
[0038] a prediction result display module for displaying the predicted sirolimus blood concentration results.
[0039] Optionally, it further comprises a mobile device application connected with the prediction system through wireless communication, realizing remote data acquisition and prediction result display.
[0040] Optionally, the prediction result display module further comprises a warning function, which automatically sends warning information when the predicted sirolimus blood concentration exceeds the safe range.
[0041] Beneficial effects: First, by combining liquid chromatography tandem mass spectrometry with high-resolution dissolution curve technology, accurate detection of sirolimus blood concentration and key genetic polymorphisms is realized, providing a reliable data basis for model construction.
[0042] Secondly, the prediction model constructed by using the extreme gradient boosting algorithm can efficiently and accurately predict the sirolimus blood concentration, significantly reducing the dependence on expensive detection equipment, thereby greatly reducing the treatment cost of patients.
[0043] In addition, through the application of the intelligent mobile device, the model realizes real-time dynamic monitoring and early warning of blood concentration, providing timely and accurate treatment adjustment basis for clinicians, which helps to improve the treatment effect and reduce the occurrence of adverse reactions.
[0044] Finally, the implementation of the present application not only optimizes the treatment plan for children with KHE, but also provides new ideas and methods for the treatment of other similar diseases, with wide application prospects and social and economic benefits. BRIEF DESCRIPTION OF DRAWINGS
[0045] Figure 1 The data collection and detection flowchart of the embodiment of the present application is shown in the figure.
[0046] Figure 2 The database construction and preprocessing flowchart of the embodiment of the present application is shown in the figure.
[0047] Figure 3 The model training and optimization flowchart of the embodiment of the present application is shown in the figure.
[0048] Figure 4 The prediction and application system workflow chart of the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION
[0049] The preferred embodiments of the present application are described below with reference to the accompanying drawings, which make the technical content of the present application clearer and easier to understand. The present application can be embodied in many different forms, and the scope of protection of the present application is not limited to the embodiments described herein.
[0050] The present application relates to the cross field of biological medicine and artificial intelligence, and proposes an innovative blood drug concentration prediction model based on artificial intelligence technology, especially for real-time dynamic monitoring and accurate prediction of blood drug concentration of children with Kaposiform hemangioendothelioma (KHE) during the treatment with sirolimus. This model aims to solve the problems of current sirolimus blood drug concentration monitoring technology, such as high difficulty, expensive equipment and difficulty in popularization, and provides a more convenient, economical and efficient solution for clinical treatment.
[0051] Kaposiform hemangioendothelioma is a rare and invasive vascular tumor, mainly affecting infants and children. Although its incidence is low, the disease is serious and often accompanied by Kasabach-Merritt phenomenon (KMP), resulting in high morbidity and mortality. Sirolimus is an effective drug for treating KHE, and its efficacy is significant, but the side effects of the drug cannot be ignored, and the incidence and severity of side effects are closely related to blood drug concentration. At present, the monitoring of sirolimus blood drug concentration mainly relies on high-end technologies such as liquid chromatography tandem mass spectrometry (LC / MS-MS). Although these technologies are accurate and reliable, they are expensive and complex to operate, and are difficult to popularize in primary hospitals, resulting in high cost of patient treatment and reduced treatment compliance.
[0052] In view of the above background, the present application proposes a blood concentration prediction model combining genetic polymorphism detection and artificial intelligence algorithm. The model collects whole blood samples of KHE children after taking sirolimus, detects blood drug concentration using LC / MS-MS technology, and detects polymorphism of CYP 3A4 / 5, CYP 2C8, P-gp and other key genes using high-resolution melting curve (HRM) technology. These genetic polymorphisms are known to have a significant impact on the pharmacokinetics of sirolimus, so they are included in the prediction model, which helps improve the accuracy of the prediction.
[0053] In the data collection and preprocessing stage, the present application first establishes a standardized sample collection and processing procedure. After the whole blood sample is collected, blood drug concentration detection and genetic polymorphism analysis are immediately performed. Blood drug concentration detection uses LC / MS-MS technology, which includes preparing standard stock solution and working solution, preparing whole blood standard curve and quality control solution, sample pretreatment (such as protein precipitation, organic solvent extraction), etc. to ensure the accuracy and reliability of the detection results. Genetic polymorphism detection is performed by extracting whole blood DNA, designing and synthesizing primers for target genes, performing PCR amplification, verifying PCR products using gel electrophoresis, and finally analyzing melting curves using HRM technology to determine genotypes.
[0054] After completing data collection, the present application organizes and stores the data, establishing a structured database. The database not only contains blood drug concentration and genetic polymorphism data, but also integrates basic information of the children (such as age, gender, weight) and treatment records (such as drug dosage, drug administration time), providing comprehensive data support for subsequent feature extraction and model construction.
[0055] Feature extraction and dimensionality reduction are key steps in model construction. The present application extracts feature variables related to sirolimus blood concentration from the database, including genetic polymorphism, age, weight, drug dosage, and drug administration time. Considering the overfitting problem that may be caused by high-dimensional data, the present application applies the least absolute shrinkage and selection operator (LASSO) algorithm to reduce the dimensionality of the feature data. LASSO algorithm can automatically select feature variables that have a significant impact on blood drug concentration prediction by introducing L1 regularization term, reducing model complexity and improving prediction accuracy.
[0056] In the prediction model construction stage, the present application selects the extreme gradient boosting algorithm (XGBoost) to construct a regression prediction model. XGBoost is an ensemble learning algorithm based on decision trees, which has the advantages of high efficiency, accuracy, and anti-overfitting, and is especially suitable for processing large-scale data sets. The present application uses Python 3.6 as the platform, based on the scikit-learn machine learning algorithm library, and uses the XGBRegressor function to construct an XGBoost regression prediction model. In the model training process, the five-fold cross-validation method is used to divide the data set into training set and validation set, and by continuously adjusting the model parameters (such as learning rate, maximum depth of tree, and proportion of subsample, etc.), the model performance is optimized. At the same time, the present application uses mean square error (MSE), square root mean square error (RMSE) and R² coefficient as evaluation indexes to evaluate the model performance, to ensure the accuracy and reliability of the prediction results.
[0057] In order to reveal the key factors affecting the blood concentration of sirolimus, the present application also analyzes the importance of each feature in the model by using SHAP value (SHapley Additive exPlanation value). SHAP value is based on the idea of game theory, which can quantify the contribution of each feature to the prediction result, and provide a scientific basis for clinical treatment. Through feature importance analysis, the present application finds that genetic polymorphism, dosage and body weight are the main factors affecting the blood concentration of sirolimus, which is consistent with the existing research results, further verifying the accuracy of the model.
[0058] In the prediction software design and development stage, the present application designs a set of blood concentration prediction software architecture based on artificial intelligence. The architecture includes data acquisition module, data processing module, prediction model module and prediction result display module. The data acquisition module is responsible for reading data from the database, the data processing module performs cleaning, desensitization and standardization processing on the data, the prediction model module loads the trained XGBoost model for prediction, and the prediction result display module presents the prediction result to the user in a visual form. In order to facilitate the operation and use of medical staff, the present application uses Visual Studio C++ to develop a human-computer interaction interface, which is simple and clear in design, and reasonable in function layout. At the same time, the present application also develops a mobile device end application, which is connected with the prediction system through wireless communication, realizing remote data acquisition and prediction result display. Medical staff can check the blood concentration prediction results of patients at any time and anywhere through mobile devices such as mobile phones or tablets, and adjust the treatment plan in time.
[0059] The early warning function is a highlight of the present application. In the prediction software, the present application sets up an early warning mechanism. When the predicted sirolimus blood concentration exceeds the safe range, the system automatically sends an early warning message to remind medical staff to adjust the dosage or take other measures in time. The early warning message is pushed to medical staff through short message, email or application notification, etc. to ensure the timely transmission and processing of the early warning information. This function helps to discover and handle potential drug overdose or deficiency in time, reduces the occurrence of adverse reactions, and improves the safety of patient treatment.
[0060] The technical innovations of the present application mainly lie in the following aspects: 1. Multi-dimensional data fusion. The present application fuses blood concentration detection data with genetic polymorphism data, patient basic information and other multi-dimensional data, improving the accuracy and reliability of the prediction model. 2. Application of artificial intelligence algorithm. The present application introduces the extreme gradient boosting algorithm (XGBoost) to build the prediction model, which automatically learns the data features by using machine learning technology, realizing the accurate prediction of sirolimus blood concentration. 3. Real-time dynamic monitoring. Through the mobile device application, real-time dynamic monitoring and early warning of sirolimus blood concentration are realized, providing timely and accurate data support for clinical treatment. 4. Low cost and easy to promote. The prediction model of the present application is based on conventional detection equipment and artificial intelligence algorithm, without the need for expensive special equipment, which is low in cost and easy to promote and apply in primary hospitals.
[0061] In the implementation process, the present application relies on the experimental base of West China Hospital of Sichuan University, equipped with conventional experimental equipment such as liquid chromatography tandem mass spectrometer, PCR amplifier, gel electrophoresis instrument, etc. A multidisciplinary team composed of biomedical basic experimental personnel, computer professionals and medical related personnel is formed. Team members perform their respective duties and work together to complete data collection, model building, software development and optimization, etc. At the same time, the present application selects mature database management systems (such as MySQL) and artificial intelligence algorithm libraries (such as scikit-learn), ensuring the smooth progress of data storage, query and analysis, and algorithm development and verification.
[0062] Through the implementation of the present application, real-time dynamic monitoring and prediction of blood concentration of KHE children during the treatment with sirolimus can be realized, providing timely and accurate data support for clinical treatment. This achievement not only helps to improve the treatment effect, reduce the treatment cost and reduce the occurrence of adverse reactions, but also promotes the development of personalized treatment. In the future, the present application will further optimize the model performance, expand the application range, and provide strong support for the treatment of more diseases.
[0063] The foregoing merely illustrates the principles of the application and application of its leading features. This application is not limited to the illustrative embodiments shown and described herein. Rather, this application is capable of operating within a further range of conditions and environments than those specifically described herein, and further modifications can be made without departing from the spirit or scope of the application. Accordingly, the description is to be construed as illustrative only and not restrictive of the broad disclosure or application of the application. The specification and drawings are, accordingly, to be regarded simply as illustrative and with the scope of the application being measured by the appended claims, and not with the specification. No admission is made that any reference constitutes prior art. It is my intent, therefore, to be limited only as appears in the following claims.
[0064] Furthermore, it should be understood that although the description above relates to embodiments, not every embodiment contains only one independent technical solution, and the description above is only for the sake of clarity, and those skilled in the art should understand the description as a whole, and the technical solutions in each embodiment can be appropriately combined to form other embodiments that those skilled in the art can understand.
Claims
1. A method of constructing an artificial intelligence-based blood drug concentration prediction model, characterized by, The method comprises the following steps: Collecting whole blood samples of patients after taking sirolimus, and detecting the sirolimus blood concentration by using liquid chromatography tandem mass spectrometry; Extracting whole blood DNA, and detecting the CYP 3A4 / 5, CYP 2C8 and P-gp gene polymorphisms by using high-resolution melting curve technology; Constructing a database containing the basic information of patients, the sirolimus blood concentration and the gene polymorphism data; Training the data in the database by using the extreme gradient boosting algorithm to establish an artificial intelligence prediction model of the sirolimus blood concentration; Verifying and optimizing the prediction model to ensure its prediction accuracy.
2. The construction method of claim 1, wherein, The step of detecting the sirolimus blood concentration by using liquid chromatography tandem mass spectrometry comprises: Preparing standard stock solution and working solution; Preparing whole blood standard curve and quality control solution; Pretreating the whole blood sample, including the steps of protein precipitation and organic solvent extraction; Using LC / MS-MS for detection to obtain the sirolimus blood concentration data.
3. The construction method of claim 1, wherein, The step of detecting the gene polymorphism by using high-resolution melting curve technology comprises: Extracting whole blood DNA; Designing and synthesizing primers for CYP 3A4 / 5, CYP 2C8 and P-gp genes; Performing PCR amplification to obtain target gene fragments; Using HRM technology to analyze the melting curve of PCR products to determine the gene polymorphism.
4. The construction method of claim 1, wherein, The construction of the database comprises: Collecting and organizing the basic information of patients, treatment records, sirolimus blood concentration and gene polymorphism data; Importing the data into a database management system for data cleaning, desensitization and labeling; Establishing a structured database to facilitate subsequent data analysis and model construction.
5. The construction method of claim 1, wherein, The step of training the prediction model by using the extreme gradient boosting algorithm comprises: Pretreating and extracting features from the data in the database; Using the XGBoost algorithm to construct a regression prediction model and setting relevant parameters; Training and verifying the model by using five-fold cross-validation method; Adjusting the model parameters according to the verification results to optimize the model performance. 6.A blood concentration prediction system based on artificial intelligence, characterized by, It comprises: A data acquisition module for acquiring the basic information of patients, the sirolimus blood concentration and the gene polymorphism data; A data processing module for pretreating and extracting features from the acquired data; A prediction model module for loading the sirolimus blood concentration artificial intelligence prediction model constructed according to the method of any one of claims 1 to 5; A prediction result display module for displaying the predicted sirolimus blood concentration results.
7. The prediction system of claim 6, wherein, It further comprises a mobile device application connected with the prediction system through wireless communication to realize remote data acquisition and prediction result display.
8. The prediction system of claim 6, wherein, The prediction result display module further comprises a warning function for automatically sending warning information when the predicted sirolimus blood concentration exceeds the safe range.