Method and model for predicting blood concentration of personalized medication and electronic equipment
The blood drug concentration prediction model constructed by various optimization algorithms and Gibbs sampling solves the problem of inaccurate blood drug concentration prediction in personalized medicine, and achieves higher prediction accuracy and medication safety.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-03
- Publication Date
- 2026-03-10
Smart Images
Figure CN121641501A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of blood concentration prediction, in particular to a blood concentration prediction method and model for personalized medication and an electronic device. BACKGROUND
[0002] Personalized medication research needs to be deeply mined and innovated in data collection, analysis methods, model construction, and special population research, etc. to overcome the inaccuracy of personalized medication and promote the development of precision medicine.
[0003] The technical challenges in the current personalized medication research field are quite complex. The physical indicators, medical history and combined medication of patients are different, so the blood concentrations of different patients are different. The accuracy of the prediction of the blood concentration by the doctor's experience and the existing blood concentration model is limited, and more accurate blood concentration prediction is needed to improve the accuracy of the doctor's medication. SUMMARY
[0004] The purpose of the present application is to solve the above problems, provide a blood concentration prediction method and model for personalized medication and an electronic device to improve the stability and accuracy of individual pharmacokinetic parameters, and ultimately improve the accuracy of blood concentration prediction.
[0005] To solve the above problems, the present application provides the following technical scheme: In a first aspect, a blood concentration prediction method for personalized medication includes: Obtain a clinical data set, which is divided into a test data set and a training set; The training set uses at least two optimization algorithms to calculate pharmacokinetic parameters; According to the blood concentration prediction values obtained by each optimization algorithm and the actual blood concentration values, the prediction effect accuracy corresponding to each optimization algorithm is determined; According to the weight of each prediction effect accuracy in the sum of all prediction effect accuracies and the pharmacokinetic parameters of the corresponding optimization algorithm, the individual pharmacokinetic parameters are determined; and the blood concentration prediction model is trained according to the individual pharmacokinetic parameters; The test data set is input into the trained blood concentration prediction model to obtain individual blood concentration prediction values.
[0006] In related embodiments, before the clinical data set is used, the core indicators of the clinical data set are optimized, and the optimization method includes: The data in the clinical data set is selected by random forest feature importance screening, and the data with importance greater than a preset importance threshold is selected; The data in the clinical data set is selected by correlation test with the target pharmacokinetic parameters, and the data with a correlation coefficient greater than a preset threshold is selected; The data satisfying the feature importance screening, the correlation test and the statistical result having significance are the core indicators; and the core indicators form an original data set.
[0007] In related embodiments, the optimization method further includes optimizing the dimension difference of the core indicators in the original data set, which includes: When the core indicators in the original data set are skewed, the core indicators are subjected to natural logarithmic conversion; When the core indicators in the original data set are normal, the core indicators are subjected to standardization processing to eliminate the dimension difference.
[0008] In related embodiments, the optimization method further includes optimizing the singular value of the core indicators in the original data set, which includes: The core indicators in the original data set are subjected to Z-score detection to obtain the singular value, and the representative data set is obtained after removing the singular value.
[0009] In related embodiments, the optimization method further includes optimizing the missing value of the original data set or the representative data set, which includes: When the data distribution trend in the original data set or the representative data set is significant, a fitting equation is constructed, the predicted value is determined according to the missing indicator value and the fitting equation, and the predicted value is used as the interpolation value to fill the missing place; When the data distribution trend in the original data set or the representative data set is not significant, but the correlation coefficient between the indicators is greater than or equal to a preset value, a regression prediction equation is constructed, the prediction accuracy is calculated; when the prediction accuracy meets the preset condition, the interpolation value is calculated according to the missing indicator and the regression prediction equation, and the interpolation value is filled in the missing place.
[0010] In related embodiments, before the clinical data set is used, the sample category number balance optimization of the clinical data set is further included, which includes generating new samples by interpolation in the minority category samples to match the number of samples of each category.
[0011] In related embodiments, the clinical data set includes pharmacokinetic parameters, patient numbers, demographic information data, drug measurement data, medical history data, test indicator data, combined medication data, genotype data.
[0012] In related embodiments, the prediction effect accuracy is the average percentage error of the blood drug concentration predicted value and the actual blood drug concentration value.
[0013] The second aspect is a blood drug concentration prediction model for individualized medication, which includes: A data acquisition module is configured to acquire a clinical data set; A data optimization module is configured to optimize the clinical data set and execute the prediction method; The pharmacokinetic parameter calculation module is configured to calculate individual pharmacokinetic parameters, and the training set is configured to calculate the pharmacokinetic parameters by using at least two optimization algorithms; the prediction effect accuracy of each optimization algorithm is determined according to the blood drug concentration predicted value obtained by each optimization algorithm and the actual blood drug concentration value; and the individual pharmacokinetic parameters are determined according to the pharmacokinetic parameters of the corresponding optimization algorithm in the weight of all prediction effect accuracies and the prediction effect accuracy. The blood drug concentration prediction module is configured to input the to-be-tested data set into the trained blood drug concentration prediction model to obtain individual blood drug concentration predicted values.
[0014] In a third aspect, an electronic device includes a memory and a processor, the memory stores computer readable instructions, and the computer readable instructions are executed by the processor to make the processor execute the blood drug concentration prediction method for personalized medication.
[0015] Compared with the prior art, the present application has the following beneficial effects: The present application uses at least two optimization algorithms to calculate the pharmacokinetic parameters and the prediction effect accuracy of the blood drug concentration predicted value calculated by each optimization algorithm and the actual value, and the prediction effect accuracy is used as a weight to optimize different pharmacokinetic parameters, so as to avoid the volatility and local optimal solution of a single method, which helps to improve the stability and accuracy of individual pharmacokinetic parameters, and finally improves the accuracy of blood drug concentration prediction. BRIEF DESCRIPTION OF DRAWINGS
[0016] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the drawings needed in the specific embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor. Figure 1 It is a flowchart of the embodiment.
[0017] Figure 2 It is a flowchart of individual pharmacokinetic parameter calculation.
[0018] Figure 3 It is a specific process of Gibbs sampling.
[0019] Figure 4 It is a schematic diagram of the blood drug concentration prediction model for personalized medication.
[0020] Figure 5 It is a schematic diagram of the electronic device. DETAILED DESCRIPTION
[0021] In order to make the purpose, technical scheme and advantages of the present application clearer, the following will combine the present application with specific embodiments and prior art. Figures 1 to 5Further detailed description of the present application, the described embodiments should not be considered as limiting the present application, all other embodiments obtained by those skilled in the art without doing creative work, within the scope of the present application.
[0022] The present application is described below Figure 1 And Figure 2 The personalized medicine blood concentration prediction method of the present application is described, comprising: Step 100, obtaining a clinical data set, divided into a test data set and a training set; The clinical data set includes pharmacokinetic parameters, patient number, demographic information data, drug measurement data, medical history data, test index data, combined drug data, and genotype data; can be obtained from clinical data or database.
[0023] For example, the clinical data set of escitalopram personalized medicine is as follows:
[0024] The fields, field names and explanations in the above table are selected according to the specific medication; the above table lists the clinical data set of escitalopram personalized medicine; whether the infection is specific to whether there is an accompanying infectious disease, such as upper respiratory tract infection, C-reactive protein (CRP) elevation, white blood cell elevation or fever, etc.; the pharmacokinetic parameters in this embodiment include clearance CL and apparent volume of distribution V.
[0025] Step 200, the training set is calculated by at least two optimization algorithms; optionally, differential evolution algorithm (Differential Evolution, DE), genetic algorithm (Genetic Algorithm, GA), particle swarm optimization (Particle Swarm Optimization, PSO) are used to calculate the pharmacokinetic parameters, and the objective function is set to evaluate the difference between the predicted blood concentration and the actual value; iteration process: by setting the population size and iteration times, parameter extraction and combination are carried out, until the optimal solution of the pharmacokinetic parameters is found; other evolutionary algorithms can also be selected, and two or more pharmacokinetic parameters with optimal solution obtained by the above method are selected.
[0026] Step 300, the blood concentration prediction value obtained according to each optimization algorithm is respectively determined with the actual value of blood concentration, the prediction effect precision corresponding to each optimization algorithm is determined; in order to evaluate the parameter prediction precision of the three algorithms of differential evolution algorithm, genetic evolution algorithm and particle swarm algorithm obtained in step 200, and calculate the weight of each algorithm, the fitness function is used to take the difference between the blood concentration prediction value and the actual measured value as the evaluation standard, the target is to minimize the difference, according to this requirement, the mean absolute percentage error (MAPE) is selected to measure the prediction effect precision of each patient, the parameters calculated by iteration are substituted into the pharmacokinetic model; the mean absolute percentage error calculation formula is as follows: MAPE= ×100% (1-1) In the formula, is the true value of the ith blood concentration, is the prediction value of the ith blood concentration.
[0027] Step 400, according to the weight of all prediction effect precisions and the pharmacokinetic parameters corresponding to the optimization algorithm, the individual pharmacokinetic parameters are determined; specifically, the mean absolute percentage error (MAPE) method is used to calculate the iteration precision of the three algorithms of differential evolution algorithm, genetic evolution algorithm and particle swarm algorithm in step 200, respectively, MAPE DE , MAPE GA , MAPE PSO , which is used as the weight of the three ways of iteration results in the final solution: (1-2) In the formula, MAPE i is the mean absolute percentage error of the ith method, W i represents the weight of the iteration result of the ith way in the final solution, i=DE, GA, PSO.
[0028] The individual pharmacokinetic parameters are: P=W DE P DE +W GA P GA +W PSO P PSO (1-3) In the formula, P is the individual pharmacokinetic parameter, P DE , P GA , P PSO are the iteration results of the three algorithms of differential evolution algorithm, genetic evolution algorithm and particle swarm algorithm, , , The weight of the iteration result of each of the differential evolution algorithm, the genetic evolution algorithm, and the particle swarm algorithm is in the individual pharmacokinetic parameter.
[0029] The blood drug concentration prediction model is trained according to the individual pharmacokinetic parameter P, and the blood drug concentration prediction model is determined according to the individual pharmacokinetic parameter P.
[0030] The blood drug concentration prediction model is trained according to the individual pharmacokinetic parameter P, and the blood drug concentration prediction model is determined according to the individual pharmacokinetic parameter P. In an embodiment, as shown in Figure 3 , the random effect model is determined by sampling from the random effect multi-dimensional probability distribution, and specifically as follows: Gibbs sampling (Gibbs sampling), the result analysis: the combination of random disturbance items is determined by Gibbs sampling, and the mean and variance of the sampling result are analyzed. As shown in Figure 3 , the specific process of Gibbs sampling is as follows: (1) Select the initial state: clear the rate random effect item η CL , the apparent distribution volume random effect item η V , and the residual random effect item ε for each variable, and select an initial value, which constitutes the initial sample point set.
[0031] (2) Iterative update: a. Loop through each variable: for each variable, update in a certain order (which can be fixed or random). In each update, fix the values of other variables and only update the current variable.
[0032] b. Sample the current variable: according to the current value of the other fixed variable, draw a new value from the conditional distribution p(η CL ,η V ,ε), which is the core of Gibbs sampling, which uses conditional distribution to sample.
[0033] c. Repeat until convergence: repeat step (2) until the result is stable, i.e. reach the required convergence condition, the distribution of the sample no longer changes significantly, or reach the predetermined number of iterations.
[0034] (3) Convergence test: check whether the algorithm has converged by various diagnostic methods. If it does not converge, more iterations are needed until the preset convergence condition is met.
[0035] (4) Remove the burn-in period: the initial stage of Gibbs sampling may be very dependent on the initial value, and the initial samples may not be representative of the target distribution. Discard them, which is called "burn-in period" processing, which makes the algorithm "forget" its initial state.
[0036] (5) Distribution characteristics: the various statistical quantities of the target distribution estimated by the converged samples, such as mean, variance, probability, etc.
[0037] Evaluation of the effect of the trained model For example, referring to the above individual pharmacokinetic parameters and the eslicarbazepine blood concentration prediction model preset by Gibbs training, the in vivo blood concentration is predicted, and the model effect is evaluated by comparing the predicted value with the actual value. In order to comprehensively evaluate the performance of the model under different accuracy standards, the distribution proportion of the model results under different accuracy thresholds can be calculated respectively. This method has a double effect: on the one hand, it helps to understand the performance of the model under different accuracy requirements, thereby evaluating its prediction accuracy; on the other hand, by observing the behavior of the model under different accuracy standards, it can conduct a more in-depth error analysis. For example, if the performance of the model decreases sharply under a higher accuracy threshold, it may indicate that the model has problems in processing certain specific patient data or situations, which needs to be further explored and improved. At the same time, it needs to be noted that improving the accuracy of the model may bring additional costs, such as increased computing resources and time. By analyzing the result proportion under different accuracy, we can balance the relationship between performance improvement and cost increase, providing decision support for the further optimization and application of the model.
[0038] Step 500, input the to-be-tested data set into the trained blood concentration prediction model to obtain individual blood concentration prediction values.
[0039] This embodiment uses three algorithms to calculate pharmacokinetic parameters and the prediction effect precision of the blood concentration prediction values calculated by each algorithm, and uses the prediction effect precision as a weight to weight different pharmacokinetic parameter optimization algorithms, in order to avoid the volatility and local optimal solution of a single method, which helps to improve the stability and accuracy of individual pharmacokinetic parameters, and ultimately improve the accuracy of concentration prediction.
[0040] In an embodiment, before using the clinical data set, the core indicators of the clinical data set are also optimized, and the optimization method includes: Step 111, the data in the clinical data set is screened by random forest feature importance, and the data with importance greater than the preset importance threshold is selected; Step 112, the data in the clinical data set is tested for correlation with the target pharmacokinetic parameter, and the data with a correlation coefficient greater than the preset threshold is selected; Step 113, the data that simultaneously satisfies the feature importance screening, the correlation test and the statistical result with significance is the core indicator; the core indicators constitute the original data set.
[0041] Core indicators refer to biomedical indicators that are highly correlated with key pharmacokinetic parameters (such as clearance rate CL, distribution volume V, etc.) and play a decisive role in the prediction accuracy of the model. For example, the screening criteria of the present embodiment can be feature importance > 0.1, which screens out core indicators with a contribution to key pharmacokinetic parameters of more than 10%, and a correlation coefficient (|r| > 0.3, indicating a certain degree of linear correlation between two variables) with the target pharmacokinetic parameter and significant statistical results (P < 0.05) of the core indicators, so as to screen out core indicators In an embodiment, the optimization method further comprises optimizing the dimension difference of the core indicators in the original data set, which comprises: Step 121, when the core indicators in the original data set are skewed, the core indicators are subjected to natural logarithmic transformation; the specific process is as follows: Determine the distribution type of the indicators by Shapiro-Wilk normality test: if the index distribution is skewed (P < 0.05), such as alanine aminotransferase (ALT) and aspartate aminotransferase (AST), natural logarithmic transformation is performed to make the data closer to normal.
[0042] Step 122, when the core indicators in the original data set are normally distributed, the core indicators are subjected to standardization processing to eliminate the dimension difference; the specific process is as follows: Determine the distribution type of the indicators by Shapiro-Wilk normality test: if the index distribution is approximately normal (P ≥ 0.05), such as age and weight, standardization processing is performed to eliminate the dimension difference.
[0043] In an embodiment, the optimization method further comprises optimizing the singular value of the core indicators in the original data set, which comprises: Step 131, the core indicators in the original data are subjected to Z-score detection to determine the singular value, and after removing the singular value, a representative data set is obtained; the specific process is as follows: Singular value discrimination based on Z-score method: Z-score (standard score) is used to quantify the deviation of data points from the mean, and the formula is: Wherein, x is a single observation value of the core indicator, is the mean of all valid observation values of the index, is the standard deviation of all valid observation values of the index. Discrimination rule: if the absolute value of Z of a certain observation value is greater than 3 (|Z| > 3), it is determined as a singular value (an outlier), which is removed from the data set; if the absolute value of Z is less than or equal to 3 (|Z| ≤ 3), it is retained as valid data.
[0044] After Z-score test is performed on all core indicators one by one, samples with no singular value in all core indicators are screened out to form a representative data set.
[0045] In an embodiment, the optimization method further comprises missing value optimization of the original data set or the representative data set, which comprises: Step 141, when the data distribution trend in the original data set or the representative data set is significant, a fitting equation is constructed, a predicted value is determined according to the missing index value and the fitting equation, and the predicted value is filled in the missing place as an imputation value; the specific process is as follows: Trend significance judgment: for the effective data (without missing and non-singular value) of the core index, the time / dose number is taken as the independent variable t, and the index value is taken as the dependent variable y, and a trend curve (such as linear, quadratic function) is fitted; the determination coefficient R2 is calculated Judge the trend, if ≥0.6, it is considered that “the data distribution trend is significant”, otherwise it is considered not significant.
[0046] Constructing trend extrapolation equation: if it is a linear trend, the fitting equation is: (a is the slope, b is the intercept, and t is the dependent variable); if it is a nonlinear trend, a quadratic equation is fitted: . The equation parameters (a, b, c) are fitted by the least square method (to make the residual sum of squares of the actual value and the predicted value minimum).
[0047] Missing value imputation: for the missing index value, the corresponding t is substituted into the trend equation to calculate the predicted value , and the is taken as the imputation value to fill in the missing place.
[0048] Step 142, when the data distribution trend in the original data set or the representative data set is not significant, but the correlation coefficient between indexes is greater than or equal to a preset value, a regression prediction equation is constructed, the prediction accuracy is calculated, and when the prediction accuracy meets the preset condition, the imputation value is calculated according to the missing index and the regression prediction equation, and the imputation value is filled in the missing place; the specific process is as follows: Correlation test: for the core index Y with “not significant distribution trend <0.6)”, the Pearson correlation coefficient of the core index Y with other core indexes is calculated, here the core index is any one or more of pharmacokinetic parameters, patient number, demographic information data, drug measurement data, medical history data, test index data, combined drug data, and genotype data. If there is a core index with Y (|r|≥0.5) (and P<0.05), it is considered that “Y has strong correlation with ”.
[0049] Constructing regression prediction equation: taking as the independent variable and Y as the dependent variable, a linear regression equation is constructed: (c is the regression coefficient, d is the intercept). Fitted by least squares, and calculate the prediction accuracy (such as the coefficient of determination ), the mean absolute percentage error MAPE).
[0050] Interpolation judgment: if the regression prediction accuracy meets ≥ 0.6 (or MAPE ≤ 20%), equivalent to "prediction accuracy greater than 85%", the missing value of Y is substituted into the corresponding observation value in the equation, and the calculated is taken as the interpolation value; if the accuracy is insufficient, it is not interpolated.
[0051] Step 143, when the distribution trend is not significant, and the correlation between indicators is weak, then do not interpolate if the core indicator is both "distribution trend is not significant < 0.6 ", and "correlation with other core indicators is weak (all (| r | < 0.5) or (P ≥ 0.05)) ", do not interpolate the missing value, directly keep the null value, and the subsequent pharmacokinetic equation training is completed by the built-in missing value processing mechanism of the model (such as Gibbs sampling).
[0052] The above embodiment designs a method of using Gibbs sampling to construct a random effect model to handle unmeasurable factors in clinical data. The joint posterior distribution is explored by Gibbs three-dimensional sampling, and a sample set is generated therefrom to determine the best combination of random disturbance terms. This method is suitable for handling the nested random effect structure in pharmacokinetic models.
[0053] The above embodiment uses random forest to screen possible prediction factors, and combines the actual situation of high-dimensional sparse data to realize individualized blood concentration prediction, making drug recommendations more accurate. The influence of classification variables such as gender on blood concentration is determined by variance analysis to reasonably divide the population and balance the samples, improve the prediction accuracy, and at the same time ensure the medical interpretability and statistical rationality.
[0054] In an embodiment, before using the clinical data set, the sample category number balance optimization of the clinical data set is also included, which includes generating new samples by interpolation in the minority class samples to match the number of samples in each category; the SMOTE algorithm can be selected.
[0055] The core indicator screening optimization, core indicator dimension difference optimization, Z-score detection of singular values, missing value optimization, and sample category number balance optimization in the above optimization method can be selected in combination with two or more to optimize the sampling data; through each optimization method, the quality of the final data in the clinical data set is greatly improved, and the accuracy of the model measurement is improved.
[0056] The blood drug concentration prediction method for personalized medicine has the following advantages: 1. Improve medication accuracy: By combining clinical pharmacology and statistical methods, the constructed personalized medication model can more accurately predict the effects of drugs in different populations, thereby improving the accuracy and reliability of medication.
[0057] 2. Optimize initial medication dosage: The model can determine the initial medication dosage based on patient characteristics, reducing reliance on physician experience, improving computational efficiency, and providing patients with a more reasonable initial medication dosage.
[0058] 3. Adjusting personalized medication: The model can adjust the recommended dosage for patients based on the actual measured blood drug concentration data to ensure that the blood drug concentration is within a reasonable range, thus achieving personalized medication recommendations.
[0059] 4. Assisting clinical decision-making: Personalized dosing regimens will assist clinicians in making more accurate medication decisions, improving the safety and effectiveness of drug therapy.
[0060] In one embodiment, such as Figure 4 As shown, a blood drug concentration prediction model for personalized medication includes: The data acquisition module 601 is used to collect clinical datasets, including pharmacokinetic parameters, patient IDs, demographic information, drug measurement data, medical history data, laboratory test data, combined medication data, and genotype data.
[0061] The data optimization module 602 is used to optimize the clinical dataset and execute the prediction method; it is used to execute at least one of the optimization methods in steps 111 to 113, steps 121 to 122, step 133, and steps 141 to 143.
[0062] The pharmacokinetic parameter calculation module 603 is used to calculate individual pharmacokinetic parameters. The training set uses at least two optimization algorithms to calculate pharmacokinetic parameters and is used to execute steps 200 to 400. Based on the predicted blood drug concentration values obtained by each optimization algorithm and the actual blood drug concentration values, the prediction accuracy of each algorithm is determined. Based on the weight of each prediction accuracy in the sum of all prediction accuracy and the pharmacokinetic parameters of the corresponding optimization algorithm, the individual pharmacokinetic parameters are determined. The blood drug concentration prediction module 604 is used to input the dataset to be tested into the trained blood drug concentration prediction model to obtain the individual blood drug concentration prediction value; and to execute step 500.
[0063] In one embodiment, such as Figure 5As shown, an electronic device is provided, including a memory and a processor. The memory stores computer-readable instructions, which, when executed by the processor, cause the processor to perform a blood drug concentration prediction method for personalized medication.
[0064] Electronic devices can be desktop computers, laptops, handheld computers, and cloud servers, among other electronic devices. Electronic devices may include, but are not limited to, processors and memory. Those skilled in the art will understand that the figures are merely examples of electronic devices and do not constitute a limitation on the electronic device; it may include more or fewer components than illustrated, or different components.
[0065] The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
[0066] Memory can be an internal storage unit of an electronic device, such as a hard drive or RAM. Memory can also be an external storage device, such as a plug-in hard drive, SmartMediaCard (SMC), Secure Digital (SD) card, or FlashCard. Memory can also include both internal and external storage units. Memory is used to store computer programs and other programs and data required by the electronic device.
[0067] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0068] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0069] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0070] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the scope of protection of the invention. It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0071] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A blood drug concentration prediction method for personalized medicine, characterized by, comprising obtaining a clinical data set, which is divided into a test data set and a training set; the training set is calculated by at least two optimization algorithms to obtain pharmacokinetic parameters; the predicted blood concentration values obtained by each optimization algorithm are compared with the actual blood concentration values to determine the prediction accuracy of each optimization algorithm; based on the weight of each prediction accuracy in the sum of all prediction accuracies and the pharmacokinetic parameters of the corresponding optimization algorithm, the individual pharmacokinetic parameters are determined; and the blood concentration prediction model is trained based on the individual pharmacokinetic parameters; the test data set is input into the trained blood concentration prediction model to obtain individual blood concentration prediction values.
2. The prediction method of claim 1, wherein, Before the clinical data set is used, the core indicators of the clinical data set are screened and optimized, and the optimization method includes: the data in the clinical data set is screened by random forest feature importance, and the data with importance greater than a preset importance threshold is selected; the data in the clinical data set is tested for correlation with the target pharmacokinetic parameters, and the data with a correlation coefficient greater than a preset threshold is selected; the data that meets the feature importance screening, correlation test and statistical results with significant results at the same time is the core indicator; the core indicator constitutes an original data set.
3. The prediction method of claim 2, wherein, The optimization method further includes optimization of the dimension difference of the core indicators in the original data set, which includes: when the core indicators in the original data set are skewed, the core indicators are subjected to natural logarithmic conversion; when the core indicators in the original data set are normal, the core indicators are subjected to standardization processing to eliminate the dimension difference.
4. The prediction method of claim 2, wherein, The optimization method further includes singular value optimization of the core indicators in the original data set, which includes: singular values are detected in the core indicators in the original data set by Z-score, and after removing the singular values, a representative data set is obtained.
5. The prediction method according to any one of claims 2-4, characterized in that, The optimization method further includes missing value optimization of the original data set or the representative data set, which includes: when the data distribution trend in the original data set or the representative data set is significant, a fitting equation is constructed, the predicted value is determined according to the missing indicator value and the fitting equation, and the predicted value is used as the interpolation value to fill in the missing place; when the data distribution trend in the original data set or the representative data set is not significant, but the correlation coefficient between the indicators is greater than or equal to a preset value, a regression prediction equation is constructed, the prediction accuracy is calculated; when the prediction accuracy meets the preset condition, the interpolation value is calculated according to the missing indicator and the regression prediction equation, and the interpolation value is filled in the missing place.
6. The prediction method of claim 1, wherein, Before the clinical data set is used, the sample category number balance optimization of the clinical data set is also included, which includes generating new samples by interpolation in the minority category samples to match the number of samples of each category.
7. The prediction method of claim 1, wherein, The clinical data set includes pharmacokinetic parameters, patient number, demographic information data, drug measurement data, medical history data, test indicator data, combined drug data, genotype data.
8. The prediction method of claim 1, wherein, The prediction accuracy is the average percentage error of the predicted blood concentration value and the actual blood concentration value.
9. A blood concentration prediction model for personalized medication, characterized in that a data acquisition module for acquiring a clinical data set; a data optimization module for optimizing the clinical data set, and performing the prediction method of any one of claims 2-6; The pharmacokinetic parameter calculation module is configured to calculate individual pharmacokinetic parameters, and the training set is calculated by at least two optimization algorithms; the prediction accuracy of each algorithm is determined according to the blood concentration prediction value obtained by each optimization algorithm and the actual blood concentration value; and the individual pharmacokinetic parameters are determined according to the weight of all prediction accuracy sums and the pharmacokinetic parameters of the corresponding optimization algorithm. The blood concentration prediction module is configured to input the to-be-tested data set into the trained blood concentration prediction model to obtain individual blood concentration prediction values.
10. An electronic device comprising a memory and a processor, the memory having stored therein computer readable instructions, the computer readable instructions, when executed by the processor, causing the electronic device to, The processor is caused to perform the blood concentration prediction method for personalized medication according to any one of claims 1-8.