System and method for accurately selecting tacrolimus therapeutic dose of myasthenia gravis patient

Abstract

The invention relates to the technical field of medicine information, and discloses a system and a method for accurately selecting tacrolimus treatment dosage of a myasthenia gravis patient. The method comprises the following steps: acquiring multi-cycle administration records and corresponding blood concentration of a patient to form an initial sequence; establishing a theoretical contribution degree matrix through lagging compensation, and reversely decomposing a pure single administration concentration curve; matching the curve with an individual pharmacokinetic template, and identifying an abnormal metabolic segment; and associating the same-period clinical records to generate clinical metabolism association pairs, establishing a mapping relation between concentration fluctuation and clinical symptoms, and classifying abnormal drug effect modes. According to the method, the independent metabolism trajectory of single administration can be analyzed from the mixed monitoring data, and the drug metabolism dynamics and the clinical effect are accurately associated, so that a technical basis is provided for realizing prospective individualized accurate administration.

Description

Technical Field

[0001] This invention relates to the field of pharmaceutical information technology, specifically to a system and method for precise selection of tacrolimus treatment dosage in patients with myasthenia gravis. Background Technology

[0002] In the clinical treatment of myasthenia gravis, individualized tacrolimus dosing relies on therapeutic drug monitoring. Current technologies primarily adjust dosage by detecting trough blood drug concentrations or by correlating concentration data with the patient's genotype. These methods treat blood drug concentration values ​​measured at different time points as independent observations. Blood drug concentration values ​​are a mixture of the cumulative effects of multiple doses in the body and cannot reflect the true metabolic process of a single dose in an individual. Current technologies lack methods for quantifying and separating the residual contribution of historical dosing, leading to confounding biases in dosage adjustment.

[0003] Existing technologies for analyzing the correlation between blood drug concentration and clinical symptoms typically employ time-asynchronous retrospective comparisons. There is often a time lag between muscle strength scores, complication records, and blood drug concentration monitoring, making it difficult to precisely synchronize clinical assessments with drug metabolism events. This makes it impossible to effectively identify and confirm the causal relationship between specific patterns of blood drug concentration fluctuations and changes in clinical symptoms. Dosage adjustments therefore rely primarily on the single dimension of whether blood drug concentration falls within the target range, lacking a quantitative analytical model that directly correlates dynamic changes in concentration with clinical efficacy.

[0004] The technical problem that this invention aims to solve is how to extract the independent metabolic trajectory of a single drug administration from mixed blood drug concentration data, and how to establish a precise mapping relationship between dynamic fluctuations in blood drug concentration and real-time changes in clinical symptoms. Summary of the Invention

[0005] The purpose of this invention is to provide a system and method for precise selection of tacrolimus treatment dosage in patients with myasthenia gravis, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a method for accurately selecting the tacrolimus treatment dose in patients with myasthenia gravis, the method comprising: Obtain tacrolimus dosing records for a patient with myasthenia gravis to be evaluated over multiple independent cycles, as well as blood drug concentration values ​​measured at the corresponding blood collection points after each dosing, to form an initial dosing concentration sequence; The initial dosing concentration sequence was subjected to hysteresis compensation processing. Based on the absorption and distribution kinetics of tacrolimus in the human body, the theoretical contribution of each dosing record to the concentration values ​​of subsequent blood collection points was calculated, and a theoretical contribution matrix was generated. The initial drug concentration sequence was decomposed in reverse using the theoretical contribution matrix to extract the theoretical blood drug concentration curves generated independently by a single dose, resulting in a set of pure single-dose concentration curves. Each pure single-dose concentration curve is matched with a preset individual pharmacokinetic parameter template to identify abnormal curve segments that deviate from the expected metabolic pattern, and all abnormal curve segments are merged into an abnormal metabolic event record set. The system associates and retrieves the patient's concurrent clinical assessment records, which include muscle strength scores and complication logs. Based on the timestamps, each clinical assessment record is paired with the corresponding abnormal metabolic event record to generate clinical metabolic association pairs. By analyzing each clinical metabolic correlation pair, a mapping relationship between abnormal fluctuations in blood drug concentration and the appearance of specific clinical symptoms was established, and several clinically significant abnormal pharmacodynamic patterns were classified.

[0007] Preferably, the method further includes: Based on the categorized abnormal pharmacodynamic patterns, the individual pharmacokinetic parameter template is reverse-corrected to form a personalized pharmacokinetic parameter model adapted to the patient's current physiological state. The system receives the patient's latest real-time tacrolimus blood concentration monitoring value, inputs it into a personalized pharmacokinetic parameter model, performs short-term concentration prediction, and obtains the predicted trajectory of blood concentration within a preset time period. The predicted blood drug concentration trajectory is compared with the preset safe and effective concentration window. If the predicted trajectory exceeds the window, a dose adjustment warning is triggered. Based on the dose adjustment warning and combined with the personalized pharmacokinetic parameter model, a tacrolimus dose adjustment scheme is calculated that can bring the future blood drug concentration prediction trajectory back to the safe and effective concentration window. The dose adjustment scheme includes specific dose change values ​​and subsequent monitoring time points.

[0008] Preferably, the hysteresis compensation processing of the initial dosing concentration sequence, based on the absorption and distribution kinetics of tacrolimus in the human body, calculates the theoretical contribution of each dosing record to the concentration values ​​of subsequent blood collection points, generating a theoretical contribution matrix, specifically including: Standard population pharmacokinetic parameters of tacrolimus were extracted, including the absorption rate constant, volume of distribution, and elimination rate constant. Based on the aforementioned standard population pharmacokinetic parameters, a fundamental function describing the change in blood drug concentration over time after a single dose is constructed. For each dosing record in the initial dosing concentration sequence, the theoretical concentration value generated by that dosing at each blood collection point time point in the sequence is calculated using the aforementioned basic function; All the calculated theoretical concentration values ​​are arranged in rows according to the dosing record number and columns according to the blood collection point number to form a theoretical contribution matrix. Each element in the matrix represents the theoretical contribution of a specific dosing to the concentration value of a specific blood collection point.

[0009] Preferably, the step of using the theoretical contribution matrix to perform inverse decomposition on the initial drug concentration sequence to extract the theoretical blood drug concentration curves independently generated by a single dose, thereby obtaining a set of pure single-dose concentration curves, specifically includes: The blood drug concentration values ​​measured at all blood sampling points in the initial drug administration concentration sequence are used to construct a concentration observation vector; Establish a system of linear equations, where the theoretical contribution matrix is ​​used as the coefficient matrix, the initial concentration peak vector generated independently by each administration is used as the unknown vector, and the concentration observation vector is used as the result vector. Solve the system of linear equations to obtain the estimated initial concentration peak value generated independently for each administration; Substituting each initial peak concentration estimate into the fundamental function describing the change in blood drug concentration over time after a single dose, a theoretical blood drug concentration change curve is generated, starting from the time of administration and covering the time range of all subsequent blood collection points. All curves together constitute a set of pure single-dose concentration curves.

[0010] Preferably, the step of matching each pure single-dose concentration curve with a preset individual pharmacokinetic parameter template to identify abnormal curve segments that deviate from the expected metabolic pattern, and merging all abnormal curve segments into an abnormal metabolic event record set, specifically includes: A preset individual pharmacokinetic parameter template is provided, which defines the standard shape and time parameters of the blood drug concentration curve in the absorption, distribution and elimination phases. For each pure single-dose concentration curve, calculate its morphological similarity and time parameter difference with the individual pharmacokinetic parameter template at each phase; Set a similarity threshold and a tolerance range for the difference, and mark the curve segments with similarity below the threshold or time parameter differences exceeding the tolerance range as deviating segments; Extract all marked deviation segments, record their corresponding drug administration number, the phase in which they occurred, the type of deviation, and the quantitative value of the deviation. Each record is recorded as an abnormal metabolic event record, and all records are summarized into an abnormal metabolic event record set.

[0011] Preferably, the step of associating and retrieving the patient's concurrent clinical assessment records, and pairing each clinical assessment record with the corresponding abnormal metabolic event record based on the timestamp to generate clinical metabolic association pairs, specifically includes: Retrieve all clinical assessment records from the patient's medical database within the same timeframe as the initial dosing concentration sequence; Extract precise timestamp information for each clinical assessment record and each abnormal metabolic event record; Using a time window matching strategy, if the timestamp of a clinical assessment record falls within the time window of the abnormal curve segment corresponding to an abnormal metabolic event record, then the clinical assessment record is considered to be associated with that abnormal metabolic event record. Each successfully associated clinical assessment record is linked to an abnormal metabolic event record to form a clinical metabolic association pair containing complete information on metabolic abnormalities and clinical status.

[0012] Preferably, the analysis of each clinical metabolic correlation pair establishes a mapping relationship between abnormal fluctuations in blood drug concentration and the appearance of specific clinical symptoms, classifying several clinically significant abnormal pharmacodynamic patterns, specifically including: Analyze each clinical metabolic association pair to extract the types of abnormal metabolic events and key clinical symptom descriptions from clinical assessment records; Perform association rule mining to find frequent combinations of specific abnormal metabolic event types and specific clinical symptom descriptions that occur simultaneously. The medical significance of the frequently discovered combinations is assessed, and those combinations that are explanatory from a pharmacological or pathological perspective are selected. Each selected combination is defined as an independent abnormal pharmacodynamic pattern, and each abnormal pharmacodynamic pattern is named and characterized to form an abnormal pharmacodynamic pattern library.

[0013] Preferably, the step of reversing the individual pharmacokinetic parameter template based on the categorized abnormal pharmacodynamic patterns to form a personalized pharmacokinetic parameter model adapted to the patient's current physiological state specifically includes: From the abnormal pharmacodynamic pattern library, feature parameters of recurring abnormal metabolic events corresponding to each pattern are extracted; Analyze the systematic shift direction and magnitude of the characteristic parameters to infer the individual pharmacokinetic parameter deviations they reflect; Based on the inferred deviation, the corresponding parameters in the preset individual pharmacokinetic parameter template are corrected, including the absorption rate constant, distribution volume, or elimination rate constant. The corrected parameters are used to generate a new pharmacokinetic model kernel, which is a personalized pharmacokinetic parameter model adapted to the patient's current physiological state. The process of receiving the patient's latest real-time tacrolimus blood concentration monitoring value and inputting it into a personalized pharmacokinetic parameter model for short-term concentration prediction, thereby obtaining a predicted blood concentration trajectory within a preset future time period, specifically includes: Obtain the patient's most recent tacrolimus administration time and dose, as well as at least one real-time blood drug concentration monitoring value collected since then; The latest real-time blood drug concentration monitoring value was used as the calibration point and substituted into the personalized pharmacokinetic parameter model for model state calibration. Using a corrected personalized pharmacokinetic parameter model, starting from the most recent dosing, we simulated the dynamic process of blood drug concentration changing over time within a preset future period, assuming that the current dosing regimen remains unchanged. The simulation results of this dynamic process are output, forming a predicted trajectory of blood drug concentration covering a preset future time period.

[0014] Preferably, the step of comparing the predicted blood drug concentration trajectory with a preset safe and effective concentration window, and triggering a dose adjustment warning if the predicted trajectory exceeds the window, specifically includes: A safe and effective concentration window for tacrolimus in patients with myasthenia gravis is preset, wherein the window has a lower concentration limit and an upper concentration limit; The predicted concentration value at each time point on the blood drug concentration prediction trajectory is compared with the lower limit and upper limit of the concentration, respectively. If the predicted concentration values ​​at multiple consecutive time points in the predicted trajectory are lower than the lower concentration limit, a low-dose warning is triggered; if the predicted concentration values ​​at multiple consecutive time points are higher than the upper concentration limit, a high-dose warning is triggered; the low-dose warning and the high-dose warning are collectively referred to as dose adjustment warnings. Record the specific time intervals that trigger dose adjustment warnings and the predicted extent to which the concentration exceeds the safe and effective concentration window; Based on the dose adjustment warning and combined with a personalized pharmacokinetic parameter model, a tacrolimus dose adjustment scheme is calculated that can bring the future blood drug concentration prediction trajectory back to the safe and effective concentration window. Specifically, this includes: Based on the type of dose adjustment alert triggered, determine whether the target direction of the dose adjustment is to increase or decrease the dose; Using a personalized pharmacokinetic parameter model as the computational engine, a new dosing dose value is searched through mathematical optimization under the constraint of adjusting the target direction; The new dosage value must meet the following conditions: when this new dosage is used for the next and subsequent administrations, the new blood drug concentration trajectory predicted by the model will be entirely within the safe and effective concentration window, and the center line of the trajectory should be as close as possible to the target optimal concentration value of the window. The new dosage value obtained from the search is output as the suggested adjustment dose, and the next recommended blood drug concentration monitoring time point for verifying the adjustment effect is calculated simultaneously, together forming the tacrolimus dosage adjustment plan.

[0015] Preferably, the present invention also includes a system for precisely selecting the tacrolimus treatment dose for patients with myasthenia gravis, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, implements the steps of the method for precisely selecting the tacrolimus treatment dose for patients with myasthenia gravis as described above.

[0016] Compared with the prior art, the beneficial effects of the present invention are: Hysteresis compensation and inverse decomposition of the initial dosing concentration sequence generate a theoretical contribution matrix, separating the theoretical blood drug concentration curve independently generated by a single dose. Based on the standard pharmacokinetic model of tacrolimus in the human body, this method quantifies the residual effect of each previous dose on the concentration values ​​at subsequent blood sampling points, extracting the pure single-dose metabolic trajectory from the mixed concentration data obtained from actual monitoring. The dimension of blood drug concentration data analysis changes from discrete, static concentration points to continuous, dynamic metabolic process waveforms. This eliminates the interpretative bias of the current concentration value caused by the superposition of effects from multiple doses in traditional methods, clearly revealing the absorption and clearance characteristics of a single dose, and providing a direct data basis for identifying individual metabolic rate abnormalities, absorption delays, or nonlinear kinetics.

[0017] By precisely pairing timestamps, abnormal metabolic events are recorded and correlated with concurrent muscle strength scores and complication logs to form clinical metabolic correlation pairs, and a systematic mapping relationship from fluctuations in blood drug concentration to the appearance of specific clinical symptoms is established. This method achieves rigorous synchronous analysis of pharmacokinetic events and clinical efficacy events over time, thereby enabling the identification of statistically significant abnormal pharmacodynamic patterns. The basis for dose adjustment decisions is thus deepened from a single pharmaceutical indicator—whether the blood drug concentration is within the target range—to a comprehensive pharmacodynamic indicator—how specific types of concentration fluctuations are associated with clinical benefit or risk. Clinical interventions can be made proactively or synchronously to address identified abnormal patterns, rather than responding with a lag, improving the foresight and precision of individualized dosing. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the working principle of the method for precise selection of tacrolimus treatment dosage in patients with myasthenia gravis as described in this invention. Figure 2 A flowchart for generating the theoretical contribution matrix; Figure 3 A flowchart for obtaining the pure single-dose concentration curve through reverse decomposition; Figure 4 A bar chart showing the comparison of pharmacokinetic parameters before and after correction; Figure 5 A comparison chart showing the duration of abnormal tacrolimus dosage in different patients. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 This invention provides a method for precise selection of tacrolimus treatment dosage in patients with myasthenia gravis. The method includes: obtaining tacrolimus dosing records of the patient to be evaluated in multiple independent cycles and the blood drug concentration values ​​at the corresponding blood sampling points after each dosing to form an initial dosing concentration sequence. Subsequently, hysteresis compensation processing is performed on the initial dosing concentration sequence, and based on the absorption and distribution kinetics of tacrolimus in the human body, the theoretical contribution of each dosing record to the concentration values ​​at subsequent blood sampling points is calculated, and a theoretical contribution matrix is ​​generated. This matrix is ​​used to perform inverse decomposition of the initial dosing concentration sequence, separating the theoretical blood drug concentration curves independently generated by a single dosing, thereby obtaining a set of pure single-dose concentration curves. These curves are waveform matched with a preset individual pharmacokinetic parameter template to identify abnormal curve segments that deviate from the expected metabolic pattern, and all abnormal curve segments are merged to form an abnormal metabolic event record set. The patient's concurrent clinical assessment records are associated and retrieved, and each clinical assessment record is paired with the corresponding abnormal metabolic event record according to the timestamp to generate clinical metabolic association pairs. By analyzing each clinical metabolic correlation pair, a mapping relationship between abnormal fluctuations in blood drug concentration and the appearance of specific clinical symptoms was established, and finally, several abnormal pharmacodynamic patterns with clinical significance were classified.

[0021] In one embodiment of the present invention, see [reference] Figure 2 The system obtains the patient's latest real-time tacrolimus blood concentration monitoring value and inputs it into a personalized pharmacokinetic parameter model for short-term concentration prediction, obtaining a predicted blood concentration trajectory within a preset time period. This predicted trajectory is compared with a preset safe and effective concentration window. If the predicted trajectory exceeds this window, a dose adjustment warning is triggered. Based on the dose adjustment warning and the personalized pharmacokinetic parameter model, a tacrolimus dose adjustment plan is calculated that will bring the future predicted blood concentration trajectory back to within the safe and effective concentration window. This dose adjustment plan includes specific dose changes and subsequent monitoring time points.

[0022] Standard population pharmacokinetic parameters of tacrolimus were extracted, including the absorption rate constant, volume of distribution, and elimination rate constant. Based on these parameters, a fundamental function describing the change in plasma concentration over time after a single dose was constructed. For each dosing record in the initial dosing concentration sequence, the theoretical concentration value generated by that dose at each blood sampling time point in the sequence was calculated using the fundamental function. All calculated theoretical concentration values ​​were arranged in rows by dosing record number and columns by blood sampling point number to form a theoretical contribution matrix. Each element in the matrix represents the theoretical contribution of a specific dose to the concentration value at a specific blood sampling point.

[0023] In practice, the establishment of a personalized pharmacokinetic parameter model relies on extracting characteristic parameters of recurring abnormal metabolic events from an abnormal pharmacodynamic pattern library. Analyzing the systematic shift direction and magnitude of the characteristic parameters can infer the deviation of individual pharmacokinetic parameters. Based on the inferred deviation, the corresponding parameters in the preset individual pharmacokinetic parameter template are corrected. The parameters include the absorption rate constant, volume of distribution, or elimination rate constant. The corrected parameters are used to generate a new pharmacokinetic model kernel. The new pharmacokinetic model kernel is the personalized pharmacokinetic parameter model adapted to the patient's current physiological state. In some embodiments, the latest real-time tacrolimus blood concentration monitoring value of the patient is received and input into the personalized pharmacokinetic parameter model for short-term concentration prediction. The time and dose of the patient's most recent tacrolimus administration and at least one latest real-time blood concentration monitoring value collected after that are obtained. The latest real-time blood concentration monitoring value is used as a correction point and substituted into the personalized pharmacokinetic parameter model for model state correction. The corrected personalized pharmacokinetic parameter model is used to simulate the dynamic process of blood concentration change over time in the future preset period, assuming that the current dosing regimen remains unchanged, starting from the most recent administration. The simulation results of the dynamic process are output to form a blood concentration prediction trajectory covering the future preset period.

[0024] In practice, when comparing the predicted blood drug concentration trajectory with the preset safe and effective concentration window, the preset safe and effective concentration window for tacrolimus for patients with myasthenia gravis has a lower limit and an upper limit. The predicted concentration value corresponding to each time point on the blood drug concentration prediction trajectory is compared with the lower limit and the upper limit respectively. If the predicted concentration value at multiple consecutive time points in the prediction trajectory is lower than the lower limit, a low-dose warning is triggered; if the predicted concentration value at multiple consecutive time points is higher than the upper limit, a high-dose warning is triggered. The low-dose warning and the high-dose warning are collectively referred to as dose adjustment warnings. The specific time interval that triggers the dose adjustment warning and the degree to which the predicted concentration exceeds the safe and effective concentration window are recorded. It is understandable that a tacrolimus dose adjustment scheme can be calculated based on dose adjustment warnings combined with personalized pharmacokinetic parameter models to ensure that the predicted trajectory of future blood drug concentrations returns to the safe and effective concentration window. The target direction of dose adjustment is determined according to the type of dose adjustment warning triggered, whether to increase or decrease the dose. Using the personalized pharmacokinetic parameter model as the calculation engine, a new dosing dose value is searched in a mathematical optimization manner under the constraint of the adjustment target direction. The new dosing dose value must satisfy the condition that when the new dose is used for the next and subsequent dosing, the new blood drug concentration trajectory predicted by the model will be entirely within the safe and effective concentration window, and the center line of the trajectory will be close to the target optimal concentration value of the safe and effective concentration window. The new dosing dose value obtained from the search is output as the suggested adjustment dose, and the next blood drug concentration monitoring time point recommended to verify the adjustment effect is calculated simultaneously to constitute the tacrolimus dose adjustment scheme.

[0025] Optionally, standard population pharmacokinetic parameters of tacrolimus are extracted to construct a theoretical contribution matrix. These parameters include the absorption rate constant, volume of distribution, and elimination rate constant. Based on these parameters, a fundamental function describing the change in blood drug concentration over time after a single dose is constructed. For each dosing record in the initial dosing concentration sequence, the theoretical concentration value generated by that dose at each blood sampling point in the sequence is calculated using the fundamental function. All calculated theoretical concentration values ​​are arranged in rows by dosing record number and columns by blood sampling point number to form a theoretical contribution matrix. Each element in the theoretical contribution matrix represents the theoretical contribution of a specific dose to the concentration value at a specific blood sampling point. In some embodiments, the construction of the fundamental function involves a pharmacokinetic model, such as a one-compartment or multi-compartment model. In specific implementations, the construction of the fundamental function begins with extracting standard population pharmacokinetic parameters of tacrolimus, including the absorption rate constant, volume of distribution, and elimination rate constant, and selecting an appropriate pharmacokinetic model framework, such as a one-compartment or multi-compartment model, based on these parameters. For a one-compartment model, the fundamental function describes the change in blood drug concentration over time after a single dose. The drug is considered uniformly distributed within a single compartment, and concentration changes follow first-order kinetics: the absorption phase is characterized by a gradual increase in concentration, while the elimination phase is characterized by an exponential decrease in concentration. For a multi-compartment model, the fundamental function considers the distribution and transfer of drug between the central and peripheral compartments. Concentration changes involve the dynamic equilibrium of multiple compartments, and the fundamental function includes multiple exponential terms to reflect more complex absorption, distribution, and elimination characteristics. The constructed fundamental function is used to calculate the theoretical concentration value generated by each dose at a specific blood sampling time point, thus providing a basis for generating the theoretical contribution matrix.

[0026] In practice, the basic function can be expressed as an exponential decay function of blood drug concentration over time. For a single oral administration, the basic function takes the following form: in: Indicates time Theoretical blood drug concentration, Indicates the dosage. Indicates bioavailability, Represents the distributed volume. Represents the absorption rate constant. This represents the elimination rate constant. It can be understood that generating the theoretical contribution matrix is ​​the core step in hysteresis compensation. In practice, the theoretical contribution matrix can be used to reverse decompose the initial drug concentration sequence to extract the theoretical blood drug concentration curves independently generated by a single dose. The accuracy of the theoretical contribution matrix directly affects the quality of the pure single-dose concentration curves.

[0027] Optionally, in practical applications, a myasthenia gravis patient follows a regular dosing regimen and periodically monitors blood drug concentration. By extracting standard population pharmacokinetic parameters and constructing a basic function, the theoretical contribution of each dose to the concentration at each blood sampling point is calculated, forming a theoretical contribution matrix. This matrix, combined with the patient's real-time blood drug concentration monitoring values, is input into a personalized pharmacokinetic parameter model for prediction. The prediction results indicate that the blood drug concentration will fall below the lower limit of the safe and effective concentration window in the future, triggering a low-dose warning. Based on the low-dose warning, the personalized pharmacokinetic parameter model calculates a recommended adjustment of the tacrolimus dose and recommends subsequent monitoring time points. In specific implementation, the predicted trajectory of another myasthenia gravis patient shows that the blood drug concentration will exceed the upper limit of the safe and effective concentration window, triggering a high-dose warning. Based on the high-dose warning, the personalized pharmacokinetic parameter model calculates a dose reduction plan.

[0028] In one embodiment of the present invention, see [reference] Figure 3 The blood drug concentration values ​​measured at all blood sampling points in the initial dosing concentration sequence are constructed as a concentration observation vector. A system of linear equations is established, where the theoretical contribution matrix is ​​used as the coefficient matrix, the initial concentration peak vector generated independently by each dosing is used as the unknown vector, and the concentration observation vector is used as the result vector. Solving this system of linear equations yields the estimated initial concentration peak value generated independently by each dosing. Substituting each estimated initial concentration peak value into the fundamental function describing the change of blood drug concentration over time after a single dosing generates a theoretical blood drug concentration change curve that starts from the dosing time and covers the time range of all subsequent blood sampling points. All curves together constitute a pure single-dose concentration curve set.

[0029] An individual pharmacokinetic parameter template is predefined, which defines the standard morphology and time parameters of the blood drug concentration curve in the absorption, distribution, and elimination phases. For each pure single-dose concentration curve, the morphological similarity and time parameter difference between it and the individual pharmacokinetic parameter template are calculated at each phase. A similarity threshold and a tolerance range for the difference are set, and curve segments with similarity below the threshold or time parameter differences exceeding the tolerance range are marked as deviation segments. All marked deviation segments are extracted, and their corresponding dosing number, the phase in which they occurred, the type of deviation, and the quantitative value of the deviation are recorded. Each record is recorded as an abnormal metabolic event record, and all records are summarized into an abnormal metabolic event record set.

[0030] In practice, the reverse decomposition requires constructing a concentration observation vector from the blood drug concentration values ​​measured at all blood collection points in the initial dosing concentration sequence. When establishing the linear equation system, the theoretical contribution matrix is ​​used as the coefficient matrix, the initial concentration peak vector generated independently by each dosing is used as the unknown vector, and the concentration observation vector is used as the result vector. Solving the linear equation system yields the estimated value of the initial concentration peak generated independently by each dosing. Substituting each estimated value of the initial concentration peak into the basic function describing the change of blood drug concentration with time after a single dosing generates a theoretical blood drug concentration change curve that covers the time range of all subsequent blood collection points starting from the dosing time. All curves together constitute a pure single-dose concentration curve set. In some embodiments, an individual pharmacokinetic parameter template is preset for waveform matching. The individual pharmacokinetic parameter template defines the standard morphology and time parameters of the blood drug concentration curve in the absorption, distribution, and elimination phases. For each pure single-dose concentration curve, the morphological similarity and time parameter difference between it and the individual pharmacokinetic parameter template in each phase are calculated. A similarity threshold and a tolerance range for the difference are set. Curve segments with similarity below the threshold or time parameter differences exceeding the tolerance range are marked as deviation segments. All marked deviation segments are extracted and their corresponding dosing sequence number, the phase in which the deviation occurred, the type of deviation, and the quantitative value of the deviation are recorded. Each record is recorded as an abnormal metabolic event record, and all records are summarized into an abnormal metabolic event record set.

[0031] In practice, the linear equations of the inverse decomposition can be represented in matrix operation form. For example, the formula shows the relationship between the theoretical contribution matrix, the initial concentration peak vector, and the concentration observation vector: Where: matrix symbol Represents the theoretical contribution matrix, vector notation. This represents the vector of initial concentration peaks generated independently by each dosing, to be solved. The vector symbol is... This represents the concentration observation vector. It can be understood that establishing and solving a system of linear equations is the core step in the inverse decomposition. The generation of the pure single-dose concentration curve set depends on the accurate solution of the initial peak concentration estimate and the correct application of the fundamental functions.

[0032] Optionally, a specific example scenario involves a myasthenia gravis patient who recorded three tacrolimus administrations and corresponding blood drug concentration values ​​at six blood sampling points during a past treatment cycle. The concentration observation vector consists of the six blood drug concentration values ​​arranged in chronological order of blood sampling time. The theoretical contribution matrix is ​​a 6x3 matrix, where each column corresponds to one administration and each row corresponds to one blood sampling point. The elements in the matrix are calculated using a basis function, representing the theoretical contribution of a particular administration to the concentration value at a given blood sampling point. Solving the linear equations yields three initial peak concentration estimates, corresponding to the three administrations respectively. Substituting these three estimates into the basis function generates three theoretical blood drug concentration change curves. These three curves constitute a set of pure single-dose concentration curves. In some embodiments, for each generated pure single-dose concentration curve, the system compares it with an individual pharmacokinetic parameter template, which specifies that the absorption response should reach its peak within a specific time range after administration. If the peak time of a curve is significantly later than the time range specified by the template, the calculated time parameter difference will exceed the preset difference tolerance range, and the absorption phase segment of the curve will be marked as a deviation segment and recorded as an abnormal metabolic event.

[0033] It is understandable that the waveform matching process involves quantitative calculations. The calculation of morphological similarity uses the curve fitting goodness index, and the calculation of the time parameter difference involves the comparison of pharmacokinetic parameters such as peak time and half-life. Optionally, in another example scenario, the set of pure single-dose concentration curves contains five curves, two of which show an abnormally slow decreasing trend in the elimination phase. Their morphological similarity with the individual pharmacokinetic parameter template in the elimination phase is lower than a preset similarity threshold. The elimination phase segments of these two curves are marked as deviation segments. The system records the occurrence phase of the abnormal metabolic event as the elimination phase and the deviation type as elimination delay. All marked deviation segment information is summarized to form an abnormal metabolic event record set.

[0034] In one embodiment of the invention, all clinical assessment records within the same time range as the initial drug concentration sequence are retrieved from the patient's medical database. Precise timestamp information is extracted for each clinical assessment record and each abnormal metabolic event record. A time window matching strategy is employed: if the timestamp of a clinical assessment record falls within the time window of the abnormal curve segment corresponding to an abnormal metabolic event record, the clinical assessment record is considered associated with that abnormal metabolic event record. Each successfully associated clinical assessment record is bound to an abnormal metabolic event record, forming a clinical metabolic association pair containing complete metabolic abnormality information and clinical status information.

[0035] Each clinical metabolic association pair is analyzed to extract aberrant metabolic event types and key clinical symptom descriptions from clinical assessment records. Association rule mining is performed to identify frequent combinations of aberrant metabolic event types and clinical symptom descriptions. The medical significance of these frequent combinations is assessed, and those combinations that are pharmacologically or pathologically interpretable are selected. Each selected combination is defined as an independent aberrant pharmacodynamic pattern, and each aberrant pharmacodynamic pattern is named and characterized, forming an aberrant pharmacodynamic pattern library.

[0036] In practice, all clinical assessment records within the same time range as the initial drug concentration sequence are retrieved from the patient's medical database. Precise timestamp information is extracted for each clinical assessment record and each abnormal metabolic event record. A time window matching strategy is used; if the timestamp of a clinical assessment record falls within the time window of the abnormal curve segment corresponding to an abnormal metabolic event record, the clinical assessment record is considered associated with the abnormal metabolic event record. Each successfully associated clinical assessment record and abnormal metabolic event record is bound to form a clinical metabolic association pair containing complete metabolic abnormality information and clinical status information. In some embodiments, each clinical metabolic association pair is parsed to extract the abnormal metabolic event type and key clinical symptom descriptions from the clinical assessment records. Association rule mining is performed to find frequent combinations of abnormal metabolic event types and clinical symptom descriptions. The frequently mined combinations are evaluated for medical significance, and those combinations that are pharmacologically or pathologically interpretable are selected. Each selected combination is defined as an independent abnormal pharmacodynamic pattern, and each abnormal pharmacodynamic pattern is named and characterized to form an abnormal pharmacodynamic pattern library.

[0037] In practice, the time window matching strategy requires defining the start and end times of the corresponding abnormal curve segment for each abnormal metabolic event record, thus forming a time window. A match is considered successful when the timestamp of a clinical assessment record falls within this time window. For example, a record eliminating a phase-delayed abnormal metabolic event has a time window covering the period from 8 hours to 24 hours after drug administration. If a clinical assessment record showing worsened ptosis has a timestamp at 12 hours after drug administration, then the clinical assessment record is successfully associated with the abnormal metabolic event record. Association rule mining aims to discover stable co-occurrence relationships between metabolic events and clinical symptoms. A commonly used quantification method is to calculate the support of association rules. For example, a formula defines how to calculate the support for a specific combination: Where: symbol Represents an abnormal metabolic event type, symbol Represents a clinical symptom description, symbol Represents the types of abnormal metabolic events in all clinical metabolic association pairs. Compared with clinical symptom description The number of simultaneous association pairs, and their symbols. The symbol represents the total number of clinical metabolic association pairs involved in the analysis. This represents the calculated support value.

[0038] Optionally, a specific example scenario involves a patient who generated fifteen abnormal metabolic event records and nine clinical assessment records over four consecutive treatment cycles. The system defines a time window for each abnormal metabolic event record and compares the timestamps of the nine clinical assessment records with it, ultimately generating ten clinical metabolic association pairs. Analysis of these ten clinical metabolic association pairs extracts abnormal event types such as "delayed absorption peak" and "accelerated elimination," as well as clinical symptom descriptions such as "decreased upper limb muscle strength score" and "dysphagia." Association rule mining reveals that the combination of "accelerated elimination" and "worsening muscle weakness the next morning after administration" appears four times in the ten association pairs. In some embodiments, the medical significance assessment is completed by the system in conjunction with a knowledge base or by physician review. After confirming the pharmacological explanation that "accelerated elimination leads to a blood drug concentration below the treatment window the next morning, thus causing symptom fluctuations" is reasonable, this combination is defined and named the "morning failure" pattern and entered into the abnormal pharmacodynamic pattern library.

[0039] Optionally, another example illustrates the data comparison: for the same patient, routine assessment based solely on the trough of blood drug concentration only reveals a low concentration. However, by precisely correlating the abnormal concentration curve morphology with clinical symptoms at specific time points, the specific abnormal efficacy pattern of "morning failure" was identified, which contrasts with viewing concentration values ​​or symptom records in isolation.

[0040] In one embodiment of the present invention, based on the categorized abnormal pharmacodynamic patterns, the individual pharmacokinetic parameter template is reverse-corrected to form a personalized pharmacokinetic parameter model adapted to the patient's current physiological state. Characteristic parameters of recurring abnormal metabolic events corresponding to each pattern are extracted from the abnormal pharmacodynamic pattern library. The systematic shift direction and magnitude of the characteristic parameters are analyzed to infer the individual pharmacokinetic parameter deviations they reflect. Based on the inferred deviations, the corresponding parameters in the preset individual pharmacokinetic parameter template are corrected; these parameters include absorption rate constants, volume of distribution, or elimination rate constants. A new pharmacokinetic model kernel is generated using the corrected parameters; this kernel is the personalized pharmacokinetic parameter model adapted to the patient's current physiological state.

[0041] The system receives the patient's latest real-time tacrolimus blood concentration monitoring value and inputs it into a personalized pharmacokinetic parameter model for short-term concentration prediction, obtaining a predicted blood concentration trajectory over a preset future time period. It acquires the patient's most recent tacrolimus administration time and dose, as well as at least one recently acquired real-time blood concentration monitoring value. Using the most recently acquired real-time blood concentration monitoring value as a correction point, it performs model state correction within the personalized pharmacokinetic parameter model. Using the corrected personalized pharmacokinetic parameter model, starting from the most recent administration, it simulates the dynamic process of blood concentration changes over time within a preset future time period, assuming the current dosing regimen remains unchanged. The simulation results of this dynamic process are output, forming a predicted blood concentration trajectory covering the preset future time period.

[0042] In practice, characteristic parameters of recurring abnormal metabolic events corresponding to each pattern are extracted from the abnormal pharmacodynamic pattern library. The systematic shift direction and magnitude of the characteristic parameters can be analyzed to infer the individual pharmacokinetic parameter deviations they reflect. Based on the inferred deviations, the corresponding parameters in the preset individual pharmacokinetic parameter template are corrected. The parameters include the absorption rate constant, distribution volume or elimination rate constant. The corrected parameters are used to generate a new pharmacokinetic model kernel. The new pharmacokinetic model kernel is a personalized pharmacokinetic parameter model adapted to the patient's current physiological state. In some embodiments, the latest real-time tacrolimus blood concentration monitoring value of the patient is received and input into a personalized pharmacokinetic parameter model to perform short-term concentration prediction to obtain a blood concentration prediction trajectory within a future preset period. The time and dose of the patient's most recent tacrolimus administration and at least one latest real-time blood concentration monitoring value collected after that are obtained. The latest real-time blood concentration monitoring value is used as a correction point and substituted into the personalized pharmacokinetic parameter model for model state correction. The corrected personalized pharmacokinetic parameter model is used to simulate the dynamic process of blood concentration change over time within a future preset period, assuming that the current dosing regimen remains unchanged, starting from the most recent administration. The simulation results of the dynamic process are output to form a blood concentration prediction trajectory covering the future preset period.

[0043] In practice, the reverse correction of individual pharmacokinetic parameter templates relies on quantitative analysis of an anomalous pharmacodynamic pattern library. For example, an "absorption delay" pattern is identified from the anomalous pharmacodynamic pattern library. Characteristic parameters regarding "time to peak concentration" are extracted from all anomalous metabolic events associated with this pattern. The average value of these characteristic parameters is calculated, revealing a systematic lag behind the nominal time to peak concentration defined in the individual pharmacokinetic parameter template. This deviation is inferred to reflect that the individual's absorption rate constant is lower than the standard population parameter. Based on this inference, the absorption rate constant in the individual pharmacokinetic parameter template is numerically down-corrected. It can be understood that the correction process can be performed simultaneously on multiple parameters; see Table 1, which shows a comparison of parameter corrections.

[0044] Table 1: Template Correction Table for Individual Pharmacokinetic Parameters Optionally, a specific example scenario illustrates the model state correction and prediction process: A patient orally takes 3 mg of tacrolimus at 8:00 AM. Two hours after administration, the real-time blood drug concentration is measured at 12 ng / mL. The system inputs this post-administration time and concentration value as a correction point into the personalized pharmacokinetic parameter model, and updates the initial values ​​of the model's internal state variables through the model state correction operation shown in the formula: Where: vector symbol Represents the corrected model state vector, vector symbol This represents the state vector predicted by the model before correction at the monitoring time point; matrix symbol. This represents the correction gain matrix calculated based on the personalized pharmacokinetic parameter model, with scalar symbols. Represents the actual observed blood drug concentration, scalar symbol. This represents the blood drug concentration value predicted by the model at the monitoring time point. The corrected model state vector is used. With personalized pharmacokinetic parameters, the system simulates changes in blood drug concentration over the next 24 hours starting from the most recent dosing, and outputs a predicted trajectory of blood drug concentration extending from the current moment to 8:00 AM the next day.

[0045] In some embodiments, the data comparison is reflected in the prediction differences using different parameter models. If the standard population pharmacokinetic parameter template before correction is used for prediction, the predicted blood drug concentration trajectory is consistently higher than the actual monitored value. However, after using the personalized pharmacokinetic parameter model generated by reverse correction and combined with the real-time monitoring value for correction, the predicted blood drug concentration trajectory can more closely match the patient's subsequent actual blood drug concentration monitoring points.

[0046] See Figure 4This is a bar chart comparing pharmacokinetic parameters before and after correction. It reflects the correction process from "population template parameters" to "individual precise parameters." For example, an increase in Vd corresponds to an abnormal pharmacodynamic pattern of "increased volume of distribution," which is a key step in building a personalized pharmacokinetic model. The corrected parameters are closer to the patient's actual metabolic characteristics, improving the accuracy of subsequent blood drug concentration prediction and dose adjustment. Parameter differences are directly related to individual differences in drug absorption and elimination, providing pharmacokinetic support for the precise selection of tacrolimus dosage. By comparing the parameters before and after correction, the rationality of the personalized pharmacokinetic model is demonstrated, providing a reliable parameter basis for subsequent blood drug concentration prediction and dose adjustment plans.

[0047] In one embodiment of the present invention, the predicted blood drug concentration trajectory is compared with a preset safe and effective concentration window. If the predicted trajectory exceeds the window, a dose adjustment warning is triggered. The preset safe and effective concentration window for tacrolimus in patients with myasthenia gravis has a lower limit and an upper limit. The predicted concentration value at each time point on the predicted blood drug concentration trajectory is compared with the lower limit and the upper limit, respectively. If the predicted concentration values ​​at multiple consecutive time points are lower than the lower limit, a low-dose warning is triggered; if the predicted concentration values ​​at multiple consecutive time points are higher than the upper limit, a high-dose warning is triggered. Both low-dose and high-dose warnings are collectively referred to as dose adjustment warnings. The specific time interval triggering the dose adjustment warning and the degree to which the predicted concentration exceeds the safe and effective concentration window are recorded.

[0048] Based on dose adjustment alerts and a personalized pharmacokinetic parameter model, a tacrolimus dose adjustment scheme is calculated to ensure that the predicted future blood drug concentration trajectory falls within the safe and effective concentration window. Depending on the type of dose adjustment alert triggered, the target direction of the dose adjustment is determined as either increasing or decreasing the dose. Using the personalized pharmacokinetic parameter model as the computational engine, a new dosing dose value is searched mathematically under the constraint of the adjustment target direction. The new dosing dose value must meet the following conditions: when this new dose is used for the next and subsequent administrations, the new blood drug concentration trajectory predicted by the model will be entirely within the safe and effective concentration window, and the center line of the trajectory will be close to the target optimal concentration value within that window. The newly obtained dosing dose value is output as the suggested adjustment dose, and the recommended next blood drug concentration monitoring time point for verifying the adjustment effect is calculated simultaneously, together forming the tacrolimus dose adjustment scheme.

[0049] In practice, a lower limit and an upper limit are pre-defined as the safe and effective concentration window for tacrolimus in patients with myasthenia gravis. The predicted concentration value at each time point on the blood drug concentration prediction trajectory is compared with the lower limit and the upper limit. If the predicted concentration value at multiple consecutive time points is lower than the lower limit, a low-dose warning is triggered; if the predicted concentration value at multiple consecutive time points is higher than the upper limit, a high-dose warning is triggered. The low-dose warning and the high-dose warning are collectively referred to as dose adjustment warnings. The specific time interval that triggers the dose adjustment warning and the degree to which the predicted concentration exceeds the safe and effective concentration window are recorded. In some embodiments, a tacrolimus dose adjustment scheme is calculated based on dose adjustment alerts and a personalized pharmacokinetic parameter model, which enables the predicted trajectory of future blood drug concentrations to return to the safe and effective concentration window. The target direction of dose adjustment is determined according to the type of dose adjustment alert triggered, whether it is to increase or decrease the dose. Using the personalized pharmacokinetic parameter model as the calculation engine, a new dosing dose value is searched in a mathematical optimization manner under the constraint of the adjustment target direction. The new dosing dose value must satisfy the condition that when the new dose is used for the next and subsequent dosing, the new blood drug concentration trajectory predicted by the model will be entirely within the safe and effective concentration window, and the center line of the trajectory will be close to the target optimal concentration value of the safe and effective concentration window. The new dosing dose value obtained by the search is output as the suggested adjustment dose, and the recommended time point for the next blood drug concentration monitoring to verify the adjustment effect is calculated simultaneously, which together constitute the tacrolimus dose adjustment scheme.

[0050] In practical implementation, a specific scenario involves the predicted trajectory analysis of patient A. Patient A's personalized pharmacokinetic parameter model outputs the predicted blood drug concentration trajectory for the next 24 hours. The preset safe and effective concentration window has a lower limit of 5 ng / mL, an upper limit of 15 ng / mL, and a target optimal concentration of 10 ng / mL. Comparing the predicted concentration values ​​for each hour on the predicted trajectory with the upper and lower limits of the window, it is found that the predicted concentration values ​​for the next 6 hours to 18 hours are continuously below 5 ng / mL. The system triggers a low-dose warning and records the extent of exceeding the lower limit as 2.1 ng / mL to 3.8 ng / mL. Similarly, another scenario involves patient B. Patient B's predicted blood drug concentration trajectory shows that the predicted concentration values ​​for the next 12 hours to 24 hours are continuously above 15 ng / mL. The system triggers a high-dose warning and records the extent of exceeding the upper limit as 1.5 ng / mL to 4.0 ng / mL. For the low-dose warning triggered by patient A, the target direction for dose adjustment is to increase the dose; for the high-dose warning triggered by patient B, the target direction for dose adjustment is to decrease the dose.

[0051] Optionally, searching for new dosing dose values ​​in a mathematical optimization manner can be done by defining an objective function. For example, the formula shows an optimization objective aimed at making the new predicted trajectory approach the target optimal concentration: Where: symbol Representatives with candidate doses The objective function value of the variable, sign Representative uses candidate dose Post-personalized pharmacokinetic parameter model predictions for future time points Blood drug concentration values, symbol Represents the preset optimal target concentration value within the safe and effective concentration window, symbol Represents a point in time The weighting coefficients are used to emphasize the importance of fitting in certain key time regions. The optimization process, constrained by the target direction of dose adjustment, seeks to achieve the objective function... Minimize the predicted concentration at all time points. The dose values ​​are all within the upper and lower limits of the safe and effective concentration window. In some embodiments, the optimized search for patient A ultimately resulted in a recommended dose adjustment to increase the next dose from 3 mg to 4 mg, and recommended blood drug concentration monitoring at 3 hours and 12 hours after the first dose following the dose adjustment. These two monitoring time points constitute part of the dose adjustment protocol to verify whether the effective concentration is reached rapidly after the dose increase and whether the peak concentration is safe.

[0052] Understandably, data comparison reflects the results of different adjustment strategies. Empirically increasing or decreasing the dosage based solely on the concentration value at a single time point cannot ensure that the blood drug concentration returns to the therapeutic window throughout the entire dosing interval. Optimizing the predicted trajectory over the entire future time period to determine the dosage aims to ensure concentration coverage throughout the adjusted cycle. In practice, the recommended time point for the next blood drug concentration monitoring is calculated based on the model's predicted concentration change dynamics. For example, monitoring may be conducted during periods with a relatively rapid predicted concentration change rate or near the window boundary to improve validation efficiency.

[0053] See Figure 5This is a comparison chart of the duration of tacrolimus dose abnormalities in different patients. It visually distinguishes the duration of dose abnormality risk in different patients, helping clinicians to focus on monitoring high-risk patients. Differences in the duration of abnormalities reflect the individuality of patients' metabolic capacity, providing a quantitative reference for personalized dosing intervals / dosage. It can also be correlated with patients' pharmacokinetic parameters (such as elimination rate) to uncover the metabolic characteristics corresponding to "long duration of abnormalities." It quickly identifies patients with longer durations of dose abnormalities, providing a basis for developing more intensive blood drug concentration monitoring plans for these patients, reducing the risk of insufficient efficacy or toxic reactions. Statistical analysis of the distribution of abnormal durations across different patients summarizes common patterns of abnormality persistence.

[0054] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for precise selection of tacrolimus treatment dosage in patients with myasthenia gravis, characterized by: The method includes: Obtain tacrolimus dosing records for a patient with myasthenia gravis to be evaluated over multiple independent cycles, as well as blood drug concentration values ​​measured at the corresponding blood collection points after each dosing, to form an initial dosing concentration sequence; The initial dosing concentration sequence was subjected to hysteresis compensation processing. Based on the absorption and distribution kinetics of tacrolimus in the human body, the theoretical contribution of each dosing record to the concentration values ​​of subsequent blood collection points was calculated, and a theoretical contribution matrix was generated. The initial drug concentration sequence was decomposed in reverse using the theoretical contribution matrix to extract the theoretical blood drug concentration curves generated independently by a single dose, resulting in a set of pure single-dose concentration curves. Each pure single-dose concentration curve is matched with a preset individual pharmacokinetic parameter template to identify abnormal curve segments that deviate from the expected metabolic pattern, and all abnormal curve segments are merged into an abnormal metabolic event record set. The system associates and retrieves the patient's concurrent clinical assessment records, which include muscle strength scores and complication logs. Based on the timestamps, each clinical assessment record is paired with the corresponding abnormal metabolic event record to generate clinical metabolic association pairs. By analyzing each clinical metabolic correlation pair, a mapping relationship between abnormal fluctuations in blood drug concentration and the appearance of specific clinical symptoms was established, and several clinically significant abnormal pharmacodynamic patterns were classified.

2. The method for accurately selecting the tacrolimus treatment dose for patients with myasthenia gravis according to claim 1, characterized in that, The method further includes: Based on the categorized abnormal pharmacodynamic patterns, the individual pharmacokinetic parameter template is reverse-corrected to form a personalized pharmacokinetic parameter model adapted to the patient's current physiological state. The system receives the patient's latest real-time tacrolimus blood concentration monitoring value, inputs it into a personalized pharmacokinetic parameter model, performs short-term concentration prediction, and obtains the predicted trajectory of blood concentration within a preset time period. The predicted blood drug concentration trajectory is compared with the preset safe and effective concentration window. If the predicted trajectory exceeds the window, a dose adjustment warning is triggered. Based on the dose adjustment warning and combined with the personalized pharmacokinetic parameter model, a tacrolimus dose adjustment scheme is calculated that can bring the future blood drug concentration prediction trajectory back to the safe and effective concentration window. The dose adjustment scheme includes specific dose change values ​​and subsequent monitoring time points.

3. The method for precisely selecting the tacrolimus treatment dose for patients with myasthenia gravis according to claim 1, characterized in that, The initial dosing concentration sequence is subjected to hysteresis compensation processing. Based on the absorption and distribution kinetics of tacrolimus in the human body, the theoretical contribution of each dosing record to the concentration values ​​of subsequent blood collection points is calculated, generating a theoretical contribution matrix, which specifically includes: Standard population pharmacokinetic parameters of tacrolimus were extracted, including the absorption rate constant, volume of distribution, and elimination rate constant. Based on the aforementioned standard population pharmacokinetic parameters, a fundamental function describing the change in blood drug concentration over time after a single dose is constructed. For each dosing record in the initial dosing concentration sequence, the theoretical concentration value generated by that dosing at each blood collection point time point in the sequence is calculated using the aforementioned basic function; All the calculated theoretical concentration values ​​are arranged in rows according to the dosing record number and columns according to the blood collection point number to form a theoretical contribution matrix. Each element in the matrix represents the theoretical contribution of a specific dosing to the concentration value of a specific blood collection point.

4. The method for precisely selecting the tacrolimus treatment dose for patients with myasthenia gravis according to claim 1, characterized in that, The method involves using the theoretical contribution matrix to perform inverse decomposition on the initial drug concentration sequence, extracting the theoretical blood drug concentration curves independently generated by a single dose, and obtaining a set of pure single-dose concentration curves, specifically including: The blood drug concentration values ​​measured at all blood sampling points in the initial drug administration concentration sequence are used to construct a concentration observation vector; Establish a system of linear equations, where the theoretical contribution matrix is ​​used as the coefficient matrix, the initial concentration peak vector generated independently by each administration is used as the unknown vector, and the concentration observation vector is used as the result vector. Solve the system of linear equations to obtain the estimated initial concentration peak value generated independently for each administration; Substituting each initial peak concentration estimate into the fundamental function describing the change in blood drug concentration over time after a single dose, a theoretical blood drug concentration change curve is generated, starting from the time of administration and covering the time range of all subsequent blood collection points. All curves together constitute a set of pure single-dose concentration curves.

5. The method for precisely selecting the tacrolimus treatment dose for patients with myasthenia gravis according to claim 1, characterized in that, The process involves matching each pure single-dose concentration curve with a preset individual pharmacokinetic parameter template to identify abnormal curve segments that deviate from the expected metabolic pattern, and merging all abnormal curve segments into an abnormal metabolic event record set. Specifically, this includes: A pre-defined individual pharmacokinetic parameter template is provided, which defines the standard shape and time parameters of the blood drug concentration curve in the absorption, distribution, and elimination phases. For each pure single-dose concentration curve, calculate its morphological similarity and time parameter difference with the individual pharmacokinetic parameter template at each phase; Set a similarity threshold and a tolerance range for the difference, and mark the curve segments with similarity below the threshold or time parameter differences exceeding the tolerance range as deviating segments; Extract all marked deviation segments, record their corresponding drug administration number, the phase in which they occurred, the type of deviation, and the quantitative value of the deviation. Each record is recorded as an abnormal metabolic event record, and all records are summarized into an abnormal metabolic event record set.

6. The method for precisely selecting the tacrolimus treatment dose for patients with myasthenia gravis according to claim 1, characterized in that, The process of associating and retrieving the patient's concurrent clinical assessment records involves pairing each clinical assessment record with its corresponding abnormal metabolic event record based on a timestamp to generate clinical metabolic association pairs. Specifically, this includes: Retrieve all clinical assessment records from the patient's medical database within the same timeframe as the initial dosing concentration sequence; Extract precise timestamp information for each clinical assessment record and each abnormal metabolic event record; Using a time window matching strategy, if the timestamp of a clinical assessment record falls within the time window of the abnormal curve segment corresponding to an abnormal metabolic event record, then the clinical assessment record is considered to be associated with the abnormal metabolic event record. Each successfully associated clinical assessment record is linked to an abnormal metabolic event record to form a clinical metabolic association pair containing complete information on metabolic abnormalities and clinical status.

7. The method for precisely selecting the tacrolimus treatment dose for patients with myasthenia gravis according to claim 1, characterized in that, The analysis of each clinical metabolic correlation pair establishes a mapping relationship between abnormal fluctuations in blood drug concentration and the appearance of specific clinical symptoms, classifying several clinically significant abnormal pharmacodynamic patterns, specifically including: Analyze each clinical metabolic association pair to extract the types of abnormal metabolic events and key clinical symptom descriptions from clinical assessment records; Perform association rule mining to find frequent combinations of specific abnormal metabolic event types and specific clinical symptom descriptions that occur simultaneously. The medical significance of the frequently discovered combinations is assessed, and those combinations that are explanatory from a pharmacological or pathological perspective are selected. Each selected combination is defined as an independent abnormal pharmacodynamic pattern, and each abnormal pharmacodynamic pattern is named and characterized to form an abnormal pharmacodynamic pattern library.

8. The method for precisely selecting the tacrolimus treatment dose for patients with myasthenia gravis according to claim 2, characterized in that, The process of reversing the individual pharmacokinetic parameter template based on the identified abnormal pharmacodynamic patterns to form a personalized pharmacokinetic parameter model adapted to the patient's current physiological state specifically includes: From the abnormal pharmacodynamic pattern library, feature parameters of recurring abnormal metabolic events corresponding to each pattern are extracted; Analyze the systematic shift direction and magnitude of the characteristic parameters to infer the individual pharmacokinetic parameter deviations they reflect; Based on the inferred deviation, the corresponding parameters in the preset individual pharmacokinetic parameter template are corrected, including the absorption rate constant, distribution volume, or elimination rate constant. The corrected parameters are used to generate a new pharmacokinetic model kernel, which is a personalized pharmacokinetic parameter model adapted to the patient's current physiological state. The process of receiving the patient's latest real-time tacrolimus blood concentration monitoring value and inputting it into a personalized pharmacokinetic parameter model for short-term concentration prediction, thereby obtaining a predicted blood concentration trajectory within a preset future time period, specifically includes: Obtain the patient's most recent tacrolimus administration time and dose, as well as at least one real-time blood drug concentration monitoring value collected since then; The latest real-time blood drug concentration monitoring value was used as the calibration point and substituted into the personalized pharmacokinetic parameter model for model state calibration. Using a corrected personalized pharmacokinetic parameter model, starting from the most recent dosing, we simulated the dynamic process of blood drug concentration changing over time within a preset future period, assuming that the current dosing regimen remains unchanged. The simulation results of this dynamic process are output, forming a predicted trajectory of blood drug concentration covering a preset future time period.

9. The method for precisely selecting the tacrolimus treatment dose for patients with myasthenia gravis according to claim 8, characterized in that, The process involves comparing the predicted blood drug concentration trajectory with a preset safe and effective concentration window. If the predicted trajectory exceeds this window, a dose adjustment warning is triggered. Specifically, this includes: A safe and effective concentration window for tacrolimus in patients with myasthenia gravis is preset, wherein the window has a lower concentration limit and an upper concentration limit; The predicted concentration value at each time point on the blood drug concentration prediction trajectory is compared with the lower limit and upper limit of the concentration, respectively. If the predicted concentration values ​​at multiple consecutive time points in the predicted trajectory are lower than the lower concentration limit, a low-dose warning is triggered; if the predicted concentration values ​​at multiple consecutive time points are higher than the upper concentration limit, a high-dose warning is triggered; the low-dose warning and the high-dose warning are collectively referred to as dose adjustment warnings. Record the specific time intervals that trigger dose adjustment warnings and the predicted extent to which the concentration exceeds the safe and effective concentration window; Based on the dose adjustment warning and combined with a personalized pharmacokinetic parameter model, a tacrolimus dose adjustment scheme is calculated that can bring the future blood drug concentration prediction trajectory back to the safe and effective concentration window. Specifically, this includes: Based on the type of dose adjustment alert triggered, determine whether the target direction of the dose adjustment is to increase or decrease the dose; Using a personalized pharmacokinetic parameter model as the computational engine, a new dosing dose value is searched through mathematical optimization under the constraint of adjusting the target direction; The new dosage value must meet the following conditions: when this new dosage is used for the next and subsequent administrations, the new blood drug concentration trajectory predicted by the model will be entirely within the safe and effective concentration window, and the center line of the trajectory should be as close as possible to the target optimal concentration value of the window. The new dosage value obtained from the search is output as the suggested adjustment dose, and the next recommended blood drug concentration monitoring time point for verifying the adjustment effect is calculated simultaneously, together forming the tacrolimus dosage adjustment plan.

10. A system for precise dose selection of tacrolimus treatment in patients with myasthenia gravis, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for accurately selecting the tacrolimus treatment dose for patients with myasthenia gravis as described in any one of claims 1 to 9.

Images