A method and system for constructing a multi-dimensional fingerprint spectrum of mulberry organic acid components

Abstract

This invention belongs to the technical field of chromatographic data processing, and relates to a method and system for constructing a multidimensional fingerprint spectrum of mulberry organic acid components. The method includes the following steps: preparing a mulberry sample solution containing the target active ingredient using solvent extraction technology; establishing a standard organic acid spectral feature template library and generating a correlation wavelength weighting coefficient table; generating a full-spectrum monitoring data stream with trigger markers; dividing and extracting a three-dimensional signal matrix of the feature region and a one-dimensional signal vector of the non-feature region; retrieving dynamic specific weighting coefficient values ​​from the wavelength weighting coefficient table based on predicted chemical identity attributes; and generating a multidimensional fingerprint spectrum of mulberry organic acids. This invention solves the problem that existing fingerprint spectrum construction technologies for traditional Chinese medicine lack the ability to dynamically integrate and specifically weight multidimensional spectral information, making it difficult to comprehensively characterize the features of complex active ingredients such as mulberry organic acids.

Description

Technical Field

[0001] This invention belongs to the technical field of chromatographic data processing and relates to a method and system for constructing multidimensional fingerprint spectra of organic acid components in mulberry. Background Technology

[0002] Mulberry leaves, mulberries, and other medicinal and edible herbs have complex compositions. Their quality evaluation needs to take into account both the integrity and specificity of the components. Traditional single content determination or conventional detection methods are difficult to fully reflect their chemical composition and quality differences. Constructing a comprehensive fingerprint spectrum has become a key requirement for the quality control of medicinal herbs. In particular, for active ingredients such as organic acids in mulberries, there is an urgent need for an efficient fingerprint spectrum construction scheme adapted to their characteristics.

[0003] To achieve comprehensive quality evaluation of Chinese medicinal materials, relevant fingerprint chromatogram construction technologies have emerged in the industry. For example, the application of the high-performance liquid chromatography fingerprint chromatogram of mulberry leaves combined with multi-component content determination in the quality evaluation of mulberry leaves (application number 202110651660.6) uses a specific chromatographic column and gradient elution conditions to establish a fingerprint chromatogram containing 31 common peaks and identify 6 characteristic peaks. By combining multi-component quantification with fingerprint chromatograms, the limitations of single-component determination are overcome, providing technical support for the quality evaluation of mulberry leaves.

[0004] The aforementioned existing technologies, based on single-dimensional detection and fixed characteristic peak analysis using high-performance liquid chromatography, can achieve the separation and quantification of some components, but they do not involve the integration and utilization of multi-dimensional spectral information. It is difficult to dynamically match the chemical identity of the analyte and perform targeted signal weighting. For components such as mulberry organic acids, which have similar structures and easily overlapping spectral features, there are problems such as insufficient fingerprint spectrum specificity and limited component resolution. They cannot uncover the deep correlation information of components and cannot meet the quality evaluation requirements for accurate characterization of complex active ingredients.

[0005] Existing fingerprinting techniques for Chinese medicinal materials lack the ability to dynamically integrate and specifically weight multidimensional spectral information, making it difficult to comprehensively characterize the features of complex active ingredients such as mulberry organic acids. This is the core technical problem that this invention aims to solve. Summary of the Invention

[0006] In a first aspect, the present invention provides a method for constructing a multidimensional fingerprint spectrum of organic acid components in mulberry, comprising the following steps:

[0007] S1. Prepare a mulberry test solution containing the target active ingredient using solvent extraction technology;

[0008] S2. Obtain a series of typical organic acid standard substances from mulberry, extract the absorption spectral characteristic parameters of the substances by scanning and detection, establish a standard organic acid spectral characteristic template library and generate a table of associated wavelength weighting coefficients;

[0009] S3. Input the mulberry test solution into the chromatographic flow path to perform separation. Generate a trigger marker by comparing the absorbance intensity of the reference wavelength with the preset occurrence threshold. Constrain the multi-channel detection hardware acquisition status according to the preset occurrence threshold and generate a full-spectrum monitoring data stream with the trigger marker.

[0010] S4. Analyze the full-spectrum monitoring data stream with trigger markers, and extract the three-dimensional signal matrix of the feature region and the one-dimensional signal vector of the non-feature region.

[0011] S5. Perform wavelength photoelectric information matching and comparison optimization operation on each acquisition time profile in the three-dimensional signal matrix of the feature region to obtain the predicted chemical identity attribute. Based on the predicted chemical identity attribute, retrieve the dynamic specific weight coefficient value from the wavelength weight coefficient table.

[0012] S6. Using the dynamic specific weight coefficient value, perform forward dimensionality reduction and weighting calculation on the three-dimensional signal matrix of the feature region to obtain the comprehensive response value sequence of the feature region. Perform fusion and splicing processing on the comprehensive response value sequence of the feature region and the one-dimensional signal vector of the non-feature region to generate a multidimensional fingerprint spectrum of mulberry organic acids.

[0013] A further aspect of the present invention involves preparing a mulberry test solution containing the target active ingredient using solvent extraction technology, comprising the following steps:

[0014] Take pure mulberry product, pulverize it to a preset mesh size, add extraction solvent, and perform ultrasonic extraction to obtain a liquid-phase extraction mixture;

[0015] The obtained liquid-phase extraction mixture was subjected to centrifugation to remove impurities and microporous membrane filtration to produce a clear mulberry test solution.

[0016] A further aspect of this invention involves establishing a standard organic acid spectral characteristic template library, comprising the following steps:

[0017] Typical organic acid standards of mulberry were injected into a chromatographic system with multi-channel detection capability to obtain standard full-wavelength spectral data of each typical organic acid standard of mulberry.

[0018] The characteristic absorption band information in the standard full-wavelength spectral data is sorted out, and the one-dimensional photometric numerical vector of the spectral profile formed after vector normalization is extracted to generate a standard identification sequence. The standard identification sequence is then used to construct a standard organic acid spectral feature template library.

[0019] A further aspect of the present invention involves generating a correlation wavelength weighting coefficient table, comprising the following steps:

[0020] The representative contribution rate parameter set covering all characteristic wavelengths is organized into a numerical vector. The representative contribution rate parameter is normalized with the sum of absorbance of the selected characteristic wavelength sequence as the denominator. The normalized numerical vector is associated with the corresponding mulberry typical organic acid standard material identity to generate a wavelength weight coefficient table.

[0021] A further aspect of the present invention involves generating a full-spectrum monitoring data stream with trigger markers, comprising the following steps:

[0022] The infusion pump driving the chromatography system delivers the mulberry test solution for column elution and separation. The absorbance intensity of the elution effluent is continuously recorded under preset reference wavelength parameters to generate a reference wavelength chromatographic signal.

[0023] The current reading value of the reference wavelength chromatographic signal is tracked in real time, and the difference between the current reading value and the preset occurrence threshold is compared.

[0024] When the comparison determines that the current reading value is greater than the preset occurrence threshold, a channel trigger command is generated to control the multi-channel detection hardware to start a high-speed polling synchronous acquisition task of a set of characteristic wavelength sequences, and record and retain the instantaneous multi-wavelength spectral segment of the clustered peak intensity that occurs within the retention time window.

[0025] When the comparison determines that the current reading value is not greater than the preset occurrence threshold, the multi-channel detection hardware is controlled to maintain the basic monitoring of the reference wavelength parameters, splicing the previously generated signals and command status records, and merging and exporting the full-spectrum monitoring data stream with trigger markers.

[0026] A further aspect of this invention involves dividing and extracting a three-dimensional signal matrix from the feature region and a one-dimensional signal vector from the non-feature region, comprising the following steps:

[0027] The high-intensity state time range containing channel trigger commands is retrieved from the full-spectrum monitoring data stream with trigger markers. Each instantaneous multi-wavelength spectral segment within the high-intensity state time range is segmented and extracted. The three-dimensional signal matrix of the feature region is reconstructed using the chromatographic retention time coordinates and the detection wavelength channel number.

[0028] The background baseline state time range in the full-spectrum monitoring data stream with trigger markers that does not contain channel trigger commands is selected, and the cross-sectional coordinates of the retention time and the corresponding basic absorbance values ​​are extracted to construct a one-dimensional signal vector in the non-feature region.

[0029] A further aspect of this invention involves retrieving dynamic specificity weighting coefficient values ​​from a wavelength weighting coefficient table based on predicted chemical identity attributes, comprising the following steps:

[0030] Along the direction of the chromatographic retention time coordinate axis, slice one by one to extract the multi-channel numerical set of the detection light intensity contained in the three-dimensional signal matrix of the feature region. Input the multi-channel numerical set of the detection light intensity into the standard organic acid spectral feature template library to initiate the cosine similarity measurement operation.

[0031] In the comparison calculation set of a single slice, the target standard organic acid identifier ranked first by cosine similarity value is locked, and the target standard organic acid identifier is used as the predicted chemical identity attribute.

[0032] Based on the predicted chemical identity attributes, the wavelength weight coefficient table is retrieved through the chemical identity index to extract the corresponding dynamic specificity weight coefficient value.

[0033] A further aspect of this invention involves inputting the multi-channel numerical set of detected light intensity into a standard organic acid spectral feature template library to initiate a cosine similarity measurement operation, including the following steps:

[0034] The three-dimensional signal matrix containing the detection light intensity multi-channel numerical set is extracted by cutting slices one by one along the direction of the chromatographic retention time coordinate axis;

[0035] Preprocessing is performed on the multi-channel numerical set of detected light intensity to generate a query vector;

[0036] Input the query vector into the standard organic acid spectral feature template library, and calculate the cosine similarity value between the query vector and each standard identification sequence stored in the standard organic acid spectral feature template library.

[0037] A further aspect of this invention involves generating a multidimensional fingerprint spectrum of mulberry organic acids, comprising the following steps:

[0038] The dynamic specific weight coefficient values ​​read are applied to the independent response parameters of each band of the three-dimensional signal matrix of the feature area on the same time slice profile by vector cross-multiplication and summation algorithm to output the comprehensive response value sequence of the feature area;

[0039] The time-stream synchronization algorithm tool is used to read the absolute chromatographic time series scale, and the coordinates of the data points of the comprehensive response value sequence of the characteristic region are embedded into the one-dimensional signal vector of the non-characteristic region to fill the original jump gaps. The replacement and splicing operation at the timestamp level is completed to generate the overall discontinuous sequence.

[0040] Boundary transition alignment processing is performed on the overall discontinuous sequence. After anchoring and flattening the baseline points of the integrated response curve of the characteristic region to the baseline height of the non-characteristic region, the envelope re-smoothing processing of the overall discontinuous sequence is performed using the spline interpolation function component to eliminate the hard steps of the stages and output the multidimensional fingerprint spectrum of mulberry organic acids.

[0041] Secondly, the present invention provides a multidimensional fingerprinting system for mulberry organic acid components, comprising the following modules:

[0042] The sample solution preparation module uses solvent extraction technology to prepare a mulberry test solution containing the target active ingredients;

[0043] The formulation dynamic analysis module obtains a series of typical organic acid standard substances from mulberry, extracts the absorption spectral characteristic parameters of the substances by scanning and detection, establishes a standard organic acid spectral characteristic template library and generates a table of associated wavelength weighting coefficients;

[0044] The trigger-type data acquisition module inputs the mulberry test solution into the chromatographic flow path to perform separation. It generates a trigger marker by comparing the absorbance intensity of the reference wavelength with the preset occurrence threshold. Based on the preset occurrence threshold, it constrains the acquisition status of the multi-channel detection hardware and generates a full-spectrum monitoring data stream with the trigger marker.

[0045] The monitoring data stream parsing module parses the full-spectrum monitoring data stream with trigger markers, and extracts the three-dimensional signal matrix of the feature region and the one-dimensional signal vector of the non-feature region.

[0046] The identity prediction and weight retrieval module performs wavelength photoelectric information matching and comparison optimization calculations for each acquisition time profile in the three-dimensional signal matrix of the feature region to obtain the predicted chemical identity attribute. Based on the predicted chemical identity attribute, the module retrieves the dynamic specific weight coefficient value from the wavelength weight coefficient table.

[0047] The fingerprint spectrum fusion generation module uses dynamic specific weight coefficients to perform forward dimensionality reduction and weighting calculation on the three-dimensional signal matrix of the feature region to obtain the comprehensive response value sequence of the feature region. The comprehensive response value sequence of the feature region is then fused and spliced ​​with the one-dimensional signal vector of the non-feature region to generate a multi-dimensional fingerprint spectrum of mulberry organic acids.

[0048] In summary, the present invention has the following beneficial technical effects:

[0049] 1. By setting signal thresholds to regulate the acquisition status of multi-channel detection hardware, high-speed synchronous acquisition of preset characteristic wavelength sequences is initiated only within the time window when the chromatographic peak signal is detected during chromatographic analysis. During the baseline stage, a basic monitoring mode with a single reference wavelength is maintained. This selective data acquisition method can reduce the generation of redundant high-dimensional data during the baseline stage when no sample components are flowing out. While retaining the spectral information of key components, it reduces the size of the original data file generated in a single analysis, alleviates the load on subsequent data storage and transmission, and improves the computational efficiency of data processing.

[0050] 2. Within the feature region, the spectral vectors of each time slice are matched with the standard organic acid spectral feature template library using cosine similarity. Based on the similarity of the spectral shape, the chemical identity of the current eluting component is dynamically determined. This method utilizes the structural information of multidimensional spectral data. Compared with methods that rely solely on retention time for component identification, this method has better adaptability when retention time drift is caused by factors such as decreased column efficiency and flow rate fluctuations. It can achieve real-time prediction of the identity of each data point within the chromatographic peak, providing a technical approach to solve the problem of chromatographic peak overlap caused by isomers and chemically similar components in complex samples, and improving the accuracy of component identification.

[0051] 3. Based on the real-time predicted chemical identity of the components, the corresponding specific weight coefficients are dynamically retrieved from the weight coefficient table and used to perform weighted fusion of multi-wavelength absorbance data at the current time point, introducing chemical information guidance into the dimensionality reduction calculation. This method can selectively enhance the signal weight of characteristic wavelengths based on the spectral contribution characteristics of the current effluent substances, while suppressing the weight of non-characteristic wavelengths or noise signals. The final generated two-dimensional fingerprint spectrum is not a simple signal superposition of data, but rather a refinement and reconstruction of species-related information in the three-dimensional data, which can improve the signal-to-noise ratio of the target components in the spectrum, while also improving the overall specificity and resolution of the fingerprint spectrum. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The drawings are used to provide a further understanding of the present invention.

[0053] Figure 1 A flowchart illustrating an embodiment of this application is disclosed.

[0054] Figure 2 A schematic diagram of the framework in the embodiments of this application is disclosed.

[0055] Figure 3 The multidimensional fingerprint spectrum of mulberry organic acids in the embodiments of this application is disclosed. Detailed Implementation

[0056] The following is in conjunction with the appendix Figures 1-2 A preferred description of the present invention is provided below.

[0057] See attached document Figure 1 This invention proposes a method for constructing a multidimensional fingerprint spectrum of organic acid components in mulberry, comprising the following steps:

[0058] S1. Prepare a mulberry test solution containing the target active ingredient using solvent extraction technology;

[0059] S2. Obtain a series of typical organic acid standard substances from mulberry, extract the absorption spectral characteristic parameters of the substances by scanning and detection, establish a standard organic acid spectral characteristic template library and generate a table of associated wavelength weighting coefficients;

[0060] S3. Input the mulberry test solution into the chromatographic flow path to perform separation. Generate a trigger marker by comparing the absorbance intensity of the reference wavelength with the preset occurrence threshold. Constrain the multi-channel detection hardware acquisition status according to the preset occurrence threshold and generate a full-spectrum monitoring data stream with the trigger marker.

[0061] S4. Analyze the full-spectrum monitoring data stream with trigger markers, and extract the three-dimensional signal matrix of the feature region and the one-dimensional signal vector of the non-feature region.

[0062] S5. Perform wavelength photoelectric information matching and comparison optimization operation on each acquisition time profile in the three-dimensional signal matrix of the feature region to obtain the predicted chemical identity attribute. Based on the predicted chemical identity attribute, retrieve the dynamic specific weight coefficient value from the wavelength weight coefficient table.

[0063] S6. Using the dynamic specific weight coefficient value, perform forward dimensionality reduction and weighting calculation on the three-dimensional signal matrix of the feature region to obtain the comprehensive response value sequence of the feature region. Perform fusion and splicing processing on the comprehensive response value sequence of the feature region and the one-dimensional signal vector of the non-feature region to generate a multidimensional fingerprint spectrum of mulberry organic acids.

[0064] In one embodiment of the present invention, step S1 includes the following steps:

[0065] Take pure mulberry product, pulverize it to a preset mesh size, add extraction solvent and perform ultrasonic extraction to obtain a liquid phase extraction mixture; perform centrifugation to remove impurities and microporous membrane filtration on the obtained liquid phase extraction mixture to generate a clear mulberry test solution.

[0066] Specifically, this embodiment first performs a sample pretreatment process for the mulberry test entity. This process is scheduled by the central control system according to a preset program. The system first instructs the sample processing device to obtain the mulberry test entity, which usually refers to dried mulberry medicinal material or its products. The sample is then placed into the pulverizing module for mechanical pulverization until the particle size distribution of the sample reaches the preset mesh size requirement. It should be understood that the preset mesh size is a pulverization particle size standard set to ensure extraction efficiency. It is usually set between 40 mesh and 60 mesh. This range is determined based on an empirical balance of increasing the contact surface area between the sample and the solvent while avoiding excessively fine powder that would lead to filtration difficulties. This process generates mulberry test entity powder.

[0067] The system controls the weighing and transfer unit to transfer the mulberry powder to be tested into an ultrasonic extraction vessel equipped with a temperature-controlled jacket, and pumps in an extraction solvent of a specific component according to the set material-to-liquid ratio. The extraction solvent is selected to efficiently dissolve the target organic acid components. For example, a methanol aqueous solution with a volume fraction of 60% to 80%, preferably 70%, is used. This concentration is determined based on the miscibility of methanol and water, which can dissolve water-soluble organic acids such as malic acid and citric acid in mulberry, and avoids the precipitation of organic acids caused by high concentrations of methanol.

[0068] Subsequently, the system activates the ultrasonic generator to apply continuous ultrasonic extraction to the solid-liquid mixture in the extraction tank. Key process parameters for ultrasonic extraction are optimized based on extensive experimental data. For example, the ultrasonic frequency is set to 40kHz, a common frequency for organic acid extraction in this field, ensuring effective dispersion of the solid matrix. The power is set to the range of 200W to 300W, the extraction time to 20min to 40min, and the extraction temperature is maintained at 20℃ to 30℃ using a temperature control jacket to prevent decomposition of heat-sensitive components. The power, frequency, and duration of this treatment are all controlled by a preset program to promote the dissolution of the target active ingredient from the solid matrix, thereby obtaining a liquid-phase extraction mixture. To remove insoluble impurities, the system pumps the liquid-phase extraction mixture into a high-speed centrifuge for centrifugation. The centrifugation speed is typically set to 10000r / min to 15000r / min for 10min to 20min, aiming to effectively settle most solid residues and suspended matter, separating the supernatant.

[0069] The supernatant is transported through a series of filters containing microporous membranes. The pore size of the microporous membranes is selected according to the requirements of the subsequent chromatographic analysis system, typically 0.22 μm or 0.45 μm, to ensure that the liquid injected into the column does not cause column head blockage. The liquid is driven by a pressure pump to pass through the filter membrane to complete the precision filtration process, ultimately producing a clear mulberry test solution free of visible particles. This solution is collected in an autosampler vial and associated with a unique sample identifier for subsequent analysis.

[0070] For example, to prepare a sample for analysis, the system first weighs 1.0 g of pre-dried mulberry fruit and pulverizes it using a pulverizer until all the sample can pass through a 50-mesh sieve, obtaining mulberry fruit powder. The mulberry fruit powder is then completely transferred to a 50 mL extraction container, and 25 mL of 70% (v / v) methanol aqueous solution is added as the extraction solvent. Subsequently, the extraction container is placed in an ultrasonic cleaning tank with a power of 250 W and a frequency of 40 kHz, and ultrasonic-assisted extraction is performed at a constant water temperature of 25 °C for 30 min. After this process, a liquid-phase extraction mixture is obtained.

[0071] The entire liquid extraction mixture was transferred to centrifuge tubes and centrifuged in a high-speed refrigerated centrifuge at 12000 rpm for 15 min. After centrifugation, the supernatant was carefully aspirated. Finally, the obtained supernatant was extracted with a syringe and injected into a 2 mL chromatographic vial through a 0.22 μm pore size polyvinylidene fluoride syringe filter, thus generating the final mulberry test solution for subsequent analysis.

[0072] In one embodiment of the present invention, step S2 includes the following steps:

[0073] Typical organic acid standards from mulberry were injected into a chromatographic system with multi-channel detection capabilities to obtain standard full-wavelength spectral data for each standard. Characteristic absorption band information from the standard full-wavelength spectral data was analyzed, and a one-dimensional photometric numerical vector of the spectral profile, after vector normalization, was extracted to generate a standard identification sequence. This standard identification sequence was then used to construct a standard organic acid spectral feature template library. Representative contribution rate parameters covering all characteristic wavelengths were organized into numerical vectors. These representative contribution rate parameters were normalized using the sum of absorbance of the selected characteristic wavelength sequences as the denominator. The normalized numerical vectors were then correlated with the corresponding typical organic acid standard from mulberry to generate a wavelength weighting coefficient table.

[0074] Specifically, this embodiment initiates the process of constructing a standard reference database, which is used to generate the comparison benchmarks required for subsequent sample analysis. The system obtains the chemical names and purity information of a series of typical organic acid standard references for mulberry from a pre-set list of standard references. Typical organic acid standard references for mulberry refer to pure organic acid compounds that are commonly found in the traditional Chinese medicine mulberry and have high content, such as malic acid, citric acid, tartaric acid, succinic acid, and quinic acid. Their purity requirement is usually above 98%, and the experimenters prepare a series of single-component standard reference solutions based on this. These solutions are then sequentially injected into a chromatographic system with multi-channel detection capability. In this embodiment, the chromatographic system with multi-channel detection capability specifically refers to a high-performance liquid chromatography system equipped with a diode array detector, which can achieve simultaneous and rapid spectral acquisition of the eluent in the spectral range of 200 nm to 400 nm.

[0075] The chromatographic system operates under preset chromatographic conditions. When each standard organic acid substance flows through the detector, the system controls the diode array detector to continuously scan within the full ultraviolet-visible spectral range, thereby acquiring and recording the standard full-wavelength spectral data corresponding to the time point of each chromatographic peak. It should be noted that the standard full-wavelength spectral data refers to a two-dimensional dataset, which records the absorbance value of a single pure substance at each wavelength point within a specified spectral range.

[0076] After acquiring the spectra of all standard substances, the data processing module performs standardization preprocessing on each standard full-wavelength spectral data. For example, vector normalization is used to eliminate the influence of concentration differences. Then, the standardized spectral profile is extracted as a standard identification sequence. The standard identification sequence is a one-dimensional numerical vector extracted from the standard full-wavelength spectral data and normalized, which is used as the spectral identity fingerprint of the substance.

[0077] The system binds the chemical identifier of each standard organic acid substance to its corresponding standard identification sequence and stores these data pairs in a structured database, thereby constructing a standard organic acid spectral feature template library. Next, for each type of standard organic acid substance in the standard organic acid spectral feature template library, the system calls a preset set of characteristic wavelength sequences. These preset characteristic wavelength sequences are a small set of specific wavelengths pre-selected based on the general spectral absorption characteristics of organic acid compounds and considerations for subsequent analytical efficiency. The characteristic wavelength sequences are selected from ultraviolet characteristic wavelengths in the range of 200nm to 300nm. This range is based on the fact that mulberry organic acids, such as malic acid, citric acid, and tartaric acid, all have characteristic absorption at the ultraviolet end of 200nm to 300nm. For example, it can be set to {210nm, 225nm, 245nm, 280nm}.

[0078] This characteristic wavelength sequence will be used for high-speed polling acquisition in subsequent real-time detection steps. For each characteristic wavelength in the sequence, the ratio of its absorbance signal value in the standard full-wavelength spectral data to the total signal intensity of that spectral data is calculated. This ratio is defined as the representative contribution rate parameter. The system organizes the calculated set of representative contribution rate parameters covering all characteristic wavelengths into a numerical vector and associates it with the corresponding standard organic acid substance identity, ultimately generating a series of specific wavelength weight coefficient tables for the representative contribution rate parameters. The calculation method is as follows:

[0079] in, Indicates the first Standard organic acids in the first Specific wavelength weighting coefficient values ​​for each characteristic wavelength; Indicates the first The standard organic acids, in their standard full-wavelength spectral data, correspond to the first... Characteristic wavelengths The absorbance reading; This indicates that the standard organic acid is in the selected The absorbance of each characteristic wavelength constitutes the sum of the absorbance across the characteristic wavelength sequence, used to achieve weighted normalization within the characteristic wavelength sequence. To avoid the sum of absorbance of the selected characteristic wavelength sequence approaching 0 or becoming negative due to extremely low concentrations, noise interference, or baseline drift, a preset minimum absorbance threshold of 0.0001 absorbance units (AU) is used. When the sum of absorbance is less than this minimum threshold, the contribution rate parameter is calculated using an equal-weighted allocation method for all characteristic wavelengths. When the absorbance value of a single characteristic wavelength is negative, it is set to 0 before participating in the calculation of the sum of absorbance, ensuring the physical meaning of the weight coefficient calculation and the robustness of the algorithm.

[0080] For example, to construct the template and weighting information for malic acid, the system first instructs the operator to prepare a 50 μg / mL malic acid standard solution. After injecting this solution into a High Performance Liquid Chromatography-Diode Array Detector (HPLC-DAD) system, standard full-wavelength spectral data is acquired at the retention time of 4.5 min for malic acid. This data shows that malic acid has a maximum absorption at 210 nm, with an absorbance of 0.65 AU. Subsequently, the system normalizes this spectral data, generating a one-dimensional vector containing 201 data points as the standard identification sequence for malic acid, and stores it along with the "malic acid" identifier in the standard organic acid spectral feature template library.

[0081] The system invokes a preset characteristic wavelength sequence {210nm, 225nm, 245nm, 280nm}. After reviewing standard full-wavelength spectral data, the absorbance of malic acid at these four wavelengths is... The absorbance values ​​are {0.65, 0.32, 0.11, 0.02} AU, respectively. Simultaneously, the system calculates the total absorbance of this spectrum at all discrete wavelengths within the 200nm to 400nm range to be 1.10 AU. The specific wavelength weighting coefficient corresponding to malic acid is calculated using the following formula: , =0.2909, = ≈0.1000, = ≈0.0182. This group of weighting coefficients { The values ​​{0.2909, 0.1000, 0.0182} are stored in a specific wavelength weighting coefficient table and bound to the identity information of malic acid. This process will be repeated for all subsequent listed standard organic acids.

[0082] In one embodiment of the present invention, step S3 includes the following steps:

[0083] The infusion pump of the driving chromatography system delivers the mulberry sample solution for column elution and separation. Under preset reference wavelength parameters, the absorbance intensity of the eluted effluent is continuously recorded to generate a reference wavelength chromatographic signal. The current reading of the reference wavelength chromatographic signal is tracked and calculated in real time, and a difference comparison is performed between the current reading and a preset occurrence threshold. When the comparison determines that the current reading is greater than the preset occurrence threshold, a channel trigger command is generated to control the multi-channel detection hardware to initiate a high-speed polling synchronous acquisition task of a set of characteristic wavelength sequences, recording the instantaneous multi-wavelength spectral fragments of the peak intensity occurring within the retention time window. When the comparison determines that the current reading is not greater than the preset occurrence threshold, the multi-channel detection hardware is controlled to maintain basic monitoring of the reference wavelength parameters, splicing the previously generated signals and command status records, and merging and exporting a full-spectrum monitoring data stream with trigger markers.

[0084] Specifically, this embodiment performs chromatographic separation and selective data acquisition of the mulberry sample solution. The system controls the autosampler to inject the mulberry sample solution prepared in the previous step into the flow path of the chromatographic system and drives the infusion pump of the chromatographic system to pump the mobile phase at a constant flow rate, pushing the sample through the chromatographic column for component elution and separation. During this process, the system instructs the multi-channel detection hardware to continuously monitor the absorbance of the elution effluent from the chromatographic column using only a single reference wavelength parameter in the initial stage. The reference wavelength is a universal monitoring wavelength selected based on the characteristic of terminal absorption of organic acid compounds, aiming to produce a basic response for all possible eluting organic acid components. It is determined that it can produce a basic response for all mulberry organic acids and can be adjusted from 205nm to 215nm according to the type of mulberry sample, with a typical setting of 210nm. This wavelength is based on the ultraviolet terminal absorption characteristics of organic acid compounds, and the basic response rate of the reference wavelength within the adjustment range to organic acids is not less than 80%. A one-dimensional reference wavelength chromatographic signal data stream is generated in real time.

[0085] An embedded real-time monitoring module in the control software continuously tracks the current reading of the chromatographic signal at the reference wavelength with a microsecond-level response speed. This monitoring module also retrieves a preset baseline signal occurrence threshold from the system configuration. The occurrence threshold is a critical absorbance value used to distinguish the true chromatographic peak signal from the instrument baseline noise. This threshold is set based on statistical analysis of the baseline signal during system blank operation, typically set to 3 to 10 times the baseline noise standard deviation. This multiple is determined to distinguish the true chromatographic peak signal from the instrument baseline noise, avoiding false triggers or missed triggers. In this embodiment, the baseline noise standard deviation of the high-performance liquid chromatography-diode array detector is 0.0003 AU to 0.0005 AU at a 210 nm reference wavelength, corresponding to an occurrence threshold range of 0.0009 AU to 0.0050 AU, with a preferred threshold of 0.0020 AU, or 5 times the baseline noise standard deviation of 0.0004 AU. To ensure detection sensitivity and avoid false triggers caused by baseline drift or random noise, a comparison is performed between the current signal reading and the occurrence threshold.

[0086] When the current reading of the reference wavelength chromatographic signal is greater than the threshold, the monitoring module immediately generates and sends a channel trigger command to the multi-channel detection hardware. The channel trigger command is an internal logic signal generated by the control software, and its state, such as binary 1 or 0, determines the data acquisition mode of the detection hardware.

[0087] The channel trigger command switches its working mode from single-wavelength monitoring to multi-wavelength polling acquisition, starts a high-speed synchronous acquisition task of a set of characteristic wavelength sequences defined in the previous steps, and records the instantaneous multi-wavelength spectral segments that represent the intensity of the clustered peak within the retention time window in the form of data packets. The retention time window refers to the duration from when the reference wavelength chromatographic signal exceeds the occurrence threshold to when it falls back below it.

[0088] A transient multi-wavelength spectral segment is a data structure that contains all absorbance values ​​synchronously acquired at a single time point under a set of characteristic wavelength sequences. Conversely, when the current reading of the reference wavelength chromatographic signal is not greater than the occurrence threshold, the system controls the multi-channel detection hardware to maintain a basic monitoring state under the reference wavelength parameters.

[0089] After the entire chromatographic analysis is completed, the data acquisition unit splices and formats the basic monitoring signals recorded at all time points, the instantaneous multi-wavelength spectral segments triggered by acquisition, and the corresponding channel trigger command status records in chronological order. Finally, it merges and exports a full-spectrum monitoring data stream with trigger markers and a non-uniform data structure. The full-spectrum monitoring data stream with trigger markers is a structured time-series data file. Each timestamp is associated with a data field. The content of this field varies depending on the trigger status. It may be a single absorbance value or a vector containing multiple absorbance values, with an additional Boolean trigger status marker.

[0090] For example, the system injects the mulberry sample solution into the chromatographic system, and sets the detector reference wavelength to 210 nm. Analysis of the baseline data during the initial 5 minutes of operation revealed a baseline noise standard deviation of 0.0004 AU, thus setting the occurrence threshold to 0.0020 AU. During the chromatographic run from 0 to 4.3 minutes, the chromatographic signal recorded at the reference wavelength of 210 nm fluctuated below 0.0015 AU, which is less than the occurrence threshold. Therefore, the system continuously recorded single-wavelength data, and the corresponding channel trigger command status was 0.

[0091] At 4.4 minutes, the signal rises to 0.0025 AU, which is greater than 0.0020 AU. The system generates a channel trigger command with a state of 1 and initiates high-speed acquisition of a set of characteristic wavelength sequences {210 nm, 225 nm, 245 nm, 280 nm}. At this time point of 4.4 minutes, the system records a momentary multi-wavelength spectral segment with absorbance readings of {0.0025, 0.0012, 0.0006, 0.0001} AU for its four channels. This high-speed acquisition mode continues until 4.8 minutes, at which point the 210 nm signal drops back to 0.0018 AU, below the trigger threshold. The channel trigger command state returns to 0, and the detector switches back to monitoring only 210 nm. Finally, the system integrates the data from the entire operation into a single file, which is the full-spectrum monitoring data stream with trigger markers. The data points between 4.4 and 4.8 minutes contain multi-dimensional spectral information and trigger markers.

[0092] In one embodiment of the present invention, step S4 includes the following steps:

[0093] The high-intensity state time range containing channel trigger commands is retrieved from the full-spectrum monitoring data stream with trigger markers. Each instantaneous multi-wavelength spectral fragment within the high-intensity state time range is segmented and extracted. A three-dimensional signal matrix of the feature region is reconstructed using the chromatographic retention time coordinates and the detection wavelength channel number. The background baseline state time range without channel trigger commands in the full-spectrum monitoring data stream with trigger markers is screened. The retention time cross-sectional coordinates and the corresponding basic absorbance values ​​are extracted to construct a one-dimensional signal vector of the non-feature region.

[0094] Specifically, this implementation performs deep analysis and data reconstruction on the full-spectrum monitoring data stream with trigger markers generated in the preceding steps. First, the data parsing module is invoked. This module uses timestamps as indexes to retrieve data entries one by one from the full-spectrum monitoring data stream with trigger markers. For each data entry, the parsing module first checks the status marker of its associated channel trigger command. When a time range with a high-level or active status marker is found, the system determines it as a high-intensity state time range. It should be noted that the high-intensity state time range is the reserved time window defined in S3.

[0095] Subsequently, each instantaneous multi-wavelength spectral fragment completely covered by this high-intensity state time range was segmented and extracted. The system uses chromatographic retention time as the first dimension and the detection wavelength channel number as the second dimension. The detection wavelength channel number is an integer index starting from 0 or 1, corresponding one-to-one with wavelength values ​​in a set of characteristic wavelength sequences. The absorbance values ​​from all extracted instantaneous multi-wavelength spectral fragments are used as matrix elements to fill the matrix, thereby reconstructing and generating a three-dimensional signal matrix of the characteristic region. This three-dimensional signal matrix of the characteristic region is a discrete data cube with time, wavelength, and absorbance as axes, and its structure is M(t, , Where t represents the chromatographic retention time point within the trigger window, λ is a specific wavelength channel in a set of characteristic wavelength sequences, and A is the absorbance value measured on the λ channel at time t. This three-dimensional signal matrix centrally embodies the complete spectral and temporal evolution information within the chromatographic peak region.

[0096] Simultaneously, the data analysis module also filters out the background baseline state time range in the full-spectrum monitoring data stream with trigger markers, where the channel trigger command status is marked as low or inactive. The background baseline state time range refers to the time segment excluding all high-intensity state time ranges within the entire chromatographic operation time. For each time point within this range, the system extracts its retention time cross-sectional coordinates and its unique corresponding baseline absorbance value. The retention time cross-sectional coordinates represent specific time values ​​on the chromatographic analysis time axis. These data pairs are organized into an ordered sequence to construct a non-characteristic region one-dimensional signal vector characterizing the eluted substances of the mobile phase. It should be noted that the non-characteristic region one-dimensional signal vector is a two-dimensional array or equivalent data structure that records the correspondence between all time points and their absorbance values ​​at the reference wavelength within the non-chromatographic peak region. This vector primarily characterizes the system's baseline state.

[0097] For example, the system receives and processes the full-spectrum monitoring data stream with trigger markers generated in the example of the preceding step S3. When scanning the data stream, the data parsing module identifies that the channel trigger command status marker remains at 1 for the period from 4.4 min to 4.8 min. Therefore, this period is identified as the high-intensity state time range, and the system then extracts all recorded instantaneous multi-wavelength spectral fragments within this time range. Assuming a data acquisition time interval of 0.1 min, the system will extract multi-wavelength data at five time points: 4.4, 4.5, 4.6, 4.7, and 4.8 min. For example, the absorbance vector at 4.5 min is {0.0080, 0.0040, 0.0020, 0.0005} AU, and at the peak at 4.6 min, it is {0.0150, 0.0075, 0.0035, 0.0008} AU. The system uses these five data vectors as the rows of a matrix and the four characteristic wavelength channels {210nm, 225nm, 245nm, 280nm} as the columns to construct a 5-row, 4-column matrix, which is part of the three-dimensional signal matrix of the characteristic region obtained in this analysis.

[0098] Simultaneously, the system selected two time periods—0 to 4.3 min and 4.9 min to the end of the analysis—as background baseline time ranges, and extracted the absorbance readings at the corresponding reference wavelength of 210 nm at these time points. For example, the absorbance was 0.0014 AU at 2.5 min and 0.0016 AU at 6.1 min. All these (time, single-wavelength absorbance) data pairs were sequentially arranged to form a one-dimensional signal vector of the non-feature region containing hundreds of data points.

[0099] In one embodiment of the present invention, step S5 includes the following steps:

[0100] Along the chromatographic retention time axis, slices are cut one by one to extract the multi-channel numerical set of detection light intensity contained in the three-dimensional signal matrix of the feature region. The multi-channel numerical set of detection light intensity is input into the standard organic acid spectral feature template library to initiate a cosine similarity measurement operation. In the comparison calculation set of a single slice, the target standard organic acid identifier ranked first in cosine similarity value is locked, and the target standard organic acid identifier is used as the predicted chemical identity attribute. Based on the predicted chemical identity attribute, the wavelength weight coefficient table is retrieved through the chemical identity index identifier, and the corresponding dynamic specific weight coefficient value is extracted.

[0101] Specifically, this implementation initiates a dynamic parameter optimization process based on spectral matching, which is specifically designed to process the three-dimensional signal matrix of the feature region generated in the preceding steps. The system extracts the set of multi-channel detection light intensity values ​​at each time point by slicing along the chromatographic retention time axis of the three-dimensional signal matrix of the feature region, according to the timestamp. The set of multi-channel detection light intensity values ​​refers to the set of values ​​at a selected time cross-section of the three-dimensional signal matrix of the feature region. The system extracts an ordered set of absorbance values ​​from all characteristic wavelength channels. The system uses this set of multi-channel values ​​of the detected light intensity as a query vector and inputs it into the standard organic acid spectral feature template library to perform cosine similarity matching to determine the chemical identity of the organic acid corresponding to the current time slice.

[0102] Because current high-speed synchronous acquisition of instantaneous multi-wavelength spectral fragments only contains a limited number of... To ensure the mathematical stability of low-dimensional vector matching and the accuracy of spectral fingerprint shape discrimination, the system employs a cosine similarity algorithm to calculate the matching degree between the query vector and the standard organic acid spectral feature template library, using a cosine similarity metric to measure spectral shape similarity. ( The calculation formula is as follows:

[0103]

[0104] in, Represents the spectral vector to be measured With standard template vector Cosine similarity between them; Refers to the first time slice in the current time slice The absorbance values ​​of each characteristic wavelength channel; Refers to the corresponding data extracted from the standard template data. Standard absorbance values ​​for each characteristic wavelength; This represents the number of wavelengths in the characteristic wavelength sequence. This formula is the standard cosine similarity formula in the field of spectral matching. It measures the directional similarity of two spectral vectors through the vector dot product, and can effectively characterize the spectral profile features of mulberry organic acids.

[0105] After all alignment calculations for a time slice are completed, the system sorts the resulting set of cosine similarity values ​​and locks the target standard organic acid identifier corresponding to the highest value. This identifier is then confirmed as the predicted chemical identity attribute of the current eluted component. The predicted chemical identity attribute is a string identifier, such as "malic acid" or "citric acid," which is the most likely compound identity given by the spectral matching algorithm at the current time point.

[0106] Based on the established predicted chemical identity attributes, the system generates a form-related retrieval password and uses this password to point to a specific wavelength weight coefficient table. The system retrieves and extracts entries from the table that perfectly match the predicted chemical identity attributes. The set of weight coefficient values ​​stored in these entries is used as the necessary calculation factors for subsequent numerical transformations. This set of dynamic-specific weight coefficient values ​​is a set of values ​​extracted from the specific wavelength weight coefficient table and bound to the predicted chemical identity attributes. Its "dynamic" nature is reflected in the fact that different weight sets may be invoked at different times due to different predicted identities, while its "specificity" means that each set of weights is calculated for a specific chemical identity.

[0107] To ensure the reliability of the predictions, the system can set a correlation threshold, such as 0.95. This threshold is based on ensuring the accuracy of the predicted chemical identity attributes and avoiding false matches. When the coefficient is ≥0.95, the spectral matching accuracy reaches over 99%. The threshold can be adjusted to 0.90 to 0.98 depending on the complexity of the mulberry sample, such as mulberry products containing additives. Within this adjustment range, the matching accuracy is no less than 90%. Only when the calculated maximum cosine similarity value exceeds this correlation threshold can the predicted chemical identity attribute be confirmed; otherwise, it can be marked as "unknown component." This process will be repeated for each time slice in the three-dimensional signal matrix of the feature region.

[0108] For example, the system processes the 3D signal matrix of the feature region generated in example S4. The system extracts a slice with a retention time of 4.6 min, whose corresponding multi-channel detection intensity values ​​are {0.0150, 0.0075, 0.0035, 0.0008}. After normalization, the query vector is obtained. The system then retrieves the standard identification sequences for malic acid and citric acid from the standard organic acid spectral feature template library, and denotes them as vectors respectively. and .

[0109] The system calculates the cosine similarity based on the formula, and then obtains the result. The result was 0.992, while The result was 0.851. Comparing these two coefficient values, 0.992 was the maximum value, and this value was greater than the preset judgment threshold of 0.95. Therefore, the system confirmed the predicted chemical identity attribute at the time point of 4.6 minutes as "malic acid".

[0110] The system uses "malic acid" as a retrieval password to search the specific wavelength weight coefficient table constructed in example S2. If the search is successful, a set of dynamic specific weight coefficient values ​​associated with "malic acid" is extracted, namely { The weighted coefficients {0.2909, 0.1000, 0.0182} will be used to process the data at time point 4.6 min. The system then processes the next time slice, for example, 4.7 min, repeating the matching and extraction process described above.

[0111] In one embodiment of the present invention, step S6 includes the following steps:

[0112] The dynamic specific weighting coefficients read are used to perform vector cross-multiplication and summation algorithms on the independent response parameters of each band in the three-dimensional signal matrix of the feature region on the same time slice profile, outputting a sequence of comprehensive response values ​​for the feature region. The time-stream synchronization algorithm is used to read the absolute chromatographic time series scale, and the coordinates of the data points of the comprehensive response value sequence of the feature region are embedded into the one-dimensional signal vector of the non-feature region to fill the original jump gaps, completing the timestamp-level replacement and splicing operation to generate an overall discontinuous sequence. Boundary transition alignment processing is performed on the overall discontinuous sequence, and the base points before and after the comprehensive response curve of the feature region are anchored and flattened to the baseline height of the non-feature region. Then, the envelope re-smoothing processing of the overall discontinuous sequence is performed using the spline interpolation function component to eliminate the hard steps of the stages, and outputting a multidimensional fingerprint spectrum of mulberry organic acids.

[0113] Specifically, this implementation aims to fuse the feature region and non-feature region data separated in the previous steps to generate the final analysis spectrum. The process first performs a forward dimensionality reduction and weighting calculation on the three-dimensional signal matrix of the feature region. The data fusion module reads a set of dynamic specific weight coefficient values ​​determined for each time slice in S5, and applies a vector cross-dot product combined with a summation algorithm to each band of the independent response parameters existing in the three-dimensional signal matrix of the feature region on the same time slice profile. In this embodiment, the vector cross-dot product combined with a summation algorithm specifically refers to performing the dot product operation of vectors, that is, multiplying the corresponding elements of two vectors with the same dimension one by one and then summing all the multiplications. This operation condenses the multidimensional spectral vector at each time point into a single scalar value, namely the comprehensive response value. After this process has traversed all time points within the feature region, a sequence of comprehensive response values ​​representing the high information density area is generated. This sequence is a one-dimensional time series array whose timestamp range coincides with the time axis of the three-dimensional signal matrix of the feature region. Its value represents the response intensity after weighted fusion, and the comprehensive response value satisfies the following formula:

[0114]

[0115] in, For a specific retention time within the feature region The comprehensive response value calculated at the point is based on the three-dimensional signal matrix of the feature region in time. The slices were prepared; In time The point is dynamically retrieved from the specific wavelength weighting coefficient table based on the prediction results of S5, corresponding to the [number]th [wavelength]. The specific weighting coefficient value of each characteristic wavelength; To be at the same time The point, read from the three-dimensional signal matrix of the feature region, corresponds to the first... Absorbance response parameters of each characteristic wavelength channel; The formula represents the total number of wavelengths contained in a set of characteristic wavelength sequences; in a physical sense, it is a weighted summation.

[0116] The system invokes a time-stream synchronization algorithm tool, which is the data processing logic used to ensure that signal points from different data sources can be accurately aligned to a unified absolute time axis to construct a complete signal stream. This tool uses the absolute occurrence time series of the entire chromatographic analysis process as the baseline scale. It embeds each data point in the dimensionality-reduced and restructured characteristic region comprehensive response value sequence into the corresponding position on the complete time axis according to its corresponding timestamp. This process essentially replaces and fills in the jump gaps in the baseline signal originally described by the non-characteristic region one-dimensional signal vector within the characteristic time period with the characteristic region comprehensive response value sequence, thus completing a timestamp-level replacement and splicing operation.

[0117] Since replacement splicing may create hard steps in signal strength at the boundary between the feature area and the non-feature area, the system first performs boundary transition alignment processing to ensure signal continuity. It reads the comprehensive response values ​​of the time nodes at both ends of the feature area and calculates the system-level offset with the one-dimensional signal vector of the adjacent non-feature area, such as the baseline value of the reference wavelength. The offset is compensated to the data points at the front and rear edges of the feature area using linear attenuation weighting, so that the front and rear base points of the comprehensive response curve of the feature area are anchored and flattened to the baseline height of the non-feature area.

[0118] The system calls the spline interpolation function component, which is a numerical analysis algorithm module. For example, cubic spline interpolation is used to smooth the baseline-aligned overall discontinuous sequence. This interpolation type is based on the continuous fitting curve and the continuity of both its first and second derivatives, effectively eliminating hard steps at the boundaries between characteristic and non-characteristic regions without distorting the chromatographic peak morphology. Alternatively, natural spline interpolation can be used, suitable for mulberry samples with broad chromatographic peaks. The smoothness of the curve after replacement is consistent with cubic spline interpolation. It fits a smooth curve passing through all data points by constructing a piecewise polynomial between the data points, performing a global envelope re-smoothing process on the overall discontinuous sequence, including the replacement splicing results, generated in the previous steps. This processing effectively eliminates intermittent hard steps, generating a continuous and smooth curve.

[0119] See attached document Figure 3 The system finally outputs and displays this curve, which is the multidimensional fingerprint spectrum of mulberry organic acids that can express the distribution pattern of trace effective acid community information. The multidimensional fingerprint spectrum of mulberry organic acids is the final product of this method and is a two-dimensional graphic. Its horizontal axis is the retention time and the vertical axis is the comprehensive response signal intensity after multidimensional information fusion and smoothing.

[0120] For example, the system continues processing the sample result from S5. At a retention time of 4.6 minutes, the system retrieves a set of dynamic specific weight coefficient values ​​associated with the identity of "malic acid". The absorbance vector {0.0150, 0.0075, 0.0035, 0.0008}AU at that moment was read.

[0121] The system performs a dot product operation to calculate the comprehensive response value of the feature region at that point.

[0122]

[0123] This calculation will be repeated for all time points between 4.4 and 4.8 minutes to generate a sequence of comprehensive response values ​​for the characteristic region.

[0124] The system constructs the final chromatographic signal sequence. For time points before 4.4 min and after 4.8 min, the system uses data from the non-feature region one-dimensional signal vector generated in example S4, for example... For the time period of 4.4 to 4.8 minutes, a newly calculated comprehensive response value is used, for example... In this way, a complete sequence is created, although it may be discontinuous at boundary points. The system calls a cubic spline interpolation function to smooth this complete sequence, eliminating signal abrupt changes that may occur, for example, between 4.3 min and 4.4 min. The smoothed curve obtained after processing is plotted and output, which is the final multidimensional fingerprint spectrum of mulberry organic acids.

[0125] See appendix Figure 2 This invention also proposes a multidimensional fingerprinting system for mulberry organic acid components, comprising the following modules:

[0126] The sample solution preparation module uses solvent extraction technology to prepare a mulberry test solution containing the target active ingredients;

[0127] The formulation dynamic analysis module obtains a series of typical organic acid standard substances from mulberry, extracts the absorption spectral characteristic parameters of the substances by scanning and detection, establishes a standard organic acid spectral characteristic template library and generates a table of associated wavelength weighting coefficients;

[0128] The trigger-type data acquisition module inputs the mulberry test solution into the chromatographic flow path to perform separation. It generates a trigger marker by comparing the absorbance intensity of the reference wavelength with the preset occurrence threshold. Based on the preset occurrence threshold, it constrains the acquisition status of the multi-channel detection hardware and generates a full-spectrum monitoring data stream with the trigger marker.

[0129] The monitoring data stream parsing module parses the full-spectrum monitoring data stream with trigger markers, and extracts the three-dimensional signal matrix of the feature region and the one-dimensional signal vector of the non-feature region.

[0130] The identity prediction and weight retrieval module performs wavelength photoelectric information matching and comparison optimization calculations for each acquisition time profile in the three-dimensional signal matrix of the feature region to obtain the predicted chemical identity attribute. Based on the predicted chemical identity attribute, the module retrieves the dynamic specific weight coefficient value from the wavelength weight coefficient table.

[0131] The fingerprint spectrum fusion generation module uses dynamic specific weight coefficients to perform forward dimensionality reduction and weighting calculation on the three-dimensional signal matrix of the feature region to obtain the comprehensive response value sequence of the feature region. The comprehensive response value sequence of the feature region is then fused and spliced ​​with the one-dimensional signal vector of the non-feature region to generate a multi-dimensional fingerprint spectrum of mulberry organic acids.

[0132] Each of the modules can be implemented in whole or in part through software, hardware, or a combination thereof. It supports hardware embedded in or independent of the processor in the computer device, and also supports software stored in the memory of the computer device, so that the processor can call and execute the operations corresponding to each of the above modules.

[0133] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for constructing a multi-dimensional fingerprint of a mulberry organic acid component, characterized in that, Includes the following steps: S1. Prepare a mulberry test solution containing the target active ingredient using solvent extraction technology; S2. Obtain a series of typical organic acid standard substances from mulberry, extract the absorption spectral characteristic parameters of the substances by scanning and detection, establish a standard organic acid spectral characteristic template library and generate a table of associated wavelength weighting coefficients; S3. Input the mulberry test solution into the chromatographic flow path to perform separation. Generate a trigger marker by comparing the absorbance intensity of the reference wavelength with the preset occurrence threshold. Constrain the multi-channel detection hardware acquisition status according to the preset occurrence threshold and generate a full-spectrum monitoring data stream with the trigger marker. S4. Analyze the full-spectrum monitoring data stream with trigger markers, and extract the three-dimensional signal matrix of the feature region and the one-dimensional signal vector of the non-feature region; specifically including: The high-intensity state time range containing channel trigger commands is retrieved from the full-spectrum monitoring data stream with trigger markers. Each instantaneous multi-wavelength spectral segment within the high-intensity state time range is segmented and extracted. The three-dimensional signal matrix of the feature region is reconstructed using the chromatographic retention time coordinates and the detection wavelength channel number. The background baseline state time range in the full-spectrum monitoring data stream with trigger markers that does not contain channel trigger commands is selected, and the cross-sectional coordinates of the retention time and the corresponding basic absorbance values ​​are extracted to construct a one-dimensional signal vector in the non-feature region. S5. Perform wavelength photoelectric information matching and comparison optimization calculations for each acquisition time profile within the three-dimensional signal matrix of the feature region to obtain predicted chemical identity attributes. Based on the predicted chemical identity attributes, retrieve the dynamic specificity weight coefficient value from the wavelength weight coefficient table. Specifically, this includes: Along the direction of the chromatographic retention time coordinate axis, slice one by one to extract the multi-channel numerical set of the detection light intensity contained in the three-dimensional signal matrix of the feature region. Input the multi-channel numerical set of the detection light intensity into the standard organic acid spectral feature template library to initiate the cosine similarity measurement operation. In the comparison calculation set of a single slice, the target standard organic acid identifier ranked first by cosine similarity value is locked, and the target standard organic acid identifier is used as the predicted chemical identity attribute. Based on the predicted chemical identity attributes, the wavelength weight coefficient table is retrieved through the chemical identity index to extract the corresponding dynamic specificity weight coefficient value. S6. Using the dynamic specific weight coefficient value, perform forward dimensionality reduction and weighting calculation on the three-dimensional signal matrix of the feature region to obtain the comprehensive response value sequence of the feature region. Perform fusion and splicing processing on the comprehensive response value sequence of the feature region and the one-dimensional signal vector of the non-feature region to generate a multidimensional fingerprint spectrum of mulberry organic acids.

2. The method for constructing a multidimensional fingerprint spectrum of mulberry organic acid components according to claim 1, characterized in that, The preparation of a mulberry test solution containing the target active ingredient using solvent extraction technology includes the following steps: Take pure mulberry product, pulverize it to a preset mesh size, add extraction solvent, and perform ultrasonic extraction to obtain a liquid-phase extraction mixture; The obtained liquid-phase extraction mixture was subjected to centrifugation to remove impurities and microporous membrane filtration to produce a clear mulberry test solution.

3. The method for constructing a multidimensional fingerprint spectrum of mulberry organic acid components according to claim 1, characterized in that, Establishing a standard organic acid spectral characteristic template library includes the following steps: Typical organic acid standards of mulberry were injected into a chromatographic system with multi-channel detection capability to obtain standard full-wavelength spectral data of each typical organic acid standard of mulberry. The characteristic absorption band information in the standard full-wavelength spectral data is sorted out, and the one-dimensional photometric numerical vector of the spectral profile formed after vector normalization is extracted to generate a standard identification sequence. The standard identification sequence is then used to construct a standard organic acid spectral feature template library.

4. The method for constructing a multidimensional fingerprint spectrum of mulberry organic acid components according to claim 1, characterized in that, Generating a table of associated wavelength weighting coefficients includes the following steps: The representative contribution rate parameter set covering all characteristic wavelengths is organized into a numerical vector. The representative contribution rate parameter is normalized with the sum of absorbance of the selected characteristic wavelength sequence as the denominator. The normalized numerical vector is associated with the corresponding mulberry typical organic acid standard material identity to generate a wavelength weight coefficient table.

5. The method for constructing a multidimensional fingerprint spectrum of mulberry organic acid components according to claim 1, characterized in that, Generating a full-spectrum monitoring data stream with trigger markers includes the following steps: The infusion pump driving the chromatography system delivers the mulberry test solution for column elution and separation. The absorbance intensity of the elution effluent is continuously recorded under preset reference wavelength parameters to generate a reference wavelength chromatographic signal. The current reading value of the reference wavelength chromatographic signal is tracked in real time, and the difference between the current reading value and the preset occurrence threshold is compared. When the comparison determines that the current reading value is greater than the preset occurrence threshold, a channel trigger command is generated to control the multi-channel detection hardware to start a high-speed polling synchronous acquisition task of a set of characteristic wavelength sequences, and record and retain the instantaneous multi-wavelength spectral segment of the clustered peak intensity that occurs within the retention time window. When the comparison determines that the current reading value is not greater than the preset occurrence threshold, the multi-channel detection hardware is controlled to maintain the basic monitoring of the reference wavelength parameters, splicing the previously generated signals and command status records, and merging and exporting the full-spectrum monitoring data stream with trigger markers.

6. The method for constructing a multidimensional fingerprint spectrum of mulberry organic acid components according to claim 1, characterized in that, The multi-channel numerical set of detected light intensity is input into the standard organic acid spectral feature template library to initiate a cosine similarity measurement operation, including the following steps: The three-dimensional signal matrix containing the detection light intensity multi-channel numerical set is extracted by cutting slices one by one along the direction of the chromatographic retention time coordinate axis; Preprocessing is performed on the multi-channel numerical set of detected light intensity to generate a query vector; Input the query vector into the standard organic acid spectral feature template library, and calculate the cosine similarity value between the query vector and each standard identification sequence stored in the standard organic acid spectral feature template library.

7. The method for constructing a multidimensional fingerprint spectrum of mulberry organic acid components according to claim 1, characterized in that, The generation of a multidimensional fingerprint of organic acids from mulberries includes the following steps: The dynamic specific weight coefficient values ​​read are applied to the independent response parameters of each band of the three-dimensional signal matrix of the feature area on the same time slice profile by vector cross-multiplication and summation algorithm to output the comprehensive response value sequence of the feature area; The time-stream synchronization algorithm tool is used to read the absolute chromatographic time series scale, and the coordinates of the data points of the comprehensive response value sequence of the characteristic region are embedded into the one-dimensional signal vector of the non-characteristic region to fill the original jump gaps. The replacement and splicing operation at the timestamp level is completed to generate the overall discontinuous sequence. Boundary transition alignment processing is performed on the overall discontinuous sequence. After anchoring and flattening the baseline points of the integrated response curve of the characteristic region to the baseline height of the non-characteristic region, the envelope re-smoothing processing of the overall discontinuous sequence is performed using the spline interpolation function component to eliminate the hard steps of the stages and output the multidimensional fingerprint spectrum of mulberry organic acids.

8. A multidimensional fingerprinting system for mulberry organic acid components, applied to the method for constructing a multidimensional fingerprinting of mulberry organic acid components as described in any one of claims 1-7, characterized in that, Includes the following modules: The sample solution preparation module uses solvent extraction technology to prepare a mulberry test solution containing the target active ingredients; The formulation dynamic analysis module obtains a series of typical organic acid standard substances from mulberry, extracts the absorption spectral characteristic parameters of the substances by scanning and detection, establishes a standard organic acid spectral characteristic template library and generates a table of associated wavelength weighting coefficients; The trigger-type data acquisition module inputs the mulberry test solution into the chromatographic flow path to perform separation. It generates a trigger marker by comparing the absorbance intensity of the reference wavelength with the preset occurrence threshold. Based on the preset occurrence threshold, it constrains the acquisition status of the multi-channel detection hardware and generates a full-spectrum monitoring data stream with the trigger marker. The monitoring data stream parsing module parses the full-spectrum monitoring data stream with trigger markers, and extracts the three-dimensional signal matrix of the feature region and the one-dimensional signal vector of the non-feature region. The identity prediction and weight retrieval module performs wavelength photoelectric information matching and comparison optimization calculations on each acquisition time profile within the three-dimensional signal matrix of the feature region to obtain predicted chemical identity attributes. Based on the predicted chemical identity attributes, it retrieves the dynamic specific weight coefficient value from the wavelength weight coefficient table. The fingerprint spectrum fusion generation module uses the dynamic specific weight coefficient value to perform forward dimensionality reduction and weighting calculations on the three-dimensional signal matrix of the feature region to obtain the comprehensive response value sequence of the feature region. It then performs fusion and splicing processing with the one-dimensional signal vector of the non-feature region to generate a multi-dimensional fingerprint spectrum of mulberry organic acids.

Images