Method for rapidly detecting content of total flavonoids in ginkgo biloba extract based on ultraviolet spectrum

Abstract

The invention provides a method for rapidly detecting the content of total flavonoids in ginkgo biloba extract based on ultraviolet spectrum, which comprises the following steps: taking a ginkgo biloba extract sample, and treating by adopting an ultrasonic-assisted extraction method to prepare flavone extract liquor; scanning the flavone extracting solution by using an ultraviolet-visible spectrophotometer to obtain ultraviolet absorption spectrum data of the flavone extracting solution in a characteristic wavelength range; and inputting the obtained ultraviolet absorption spectrum data into a pre-generated general flavone content prediction model, and directly outputting to obtain the content value of general flavone in the ginkgo leaf extract. According to the method, interference of coexisting substances in a traditional ultraviolet method is effectively overcome through a chemometrics algorithm, detection is completed within a few minutes, and the accuracy of the method is equivalent to that of a high performance liquid chromatography. The method is easy and convenient to operate, low in cost, high in efficiency and particularly suitable for rapid quality control and raw material screening in the production process of the ginkgo leaf extract.

Description

Technical Field

[0001] This invention relates to the field of chemical detection, and more specifically, to a rapid method for detecting the total flavonoid content in Ginkgo biloba leaf extract based on ultraviolet spectroscopy. Background Technology

[0002] Ginkgo biloba extract is an important raw material for drugs and health products used in the prevention and treatment of cardiovascular and cerebrovascular diseases, and its core active ingredient is flavonoids. Currently, existing technologies all use high-performance liquid chromatography (HPLC) as the legal method for determining the total flavonoid content in ginkgo biloba leaves and their preparations. Although this method yields accurate results, its inherent drawbacks severely restrict its application in modern production processes: First, sample pretreatment is extremely cumbersome, typically involving multiple solvent extractions, transfers, volume adjustments, and complex acid hydrolysis steps, with the entire process taking 2 to 4 hours or more; second, the analysis process itself is lengthy, with chromatographic separation and detection of a single sample usually requiring 30 to 60 minutes, making it difficult to meet the needs of rapid analysis of large batches of samples; third, this method consumes large amounts of high-purity organic solvents (such as methanol, acetonitrile, chloroform, etc.), resulting in high detection costs and potential environmental pollution; finally, HPLC instruments are expensive, have high maintenance costs, and require operation by specially trained technicians. All of these factors make it difficult for this method to be widely applied in production workshops, online quality control, and other similar scenarios.

[0003] To address the aforementioned issues, ultraviolet (UV) spectrophotometry has been explored for rapid screening of total flavonoids due to its speed and simplicity. However, traditional single-wavelength UV methods, which typically measure absorbance directly at 360 nm, reduce detection time to the minute level. This is because Ginkgo biloba extract has a complex composition, with numerous impurities such as biflavonoids (e.g., ginkgolic acid, isoginkgolide), phenolic acids (e.g., protocatechuic acid, gallic acid), tannins, and chlorophyll degradation products exhibiting strong and overlapping UV absorption in the 250–400 nm wavelength range. This leads to systematically high results, poor accuracy, and low repeatability, failing to meet stringent quality control requirements and only suitable for extremely rough estimations. Therefore, there is an urgent need in this field for a novel analytical method that balances detection speed and accuracy to achieve efficient and reliable quality control of the total flavonoid content in Ginkgo biloba extract. Summary of the Invention

[0004] The main objective of this invention is to provide a rapid method for detecting the total flavonoid content of Ginkgo biloba leaf extract based on ultraviolet spectroscopy, so as to at least solve the problem of insufficient accuracy of the existing ultraviolet spectrophotometric method due to severe spectral interference.

[0005] To achieve the above objectives, this invention provides a rapid method for detecting the total flavonoid content in Ginkgo biloba leaf extract based on ultraviolet spectroscopy, comprising the following steps: Ginkgo biloba leaf extract samples were processed using ultrasonic-assisted extraction to prepare flavonoid extract; The flavonoid extract was scanned using a UV-Vis spectrophotometer to obtain its UV absorption spectrum data in the characteristic wavelength range; The obtained ultraviolet absorption spectral data are input into a pre-generated total flavonoid content prediction model, and the total flavonoid content value in Ginkgo biloba extract is directly output.

[0006] Furthermore, the preparation steps of the flavonoid extract include: Weigh 0.5g~1.5g of Ginkgo biloba extract powder and place it in an Erlenmeyer flask; Add 20 mL to 60 mL of 60% to 80% methanol aqueous solution, seal tightly, and weigh. Ultrasonic treatment is performed at a power of 500W~800W, a frequency of 35kHz~45kHz, and a treatment time of 2 minutes~10 minutes. After cooling, weigh the product again and replenish the lost weight with the appropriate solvent. Shake well, filter, and collect the filtrate to obtain the flavonoid extract.

[0007] Furthermore, the characteristic wavelength range for spectral acquisition in the ultraviolet absorption spectroscopy data acquisition step is 250nm~600nm.

[0008] Furthermore, before inputting the spectral data into the prediction model, a spectral preprocessing step is included; the spectral preprocessing step includes one or more of the following: multivariate scattering correction, standard normal variable transformation, derivative method, or smoothing.

[0009] Furthermore, the total flavonoid content prediction model was obtained by establishing a mathematical relationship between the ultraviolet spectra of the Ginkgo biloba extract calibration set samples and their actual total flavonoid content values ​​measured by standard methods using partial least squares regression.

[0010] Furthermore, the standard method is high performance liquid chromatography, and the total flavonoid content refers to the total flavonol glycoside content with quercetin, kaempferol, and isorhamnetin as aglycones.

[0011] Furthermore, the methods for generating the total flavonoid content prediction model include: Model building sample set preparation: Collect multiple Ginkgo biloba extract samples with content gradients as calibration sets; Reference value determination: The true value of total flavonoid content in each sample in the calibration set was determined by high performance liquid chromatography. Spectral acquisition and preprocessing: Acquire the ultraviolet spectra of the calibration set samples and process the spectral data; Model establishment: The processed spectral data and the measured true content values ​​are used as inputs, and the partial least squares algorithm is used for training and cross-validation to generate a total flavonoid content prediction model.

[0012] Furthermore, the true value of total flavonoid content was determined using high-performance liquid chromatography (HPLC), the steps of which included: The calibration set samples were subjected to acid hydrolysis to convert flavonol glycosides into aglycones; The contents of quercetin, kaempferol, and isorhamnetin generated by hydrolysis were determined using high performance liquid chromatography. The true value of total flavonoid content can be calculated using the formula: Total flavonoid glycoside content = (quercetin content + kaempferol content + isorhamnetin content) × 2.51.

[0013] Furthermore, the method for processing the calibration set spectral data when constructing the total flavonoid content prediction model is the same as the spectral preprocessing method used for the rapid detection of total flavonoid content in Ginkgo biloba extract.

[0014] Furthermore, the performance of the total flavonoid content prediction model must meet the following requirements: validation set determination coefficient R² ≥ 0.97, and prediction root mean square error RMSEP ≤ 0.60%.

[0015] This invention provides a rapid method for detecting the total flavonoid content in Ginkgo biloba leaf extract based on ultraviolet spectroscopy, comprising the following steps: taking a Ginkgo biloba leaf extract sample, processing it using ultrasonic-assisted extraction to prepare a flavonoid extract; scanning the flavonoid extract using a UV-Vis spectrophotometer to obtain its UV absorption spectrum data within a characteristic wavelength range; inputting the obtained UV absorption spectrum data into a pre-generated total flavonoid content prediction model, directly outputting the total flavonoid content value in the Ginkgo biloba leaf extract. This invention creatively integrates efficient and green ultrasonic-assisted extraction pretreatment technology with full-band UV spectroscopy scanning and chemometric algorithms into a cohesive whole, establishing a new rapid quantitative analysis method for the total flavonoid content in Ginkgo biloba leaf extract. Its beneficial effects are significant in three aspects: First, in terms of efficiency, this method drastically reduces the analysis time for a single sample from several hours in high-performance liquid chromatography to several minutes, and significantly reduces the use and consumption of organic solvents, achieving a leapfrog improvement in analytical efficiency and a significant reduction in cost. Second, in terms of performance, its core lies in using partial least squares algorithms to extract characteristic signals related to the target components from complex spectral information, thereby intelligently correcting and eliminating spectral overlap interference caused by coexisting substances such as biflavonoids and phenolic acids. This fundamentally solves the inherent defects of poor accuracy and low specificity of traditional single-wavelength or dual-wavelength ultraviolet methods, making the analytical results highly consistent with authoritative HPLC methods. Third, in terms of application, this method is simple to operate and easy to automate, providing an unprecedentedly efficient and reliable tool for the whole-chain quality control of Ginkgo biloba leaves, from raw material warehousing and production process monitoring to rapid release of finished products, greatly promoting the modernization and intelligentization of traditional Chinese medicine quality control. Detailed Implementation

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the embodiments.

[0017] This invention aims to provide a rapid, accurate, and suitable method for detecting the total flavonoid content in Ginkgo biloba leaf extract, applicable to production process control. Its core lies in combining efficient ultrasonic extraction, rapid ultraviolet spectroscopy scanning, and a robust chemometric model.

[0018] Example 1: Preparation of flavonoid extract This embodiment details the preparation process of flavonoid extract, which significantly improves extraction efficiency by optimizing ultrasonic extraction parameters.

[0019] Accurately weigh 1.0 g of dried Ginkgo biloba extract powder (passed through a No. 4 sieve) and place it in a 100 mL Erlenmeyer flask.

[0020] Add 40 mL of 70% methanol aqueous solution precisely (i.e., the material-to-liquid ratio is 1:40 g / mL), seal tightly, and weigh.

[0021] The ultrasonic-assisted extraction step is performed in a probe-type ultrasonic disruptor or a water bath ultrasonic cleaner. When using a water bath ultrasonic cleaner, the conical flask is made of borosilicate glass, has a capacity of 100 mL, and is fixed at the same depth of 2 cm to 5 cm below the surface of the ultrasonic water bath to ensure uniformity and reproducibility of ultrasonic energy transfer. The conical flask is placed in the ultrasonic cleaner for ultrasonic treatment. The ultrasonic treatment parameters are: power 600 W, frequency 40 kHz, and treatment time 5 minutes. The water temperature should be controlled below 40℃ during the ultrasonic process, and excessive solvent evaporation can be prevented by water bath cooling.

[0022] After the ultrasound is complete, remove the conical flask and allow it to cool to room temperature before weighing it again. Make up the lost weight with a 70% methanol aqueous solution.

[0023] Shake the solution well, filter it through a 0.45μm microporous membrane, discard the initial filtrate, and collect the subsequent filtrate to obtain the flavonoid extract.

[0024] This invention involves extensive optimization experiments on key parameters of ultrasound-assisted extraction. The power range of 500W to 800W is based on the following findings: below 500W, the cavitation effect is insufficient, leading to a decrease in extraction efficiency; while above 800W, although the extraction time can be further shortened, it may result in violent solvent evaporation, temperature runaway, and degradation of certain heat-sensitive components, leading to poor reproducibility. Similarly, a processing time of 2 to 10 minutes is the optimal time window within the above power range, ensuring a continuous and stable cavitation effect sufficient to rupture plant cell walls and fully dissolve flavonoids. The solvent is a 60% to 80% methanol aqueous solution: this concentration range of methanol aqueous solution has good solubility for flavonoids while also considering safety and environmental friendliness.

[0025] In a preferred embodiment of the present invention, the ultrasonic treatment power is 600W and the treatment time is 5 minutes. Under these parameters, the optimal balance between energy consumption and time can be achieved while ensuring extraction efficiency.

[0026] As another alternative embodiment of the present invention, the flavonoid extract can also be prepared by microwave-assisted extraction. Specific parameters are as follows: accurately weigh 1.0 g of Ginkgo biloba extract powder, place it in a specially designed microwave digestion vessel, add 40 mL of 70% methanol solution, and extract for 3 minutes under microwave power of 600 W and temperature control of 60°C. After cooling, transfer to a volumetric flask, dilute to volume, and filter to obtain the final product. This method can further shorten the extraction time.

[0027] Example 2: Establishment of a prediction model for total flavonoid content This embodiment details the process of establishing a prediction model for total flavonoid content for rapid prediction.

[0028] (I) Model Building and Sample Set Preparation Eighty Ginkgo biloba extract samples with broad content gradients were collected as the calibration set. The calibration set for the total flavonoid content prediction model consisted of no fewer than 80 representative Ginkgo biloba extract samples. Representativeness was demonstrated by the following: sample sources covering major Ginkgo biloba producing areas in China (such as Pizhou in Jiangsu, Tancheng in Shandong, and Anlu in Hubei); collection time covering spring, summer, and autumn; production processes including water extraction, alcohol extraction, and macroporous resin purification; and total flavonol glycoside content measured by high-performance liquid chromatography (HPLC) uniformly distributed within the range of 0.8% to 26.0%, ensuring the model's broad applicability.

[0029] (ii) Determination of reference values ​​(using high performance liquid chromatography as the gold standard) Referring to the determination method for Ginkgo biloba leaf content in Part I of the 2020 edition of the Chinese Pharmacopoeia, the true value of total flavonol glycoside content of each sample in the calibration set was accurately determined.

[0030] The steps are as follows: Accurately weigh appropriate amounts of quercetin, kaempferol, and isorhamnetin reference standards that have been dried overnight with phosphorus pentoxide, and prepare single reference standard stock solutions containing 30 μg of each per mL with methanol. Then, accurately measure appropriate amounts of each of the above three reference standard stock solutions, place them in the same volumetric flask, dilute with methanol and make up to volume, shake well, and prepare a mixed reference standard working solution containing 12 μg of quercetin, 12 μg of kaempferol, and 6 μg of isorhamnetin per mL.

[0031] Accurately weigh approximately 0.1 g of each calibration set sample powder (passed through a No. 3 sieve) and place it in a 150 mL stoppered conical flask. Accurately add 25 mL of a methanol-25% hydrochloric acid solution (4:1, v / v), seal tightly, and weigh. Place the conical flask in a 70°C constant temperature water bath and heat under reflux for 30 minutes to carry out the acid hydrolysis reaction, ensuring that the flavonol glycosides are fully hydrolyzed to the corresponding aglycones. After the reaction is complete, rapidly cool to room temperature, remove the stopper, weigh again, and replenish the lost weight with methanol solution. Shake well. Filter through a 0.45 μm microporous membrane and use the filtrate as the test solution.

[0032] The determination was performed using high-performance liquid chromatography (HPLC), and the specific chromatographic conditions are as follows: Chromatographic column: reversed-phase C18 column packed with octadecylsilane-bonded silica gel (size: 250mm × 4.6mm, 5μm); mobile phase: methanol-0.4% phosphoric acid aqueous solution (50:50, v / v); detection wavelength: 360nm; column temperature: 30℃; flow rate: 1.0mL / min; injection volume: 10μL.

[0033] Accurately pipette 10 μL of the above mixed reference working solution and each test solution into the high-performance liquid chromatograph (HPLC), and analyze under the chromatographic conditions described above, recording the chromatograms. Measure the peak areas of quercetin, kaempferol, and isorhamnetin in the reference and test solutions, and calculate the content of the three aglycones in each test solution using the external standard method.

[0034] The total flavonol glycoside content in each calibration set sample was calculated using the following formula: Total flavonol glycoside content (%) = (quercetin content + kaempferol content + isorhamnetin content) × 2.51 In the formula, 2.51 is the conversion factor, which is based on the ratio of the average molecular weight of total flavonol glycosides in Ginkgo biloba leaves (mainly represented by quercetin-3-O-glucoside, kaempferol-3-O-rhamnoside, isorhamnetin-3-O-glucoside, etc.) to the average molecular weight of their corresponding aglycones.

[0035] The calculated total flavonol glycoside content (%) is used as the "true value" of the total flavonoid content of the sample for subsequent chemometric modeling.

[0036] (III) Spectral Acquisition and Processing For each sample in the calibration set, a flavonoid extract was prepared according to the method described in Example 1.

[0037] A UV-Vis spectrophotometer (equipped with a 1cm quartz cuvette) was used, with the corresponding solvent (70% methanol) as a blank reference. The UV absorption spectra of each flavonoid extract were scanned and collected in the wavelength range of 200nm to 600nm, with a scanning interval of 1nm. Each sample was scanned 3 times and the average spectrum was taken to reduce the error.

[0038] The acquired raw spectral data is exported and processed using chemometric software (such as The Unscrambler®). Spectral processing is a crucial step in ensuring the robustness and prediction accuracy of the subsequently established partial least squares (PLS) model. Its core purpose is to minimize the impact of physical effects (such as light scattering and baseline drift) and random noise on the spectrum, thereby extracting and enhancing the chemical information truly relevant to the total flavonoid concentration.

[0039] In this invention, commonly used preprocessing methods applicable to ultraviolet spectroscopy include, but are not limited to: Multiple scattering correction (MSC) and standard normal variable transformation (SNV): Both methods aim to eliminate baseline shift and scaling effects caused by uneven distribution, particle size differences, and minute changes in optical path in the sample solution. MSC achieves this by establishing an "ideal" spectrum and linearly correcting each sample spectrum; while SNV achieves this by centering (subtracting its mean) and standardizing (dividing by its standard deviation) each spectrum, thus more effectively correcting concentration-independent multiplicative scattering interference.

[0040] Derivative processing: Derivative processing (especially first and second derivatives) can effectively eliminate linear or slow baseline drift and significantly enhance the resolution of overlapping absorption peaks. First derivative processing can eliminate baseline shift, while second derivative processing is more effective at eliminating baseline tilt. Simultaneously, the differentiation process can amplify subtle spectral features and highlight variations in the absorption peak shape of the target component.

[0041] Smoothing: Smoothing (such as Savitzky-Golay smoothing) uses a moving window to perform polynomial fitting on the data, thereby filtering out high-frequency random noise and improving the signal-to-noise ratio of the spectrum.

[0042] Spectral preprocessing is a crucial step in ensuring the robustness and prediction accuracy of the subsequently established partial least squares (PLS) model. Due to factors such as the distribution of solid particles, differences in particle size, and slight variations in optical path length, the raw UV absorption spectra of Ginkgo biloba extract samples exhibit significant baseline drift and scattering effects. Furthermore, the absorption peaks of coexisting substances such as biflavonoids and phenolic acids in the extract severely overlap with those of the target flavonol glycosides, constituting complex spectral background interference. This invention employs a series of spectral preprocessing methods to minimize these physical interferences and random noise, thereby extracting and enhancing the chemical information truly relevant to the total flavonoid concentration. The core spectral preprocessing method in this invention is the combined use of standard normal variable transformation and first-order derivative processing.

[0043] a. The principle and function of Standard Normal Variable Transform (SNV) Principle: SNV is an algorithm designed to eliminate the effects of scattering and multiplicative interference from solid particles. It performs independent correction on each sample spectrum: first, it calculates the average absorbance of the spectrum at all wavelengths; then, it subtracts this average from the original absorbance value at each wavelength (centering); finally, it divides by the standard deviation of all absorbance values ​​in the spectrum (standardization). Its mathematical expression can be simplified to:

[0044] Among them, X SNV X represents the absorbance corrected for SNV. 原始σ represents the original absorbance, μ represents the mean absorbance of this spectrum, and σ represents the standard deviation of the absorbance of this spectrum. After SNV processing, the baseline of each spectrum is flattened to the same level and has the same scale (variance of 1).

[0045] In this invention, the core function of SNV preprocessing is to eliminate the spectral baseline shift and scaling caused by differences in the physical state of the sample (such as particle scattering), so that all spectral data return to a unified comparison benchmark. On this basis, by suppressing random noise that is unrelated to the concentration of the target substance, the essential spectral differences caused by changes in total flavonoid content are highlighted, thus laying a solid foundation for the subsequent construction of a robust and accurate quantitative model.

[0046] b. The principle and function of first derivative processing Principle: Derivative processing modifies the spectrum by calculating the rate of change of absorbance with respect to wavelength. The first derivative reflects the slope of the spectrum. In practice, the Savitzky-Golay method is often used for differentiation, which also has a smoothing function and can effectively suppress noise during the differentiation process. The geometric meaning of the first derivative is to eliminate the baseline shift (constant offset) of the spectrum. The Savitzky-Golay method slides a moving window across the spectral data. For each data point within the window, the method uses a polynomial of a preset order (in this invention, a first-order polynomial, i.e., linear fitting) to fit all wavelength points within the window and their corresponding absorbance values. Then, based on this fitted local polynomial, the first derivative value at the center point of the window is directly calculated. This method cleverly combines differentiation and smoothing. During the derivative calculation (aiming to amplify variations and sharpen peak shapes), polynomial fitting naturally filters out high-frequency random noise, thereby improving spectral resolution while ensuring the signal-to-noise ratio of the data.

[0047] This invention sets the window width of the Savitzky-Golay method to 9 data points. This parameter is a optimized and verified choice, with the following technical considerations and advantages: If the window is too narrow (e.g., less than 5 points), the smoothing effect is insufficient, making it difficult to effectively filter out noise, and the derivative spectrum may still contain too many random fluctuations, interfering with model establishment. If the window is too wide (e.g., more than 13 points), the over-smoothing effect will become prominent, potentially causing the true spectral details to be "averaged" and become blurred or even lost, which would weaken the core value of derivative processing in sharpening peak shapes and enhancing resolution. Under the spectral acquisition conditions described in this invention (scanning interval of 1 nm), a window width of 9 data points corresponds to a spectral range of approximately 8 nm. This width has been proven sufficient to effectively suppress common random noise in ultraviolet spectra, while being narrow enough to accurately preserve the key details (such as inflection points and shoulders) of the complex absorption peak shapes of flavonoids in the 250–600 nm range. This ensures that the final derivative spectrum is both smooth and stable, and features distinct, providing an optimal data foundation for the subsequent PLS model to accurately capture subtle spectral changes related to total flavonoid content.

[0048] In this invention, the core function of first-order derivative processing is to completely eliminate residual baseline drift in the spectrum and effectively separate the severely overlapping ultraviolet absorption peaks of flavonol glycosides and coexisting interfering substances by sharpening the absorption peak shape. This significantly enhances the model's ability to identify the characteristic spectra of the target analytes, thereby mathematically achieving the extraction of key information and the subtraction of background interference.

[0049] (iv) Chemometric Modeling To establish the optimal quantitative analysis model, this invention systematically compares the performance of partial least squares regression (PLSR) models under different spectral preprocessing strategies. The core idea of ​​the PLSR algorithm is to find a set of latent variables (i.e., principal factors) that best explain the covariance relationship between spectral data (X matrix) and reference measurements (Y variable, i.e., flavonoid content measured by HPLC). The modeling and comparison process is as follows: Data preparation and algorithm input: All spectral data after different preprocessing paths (including: a. no preprocessing; b. SNV preprocessing only; c. SNV combined with first derivative preprocessing) are used as the X matrix, and the corresponding HPLC determination values ​​are used as the Y variable, which together serve as the input to the PLSR algorithm.

[0050] Model Training and Principal Factor Number Optimization: For each preprocessing scheme, a series of PLSR models with principal factor numbers ranging from 1 to 20 were built. Leave-One-Out Cross-Validation (LOOCV) was used for internal validation on the training set, with the root mean square error of cross-validation (RMSECV) as the evaluation metric. The optimal principal factor number was determined by selecting a value that minimizes RMSECV or prevents it from decreasing significantly for the first time, thus ensuring the model's interpretability while strictly preventing overfitting.

[0051] Systematic Comparison and Analysis of Model Performance: The systematic comparison of model performance shows that the spectral preprocessing strategy has a decisive impact on the predictive ability of the PLSR model. The unprocessed original spectral model (number of principal factors = 5), while capturing basic information, has a relatively large prediction error (RMSEP = 0.94%). After SNV preprocessing (number of principal factors = 6), the model effectively eliminates physical scattering interference, significantly reducing the prediction error (RMSEP = 0.71%) and improving robustness. The finally selected "SNV + first derivative" preprocessing combination (number of principal factors = 6) achieves a qualitative leap in model performance, with prediction errors on both the training and test sets (RMSEC = 0.32%, RMSEP = 0.38%) reduced to the lowest levels. This confirms from an algorithmic perspective that first derivative processing, by sharpening spectral features, greatly purifies the information carried by the principal factors, enabling the model to more specifically capture spectral variations related to total flavonoid content. This mathematically achieves effective separation of coexisting interfering substances, ultimately yielding an optimal prediction model that combines high accuracy and high robustness.

[0052] Final Model Determination and Validation: Based on the above system comparisons, the PLSR model, constructed with six principal factors and using "SNV combined with first derivative" as the preprocessing method, was determined as the final prediction model for total flavonoid content. This model demonstrated near-perfect goodness of fit on the training set (R²c = 0.982), and more importantly, achieved highly consistent prediction results on an independent validation set (20 samples) (R²p = 0.974, RMSEP = 0.38%).

[0053] The above process of constructing the flavonoid content prediction model fully demonstrates that the algorithm model not only has strong learning ability, but also excellent generalization ability and robustness, which fully meets the needs of rapid and accurate detection of actual samples.

[0054] Evaluation metrics for model performance: A qualified PLS model, besides having a calibration set determination coefficient (R²c) greater than 0.95 and a calibration root mean square error (RMSEC) as small as possible, is most importantly judged by its predictive ability for external samples. This is typically evaluated using the validation set determination coefficient (R²p) and the root mean square error of prediction (RMSEP). Furthermore, the relative analysis error (RPD), the ratio of sample standard deviation to RMSEP, is also an important indicator. Generally, an RPD > 2.5 is considered to indicate good predictive ability.

[0055] Model Updates and Maintenance: Over time, changes in the origin of ginkgo raw materials, or significant alterations to the production process can weaken the representativeness of the original calibration set, leading to "model drift," where the prediction accuracy for new samples decreases. To ensure the long-term, stable application of this method to production quality control, a systematic model maintenance and update mechanism needs to be established. The core of this mechanism is to dynamically optimize the model at the algorithmic level through incremental learning. The specific steps are as follows: Model performance monitoring and drift warning: In long-term applications, the deployed PLS model is periodically evaluated using a newly acquired validation sample set with total flavonoid content accurately calibrated by high-performance liquid chromatography (HPLC). A model performance warning threshold is set: when the root mean square error of prediction (RMSEP) of the validation sample set consistently exceeds 120% of the initial model's, or the coefficient of determination (R²) falls below a certain threshold... 2 When the value remains below 0.95, it is determined that the model may be drifting, triggering the model update procedure.

[0056] Construction of the incremental calibration set: Collect samples that represent the current production status (e.g., products from new production areas or under new process parameters), typically no fewer than 15, ensuring that their total flavonoid content range covers the current production reality. Using the same HPLC method as the initial calibration set, accurately determine the true total flavonoid content of these new samples to form the incremental calibration set.

[0057] Algorithm-level reconstruction and optimization of the model: The spectral data and their true content values ​​from the incremental calibration set are merged with the original calibration set data to form a new, expanded calibration set. The expanded calibration set is retrained using the same Partial Least Squares Regression (PLSR) algorithm and the same spectral preprocessing procedure as the initial model (i.e., standard normal variable transformation combined with Savitzky-Golay first derivative processing). During this process, the optimal number of principal factors is re-determined through cross-validation. Since the data distribution may have changed, the new optimal number of principal factors may differ from the original model; this step is crucial for the algorithm to adapt to the new data characteristics. Using the expanded calibration set and the re-determined number of principal factors, a new generation of PLS ​​prediction model is rebuilt.

[0058] Model Validation and Deployment: The performance of the reconstructed model is validated using a separate validation set that is not involved in modeling. Ensure that its prediction accuracy (RMSEP, R²) recovers to or exceeds the level of the initial model. Once validation is successful, the new model replaces the old online model, completing the update.

[0059] Through the above systematic maintenance process, the detection method provided by this invention can adapt to changes in the production system, forming a self-learning and self-optimizing closed-loop system, thereby ensuring its long-term reliability, accuracy and applicability in the long-term quality control application of enterprises.

[0060] Example 3: Rapid detection of total flavonoid content in unknown samples This embodiment demonstrates how to use an established model to rapidly detect unknown Ginkgo biloba extract samples.

[0061] Take the Ginkgo biloba extract sample to be tested and prepare a solution according to the exact same method as in Example 1.

[0062] The ultraviolet absorption spectrum of the flavonoid extract was collected under the same instrument conditions.

[0063] The spectral data is preprocessed in exactly the same way as when it was modeled (i.e., SNV + first derivative).

[0064] Import the preprocessed spectral data vector into the established PLS model.

[0065] The model runs automatically and directly outputs the prediction result: the total flavonol glycoside content of the unknown sample is 21.5%.

[0066] The entire detection process, from sample preparation to obtaining results, can be completed within 10 minutes, with spectral acquisition and prediction completed within 1 minute, which is much faster than traditional HPLC methods (which usually take 3 to 4 hours).

[0067] Example 4: Comparison with traditional methods and methodological verification The rapid UV-PLS detection method established in this invention was systematically validated. By comparing it with the traditional high-performance liquid chromatography (HPLC), its accuracy, precision and repeatability were comprehensively evaluated to prove that the method can be used for practical quantitative analysis.

[0068] (a) In addition to the model calibration set, 15 ginkgo leaf extract samples from different production batches were selected, with total flavonol glycoside content covering low, medium and high ranges (approximately 5% to 25%).

[0069] The procedure was strictly followed as described in Examples 1 and 3. Flavonoid extracts were prepared independently for each sample and measured in triplicate. The average value was taken as the UV-PLS determination value for that sample.

[0070] The content determination method for Ginkgo biloba leaves in Part I of the 2020 edition of the Chinese Pharmacopoeia was used to determine the content of the 15 samples mentioned above. Each sample was measured in triplicate, and the average value was taken as the "true value" of the sample. The determination results and relative errors of the 15 samples are shown in Table 1.

[0071] Table 1 Comparison of results between UV-PLS and HPLC methods (n=3)

[0072] As shown in the table above, the absolute values ​​of the relative errors for all 15 samples are between -1.1% and +2.7%, which are far lower than the ±5% error requirement that conventional analysis requires for rapid methods, proving that the method of this invention has extremely high accuracy.

[0073] (II) Recovery Experiment: To further verify the accuracy, a recovery experiment was conducted. Nine portions of Ginkgo biloba extract sample (content 15.50%), each approximately 0.5 g, were accurately weighed. Rutin reference solutions at three levels (approximately equivalent to 80%, 100%, and 120% of the analyte content in the sample), respectively, were accurately added. Flavonoid extracts were prepared according to the method in Example 1, and the recovery rates were calculated. The results are shown in Table 2. Table 2 Results of the spiking recovery test (n=9)

[0074] The average recovery rate was between 100.0% and 100.3%, and the RSD was less than 2%, indicating that the method of the present invention has good accuracy and no obvious interference or systematic error during the determination process.

[0075] (III) Precision Test: The same flavonoid extract (sample number 8, content approximately 20.86%) was repeatedly injected and measured 6 times on a UV spectrophotometer according to the method in Example 3 to examine the instrument precision. The results are as follows: Measured values ​​(%): 20.59, 20.63, 20.55, 20.68, 20.60, 20.57, average value: 20.60%.

[0076] RSD: 0.24%.

[0077] The results show that the RSD is less than 0.5%, indicating that the measurement system (instrument and model) used in this method has excellent precision.

[0078] (iv) Repeatability Test: Six flavonoid extracts were prepared independently in parallel from the same batch of Ginkgo biloba extract samples (sample number 8), and the repeatability was tested according to the methods in Examples 1 and 3. The results are as follows: Measured values ​​(%): 20.59, 20.48, 20.71, 20.55, 20.62, 20.66, average value: 20.60%.

[0079] RSD: 0.41%.

[0080] The results show that the RSD is less than 0.5%, and the method has excellent repeatability throughout the entire process from sample pretreatment to detection.

[0081] This embodiment systematically validates the method of the present invention through accuracy (relative error, recovery rate), precision, and repeatability analysis. The results show that: The results obtained by the UV-PLS method and the HPLC method of this invention are highly consistent, with no significant difference.

[0082] The method has good precision and repeatability (RSD<0.5%), and the results are stable and reliable.

[0083] The recovery rate was close to 100%, further demonstrating the accuracy of the method.

[0084] In summary, the UV-PLS rapid detection method established in this invention meets the requirements of quantitative analysis in terms of accuracy, precision, and repeatability, and can completely replace the HPLC method for rapid quality control and finished product release inspection in the production process of Ginkgo biloba extract.

[0085] Example 5: Comparative Experiment with Aluminum Nitrate Colorimetric Method This embodiment aims to compare the UV-PLS rapid detection method of the present invention with the aluminum nitrate-sodium nitrite colorimetric method commonly used in the Chinese Pharmacopoeia and literature, in order to verify the superiority of the method of the present invention in overcoming spectral interference and improving the accuracy of measurement.

[0086] Ten Ginkgo biloba extract samples with known true contents (determined by HPLC) were randomly selected from the calibration set samples established in Example 2, covering high, medium and low content levels.

[0087] (I) Experimental methods and procedures The method of the present invention is strictly operated in accordance with the methods described in Examples 1 and 3.

[0088] Preparation of flavonoid extract: Flavonoid extract was prepared by ultrasonic extraction of 10 samples according to the method in Example 1.

[0089] Content prediction: After collecting ultraviolet spectra, the predicted content values ​​are obtained by directly importing them into the established PLS model.

[0090] Comparison method (aluminum nitrate-sodium nitrite colorimetric method) Referring to the Chinese Pharmacopoeia and related literature, the specific steps are as follows: Preparation of rutin standard solution: Accurately weigh 10.0 mg of rutin reference standard dried to constant weight at 120℃, place it in a 50 mL volumetric flask, add 70% ethanol to dissolve and dilute to the mark, shake well to obtain a reference standard stock solution with a concentration of 0.2 mg / mL.

[0091] Construction of the standard curve: Accurately measure 0.0, 1.0, 2.0, 3.0, 4.0, and 5.0 mL of the above stock solution and place them into separate 10 mL volumetric flasks. Add 70% ethanol to each flask to a final volume of 5 mL. Accurately add 0.3 mL of 5% sodium nitrite solution, shake well, and let stand for 6 minutes. Then, accurately add 0.3 mL of 10% aluminum nitrate solution, shake well, and let stand for 6 minutes. Finally, accurately add 4 mL of 4% sodium hydroxide solution, dilute to the mark with 70% ethanol, shake well, and let stand for 15 minutes. Using the first tube as a blank, measure the absorbance at a wavelength of 510 nm. Plot the standard curve with rutin concentration (μg / mL) on the x-axis and absorbance on the y-axis.

[0092] Accurately measure 1.0 mL of the flavonoid extract prepared according to the method in Example 1 and place it in a 10 mL volumetric flask. Perform the following steps as described under "Plotting the Standard Curve": Add 70% ethanol to 5 mL, then add 0.3 mL of 5% sodium nitrite solution, shake well, and let stand for 6 minutes; add 0.3 mL of 10% aluminum nitrate solution, shake well, and let stand for 6 minutes; add 4 mL of 4% sodium hydroxide solution, and finally dilute to the mark with 70% ethanol, shake well, and let stand for 15 minutes. Using the corresponding reagent as a blank, measure the absorbance at a wavelength of 510 nm. Read the total flavonoid concentration (calculated as rutin) in the flavonoid extract from the standard curve, and calculate the total flavonoid content in the sample using the following formula: Total flavonoid content (mg / g) = (C×V×N) / M In the formula: C is the concentration of rutin in the flavonoid extract (mg / mL) obtained from the standard curve; V is the final volume (mL); N is the dilution factor; and M is the sample mass (g).

[0093] (II) Experimental Results and Data Analysis Using the HPLC-determined value as the "true value," the relative errors between the measured values ​​and the true values ​​of the method of this invention (UV-PLS) and the aluminum nitrate colorimetric method were calculated respectively. The results are shown in Table 3: Table 3 Comparison of results from UV-PLS method, aluminum nitrate colorimetric method and HPLC method (n=3)

[0094] The UV-PLS method of this invention yielded results highly close to the true HPLC values, with absolute relative errors for all samples less than 3% and an average relative error of 1.6%, demonstrating its excellent accuracy. The aluminum nitrate colorimetric method, however, showed a significantly higher result systematically, with positive relative errors for all samples ranging from 12.8% to 32.3%, and an average relative error as high as 21.1%.

[0095] Error Analysis: The principle of the aluminum nitrate colorimetric method is based on the complexation colorimetric reaction between flavonoid nuclei and aluminum ions. However, non-flavonoid polyphenols such as phenolic acids and tannins present in Ginkgo biloba extract can also undergo similar reactions with aluminum ions, producing additional color development and leading to artificially high absorbance values. This is the fundamental reason for the systematically high results. The UV-PLS method used in this invention analyzes the entire band of raw ultraviolet spectrum using the PLS algorithm, intelligently identifying and subtracting background absorption interference from these non-flavonoid substances, thus obtaining results closer to the true flavonoid content.

[0096] This comparative experiment strongly demonstrates that the UV-PLS rapid detection method provided by this invention is significantly superior in accuracy to the traditional aluminum nitrate-sodium nitrite colorimetric method. It not only retains the advantages of the UV method being rapid and simple, but also fundamentally solves the serious systematic error problem caused by the poor specificity of the colorimetric method by introducing a chemometric model. This provides a reliable technical solution for achieving rapid and accurate quality control of the total flavonoid content in Ginkgo biloba extract.

[0097] In summary, this invention provides a rapid method for detecting the total flavonoid content in Ginkgo biloba leaf extract based on ultraviolet spectroscopy and partial least squares (PLS) algorithm. This method achieves high efficiency and standardization of sample pretreatment through optimized ultrasonic-assisted extraction technology, acquires rich optical information by collecting full-band ultraviolet spectra, and innovatively introduces the PLS chemometric model, successfully solving the industry problem of poor accuracy caused by spectral overlap and matrix interference in traditional ultraviolet methods.

[0098] The present invention provides a rapid method for detecting the total flavonoid content in Ginkgo biloba leaf extract based on ultraviolet spectroscopy, which has the following beneficial effects: 1. Significantly Improved Detection Efficiency: This invention constructs a complete rapid analysis process by combining high-intensity ultrasonic-assisted extraction with instantaneous ultraviolet spectroscopy scanning. The ultrasonic extraction step reduces pretreatment time from several hours in traditional methods to less than 10 minutes; while ultraviolet spectroscopy scanning and model prediction can be completed within 1 minute. The entire process, from sampling to obtaining the final result, can be controlled within 15 minutes, representing an order of magnitude improvement in efficiency compared to 3-4 hours for HPLC, truly achieving rapid detection.

[0099] 2. High accuracy and strong anti-interference capability: The core innovation of this invention lies in the introduction of a chemometric algorithm (partial least squares regression, PLS). Unlike traditional ultraviolet methods that perform simple measurements at a single wavelength, this method utilizes spectral information across the entire wavelength range (200~600nm). The PLS algorithm intelligently extracts the most relevant feature information to the target analyte (total flavonoids) from complex and overlapping spectral signals, while maximally suppressing and correcting background absorption from interfering substances such as biflavonoids and phenolic acids. In this way, this invention cleverly achieves "mathematical separation" instead of "physical separation," thereby obtaining detection results that highly match those of the HPLC method (as shown in the examples, the relative error can be stabilized within ±3%) without the need for complex HPLC separation.

[0100] 3. Simple operation and low cost: The UV-Vis spectrophotometer and ultrasonic cleaner used in this invention are conventional laboratory equipment, and their purchase and maintenance costs are far lower than those of HPLC systems. During the detection process, only a small amount of methanol is consumed as the extraction solvent, avoiding the continuous consumption of expensive chromatographic columns and large amounts of high-purity mobile phase required in HPLC methods, thus significantly reducing the cost per detection. Furthermore, the reduction in solvent consumption aligns with the development concept of green chemistry.

[0101] 4. Wide Applicability: Due to its speed, accuracy, and ease of operation, this method greatly expands its application scenarios. It can be used not only for routine laboratory analysis but also for online quality control (IPC) in pharmaceutical manufacturing companies. It enables real-time monitoring of intermediate products at various process stages, such as extraction, purification, and drying, providing data support for immediate adjustments to process parameters. Furthermore, it can be used for rapid acceptance of raw material suppliers, on-site sampling by market regulators, and rapid release from finished product warehouses, building a rapid quality assurance system spanning the entire industry chain.

[0102] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A rapid detection method for the total flavonoids content of Ginkgo biloba extract based on ultraviolet spectrum, characterized in that, The method comprises the following steps: A ginkgo leaf extract sample is taken and treated by ultrasonic-assisted extraction to prepare a flavone extract liquid; The flavone extract liquid is scanned by a UV-visible spectrophotometer to obtain UV absorption spectrum data of the liquid within a characteristic wavelength range; The obtained UV absorption spectrum data is input into a pre-generated total flavone content prediction model to directly output the content value of total flavones in the ginkgo leaf extract.

2. The method for rapid detection of the total flavone content of ginkgo biloba leaves according to claim 1, characterized in that, The flavone extract liquid preparation step comprises: 0.5-1.5 g of ginkgo leaf extract powder is weighed and placed in a conical flask; 20-60 mL of 60-80% methanol aqueous solution is added, the flask is tightly capped, and the weight is determined; Ultrasonic treatment is performed, the power of the ultrasonic treatment is 500-800 W, the frequency is 35-45 kHz, and the treatment time is 2-10 minutes; After cooling, the weight is determined again, and the lost weight is made up with the corresponding solvent; The mixture is shaken, filtered, and the filtrate is obtained as the flavone extract liquid.

3. The method according to claim 1, wherein the characteristic wavelength range for spectrum acquisition in the UV absorption spectrum data acquisition step is 250-600 nm.

4. The method according to claim 1, wherein before the spectrum data is input into the prediction model, a spectrum pretreatment step is further included; the spectrum pretreatment step comprises one or more of multivariate scatter correction, standard normal variable transformation, derivative method or smoothing treatment.

5. The method according to claim 1, wherein the total flavone content prediction model is obtained by establishing a mathematical relationship between the UV spectrum of a ginkgo leaf extract calibration set sample and the true total flavone content value of the sample measured by a standard method through a partial least squares regression method.

6. The method according to claim 5, wherein the standard method is high-performance liquid chromatography, and the total flavone content refers to the total flavone glycoside content with quercetin, kaempferol and isorhamnetin as the aglycone. The method for generating the total flavone content prediction model comprises: Calibration set sample preparation: a plurality of ginkgo leaf extract samples with content gradients are collected as a calibration set; Reference value determination: the high-performance liquid chromatography method is used to determine the true value of the total flavone content of each sample in the calibration set; Spectrum acquisition and processing: the UV spectrum of the calibration set sample is collected, and the spectrum data is processed; 7. The method for rapidly detecting the total flavone content of ginkgo biloba leaves according to claim 5, characterized in that, Model establishment: the processed spectrum data and the measured true content value are used as input, the partial least squares algorithm is used for training and cross-validation, and the total flavone content prediction model is generated. The method for determining the true value of the total flavone content by the high-performance liquid chromatography method comprises: The calibration set sample is subjected to acid hydrolysis treatment to convert flavone glycosides into aglycones; The content of quercetin, kaempferol and isorhamnetin generated by hydrolysis is determined by using a high-performance liquid chromatograph. ​ 8. The method for rapid detection of total flavonoids content of ginkgo biloba leaves according to claim 7, characterized in that, ​ ​ ​ According to the formula: total flavonol glycoside content = (quercetin content + kaempferol content + isorhamnetin content) x 2.51, the true value of the total flavonoid content is calculated.

9. The rapid detection method of the total flavonoid content of ginkgo leaf extract according to claim 7, characterized in that, The method for processing the spectral data of the calibration set when the total flavonoid content prediction model is constructed is the same as the spectral pretreatment method when the total flavonoid content of the ginkgo leaf extract is rapidly detected.

10. The rapid detection method of the total flavonoid content of ginkgo leaf extract according to claim 7, characterized in that, The performance of the total flavonoid content prediction model needs to meet: the determination coefficient R² of the validation set is greater than or equal to 0.97, and the root mean square error RMSEP is less than or equal to 0.60%.

Images