A method for optimizing design of an endogenous NO donor material based on a TCGPR method
By optimizing the design of endogenous NO donor materials using the TCGPR method, the challenges of delivery and safety of endogenous nitric oxide donor materials have been addressed, achieving efficient nitric oxide generation and promoting the development of tumor-specific gas therapy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI UNIV
- Filing Date
- 2024-05-22
- Publication Date
- 2026-07-21
AI Technical Summary
Existing endogenous nitric oxide donor materials face challenges in terms of delivery and safety, particularly the toxicity of artificial materials and the complex delivery systems of natural materials, which limits their application in nitric oxide gas therapy.
An optimization design method based on TCGPR was adopted for endogenous NO donor materials. By constructing metal-doped carbon dots, machine learning models such as Gaussian process regression tree classifier TCGPR were used to optimize the feeding ratio of L-arginine and ferric chloride hexahydrate, ultrasonic time, and hydrothermal reaction conditions to prepare iron-doped carbon quantum dots Fe-Arg-CDs. The concentration of nitric oxide produced was then measured in a hydrogen peroxide environment.
This study improved the design efficiency of endogenous nitric oxide donor materials, achieved a high concentration of nitric oxide generated, greatly accelerated the discovery and clinical translation of materials, and has potential applications in tumor-specific nitric oxide gas therapy.
Smart Images
Figure CN118471386B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of pharmaceutical technology, specifically to a method for optimizing the design of endogenous NO donor materials based on the TCGPR method. Background Technology
[0002] In nitric oxide gas synergistic therapy, nitric oxide not only kills tumor cells but also acts as a sensitizer to enhance the efficacy of other treatments. Nitric oxide concentration plays a crucial role in gas therapy: low concentrations promote tumor cell growth, while high concentrations (>1 μM) effectively inhibit tumor cell proliferation. Nitric oxide can participate in gas therapy through multiple pathways, such as disrupting mitochondria, inhibiting cellular respiration, and damaging DNA. Due to its strong diffusion and short half-life, delivery is impractical. Endogenous nitric oxide donors selectively respond to endogenous stimuli, such as pH and hydrogen peroxide. This targeted response enables the release of nitric oxide from the tumor, thus achieving tumor-specific nitric oxide gas therapy. Endogenous nitric oxide donor materials can be broadly classified into natural and artificial materials. Due to the presence of toxic functional groups, artificial materials cannot be widely used due to their lower safety profile. Natural materials with high biocompatibility usually require complex delivery systems for practical application. Therefore, exploring specific endogenous nitric oxide donors in gas therapy remains a significant challenge.
[0003] Carbon dots (CDs) possess excellent luminescent properties and biocompatibility, leading to their wide application in the biological field. With their zero-dimensional structure, carbon dots retain the chemical groups of their precursors, exhibiting similar or even superior chemical properties. Metal doping (such as Fe) affects the chemical properties of carbon dots by influencing electron transfer. Furthermore, natural L-arginine provides another pathway as an endogenous nitric oxide donor. L-arginine releases nitric oxide in the presence of hydrogen peroxide without significant toxicity to normal cells. Due to the poor pharmacokinetics and complex delivery of L-arginine, developing endogenous nitric oxide donor materials with similar structures to L-arginine holds great potential. Therefore, applying endogenous nitric oxide donors to gas therapy, using L-arginine as a precursor to construct metal-doped carbon dots, represents a novel strategy.
[0004] As a branch of artificial intelligence (AI), the integration of experimentation, computation, and machine learning (ML) has significantly accelerated the discovery, development, and application of materials compared to traditional trial-and-error and empirical methods. Pre-trained machine learning models can analyze the relationship between experimental parameters and material properties, ultimately improving material performance. In cancer treatment, machine learning methods are used to simulate cancer progression, identify informational factors related to tumor detection, and predict and diagnose specific cancer types. In biomaterials, machine learning methods can explore not only material formation (peptide hydrogels, antibacterial coatings) but also external factors, such as the response of foreign substances to materials. Therefore, integrating machine learning into the development process of constructing metal-doped carbon dots using L-arginine as a precursor will be of great significance for accelerating the discovery, development, and clinical translation of novel endogenous nitric oxide donor materials. Summary of the Invention
[0005] In view of this, the purpose of this invention is to provide a method for optimizing the design of endogenous NO donor materials based on the TCGPR method.
[0006] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:
[0007] This invention provides a method for optimizing the design of endogenous NO donor materials based on the TCGPR method, the steps of which include the following:
[0008] S1. L-arginine and ferric chloride hexahydrate were dissolved in water and sonicated to prepare a precursor solution. Then, a one-pot hydrothermal reaction was carried out at a controlled temperature. After the reaction was completed, the resulting mixture was centrifuged, the supernatant was filtered through a membrane, the filtrate was dialyzed, and freeze-dried to obtain iron-doped carbon quantum dots Fe-Arg-CDs. The obtained iron-doped carbon quantum dots Fe-Arg-CDs were placed in a hydrogen peroxide environment to determine the nitric oxide production concentration (NOR).
[0009] S2. An orthogonal experiment was designed with the feeding ratio of L-arginine and ferric chloride hexahydrate, ultrasonic time, hydrothermal reaction temperature, and hydrothermal reaction time as experimental variables. Data sets were collected by collecting the nitric oxide generation concentration of each experimental variable and the corresponding product.
[0010] S3. Construct a machine learning model based on the Gaussian Process Regression Tree Classifier (TCGPR) and train it using the collected dataset. The TCGPR machine learning model separates the collected dataset into multiple subsets with enhanced consistency, and builds a Gaussian Process Regression (GPR) machine learning model for each subset. The TCGPR machine learning model uses the radial kernel function in the Gaussian Process Regression model, combined with leave-one-out cross-validation to fit the dataset and evaluate data consistency. Considering the length scale parameter in the Gaussian radial kernel function, a global Gaussian disorder factor is defined. The dataset is partitioned by minimizing the global disorder factor, and the consistency of the dataset distribution is evaluated by examining the response surface of the Gaussian Process Regression model.
[0011] S4. A virtual experimental formulation was designed with the feeding ratio of L-arginine and ferric chloride hexahydrate, ultrasonic time, hydrothermal reaction temperature, and hydrothermal reaction time as experimental variables. The virtual experimental formulation was predicted using a trained model. The potential to generate the maximum nitric oxide concentration was evaluated using the knowledge gradient utility function, and potential endogenous nitric oxide donor candidate materials were predicted.
[0012] S5. Conduct experimental verification of promising endogenous nitric oxide donor candidates and screen out the optimal solution.
[0013] Furthermore, in S1, the centrifugation conditions were: centrifugation at 13000 rpm for 10 minutes; the membrane filtration conditions were: filtration using a 0.22 μm membrane; the dialysis conditions were: dialysis in a 1000 Da dialysis bag for 3 days; and the nitric oxide formation concentration determination conditions were: reaction in a 2 mm hydrogen peroxide environment at 37°C for 48 hours, with the nitric oxide formation concentration value determined using the Griess method.
[0014] Furthermore, in S2, the specific scheme of the orthogonal experiment is as follows: the ratio of L-arginine to ferric chloride hexahydrate is selected as 50:1, 40:1, 30:1, 20:1 and 10:1, the hydrothermal reaction temperature is selected as 180℃, 200℃ and 220℃, the hydrothermal reaction time is selected as 4 h, 6 h and 8 h, and the ultrasonic time is selected as 10 minutes, 20 minutes and 30 minutes to conduct a univariate orthogonal experiment, generating a total of 74 experimental samples.
[0015] Furthermore, in S3, the TCGPR machine learning model separates the collected dataset into four subsets with enhanced consistency, and builds a GPR machine learning model on each subset for learning.
[0016] Furthermore, in S4, 24,000 virtual samples were obtained through orthogonal design, and the potential values of all virtual samples were predicted. Three potential endogenous NO donor candidate materials were predicted. The design space was as follows: L-arginine to ferric chloride hexahydrate ratio 27:1, hydrothermal reaction temperature 150℃, hydrothermal reaction time 5h, and ultrasonic time 15 minutes, 20 minutes and 25 minutes respectively.
[0017] Furthermore, in S5, the optimal selection for the preparation of the endogenous nitric oxide donor candidate material was determined by the following parameters: L-arginine to ferric chloride hexahydrate ratio 27:1, hydrothermal reaction temperature 150℃, hydrothermal reaction time 5h, and ultrasonic time 25min.
[0018] The present invention also provides an endogenous nitric oxide donor material designed and screened by the above method, the preparation reaction parameters of which are: L-arginine to ferric chloride hexahydrate ratio 27:1, hydrothermal reaction temperature 150℃, hydrothermal reaction time 5h, and ultrasonic time 25min.
[0019] The beneficial effects of the invention are as follows: This invention provides a method for optimizing the design of endogenous nitric oxide donor materials based on the Gaussian Process Regression Tree Classifier (TCGPR). The method uses the feeding ratio of L-arginine and ferric chloride hexahydrate, ultrasonic time, hydrothermal reaction temperature, and hydrothermal reaction time as experimental variables, and the nitric oxide generation concentration (NOR) value as the performance index of the endogenous nitric oxide donor to construct an orthogonal experimental dataset. An initial machine learning model is constructed based on the TCGPR classifier and trained to obtain a prediction model. Then, a virtual experimental formulation is designed using the feeding ratio of L-arginine and ferric chloride hexahydrate, ultrasonic time, hydrothermal reaction temperature, and hydrothermal reaction time as experimental variables. The trained model is used to predict the virtual experimental formulation, and the potential for generating the maximum nitric oxide generation concentration (NOR) is evaluated using a knowledge gradient utility function. Potential endogenous nitric oxide donor candidate materials are predicted. Finally, the potential endogenous nitric oxide donor candidate materials are experimentally verified, and the optimal scheme is selected. The method described in this invention effectively improves the design efficiency of endogenous nitric oxide donor materials, accelerating their discovery and clinical translation. This method uses a Gaussian Process Regression Tree (TCGPR) classifier to construct an initial machine learning model. The training dataset is divided into multiple sub-data sets with enhanced consistency. Gaussian Process Regression Tree (TPR) learning models are then built on each sub-data set for training, achieving excellent prediction results. The Pearson correlation coefficient (Pearson R) reaches 0.81, while the Pearson correlation coefficient (Pearson R) of a single GPR learning model trained on the entire training dataset is only 0.44, significantly improving the data fitting ability of the machine learning model. The endogenous nitric oxide donor materials designed and screened using this method exhibit excellent nitric oxide generation concentrations and have potential applications in nitric oxide gas synergistic therapy. Attached Figure Description
[0020] Figure 1 A schematic diagram illustrating the strategy for designing endogenous NO donor materials based on the TCGPR method.
[0021] Figure 2 For machine learning models and data analysis, A represents the machine learning model scheme, B represents the thermodynamic correlation coefficient plot, C represents the MIC correlation plot, D represents the GPR fitting result of LOOCV on the original dataset, and E represents the GPR fitting result of LOOCV on the dataset derived from the Gaussian process regression tree classifier TCGPR. The distribution of data values is estimated using a 20-bin method. UST indicates ultrasonic time, Time represents reaction time, Rat represents raw material ratio, Temp represents reaction temperature, and NOR represents nitric oxide release.
[0022] Figure 3 The images show the transmission electron microscopy (TEM) results of iron-doped carbon quantum dots (Fe-Arg-CDs), with insets showing the particle size distribution of the Fe-Arg-CDs under high-resolution TEM (HRTEM) and other TEM images.
[0023] Figure 4 The results are as follows: A represents the fluorescence characteristics of iron-doped carbon quantum dots (Fe-Arg-CDs), with an optimal excitation wavelength of 390 nm and an emission wavelength of 470 nm; B represents the emission wavelength shift of iron-doped carbon quantum dots (Fe-Arg-CDs) under different excitation wavelengths (360 ~ 420 nm); C represents the fluorescence characteristics of arginine carbon dots (Arg-CDs), with an optimal excitation wavelength of 345 nm and an emission wavelength of 445 nm; and D represents the emission wavelength shift of arginine carbon dots (Arg-CDs) under different excitation wavelengths (320 ~ 380 nm).
[0024] Figure 5 The XPS results for iron-doped carbon quantum dots (Fe-Arg-CDs) are shown below. A represents the high-resolution XPS spectrum of C1s, B the high-resolution XPS spectrum of N1s, C the high-resolution XPS spectrum of O1s, D the high-resolution XPS spectrum of Fe2p, and E the full-scan XPS spectrum of iron-doped carbon quantum dots (Fe-Arg-CDs). The high-resolution XPS spectra of C1s and N1s indicate the presence of additional binding sites, suggesting successful iron incorporation into the Fe-Arg-CDs. The high-resolution XPS spectrum of Fe2p shows two distinct peaks, corresponding to the Fe3+ and Fe2+ states, respectively. The peaks at 709.1 eV and 723.5 eV indicate the presence of Fe in the iron-doped carbon quantum dots (Fe-Arg-CDs). 2+ The peaks at 716 eV and 723 eV indicate that Fe 3+ The existence of.
[0025] Figure 6 The XPS results are for arginine carbon dots (Arg-CDs). In this figure, A is the full-scan XPS spectrum of arginine carbon dots (Arg-CDs), B is the high-resolution XPS spectrum of C1s, C is the high-resolution XPS spectrum of N1s, and D is the high-resolution XPS spectrum of O1s.
[0026] Figure 7 These are the results of the FT-IR test.
[0027] Figure 8 For zeta potential testing.
[0028] Figure 9The results are as follows: fluorescence lifetime test analysis results, where A is the fluorescence lifetime of arginine carbon dots (Arg-CDs) and B is the fluorescence lifetime of iron-doped carbon quantum dots (Fe-Arg-CDs).
[0029] Figure 10 The results are from an in vitro nitric oxide release experiment. In this experiment, A represents the nitric oxide release from L-arginine, arginine carbon dots (Arg-CDs), and iron-doped carbon quantum dots (Fe-Arg-CDs) under the same conditions. In this experiment, B represents the effect of different concentrations of hydrogen peroxide on the nitric oxide release from iron-doped carbon quantum dots (Fe-Arg-CDs).
[0030] Figure 11 The model is a DFT model, where A is the charge density difference of iron-doped carbon quantum dots Fe-Arg-CDs, and the CN* bond length of the iron-doped carbon quantum dot Fe-Arg-CDs system decreases from 1.29178 Ȧ to 1.29164 Ȧ; B is the electronic density of states of iron-doped carbon quantum dots Fe-Arg-CDs; C is the charge density difference of arginine carbon dots Arg-CDs; and D is the electronic density of states of arginine carbon dots Arg-CDs.
[0031] Figure 12 denoted as AB, where AB is the electronic density of states of iron-doped carbon quantum dots Fe-Arg-CDs and CD is the electronic density of states of arginine carbon dots Arg-CDs. Detailed Implementation
[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the following embodiments are merely illustrative and explanatory of the invention and should not be construed as limiting the scope of protection of the invention. All technologies implemented based on the above content of this invention are covered within the scope of protection intended by this invention. Unless otherwise stated, the raw materials and reagents used in the following embodiments are commercially available products or can be prepared by known methods. In this invention, TCGPR is short for Gaussian Process Regression Tree Classifier, and GPR is short for Gaussian Process Regression.
[0033] This invention discloses a method for optimizing the design of endogenous nitric oxide donor materials based on the Gaussian process regression tree classifier TCGPR method. Figures 1-2 The steps include the following:
[0034] S1. L-arginine and ferric chloride hexahydrate were dissolved in water and sonicated to prepare a precursor solution. Then, a one-pot hydrothermal reaction was carried out at a controlled temperature. After the reaction was completed, the resulting mixture was centrifuged, the supernatant was filtered through a membrane, the filtrate was dialyzed, and freeze-dried to obtain iron-doped carbon quantum dots Fe-Arg-CDs. The obtained iron-doped carbon quantum dots Fe-Arg-CDs were placed in a hydrogen peroxide environment to determine the nitric oxide production concentration (NOR).
[0035] S2. An orthogonal experiment was designed with the feeding ratio of L-arginine and ferric chloride hexahydrate, ultrasonic time, hydrothermal reaction temperature, and hydrothermal reaction time as experimental variables. Data sets were collected by collecting the nitric oxide generation concentration of each experimental variable and the corresponding product.
[0036] S3. Construct a machine learning model based on the Gaussian Process Regression Tree Classifier (TCGPR) and train it using the collected dataset. The TCGPR machine learning model separates the collected dataset into multiple subsets with enhanced consistency, and builds a Gaussian Process Regression (GPR) machine learning model for each subset. The TCGPR machine learning model uses the radial kernel function in the Gaussian Process Regression model, combined with leave-one-out cross-validation to fit the dataset and evaluate data consistency. Considering the length scale parameter in the Gaussian radial kernel function, a global Gaussian disorder factor is defined. The dataset is partitioned by minimizing the global disorder factor, and the consistency of the dataset distribution is evaluated by examining the response surface of the Gaussian Process Regression model.
[0037] S4. A virtual experimental formulation was designed with the feeding ratio of L-arginine and ferric chloride hexahydrate, ultrasonic time, hydrothermal reaction temperature, and hydrothermal reaction time as experimental variables. The virtual experimental formulation was predicted using a trained model. The potential to generate the maximum nitric oxide concentration was evaluated using the knowledge gradient utility function, and potential endogenous nitric oxide donor candidate materials were predicted.
[0038] S5. Conduct experimental verification of promising endogenous nitric oxide donor candidates and screen out the optimal solution.
[0039] In some embodiments, in S1, the centrifugation conditions are: centrifugation at 13000 rpm for 10 minutes; the membrane filtration conditions are: filtration using a 0.22 μm membrane; the dialysis conditions are: dialysis in a 1000 Da dialysis bag for 3 days; and the nitric oxide formation concentration determination conditions are: reaction at 37°C for 48 hours in a 2 mm hydrogen peroxide environment, with the nitric oxide formation concentration value determined by the Griess method.
[0040] In some embodiments, in S2, the specific scheme of the orthogonal experiment is as follows: the ratio of L-arginine to ferric chloride hexahydrate is selected as 50:1, 40:1, 30:1, 20:1 and 10:1, the hydrothermal reaction temperature is selected as 180℃, 200℃ and 220℃, the hydrothermal reaction time is selected as 4 h, 6 h and 8 h, and the ultrasonic time is selected as 10 minutes, 20 minutes and 30 minutes to conduct a univariate orthogonal experiment, generating a total of 74 experimental samples.
[0041] In some embodiments, in S3, the TCGPR machine learning model separates the collected dataset into four subsets with enhanced consistency, and builds a GPR machine learning model on each subset for learning.
[0042] In some embodiments, in S4, 24,000 virtual samples are obtained through orthogonal design, the potential values of all virtual samples are predicted, and three potential endogenous NO donor candidate materials are predicted. The design space is as follows: L-arginine to ferric chloride hexahydrate ratio 27:1, hydrothermal reaction temperature 150°C, hydrothermal reaction time 5h, and ultrasonic time 15 minutes, 20 minutes and 25 minutes respectively.
[0043] In some embodiments, in S5, the optimal selection for the preparation of the endogenous nitric oxide donor candidate material is as follows: L-arginine to ferric chloride hexahydrate ratio 27:1, hydrothermal reaction temperature 150°C, hydrothermal reaction time 5h, and ultrasonic time 25min.
[0044] The present invention also discloses the endogenous nitric oxide donor material designed and screened by the above method, the preparation reaction parameters of which are: L-arginine to ferric chloride hexahydrate ratio 27:1, hydrothermal reaction temperature 150℃, hydrothermal reaction time 5h, and ultrasonic time 25min.
[0045] To enable those skilled in the art to more clearly understand the technical solution of the present invention, the technical solution of the present invention will be described in detail below with reference to specific embodiments. The chemical materials and instruments involved in the embodiments are described below:
[0046] 1.1 Experimental Materials
[0047] L-arginine, sodium citrate, and ferric chloride hexahydrate (FeCl3·6H2O) were purchased from Aladdin Chemical Reagent Co., Ltd.; Griess reagent was purchased from Beijing Beyotime Biotechnology Co., Ltd.; all solutions were prepared using ultrapure water.
[0048] 1.2 Instruments
[0049] The morphology of the samples was observed using a Tecnai G2 F20 S-Twin TMP transmission electron microscope (TEM, FEI, USA). The absorbance of the samples was measured using a Lambda 750 spectrophotometer (Perkins-Elmer, USA). Fluorescence spectra were observed using an FS5 fluorescence spectrometer (Edinburgh, UK). X-ray photoelectron spectroscopy (XPS) was performed using a Kratos Axis Ultra DLD instrument (Kratos, UK). Fourier transform infrared (FTIR) analysis was performed using a Nicolette iS50 FTIR spectrometer (ThermoFisher Scientific, USA).
[0050] Example 1: Optimized design of endogenous nitric oxide donor material (iron-doped carbon quantum dots Fe-Arg-CDs)
[0051] S1. L-arginine and ferric chloride hexahydrate were dissolved in water and sonicated to prepare a precursor solution. A one-pot hydrothermal reaction was then performed with controlled temperature. After the reaction, the mixture was centrifuged at 13000 rpm for 10 minutes, and the supernatant was collected and filtered through a 0.22 μm membrane. The sample was then dialyzed in a 1000 Da dialysis bag for 3 days and freeze-dried to collect the sample. The collected sample was placed in a 2 mM hydrogen peroxide environment and reacted at 37°C for 48 hours. The nitric oxide production concentration (NOR) was determined using the Griess method.
[0052] S2. An orthogonal experiment was designed with the feeding ratio of L-arginine and ferric chloride hexahydrate, ultrasonic time, hydrothermal reaction temperature, and hydrothermal reaction time as experimental variables. Data sets were collected for each experimental variable and the corresponding nitric oxide concentration values of the products. The specific scheme of the orthogonal experiment was as follows: the ratio of L-arginine to ferric chloride hexahydrate was selected as 50:1, 40:1, 30:1, 20:1, and 10:1; the hydrothermal reaction temperature was selected as 180℃, 200℃, and 220℃; the hydrothermal reaction time was selected as 4h, 6h, and 8h; and the ultrasonic time was selected as 10 minutes, 20 minutes, and 30 minutes. A total of 74 experimental samples were generated.
[0053] S3. Construct an initial machine learning model based on the Gaussian Process Regression Tree Classifier (TCGPR) and train it using the collected dataset. Figure 1 The TCGPR machine learning model separates the collected dataset into multiple subsets with enhanced consistency, and builds a Gaussian process regression (GPR) machine learning model on each subset for learning.
[0054] Specifically, the Gaussian Process Regression Tree Classifier (TCGPR) considers the scale parameter in the Gaussian radial kernel function, defines a global disorder factor, and combines leave-one-out cross-validation to fit the dataset. By minimizing the global disorder factor, the dataset is partitioned, resulting in multiple subsets with enhanced consistency. The TCGPR algorithm in this model is as follows:
[0055]
[0056] Here, ‖∙‖ represents the Euclidean distance between the parameter vectors, Xi and Xj represent the baseline parameter vectors, and 𝜗 represents the scale parameter, which has a complex relationship with the data dimension. A smaller 𝜗 means that the correlation between the data is low. This low correlation corresponds to the machine learning prediction forming an unreasonable sharp fitting surface.
[0057] In this model, the global disorder factor is defined as follows:
[0058]
[0059] in, The scale parameter is represented by n, which represents the number of data points in the given dataset during leave-one-out cross-validation evaluation. This represents the average length scale in the Gaussian radial kernel function.
[0060] In each leave-one-out cross-validation process, a length scale θ of the Gaussian radial kernel is obtained through maximum likelihood estimation. i , to obtain n θ i Values. A logarithmic transformation is used to enhance reflection across different length scales. The global disorder factor assesses the consistency of the dataset by examining the correlations between data points.
[0061] In this model, by sequentially eliminating data from the input dataset, the global disorder factor among the remaining data is reduced. The algorithm stops when a chain of datasets with different sizes is constructed, where each node in the chain represents a subset of a specific data size. The data size of the subsets decreases sequentially along this chain. Each node in the chain is obtained during the sequential elimination of data, and the determination coefficient R for different datasets is calculated along the path. 2 ,by It indicates. Among them, Indicates the starting terminal. This indicates the end of the terminal. If the condition is met... Where 𝜂 is the relaxation coefficient (default is 𝜂 = 0), then The algorithm stops here, splits the input dataset in two, and recursively applies the method to the remaining dataset, resulting in multiple subsets with enhanced consistency. The TCGPR machine learning model can be implemented using the TCGPR software downloaded from https: / / github.com / Bin-Cao / TCGPR. Specifically, the TCGPR machine learning model separates the collected dataset into four subsets with enhanced consistency, and builds a Gaussian Process Machine Learning model (GPR) for learning on each subset.
[0062] S4. Using the feeding ratio of L-arginine and ferric chloride hexahydrate, ultrasonic time, hydrothermal reaction temperature, and hydrothermal reaction time as experimental variables, 24,000 virtual experimental formulations were designed. A trained model was used to predict the potential of these virtual formulations to generate the maximum nitric oxide concentration. The knowledge gradient utility function was used to evaluate the potential of these formulations, thus predicting potential endogenous nitric oxide donor candidate materials. Specifically, a Gaussian process model (GPR) was established on each of the four subsets with enhanced consistency. The four established GPR machine learning models were input into Bayesian optimization software, along with the 24,000 designed virtual samples. The knowledge gradient utility function was used to predict the most promising samples. The Bayesian optimization software can be implemented using Bgolearn software downloaded from https: / / github.com / Bin-Cao / Bgolearn. Three potential endogenous nitric oxide donor candidate materials were predicted, and the design space is as follows:
[0063] Hydrothermal reaction temperature (°C) 150 The ratio of L-arginine to ferric chloride hexahydrate 27:1 Hydrothermal reaction time (h) 5 Ultrasound time (min) 15, 20, 25
[0064] S5. Experimentally validate promising endogenous nitric oxide donor candidates and select the optimal solution. Specifically, the selected optimal solution is as follows:
[0065] Hydrothermal reaction temperature (°C) 150 The ratio of L-arginine to ferric chloride hexahydrate 27:1 Hydrothermal reaction time (h) 5 Ultrasound time (min) 25
[0066] Example 2 Synthesis of endogenous nitric oxide donor material (iron-doped carbon quantum dots Fe-Arg-CDs)
[0067] The optimal design scheme was used to synthesize endogenous nitric oxide donor materials (iron-doped carbon quantum dots Fe-Arg-CDs). The specific method was as follows: L-arginine and ferric chloride hexahydrate (the ratio of L-arginine to ferric chloride hexahydrate was 27:1) were dissolved in water and sonicated for 25 min to prepare a precursor solution. Then, a one-pot hydrothermal method was used to control the temperature at 150℃ for 5 h. After the reaction was completed, the mixture was centrifuged at 13000 rpm for 10 min, the supernatant was collected and filtered through a 0.22 μm membrane, and then the sample was dialyzed in a 1000 Da dialysis bag for 3 days and freeze-dried to collect the endogenous nitric oxide donor material (iron-doped carbon quantum dots Fe-Arg-CDs).
[0068] Comparative Example 1
[0069] Arginine carbon dots Arg-CDs (Arg-CDs) were synthesized under the reaction conditions of Example 2 without the addition of ferric chloride hexahydrate.
[0070] Test Example 1: Basic Material Properties Test
[0071] Transmission electron microscopy (TEM) revealed a precisely defined spherical structure in iron-doped carbon quantum dots (Fe-Arg-CDs) with an average diameter of 2.5 nm and a lattice spacing of 0.21 nm, corresponding to the (001) crystal plane of graphite. Figure 3 ).
[0072] Fluorescence spectroscopy analysis showed that the optimal excitation wavelength for iron-doped carbon quantum dots (Fe-Arg-CDs) was 390 nm, and the emission wavelength was 470 nm. Figure 4 AB), the optimal excitation wavelength for arginine carbon dots Arg-CDs is 345 nm, and the emission wavelength is 445 nm ( Figure 4 Both Fe-Arg-CDs and iron-doped carbon quantum dots exhibit significant fluorescence properties and typical excitation-dependent behavior. Compared to arginine-doped Arg-CDs, the introduction of metal doping into iron-doped carbon quantum dots (Fe-Arg-CDs) significantly alters their fluorescence performance due to the additional charge transfer process.
[0073] X-ray photoelectron spectroscopy (XPS) analysis of elemental distribution revealed distinct peaks corresponding to C 1s, N 1s, O 1s, and Fe 2p in the XPS spectra of iron-doped carbon quantum dots (Fe-Arg-CDs). The high-resolution XPS spectra of C 1s and N 1s indicate the presence of additional binding sites, suggesting successful iron incorporation into the Fe-Arg-CDs. The high-resolution XPS spectrum of Fe 2p showed two distinct pairs of peaks, corresponding to Fe 1s, N 1s, O 1s, and Fe 2p peaks, respectively. 3+ and Fe 2+ The peaks at 709.1 eV and 723.5 eV indicate the presence of Fe in the iron-doped carbon quantum dots Fe-Arg-CDs.2+ The peaks at 716 eV and 723 eV indicate that Fe 3+ The existence of ( Figure 5 AE). In contrast, the XPS spectra of arginine carbon dots (Arg-CDs) only show peaks at C 1s, N 1s, and O 1s (AE). Figure 6 AD).
[0074] FT-IR spectral analysis, results as follows Figure 7 As shown, 794 and 525 cm -1 The prominent peak at this point can be attributed to the stretching vibrations associated with the C-NH-C and Fe-O bonds of the guanidine group.
[0075] Zeta potential test analysis, results as follows Figure 8 As shown, the results indicate that the Zeta potential changes significantly with the introduction of iron doping, changing from a negative value to a positive value. This change can be attributed to the inherent positive charge associated with the metal element.
[0076] Fluorescence lifetime test analysis, results are as follows Figure 9 As shown, the results indicate that the fluorescence lifetime of iron-doped carbon quantum dots Fe-Arg-CDs is shortened compared to that of arginine carbon quantum dots Arg-CDs, and this change can be attributed to the inherent positive charge associated with the metal element.
[0077] Test Example 2: In vitro nitric oxide release
[0078] Equal concentrations of L-arginine, arginine carbon dots (Arg-CDs), and iron-doped carbon quantum dots (Fe-Arg-CDs) were prepared. They were placed in a solution containing 2 mM hydrogen peroxide and reacted at 37 °C for 48 hours. The supernatant was then analyzed using the Griess reagent method to determine the amount of nitric oxide generated. The results are as follows: Figure 10 As shown in Figure A, the results indicate that the nitric oxide concentration produced by all materials exceeded 1 μM. Specifically, the NOR for L-arginine was 1.098 μM, for arginine carbon dots (Arg-CDs) it was 2.41 μM, and for iron-doped carbon quantum dots (Fe-Arg-CDs) it was 28.58 μM. Compared to arginine carbon dots (Arg-CDs), the nitric oxide concentration produced by iron-doped carbon quantum dots (Fe-Arg-CDs) was approximately 10 times higher, and compared to L-arginine alone, its concentration was approximately 28 times higher.
[0079] Iron-doped carbon quantum dots (Fe-Arg-CDs) of equal concentration were placed in hydrogen peroxide solutions of different concentrations (0, 0.5, 1, 1.5, and 2 mM) and reacted at 37°C for 24 hours. The nitric oxide content was determined using the Griess reagent method, and the results are as follows: Figure 10As shown in Figure B, the results indicate that when iron-doped carbon quantum dots (Fe-Arg-CDs) are subjected to different concentrations of hydrogen peroxide, there is a significant correlation between hydrogen peroxide concentration and nitric oxide generation.
[0080] Further DFT model analysis was conducted using the following method: Since carbon dots are a class of zero-dimensional carbon-based materials, their characteristic is the sparse connectivity between their valence bands. Against this backdrop, DFT models of arginine carbon dots (Arg-CDs) and iron-doped carbon quantum dots (Fe-Arg-CDs) were constructed on a hexagonal lattice network similar to a graphene six-membered ring structure. Using the Vienna Abinitio Simulation Package (VASP), the DFT calculation of the iron-doped carbon quantum dot Fe-Arg-CDs system was performed using the Perdew, Burke parametric exchange-related interaction generalized gradient approximation (GGA). The plane wave energy cutoff for the wavefunction was set to 600 eV, and the minimum allowable spacing of the Γ-centered k-grid was 0.025 Å⁻¹. When the energy difference is less than 10... −7 At eV, self-consistent computational convergence is assumed. The Fast Inertial Relaxation Engine (FIRE) method from the Atomic Simulation Environment (ASE) library is used for structural optimization, implemented by calling the VASP program. The results are as follows... Figure 11 As shown, the results indicate that the charge density on the CN bonds inside iron-doped carbon quantum dots (Fe-Arg-CDs) increases, while the density of diffused electrons on the outside decreases. The increase in charge density implies a shortening of the bond length between the central C and N atoms; the CN* bond length in the Fe-Arg-CDs system decreases from 1.29178 Ȧ to 1.29164 Ȧ. This result is consistent with empirical trends observed in various chemical systems. Furthermore, the increase in charge density may enhance the adsorption capacity of graphene for hydrogen peroxide. To further support this explanation, the density of states (DOS) of arginine carbon dots (Arg-CDs) and iron-doped carbon quantum dots (Fe-Arg-CDs) was calculated and analyzed, with results shown below. Figure 12 As shown, this further supports the explanation.
[0081] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for optimizing the design of endogenous NO donor materials based on the TCGPR method, characterized in that, The steps include the following: S1. L-arginine and ferric chloride hexahydrate were dissolved in water and sonicated to prepare a precursor solution. Then, a one-pot hydrothermal reaction was carried out at a controlled temperature. After the reaction was completed, the mixture was centrifuged, the supernatant was filtered through a membrane, the filtrate was dialyzed, and freeze-dried to obtain iron-doped carbon quantum dots Fe-Arg-CDs. The obtained iron-doped carbon quantum dots Fe-Arg-CDs were placed in a hydrogen peroxide environment to determine the nitric oxide production concentration (NOR). S2. An orthogonal experiment was designed with the feeding ratio of L-arginine and ferric chloride hexahydrate, ultrasonic time, hydrothermal reaction temperature, and hydrothermal reaction time as experimental variables. Data sets were collected by collecting the nitric oxide generation concentration of each experimental variable and the corresponding product. S3. Construct a machine learning model based on the Gaussian Process Regression Tree Classifier (TCGPR) and train it using the collected dataset. The TCGPR machine learning model separates the collected dataset into multiple subsets with enhanced consistency, and builds a Gaussian Process Regression (GPR) machine learning model for each subset. The TCGPR machine learning model uses the radial kernel function in the Gaussian Process Regression model, combined with leave-one-out cross-validation to fit the dataset and evaluate data consistency. Considering the length scale parameter in the Gaussian radial kernel function, a global Gaussian disorder factor is defined. The dataset is partitioned by minimizing the global disorder factor, and the consistency of the dataset distribution is evaluated by examining the response surface of the Gaussian Process Regression model. S4. A virtual experimental formulation was designed with the feeding ratio of L-arginine and ferric chloride hexahydrate, ultrasonic time, hydrothermal reaction temperature, and hydrothermal reaction time as experimental variables. The virtual experimental formulation was predicted using a trained model. The potential to generate the maximum nitric oxide concentration was evaluated using the knowledge gradient utility function, and potential endogenous nitric oxide donor candidate materials were predicted. S5. Conduct experimental verification of promising endogenous nitric oxide donor candidates and screen out the optimal solution.
2. The method according to claim 1, characterized in that, In S1, the centrifugation conditions were: centrifugation at 13000 rpm for 10 minutes; the membrane filtration conditions were: filtration using a 0.22 μm membrane; the dialysis conditions were: dialysis in a 1000 Da dialysis bag for 3 days; and the nitric oxide formation concentration was determined by reacting at 37°C for 48 hours in a 2 mm hydrogen peroxide environment, using the Griess method.
3. The method according to claim 2, characterized in that, In S2, the specific scheme of the orthogonal experiment is as follows: the ratio of L-arginine to ferric chloride hexahydrate is selected as 50:1, 40:1, 30:1, 20:1 and 10:1, the hydrothermal reaction temperature is selected as 180℃, 200℃ and 220℃, the hydrothermal reaction time is selected as 4h, 6h and 8h, and the ultrasonic time is selected as 10 minutes, 20 minutes and 30 minutes to conduct a univariate orthogonal experiment, generating a total of 74 experimental samples.
4. The method according to claim 3, characterized in that, In S3, the TCGPR machine learning model separates the collected dataset into four subsets with enhanced consistency, and builds a GPR machine learning model on each subset for learning.
5. The method according to claim 4, characterized in that, In S4, 24,000 virtual samples were obtained through orthogonal design, and the potential values of all virtual samples were predicted. Three potential endogenous NO donor candidates were predicted. The design space was as follows: L-arginine to ferric chloride hexahydrate ratio 27:1, hydrothermal reaction temperature 150℃, hydrothermal reaction time 5h, and ultrasonic time 15 minutes, 20 minutes and 25 minutes respectively.
6. The method according to claim 5, characterized in that, In S5, the optimal selection for the preparation of the endogenous nitric oxide donor candidate material was as follows: L-arginine to ferric chloride hexahydrate ratio 27:1, hydrothermal reaction temperature 150℃, hydrothermal reaction time 5h, and ultrasonic time 25min.
7. The endogenous nitric oxide donor material designed and screened by the method according to any one of claims 1-6, characterized in that, The preparation reaction parameters for the endogenous nitric oxide donor material are as follows: L-arginine to ferric chloride hexahydrate ratio 27:1, hydrothermal reaction temperature 150℃, hydrothermal reaction time 5h, and ultrasonic time 25min.