Fluorescence and colorimetric bimodal detection platform based on machine learning assistance

Abstract

A fluorescence and colorimetric bimodal detection platform based on machine learning assistance is characterized in that a four-position porphyrin ligand FeTCPP is doped into Ui O-66 through a one-pot method to construct an MOF network structure, Zr-MOFFe with excellent peroxidase and fluorescence quenching ability is obtained, and Zr-MOFFe-Probes are obtained from the material of the Zr-MOFFe and a proper amount of H1 and H2 in a Tr s-HCI buffer solution; mixing with HP, continuously reacting for 1.5 hours, and acquiring a fluorescence emission spectrum by using a fluorospectro photometer; then, 3, 3 ', 5, 5'-tetramethyl benzidine is added into the Zr-MOFFe solution after the probe is released to react with H2O2 for several minutes, then colorimetric results are observed with naked eyes for qualitative analysis, and then the LPS content can be quantitatively analyzed by measuring ultraviolet absorption through an ultraviolet spectrum; constructing a back propagation neural network model of fluorescence and colorimetric modes; taking the corresponding fluorescence intensity in the fluorescence mode and the corresponding colorimetric picture information in the colorimetric mode as an input layer; wherein the image information is a corresponding LPS concentration value and a mean square error obtained by calculating the colorimetric image of the sample to be detected through an algorithm.

Description

Technical Field

[0001] This invention relates to a machine learning-assisted fluorescence and colorimetric dual-modal detection platform. Background Technology

[0002] Early diagnosis of disease helps identify symptomatic patients as quickly as possible and allows for interventions at the earliest stage to improve prognosis, making it an important strategy for promoting health and preventing disease. Sepsis is a life-threatening severe inflammatory disease primarily caused by a dysregulated immune response in the host to bacterial infection. Traditionally, bacterial infection is considered the direct cause of sepsis, with its high early mortality rate primarily driven by an excessive inflammatory response following infection. Lipopolysaccharide (LPS), an essential component of the outer membrane of Gram-negative bacteria, is defined as an endotoxin. Literature has demonstrated that LPS is a fundamental cause of sepsis; it binds to Toll-like receptors on immune cells, inducing a massive release of inflammatory cytokines, ultimately triggering a cytokine storm. Clinical data indicate that a concentration of more than 1 pg of LPS per milliliter of blood can lead to a series of inflammatory responses. Furthermore, the first hour after the onset of sepsis is the golden period for treatment. Therefore, early diagnosis of LPS is crucial for reducing the morbidity and mortality of sepsis. Since the main sources of LPS are contamination of blood products and pharmaceuticals, there is an urgent need to develop a sensitive and reliable method for LPS detection for disease treatment and drug safety. The gold standard for LPS detection has been reported to be Limulus amoebae (LAL) assay. Subsequently, many analytical methods based on advanced technologies have been proposed, including liquid chromatography / mass spectrometry (LC / MS), high-performance liquid chromatography (HPLC), electrochemical methods, and fluorescence spectroscopy, contributing to the sensitive monitoring of LPS. Nucleic acid aptamers, as next-generation molecular recognition elements, are widely used in the detection of disease biomarkers due to their high specificity and strong affinity. Currently, developed LPS aptamer-based detection methods are mainly based on fluorescence and electrochemical methods, which rely on a single signal output to quantify LPS in real samples. However, due to the low abundance of LPS, the sensitivity and accuracy requirements of detection methods are stringent. Furthermore, the complex composition of real samples increases the uncertainty of detection results. Based on these issues, a dual-mode method has been proposed to correct results and improve accuracy. It is worth noting that dual-mode detection methods based on fluorescence and colorimetry offer the advantages of both visualization and high sensitivity, thus improving the accuracy of trace biomarker detection. Therefore, dual-mode detection methods have broad application prospects in LPS determination. Typically, to achieve dual-mode detection using both fluorescence and colorimetry, the nanomaterials used need to consider both peroxidase activity and fluorescence properties.

[0003] Integrating multifunctionality into metal-organic frameworks (MOFs) can fully realize the application potential of MOF materials, and this research direction is currently receiving widespread attention. Studies have shown that MOFs, by altering the composition of the metal center or organic ligands, can serve as ideal models for studying fluorescence and colorimetric sensing behavior within systems. Notably, Wang et al. reported that Cu-MOFs were used to develop a colorimetric sensor based on peroxidase activity for dopamine detection. However, these single-function MOF materials cannot meet the requirements for multi-signal output, greatly limiting the practicality of MOF materials. To address these issues, a novel bifunctional MOF possessing both peroxidase activity and fluorescence modulation activity has attracted significant attention. Compared to single-function MOF materials, this material offers broader applications and superior performance.

[0004] For example, Fe-MOFs can be integrated into the Zr-MOF framework using a one-pot thermodynamically controlled synthesis method to synthesize bifunctional materials with good peroxidase activity and fluorescence quenching efficiency. Furthermore, studies have demonstrated that DNA adsorption onto the surface of MOF materials can inhibit Zr-MOF activity. Fe The peroxidase activity of Zr-MOF Fe Conversely, it can also quench the fluorescence signal of the DNA probe, exhibiting a bidirectional regulatory effect. Therefore, adsorbing the DNA probe onto Zr-MOF... Fe Constructing dual-mode detection methods on material surfaces is currently a hot research topic. These methods typically rely on external fluorophores or nanozymes to catalyze reactions to generate signals. Interestingly, Zr-MOF... F e binds tightly to ssDNA and hairpin DNA via electrostatics, π-stacking, hydrogen bonding, and coordination, but cannot effectively bind dsDNA, which is attributed to the rigid structure of dsDNA. Therefore, in this study, Zr-MOF was selected. Fe Materials are key components for achieving quantitative LPS analysis. Catalytic hairpin self-assembly (CHA) is a highly efficient enzyme-free signal amplification technique. Under the triggering of a specific initiator, two metastable DNA hairpins can be induced to self-assemble, forming a double-stranded DNA structure, thereby amplifying the signal. To achieve exponential signal amplification, researchers have devoted considerable effort to improving the traditional CHA method. In our previous work, we constructed the FCHA strategy by introducing H3L double strands into the traditional CHA method, achieving 2... n-1 The signal amplification was increased several times, successfully detecting trace RNA in tumor cells. However, this method relies on only a single signal output, lacking reliability for quantitative analysis of trace biomarkers in real-world samples with complex compositions. To address this, based on our previous research, a bifunctional Zr-MOF... FeThe novel strategy combining nucleic acid amplification technology not only inherits the advantages of exponential signal amplification but also exhibits dual-mode signal output due to bidirectional regulation, meeting the analytical requirements for trace LPS detection. To distinguish it from previous work, the nucleic acid amplification technology designed in this work is defined as self-feedback exponentially catalyzed hairpin self-assembly (SECHA).

[0005] Machine learning (ML), a representative strategy in artificial intelligence, plays an increasingly important role in biomarker analysis. Among them, backpropagation neural networks (BPNNs) are one of the most widely used neural networks in machine learning due to their rigorous structure, ease of operation, and fast convergence speed. The BPNN operation principle is based on gradient descent, updating the network weights by calculating the gradient of the loss function, thereby minimizing the loss function and obtaining the prediction result. Therefore, combining BPNNs with bimodal detection methods has the potential to achieve a synergistic effect. On the one hand, bimodal detection methods significantly enhance the reliability and applicability of the analysis by providing complementary analytical information, thus greatly improving detection efficiency and accuracy in practical applications. On the other hand, the powerful nonlinear fitting capability of the BPNN algorithm can effectively analyze LPS signals in complex samples, providing a theoretical basis for accurate LPS detection. However, such reports have not yet been widely reported.

[0006] Compared to previously reported LPS detection single-mode methods, this strategy cleverly incorporates Zr-MOF... Fe By combining peroxidase activity and fluorescence quenching ability with SECHA technology (exponential signal amplification), dual-mode detection of LPS was achieved, demonstrating better accuracy and sensitivity. Finally, by leveraging machine learning, LPS fingerprint information was extracted from fluorescence spectra and color photographs, respectively, and a BPNN model was successfully constructed.

[0007] By utilizing spectral / image data, the concentration of LPS can be calculated, providing strong technical support for the accurate and sensitive detection of LPS. Through fluorescence / colorimetric and spectral / image detection, our system has great potential for high-precision analysis of low-abundance LPS in complex samples. Summary of the Invention

[0008] The present invention addresses the problem that MOF materials are currently the ideal models for studying fluorescence and colorimetric sensing behavior. However, single-function MOF materials cannot simultaneously possess both peroxidase and fluorescence modulation activities, and the output of a single signal cannot guarantee the reliability and accuracy of the results. Therefore, it proposes to develop a bifunctional Zr-MOF... Fe The materials, SCHA, and BPNN are integrated into a fluorescence / colorimetric dual-modal intelligent analysis platform to address the technical challenges of rapid, accurate, and sensitive analysis of trace LPS in real samples.

[0009] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0010] A machine learning-assisted dual-modal detection platform for fluorescence and colorimetry includes the following steps:

[0011] S1: A MOF network structure was constructed by incorporating the four-position porphyrin ligand FeTCPP into UiO-66 using a one-pot method, resulting in a Zr-MOF with excellent peroxidase and fluorescence quenching capabilities. Fe

[0012] S2: Zr-MOF Fe The material and appropriate amounts of H1 and H2 were incubated in Tris-HCl buffer at 37°C for 10 minutes to obtain Zr-MOF. Fe @Probes

[0013] S3: Mix appropriate amounts of biotin-modified ST, streptavidin-modified magnetic beads, and LPS, and incubate for a period of time. Collect the supernatant containing a large amount of T probe using a magnetic separator. Then, transfer the supernatant to Zr-MOF... Fe @Probes and HP were mixed and reacted for another 1.5 hours. Fluorescence emission spectra between 505-650 nm were obtained using a fluorescence spectrophotometer with an excitation wavelength of 490 nm and a bandwidth of 5 nm. Subsequently, the Zr-MOF after probe release... Fe After adding 3,3',5,5'-tetramethylbenzidine to the solution and reacting it with H2O2 for several minutes, the colorimetric results were observed with the naked eye for qualitative analysis. Subsequently, the LPS content was quantitatively analyzed by measuring the ultraviolet absorption at 652 nm using ultraviolet spectroscopy.

[0014] S4: Construct backpropagation neural network (BPNN) models for fluorescence and colorimetric modes. LPS concentrations in fluorescence mode were set to 1, 10, 100, 500, 10000, 100000, and 1000000 fg / mL. -1 The corresponding fluorescence intensities and LPS concentrations in the colorimetric mode were 0.05, 0.1, 0.5, 1, 5, 10, 25, and 50 ng / mL. -1 The corresponding colorimetric image information is used as the input layer. This image information is obtained by converting the R, G, and B channel values ​​of the colorimetric image of the sample to be tested into values ​​for the three channels R, G, and B using ImageJ. The algorithm then calculates the corresponding LPS concentration value and mean square error.

[0015] The beneficial technical effects of adopting the above technical solution are:

[0016] This invention designs a machine learning-assisted Zr-MOF. FeThe @SECHA fluorescence / colorimetric dual-mode intelligent platform enables rapid, accurate, and sensitive detection of trace LPS in complex samples. This technology has the following advantages:

[0017] 1) Good accuracy and sensitivity. The dual-signal output not only effectively calibrates each other but also exhibits excellent amplification efficiency, with LODs of 10.254 fg / mL. -1 and 0.143 pg mL -1 .

[0018] 2) High practicality. Zr-MOF Fe The @SECHA strategy has been effectively used to analyze human serum and vancomycin, demonstrating good reproducibility (RSD, 1.5–5.1%) and accuracy (recovery, 91.2–104.1%).

[0019] 3) Intelligentization. A BNPP model was successfully constructed, which can quantify LPS concentration using a fluorescence / colorimetric dual-mode platform, providing a more efficient and intelligent strategy for real-time and accurate LPS monitoring. Attached Figure Description

[0020] Figure 1 Zr-MOF Fe @SECHA schematic diagram.

[0021] Figure 2 Zr-MOF Fe The representation diagram.

[0022] Figure 3 for (A)Zr-MOF Fe (A) HOMO and LUMO energy levels of FAM; (B) Zr-MOF Fe Electrostatic potential distribution diagram of FAM; Zr-MOF Fe DLS size distribution (C) and zeta potential (D) before and after loading nucleic acid probe.

[0023] Figure 4 Zr-MOF Fe Feasibility verification of @SECHA; polyacrylamide gel images of H1+H2, H1+H2+ST, H1+H2+ST+LPS and H1+H2+ST+HP+LPS (A); fluorescence spectra in fluorescence mode (B) and UV-Vis absorption spectra in colorimetric mode (C).

[0024] Figure 5 for (A)Zr-MOF Fe @SECHA H1 to H2 concentration ratio (1:4, 1:3, 1:2, 1:1, 2:1, 2:1, 2:1, 2:1, 3:1, 4:1, LPS concentration 10 ng / mL)-1 The reaction temperature was 37℃. (B)Zr-MOF Fe Real-time fluorescence intensity of @SECHA, reaction time 0-180 min, LPS concentration 10 ng / mL -1 The reaction temperature was 37℃. The concentrations of H1, H2, ST, and HP were 200 nM, 200 nM, 200 nM, and 100 nM, respectively. The error bars represent the results obtained from three repeated measurements.

[0025] Figure 6 for (A)Zr-MOF Fe The fluorescence spectrum of the @SECHA system for 0 to 50 ng / mL -1 (A) The reaction of LPS at different concentrations; (B) The fluorescence intensity at 520 nm showed a linear relationship with LPS concentration, ranging from 1 fg / mL. -1 Up to 1 ng mL -1 (C) Zr-MOF was studied in fluorescence mode. Fe @SECHA's specificity in LPS, PCT, AFP, CEA, HSA, and dextran reactions, ***p<0.001; (D)Zr-MOF Fe The UV-Vis absorption spectrum of the @SECHA system responds to different concentrations of LPS between 0 and 50 ng mL⁻¹; (E) absorbance at 652 nm and 5 pg mL -1 Up to 5ng mL -1 The LPS concentration within the range showed a linear relationship; (F) Zr-MOF was studied in colorimetric mode. Fe @SECHA's specificity in LPS, PCT, AFP, CEA, HSA and dextran reactions, ***p<0.001.

[0026] Figure 7 This is the BPNN mode used for fluorescence / colorimetric dual-modality detection.

[0027] Figure 8 Zr-MOF for (A) LPS fluorescence information recognition and (B) colorimetric image recognition Fe @SECHA network training curve.

[0028] Figure 9 Zr-MOF Fe The @SECHA system and commercial ELISA kits were used to detect LPS concentrations in serum and vancomycin samples. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention.

[0030] 1. Implementation Plan

[0031] This invention aims to innovatively combine bifunctional Zr-MOF Fe Using SECHA and BPNN, constructing a machine learning-assisted Zr-MOF Fe The @SECHA fluorescence / colorimetric dual-mode intelligent platform enables sensitive and accurate LPS analysis in complex samples. Firstly, a stable bifunctional Zr-MOF was synthesized using a one-pot thermal controlled synthesis method. Fe Using the crystalline stable UiO-66 as the parent framework, the four-position porphyrin ligand FeTCPP was incorporated into the UiO-66 framework. This material exhibited excellent peroxidase activity and strong fluorescence quenching ability. This Zr-MOF... Fe It can bind to nucleic acid probes with optimal adsorption force, effectively suppress signal intensity, and significantly reduce background signal leakage.

[0032] Zr-MOF Fe @SECHA principle as follows Figure 1 As shown. Through reasonable design, Zr-MOF Fe @SECHA consists of only two hairpins (H1 and H2) and two double-stranded probes (ST and HP), with probe H2 modified with a carboxyfluorescein (FAM) fluorophore. First, biotin-modified ST is coupled to the surface of streptavidin-modified magnetic beads, followed by the addition of LPS to the solution. Under affinity, LPS specifically recognizes and captures specific aptamers (S), leading to the release of initiator (T) into the supernatant. With the aid of a magnet, a large amount of T is enriched in the supernatant and used in subsequent steps. Literature has confirmed that hairpin probes can be adsorbed onto Zr-MOF via electrostatics, π-stacking, and hydrogen bonding. Fe The rigid structure of dsDNA can effectively shield nucleobases and phosphate groups, thereby reducing its interaction with Zr-MOF. Fe The binding affinity of Zr-MOF. Fe The surface can firmly adsorb H1 and H2, thereby suppressing the FAM fluorescence signal of H2. Upon addition of the supernatant enriched with T, the T probe can induce H1 and H2 to form a double-stranded structure, which can then be extracted from Zr-MOF. Fe The surface detaches. Furthermore, because the synthesized CHA product contains active chain reaction sites, it can further hybridize with P in HP, subsequently releasing H via toe-mediated chain displacement. The released H can spontaneously fold into a hairpin structure, initiating a new round of CHA reactions between H1 and H2.

[0033] Through this continuous recovery process, the fluorescence signal exhibits exponential growth. Furthermore, due to the involvement of HP, the CHA product can assemble into a more rigid Y-shaped DNA structure, which further reduces its interaction with Zr-MOF. Fe The affinity between Zr-MOF and its constituent elements is noteworthy. Fe After surface release of CHA products, Zr-MOF Fe The peroxidase activity was restored, enabling it to catalyze the oxidation of colorless TMB to blue oxTMB by H2O2, and the absorbance at 652 nm could be measured using UV-Vis spectroscopy.

[0034] 1.1 Zr-MOF Fe Preparation of;

[0035] Zr-MOF Fe The synthesis of this product was optimized based on previous research. In short, 100 mg ZrCl4, 100 mg BDC, 33 mg FeTCPP, and 1.67 g benzoic acid were added to 20 mL DMF. After stirring for 15 min, the mixture was transferred to a high-pressure reactor at 130 °C for overnight reaction. The cooled mixture was centrifuged at 120,000 rpm for 15 min to obtain a brown precipitate. Subsequently, the product was washed three times with DMF and methanol. Finally, the solid product was dried in a vacuum drying oven at 80 °C for 24 h to obtain powdered Zr-MOF. Fe .

[0036] 1.2 Preparation of Zr-MOF Fe @probes;

[0037] To prepare Zr-MOF Fe @probes, the resulting dark brown powder was redispersed in Wahaha purified water to prepare a 0.5 mg / mL solution. Simultaneously, H1 (H1-Cy5) loaded with Cy5 and H2 (H2-FAM) loaded with FAM were heated at 95 °C for 5 min, then cooled to room temperature for 30 min to allow for the formation of more hairpin structures. Next, 10 μL of H1-Cy5, 10 μL of H2-FAM (final concentration, 200 nM), and 0.5 mg / mL Zr-MOF were added... Fe Dissolve in 20 mM Tris-HCl buffer (pH 8.0, containing 100 mM NaCl). Incubate with shaking at room temperature for 10 minutes to form Zr-MOF. Fe @probes. Subsequently, unloaded probes were removed by centrifugation (120,000 rpm, 15 min), and the fluorescence intensity of the supernatant was measured to determine the probe loading in Zr-MOF. Fe Loading efficiency.

[0038] 1.3 Evaluation of Zr-MOF Fe Peroxidase activity;

[0039] To explore the peroxidase activity of the material, 500 μL of UiO-66 and Zr-MOF were added. Fe and Zr-MOF Fe @Probe s was added to an EP tube containing 300 μL acetate buffer (0.2 M, pH = 4.0), followed by the addition of 100 μL 40 mM TMB and 100 μL 80 mM H2O2. After gently shaking for 10 minutes at room temperature, the UV absorbance at 652 nm was measured.

[0040] 1.4 Gel electrophoresis analysis;

[0041] A 15% natural polyacrylamide gel electrophoresis system was prepared for system validation. Different reactants, including H1, H2, ST, HP, H1+H2, H1+H2+ST, H1+H2+ST+HP, H1+H2+ST+HP, and H1+H2+ST+HP+LPS, were reacted in 20 mM Tris-HCl buffer at 37°C for 1 h. Next, 2 μL of the mixture was added to each 200 μL LEP tube containing 8 μL of 6× loading buffer, and electrophoresis was performed at 220 V for 35 min.

[0042] 1.5 Fluorescence Experiment;

[0043] First, H1 and H2-FAM were heated at 95°C for 5 min, then rapidly cooled to room temperature for 30 min to allow for the formation of more hairpin structures. Then, 10 μL of H1, 10 μL of H2-FAM (final concentration, 200 nM), and 0.5 mg / mL were added... -1 Zr-MOF Fe Add an appropriate amount of 20mM Tris-HCl buffer (pH=8.0, containing 100mM NaCl). After incubation for 10 min, obtain the fluorescence emission spectrum between 05-650nm using a fluorescence spectrophotometer at an excitation wavelength of 490nm and a bandwidth of 5nm.

[0044] 1.6 Dual-mode detection of LPS;

[0045] Streptavidin-modified magnetic beads and biotin-modified ST (final concentration, 200 nM) were added together to 20 mM PBS buffer (pH = 7.4). After incubation at 37°C for 30 minutes, 10 ng / mL of the solution was added. -1LPS was incubated for another hour. The supernatant was retained under magnetic attraction to obtain the free T probe. The obtained free T probe was mixed with the two-mode system, and Tris buffer (20 mM, pH 7.4) containing 30 mM MgCl2 and 5 mM CaCl2 was added. The mixture was incubated at 37°C for 1.5 h. For fluorescence experiments, the fluorescence values ​​of the reaction mixture were measured using a fluorescence spectrophotometer at an excitation wavelength of 490 nm and an emission wavelength of 505-650 nm. For colorimetric analysis, Zr-MOF... Fe The Probe s solution was further mixed with TMB and H2O2, and subsequent experimental procedures were the same as in "1.3 Evaluation of Zr-MOF". Fe The peroxidase activity is consistent with that.

[0046] 1.7 Analysis of real samples;

[0047] Zr-MOF Fe The @SECHA method was used to analyze human serum and vancomycin to evaluate its practicality and reliability. The human serum used was approved by the institution's Ethics Committee and obtained from the Department of Laboratory Medicine, Nanhai District People's Hospital (Foshan, China) (Ethics Approval No.: NY KY-2025-246-01). Different concentrations of LPS solution (1 pg, 100 pg, and 400 pg) were added to diluted serum samples and vancomycin, respectively, and then subjected to Zr-MOF... Fe Incubate for 1.5 hours in the @SE CHA system. Finally, obtain fluorescence intensity using a fluorescence spectrophotometer to calculate the recovery rate. The results are then compared with ELISA measurements.

[0048] 1.8 BPNN model for intelligent LPS detection;

[0049] Backpropagation Neural Network (BPNN) is a typical nonlinear algorithm consisting of three parts: an input layer, hidden layers, and an output layer. In the BPNN model, the nodes in the input layer represent independent variables, and the nodes in the output layer represent dependent variables. The hidden layer contains a large number of neurons, which can be used to quickly analyze the nonlinear relationships in data. Data transmission between the three layers is achieved through the BPNN algorithm, which involves two steps: forward propagation and backward propagation. Forward propagation passes data from the input layer through the hidden layers to the output layer, generating prediction results. Backpropagation refers to the process of propagating back the error, where the mean squared error (MSE) of each neuron is calculated to update the weights between neurons. By training on a large amount of data, BPNN has the advantages of adaptation and self-learning, enabling it to quickly process, analyze, and understand data. Therefore, a fluorescence / colorimetric dual-modality BPNN model was constructed using the Maishi Neural Network software developed by Nanjing Maishi Software Co., Ltd., for the accurate determination of LPS. In fluorescence mode, the fluorescence intensity of LPS at different concentrations at 520 nm was used as the input layer of the BPNN model. In colorimetric mode, the Zr-MOF of LPS at different concentrations was analyzed using ImageJ. Fe The true-color image of +TMB+H2O2 was divided into RGB channels. Then, the color intensity of the R (red), G (green), and B (blue) channels at different concentrations was calculated, and this data was used as the input layer.

[0050] 2. Best Practices (Taking the detection of LPS in human serum and vancomycin as an example)

[0051] 2.1 The nucleic acid sequences used in the experiment are shown below.

[0052] ①H1

[0053] AGGAGACGAGATACCCGGACACTAGTGGTGTGTCGAGTCAGTGTCCGGGTATGGAGG

[0054] ②H1-Cy5

[0055] AGGAGACGAGATACCCGGACACTAGTGGTGTGTCGAGTCAGTGTCCGGGTATGGAGG-Cy5

[0056] ③H2-FAM

[0057] FAM-AGAGTAGGGAGACACTGACTCGACACACCACTAGTGTCCGGGTGTGTCGAGTC

[0058] ④S-biotin

[0059] biotin-CTTCTGCCCGCCTCCTTCCTAGCCGGATCGCGCTGGCCAGATGATATAAAGGGTCAGCCCCCCAGGAGACGAGATAGGCGGACACT

[0060] ⑤T

[0061] AGTGTCCGGGTATCTCGTCTCCT

[0062] ⑥H

[0063] AGGTTCCGGGGGAGAGTAGGGACGGGTACTCTCCTATCTCGTCTCCT

[0064] ⑦P

[0065] CCTCCATACCCGTCCCTACTCT

[0066] 2.2 Material Characterization

[0067] In short, Zr-MOFs were constructed by incorporating the four-position porphyrin ligand FeTCPP into UiO-66 using a solvothermal method, with the crystalline stable UiO-66 as the parent framework. Fe Network, such as Figure 2 As shown in Figure A. First, Zr-MOF was examined using transmission electron microscopy (TEM). Fe Morphology is characterized. For example... Figure 2 As shown in Figure B, Zr-MOF Fe It exhibits a tetrahedral shape and possesses the skeletal structural characteristics of MOFs. Furthermore, Zr-MOFs... Fe HAADF-STEM and element mapping images, such as Figure 2 As shown in Figure C, the uniform distribution of C, N, O, Fe, and Zr elements throughout the network structure is confirmed. Next, X-ray photoelectron spectroscopy (XPS) was used to further characterize the Zr-MOF. Fe The chemical composition of the sample was determined by spectral analysis to be primarily Fe, O, N, C, and Zr. Figure 2 As shown in D. (As shown in Figure D) Figure 2 As shown in Figure E, the Zr 3d spectrum exhibits two peaks at binding energies of 184.26 eV and 182.86 eV, corresponding to Zr 3d and 3d, respectively. 3 / 2 and Zr3d 5 / 2 The binding energy difference of 1.40 eV indicates the presence of Zr in the zirconium carbonyl cluster. 6+ In this bifunctional MOF, the two fitted peaks at 722.34 eV and 708.54 eV are attributed to Fe 2p, respectively. 1 / 2 and Fe2p 3 / 2,like Figure 2 As shown in Figure F. The O 1s spectrum is as follows: Figure 2 As shown in G, the two peaks at 530.70 eV and 529.04 eV correspond to C-OH and C=O groups, which may originate from FeTCPP.

[0068] To investigate the chemical state of the constituent elements in the prepared MOF, X-ray diffraction (XRD) analysis was used to confirm the synthesis of Zr-MOF. Fe Crystal structure, such as Figure 2 As shown in Figure I. These results indicate that Fe 2+ Successfully with Zr-MOF Fe The porphyrin coordination in Zr-MOF was then investigated using Fourier transform infrared (FT-IR) spectroscopy. Fe Functional groups. For example... Figure 2 As shown in Figure J, Zr-MOF Fe At approximately 996cm -1 The symmetric Fe–N vibration at the location is consistent with the overall FeTCPP, indicating that Zr-MOF Fe The structure contains a large number of Fe–N active sites. Compared with FeT CPP, Zr-MOF... Fe At 1685cm -1 The carboxyl vibration peak gradually red-shifted to 1583 cm⁻¹. -1 This is attributed to the synthesized Zr-MOF Fe In the structure, the free porphyrin bond is redirected to the chelating functional group of the carboxylate. These characterization results confirm that FeTCPP is anchored to the Zr6 cluster through stable chemical bonds. All experimental results consistently confirm the Zr-MOF... Fe The successful preparation of the hairpin probe H2-FAM was achieved. This study aimed to investigate how the fluorescence of the Zr-MOF material is affected by the fluorescence of the hairpin probe H2-FAM. Fe For quenching, this invention uses density functional theory (DFT) to calculate Zr-MOF. Fe The energy values ​​of the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) of FAM. Figure 3 As shown in Figure A, the HOMO and LUMO of the FAM fluorophore are -6.280 eV and -2.419 eV, respectively. Meanwhile, the Zr-MOF... Fe The HOMO and LUMO were calculated to be -5.431 eV and -3.870 eV, respectively. Since the LUMO level of FAM is higher than that of Zr-MOF... Fe The excited electrons of FAM can be transferred to Zr-MOF Fe The LUMO, leading to fluorescence quenching. Furthermore, Zr-MOF... FeThe gap was calculated to be 1.562 eV. This low gap facilitates electron transfer and promotes the adsorption of target DNA. We also calculated the Zr-MOF material... Fe Electrostatic potential energy (ESP) of the surface. ESP shows that Zr-MOF Fe With a higher charge density (shown in red), it exhibits Lewis base behavior and can form hydrogen bonds with target DNA, such as... Figure 3 As shown in Figure B. Furthermore, dynamic light scattering (DLS) and Zeta potential are used to further demonstrate the Zr-MOF... Fe The adsorption of the material onto the nucleic acid probe. For example... Figure 3 As shown in C, the synthesized Zr-MOF after adsorption of nucleic acid probes Fe The hydrated particle size increased. Zeta potential measurements showed that Zr-MOF… Fe The surface charge shifted from -18.2 mV to -37.3 mV, which is attributed to the nucleic acid probe carrying phosphate groups, such as... Figure 3 As shown in Figure D. These results demonstrate that nucleic acid probes can be successfully adsorbed onto Zr-MOF. Fe surface.

[0069] 2.3 Zr-MOF Fe Feasibility verification of the @SECHA system

[0070] To confirm the feasibility of the developed LPS detection method, polypropylene electrophoresis (PAGE) analysis in different reaction solutions was first studied. Observations from the PAGE experiments showed that... Figure 4 As shown in Figure A, H1, H2, and HP correspond to lanes 2 to 5. No hybridization products were observed only when H1 and H2 or H1, H2, and ST were mixed (lanes 6 and 7), indicating that the proposed method has high stability and no significant signal leakage. Hybridization of H1, H2, and ST begins upon the addition of LPS, producing a new band (lane 8). As expected, after incubation of H1, H2, and ST with HP, the HP band significantly weakens, and a new band (lane 9) is produced, corresponding to the H1-H2-P product. Next, we further validated the feasibility of the method through fluorescence experiments and colorimetric assays. Figure 4As shown in Figure B, the fluorescence results indicate that only a very weak FL signal was observed when H1 and H2 were mixed. Similarly, when ST or HP was incubated with H1 and H2, only a negligible signal was observed, indicating that ST and HP cannot trigger the method without LPS. When ST was added to a solution containing LPS, the supernatant was obtained using an external magnetic field. After the supernatant was incubated with H1 and H2, an enhanced FL signal was observed, indicating that LPS can release T to induce the H1 and H2 reaction. When the supernatant was incubated with H1, H2, and HP, a very strong FL signal was produced. The colorimetric results are consistent with the fluorescence experiments, as shown in Figure B. Figure 4 As shown in Figure C. These results demonstrate that the proposed method can successfully achieve high-sensitivity detection of LPS.

[0071] 2.4 Concentration and Time Optimization

[0072] First, the concentration ratio of H1 to H2 was studied, such as... Figure 5 As shown in Figure A, it can be observed that as the H2:H1 concentration ratio increases, the FL signal first increases, then decreases, and finally stabilizes, proving that the optimal H2:H1 concentration ratio is 1:1. Therefore, a 1:1 concentration ratio was used in subsequent reactions. Next, real-time fluorescence spectroscopy was used to measure the fluorescence of this method. Figure 5 As shown in Figure B, in the presence of LPS, the solution of H1+H2+HP+ST initially exhibited a rapid increase in fluorescence intensity within approximately 1.5 hours, followed by a gradual plateauing of fluorescence intensity. Notably, the signal-to-noise ratio (F / F0) of this method decreased rapidly after 2 hours, which may be attributed to the non-specific amplification of the nucleic acid probe. Therefore, 2 hours was chosen as the optimal time for the in vitro reaction.

[0073] 2.5, Zr-MOF Fe The detection and accuracy performance of the @SECHA system was studied under optimal solution conditions using two analytical modes. In fluorescence mode, a series of concentrations from 1 fg / mL were analyzed. -1 up to 50 ng / mL -1 Add to the reaction solution, such as Figure 6 As shown in Figure A. The increase in the FL signal using this method depends on the increase in LPS concentration. Furthermore, the Y value varies with LPS concentration ranging from 1 fg / mL. -1 up to 1 ng mL -1 It is proportional to the logarithm (lg), where the corresponding linear equation is Y = 1.397 + 0.995lg(C). LPS / fg (R) 2 =0.999), the limit of detection (LOD) was calculated to be 10.254 fg / mL. -1 like Figure 6As shown in Figure B. As expected, in colorimetric mode, the linear range of LPS concentration was 5 pg / mL. -1 Up to 5ng mL -1 The regression equation is described as Y = 0.231 + 0.159lg(C LPS / fg (R) 2 =0.997), its LOD is 0.143 pg mL -1 ,like Figure 6 China D and Figure 6 As shown in Figure E. Furthermore, we investigated the effects of procalcitonin (PCT), fetal protein α (AFP), carcinoembryonic antigen (CEA), human serum albumin (HSA), and dextran on Zr-MOF. Fe @SECHA system selectivity performance. Notably, in the presence of the LPS analogue, a negligible change in the FL signal was observed compared to the control group, while the LPS group exhibited a significantly enhanced FL signal, such as... Figure 6 As shown in C. In colorimetric mode, the proposed Zr-MOF Fe The @SECHA system responds significantly better to LPS than LPS analogues, such as... Figure 6 As shown in Figure F, this is consistent with the results of the fluorescence mode. Statistically significant differences were observed between the two modes for LPS and LPS analogs (P < 0.001(***)), indicating the excellent specificity of our method. The consistent specificity highlights the advantage of cross-validation between the fluorescence and colorimetric modes.

[0074] 2.6 Establishment of the BPNN model

[0075] The established BPNN model is as follows Figure 7 As shown. In the fluorescence recognition model, the fluorescence intensities at 520 nm corresponding to LPS concentrations were 1, 10, 100, 500, 10000, 100000, and 1000000 fg / mL. -1 As the input layer (1), it is then passed to the hidden layers, resulting in 7 neurons in the hidden layers (2–8). Mean squared error (MSE) is an important parameter; it evaluates the average difference between the model's predicted and actual values ​​to assess the model's accuracy. This model, using the backpropagation algorithm, reaches an MSE of 0 after 30252 iterations. Figure 8 As shown in Figure A, the obtained LPS concentration in the output layer meets the accuracy requirements of this method. In the colorimetric photograph recognition model, different concentrations (0.05, 0.1, 0.5, 1, 5, 10, 25, and 50 ng / mL) were used. -1 The R, G, and B values ​​of the solution images were transferred to the input layer (1-3), while the corresponding LPS concentrations were used as the output layer (11-18). After 547 computational iterations, the mean square error gradually approached 0, as shown in the figure. Figure 8 As shown in Figure B, the output results also meet the accuracy requirements of this method. Therefore, we have successfully constructed a BPNN model for accurately detecting LPS concentration based on a dual-mode method, contributing a new approach to the application of neural networks in the field of dual-modal sensing.

[0076] 2.7 Real Sample Testing

[0077] Zr-MOF Fe The @SECHA system was applied to the analysis of real human serum and vancomycin samples, and the results were compared with those of commercial enzyme-linked immunosorbent assays (ELISA). As shown in Table 1, the system achieved good reproducibility (RSD, 1.5–5.1%) and accuracy (recovery, 91.2–104.1%). As shown in Table 2, due to the narrow linear range of ELISA, 1 pg / mL... -1 The concentration signal could not be detected by ELISA and Zr-MOF. Fe Statistical analysis of the signals from the @SECHA system showed no significant difference in detection data between the two methods, proving the validity of the proposed Zr-MOF. Fe The @SECHA system has high accuracy in the quantitative detection of LPS, such as... Figure 9 As shown.

[0078] Table 1. Zr-MOF Fe @SECHA system detects LPS in real samples (n=3)

[0079]

[0080]

[0081] Table 2. LPS detected by ELISA in real samples (n=3)

[0082]

[0083] In the description of this invention, it should also be noted that the "@" symbol throughout the text represents probe adsorbed on the material surface.

Claims

1. A machine learning assisted based fluorescence and colorimetric dual modal detection platform, characterized in that: It comprises the following steps: S1: FeTCPP, a four-position porphyrin ligand, is incorporated into Ui O-66 by a one-pot method to construct a MOF network structure, obtaining a Zr-MOF with excellent peroxidase and fluorescence quenching capabilities Fe S2: Zr-MOF Fe Material and appropriate amount of H1, H2 were incubated in Tris-HCI buffer at 37°C for 10 minutes to obtain Zr-MOF Fe @Probes; S3: mixing the appropriate amount of biotin-modified ST, streptavidin-modified magnetic beads and LPS, and incubating for a period of time, using a magnetic separator to collect the supernatant containing a large amount of T probes; then, collecting the supernatant, Zr-MOF Fe @mixing the probes and HP, continuing the reaction for 1.5 hours, and using a fluorescence spectrophotometer to obtain the fluorescence emission spectrum between 505-650 nm at an excitation wavelength of 490 nm and a bandwidth of 5 nm; then, using the Zr-MOF Fe After adding 3,3',5,5'-tetramethylbenzidine and H2O2 in the solution and reacting for a few minutes, the colorimetric results were observed with the naked eye for qualitative analysis, and then the content of LPS was quantitatively analyzed by measuring the ultraviolet absorption at 652 nm by ultraviolet spectrum. S4: constructing a back propagation neural network (BPNN) model in fluorescence and colorimetric modes; taking the fluorescence intensity corresponding to the LPS concentration of 1, 10, 100, 500, 10000, 100000 and 1000000 fg mL-1 in the fluorescence mode and the colorimetric image information corresponding to the LPS concentration of 0.05, 0.1, 0.5, 1, 5, 10, 25 and 50 ng mL-1 in the colorimetric mode as the input layer; wherein the image information is to convert the RGB data of the colorimetric image of the sample to be tested into the values of R, G and B three different channels by using imgeJ; under the operation of the algorithm, the corresponding LPS concentration value and the mean square error are obtained.

2. The machine learning assisted based fluorescent and colorimetric dual modal detection platform according to claim 1, wherein: The Zr-MOF Fe The synthesis was optimized based on previous studies. Briefly, 100 mg of ZrCl4, 100 mg of BDC, 33 mg of FeTCPP and 1.67 g of benzoic acid were added to 20 mL of DMF. After stirring for 15 min on a stirrer, the mixture was transferred to a 130 °C high-pressure reactor for overnight reaction. The cooled mixture was centrifuged at 120,000 rpm for 15 minutes to obtain a brown precipitate. Subsequently, the product was washed 3 times with DMF and methanol. Finally, the solid product was dried in a vacuum oven at 80 °C for 24 h to obtain the powder Zr-MOF Fe . 3.The machine learning assisted based fluorescent and colorimetric dual-mode detection platform of claim 1, wherein: The Zr-MOF Fe @probes were prepared by redispersion of the obtained dark brown powder in Wahaha purified water to make a 0.5 mg / mL solution. Meanwhile, Cy5-labeled H1 (H1-Cy5) and FAM-labeled H2 (H2-FAM) were heated at 95 °C for 5 min and then cooled to room temperature for 30 min to form more hairpin structures. Next, 10 μL H1-Cy5, 10 μL H2-FAM (final concentration, 200 nM), and 0.5 mg / mL Zr-MOF Fe were dissolved in 20 mM Tris-HCl buffer (pH = 8.0, containing 100 mM NaCl). Incubation was performed at room temperature for 10 min to allow the Zr-MOF Fe @probes to form. Subsequently, the unbound probes were removed by centrifugation (120,000 rpm, 15 min), and the fluorescence intensity of the supernatant was measured to determine the loading efficiency of the probes on the Zr-MOF Fe .

Images