Whole-cell biosensor arrays and their applications

Abstract

The present invention discloses a whole-cell biosensor array, comprising a 96-well plate on which 14 whole-cell biosensors are placed, with 3-4 of each whole-cell biosensor. Each whole-cell biosensor comprises: a host cell; a recombinant vector located within the host cell, the recombinant vector comprising a vector and a promoter inserted into the vector, wherein the promoters within the 14 whole-cell biosensors are recAp, dnaKp, grpEp, katGp, fabAp, yibTp, soxSp, glgSp, oxyRp, uvrAp, ompFp, spyp, pspAp, and leuAp, as well as an application of the whole-cell biosensor array. The present invention can indicate the moldy state by monitoring the headspace organic volatile components of food raw materials without sample pretreatment, has low cost, high accuracy, and is easy to operate.

Description

Technical Field

[0001] The present invention relates to the field of biotechnology, and more particularly to a whole-cell biosensor array and applications thereof. Background Art

[0002] Food raw materials such as peanuts and corn are contaminated by mold due to improper storage, which not only causes huge economic losses to producers, food processors and consumers, but also leads to a reduction in the sensory quality of food. More importantly, mold contamination is often accompanied by the production of fungal toxins that are harmful to human or animal health, such as aflatoxin B1 (AFB1) with strong carcinogenic and immunosuppressive properties. Some current technologies for detecting mold in food include methods based on polymerase chain reaction (PCR), enzyme-linked immunosorbent assay (ELISA), and technologies based on mass spectrometry to measure organic volatile spectra such as gas chromatography-mass spectrometry (GC-MS) and proton transfer reaction mass spectrometry (PTR-MS). However, these methods require cumbersome sample pretreatment processes, expensive and bulky instruments and equipment, professionally trained technicians, etc., and are not suitable for real-time, continuous monitoring and portable deployment. For this reason, there is an urgent need to develop a low-cost, highly sensitive and highly portable mold monitoring method. The patent of this invention intends to use aflatoxin ( Aspergillus flavus ) Using mold monitoring as a demonstration, a whole-cell biosensor array based on luminescent bacteria was developed. It can monitor organic volatiles in the headspace of food in the early stages of mold non-destructively and continuously, and combine with multiple machine learning prediction models to achieve high-accuracy monitoring and early warning of early mold in food raw materials. Summary of the Invention

[0003] An object of the present invention is to solve at least the above problems and to provide at least the advantages which will be described hereinafter.

[0004] Another object of the present invention is to provide a whole-cell biosensor array and its application, which can indicate the moldy state by monitoring the headspace organic volatile components of food raw materials without sample pretreatment, and has low cost, high accuracy and convenient operation.

[0005] To achieve these objectives and other advantages according to the present invention, the present invention provides a whole-cell biosensor array comprising a 96-well plate, wherein 14 types of whole-cell biosensors are placed on the 96-well plate, with 3-4 of each type of whole-cell biosensors, and each whole-cell biosensor comprises:

[0006] host cells;

[0007] A recombinant vector, which is located in the host cell, comprises a vector and a promoter inserted into the vector, wherein,

[0008] The promoters in the 14 whole-cell biosensors are recAp, dnaKp, grpEp, katGp, fabAp, yibTp, soxSp, glgSp, oxyRp, uvrAp, ompFp, spyp, pspAp and leuAp.

[0009] Preferably, the host cell is Escherichia coli.

[0010] Preferably, the recombinant vector is a plasmid.

[0011] Preferably, the plasmid contains Photorhabdus luminescens Luciferase operon luxCDABE and ampicillin resistance-conferring plasmid pGEN-luxCDABE.

[0012] A method for constructing a whole-cell sensor array is provided, comprising the following steps:

[0013] S1. Obtain host cells and recombinant vectors, and transform the recombinant vectors into the host cells using a transformation method to obtain 14 whole-cell sensors;

[0014] S2. Each whole-cell sensor is made into calcium alginate gel microspheres, and then placed on a 96-well plate, with the number of each whole-cell sensor being 3-4, to obtain a whole-cell sensor array.

[0015] A whole-cell sensor array detection method is provided, comprising the following steps:

[0016] Sa. Place the test sample and the whole-cell sensor array in a sealed aluminum box, and monitor the headspace organic volatiles of the test sample without contact at room temperature;

[0017] Sb. Measure the response data of the whole-cell biosensor array every 1 h using a multi-wavelength microplate reader, then reseal the aluminum box and continue to monitor and analyze headspace organic volatiles for a total of 6 h.

[0018] Sc. Repeat steps Sa-Sb at least 30 times to organize the data into a 120-row × 70-column data matrix;

[0019] Sd, using the caret machine learning analysis package in R language, respectively used random forest, support vector machine linear kernel function, support vector machine radial basis kernel function, sparse partial least squares discriminant analysis, high-dimensional data discriminant analysis and other machine learning model functions to build machine learning models, screen out the model with the highest accuracy for feature value screening and optimization, and finally establish a high-accuracy classification model.

[0020] Provided is an application of a whole-cell sensor array in detecting early mildew of peanuts.

[0021] Provided is an application of a whole-cell sensor array in detecting early mildew of corn.

[0022] The present invention has at least the following beneficial effects:

[0023] This application develops a method that can indicate the moldy state by monitoring the headspace organic volatile components of food raw materials without the need for sample pretreatment. It is low-cost, highly accurate and easy to operate.

[0024] Other advantages, objectives and features of the present invention will be reflected in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is the pGEN-luxCDABE plasmid map;

[0026] Figure 2 Whole-cell biosensor preparation and whole-cell biosensor array diagram;

[0027] Figure 3 The sub-inhibitory concentrations of six organic volatiles were determined and the induced luminescence responses of 14 recombinant luminescent bacteria to them.

[0028] Figure 4 Aspergillus flavus ( Aspergillus flavus ) Diagram of the infection stages in peanuts and corn and the whole-cell biosensor array device for monitoring moldy food;

[0029] Figure 5 A comparison chart of the prediction performance of whole-cell biosensor array combined with six machine learning models for moldy peanuts;

[0030] Figure 6 This is a comparison chart of the prediction performance of moldy corn using a whole-cell biosensor array combined with six machine learning models;

[0031] Figure 7 A comparison of the specific prediction performance of whole-cell biosensor array combined with six machine learning models for moldy peanuts and corn. DETAILED DESCRIPTION

[0032] The present invention is further described in detail below with reference to the embodiments so that those skilled in the art can implement the invention with reference to the description.

[0033] An embodiment of the present application provides a whole-cell biosensor array, comprising a 96-well plate, on which 14 whole-cell biosensors are placed, with 3-4 of each whole-cell biosensor. Each whole-cell biosensor comprises: a host cell; a recombinant vector located within the host cell, the recombinant vector comprising a vector and a promoter inserted into the vector, wherein the promoters within the 14 whole-cell biosensors are recAp, dnaKp, grpEp, katGp, fabAp, yibTp, soxSp, glgSp, oxyRp, uvrAp, ompFp, spyp, pspAp and leuAp, respectively. All 14 promoters are promoters with stress responses, such as membrane damage response, oxidative damage response, intracellular stress response, etc. Each promoter has a nucleic acid sequence of about 300 bp, and the 14 promoters can have different degrees of induction response to the early characteristic organic volatiles of peanuts or corn infected with aflatoxin.

[0034] In other embodiments, the host cell is Escherichia coli.

[0035] In other embodiments, the recombinant vector is a plasmid.

[0036] In other embodiments, the plasmid is Photorhabdus luminescens The luciferase operon luxCDABE and the plasmid pGEN-luxCDABE with ampicillin resistance were cloned by PCR from the Escherichia coli DH5α genome, which contained about 300 bp of the promoter region. The promoter amplified sequence with restriction sites and the sequence with restriction sites were double-digested with BamHI and PmeI restriction endonucleases, respectively. Photorhabdus luminescens The plasmid pGEN-luxCDABE of the luciferase operon of the luminescent bacteria was then used to construct a recombinant plasmid using T4-DNA ligase, which was heat-shock transformed into Escherichia coli DH5α to construct 14 recombinant luminescent bacteria.

[0037] A method for constructing a whole-cell sensor array is provided, comprising the following steps:

[0038] S1. Obtain host cells and recombinant vectors, transform the recombinant vectors into the host cells using a transformation method, obtain 14 whole-cell sensors, and identify 30 organic volatiles in the early (1 and 2 days after infection) and late (3 and 4 days after infection) stages of Aspergillus flavus molded peanuts through headspace solid-phase microextraction and GC-MS techniques. Hierarchical cluster analysis, partial least squares discriminant analysis (PLS-DA), and variable importance (VIP) analysis were used to identify three characteristic organic volatiles in the early stage of peanut mold: 1-methyl-1H-pyrrole, 2,3-butanediol, and ethylpropionate. Dynamic detection of bacterial suspensions at OD 600nm The absorbance of volatile substances, the characteristic values ​​of corn infected with Aspergillus flavus reported in the previous article, and the sub-inhibitory concentrations of 6 organic volatiles including 3 characteristic volatiles ethyl acetate and acetic acid and ethanol that often appear under the conditions of Aspergillus flavus infection on the host bacteria Escherichia coli DH5α of the luminescent bacteria were obtained. The subinhibitory concentrations of acid and ethanol were 1 / 1000, 1 / 200, 1 / 5000, 1 / 5000, 1 / 5000, and 1 / 500 (v / v), respectively. Fourteen candidate promoters for stress responses in E. coli (such as membrane damage response, oxidative damage response, and intracellular stress response) were selected from literature reports and previous experiments. These 14 stress-responsive E. coli promoters were fused with a luciferase operon plasmid to form recombinant plasmids, which were then transformed into E. coli DH5α to construct recombinant E. coli. These recombinant E. coli were then immobilized in calcium alginate gel microspheres, forming the whole-cell biosensor.

[0039] S2. Each whole-cell sensor is made into a calcium alginate gel microsphere, and then placed on a 96-well plate. The number of each whole-cell sensor is 3-4 to obtain a whole-cell sensor array. 2% sodium alginate solution and recombinant Escherichia coli cultured in LB culture medium are mixed in equal volumes, injected into a 5% CaCl2 solution at a flow rate of 2 ml / min via a syringe pump. After fixation for 30 minutes, it is placed in a white 96-well plate to obtain a whole-cell sensor array, among which the recombinant Escherichia coli colonies picked from the LB agar plate are cultured overnight in 5 ml of LB liquid culture medium containing 100 μg / mL ampicillin for activation, and then 100 μL of the recombinant Escherichia coli suspension is re-aspirated into fresh LB liquid culture medium containing 100 μg / mL ampicillin and cultured with shaking for 2 to 3 hours until the absorbance at 600 nm is 0.2, which can be used for calcium alginate gel fixation.

[0040] A whole-cell sensor array detection method is provided, characterized in that it comprises the following steps:

[0041] Sa. Place the test sample and the whole-cell sensor array in a sealed aluminum box, and monitor the headspace organic volatiles of the test sample without contact at room temperature;

[0042] Sb. Measure the response data of the whole-cell biosensor array every 1 h using a multi-wavelength microplate reader, then reseal the aluminum box and continue to monitor and analyze headspace organic volatiles for a total of 6 h.

[0043] Sc. Repeat steps Sa-Sb at least 30 times to organize the data into a 120-row × 70-column matrix.

[0044] Sd, using the caret machine learning analysis package in R language, respectively used random forest, support vector machine linear kernel function, support vector machine radial basis kernel function, sparse partial least squares discriminant analysis, high-dimensional data discriminant analysis and other machine learning model functions to build a machine learning model, screened out the model with the highest accuracy for feature value screening and optimization, and finally established a high-accuracy classification model. Specifically, 14 strains of luminous bacteria were fixed on calcium alginate gel microspheres and arranged on 96 wells to form a whole-cell biosensor array based on luminous bacteria to monitor the organic volatiles in the headspace of peanut and corn samples in the early stage of mold. Combining 6 machine learning algorithms, random forest, support vector machine (linear kernel function), support vector machine (radial basis kernel function), sparse partial least squares discriminant analysis, artificial neural network, high-dimensional discriminant analysis, etc., through model screening and recursive feature elimination and other methods, a high-accuracy classification model was optimized. In the classification models for monitoring moldy peanuts and corn, the random forest model achieved the highest prediction accuracy of 97.5%, while in the specific classification model for distinguishing moldy peanuts and moldy corn, the artificial neural network achieved a prediction accuracy of 92.8%, confirming the high accuracy and high specificity of the whole-cell biosensor array.

[0045] <Example>

[0046] 1. Identification of characteristic organic volatiles in the early stage of peanut mildew

[0047] 1.1 GC-MS determination of organic volatiles in the headspace of peanuts infected with aflatoxin

[0048] Weigh 5g of shelled peanut seeds and sterilize them with 80% (v / v) ethanol solution for 1 minute. Rinse twice with sterile distilled water, then soak in 5% (v / v) sodium hypochlorite solution for 20 minutes, and rinse three times with sterile distilled water. Add 1mL of sterile water to the peanuts and incubate them in a 4°C refrigerator for 48 hours. Measure the water activity every 2 hours using a water activity meter until it reaches approximately 0.90. Rinse Aspergillus flavus spores cultured on potato dextrose agar for 6 days with 1× PBS solution and dilute to a concentration of 1×10 6 1 mL of spore suspension was mixed with 5 g of peanuts adjusted for water activity and placed in 20 mL headspace vials for 0, 1, 2, 3, 4, and 5 days.

[0049] Prior to GC-MS analysis, 5 μL / L of 1-pentanol was added to each headspace vial as an internal standard. GC operating conditions included the following: the inlet temperature was set to 250°C; the carrier gas was helium, the column flow rate was 1.2 mL / min; the injection volume was set to 1 μL, and splitless injection was performed. The temperature program was as follows: the starting temperature was 40°C, maintained for 3 minutes, then increased to 160°C at a rate of 3°C / min, maintained for 2 minutes, then increased to 220°C at a rate of 8°C / min, and maintained for an additional 3 minutes. MS operating conditions included the source temperature being set to 230°C and the quadrupole temperature being set to 150°C. The ionization mode was electron impact using an EI source with an electron energy of 70 eV, and the full scan mass (m / z) range was 50–550 amu. The substances were identified using a computer-automated spectral database, the National Institute of Standards and Technology (NIST) / EPA / NIH library (NIST05) and the Wiley Registry of Mass Spectral Database (Wiley 7.0). C8-C20 normal alkane standards were also analyzed under the same GC-MS conditions to calculate the retention indices of the organic volatiles.

[0050] 1.2 Partial least squares discriminant analysis to identify the characteristic organic volatiles in the early stage of peanuts infected with Aspergillus flavus

[0051] Partial least squares discriminant analysis and variable importance values ​​were used to calculate the importance of each organic volatile compound in the peanut headspace at the early stage of aflatoxin mildew (0, 1, and 2 days), and three characteristic organic volatile compounds were screened out by ranking them according to their importance: 1-methyl-1H-pyrrole, 2,3-butanediol, and ethyl propionate.

[0052] 2. Construction of a whole-cell biosensor based on a promoter-bacterial luciferase recombinant plasmid

[0053] Fourteen E. coli promoters with environmental stress responses were screened from various literature and previous studies (see Table 1), including recAp, dnaKp, grpEp, katGp, fabAp, yibTp, soxSp, glgSp, oxyRp, uvrAp, ompFp, spyp, and pspAp. Based on promoter sequences from the E. coli Promoter Database (https: / / biocyc.org / group?id=:ALL-PROMOTERS&orgid=ECOLI), primers for amplification of the 14 promoter regions were designed (with BamHI and PmeI restriction enzyme sites on the upstream and downstream primers, respectively). DNA fragments containing the promoters, approximately 300 bp in size, were amplified from genomic DNA of E. coli DH5α.

[0054] The amplified fragment containing the promoter and the fragment containing Photorhabdus luminescens ( Photorhabdus luminescens ) The backbone plasmid pGEN-luxCDABE containing the luciferase operon and ampicillin resistance is as follows Figure 1 As shown, after recovery and purification through agarose gel excision, the double-digested plasmid and promoter fragment were ligated with T4-DNA ligase to construct a promoter-bacterial luciferase recombinant plasmid. Finally, the recombinant plasmid was transformed into DH5α competent Escherichia coli using the heat shock method to construct 14 whole-cell biosensors based on the promoter-bacterial luciferase recombinant plasmid.

[0055] Table 1 Information of 14 promoters

[0056]

[0057]

[0058]

[0059] 3. Construction of a whole-cell biosensor array that responds to organic volatiles in the headspace of peanuts and corn infected with Aspergillus flavus

[0060] 3.1 Determination of subinhibitory concentrations of six organic volatiles against Escherichia coli

[0061] The six organic volatiles included three identified early-stage characteristic organic volatiles of Aspergillus flavus mold in peanuts: 1-methyl-1H-pyrrole, 2,3-butanediol, and ethyl propionate; one early-stage characteristic organic volatile of Aspergillus flavus mold in corn: ethyl acetate; and two organic volatiles frequently found under Aspergillus flavus infection conditions: acetic acid and ethanol. 100 μL of overnight cultured Escherichia coli DH5α was added to freshly prepared LB medium and cultured until the OD value of the culture reached 0. 600nm The value was about 0.1, and organic volatiles with concentrations of 1 / 5000, 1 / 2000, 1 / 1000, 1 / 500, 1 / 200 and 0 (v / v) were added respectively, and the shaking culture was continued at 37℃ for 6 hours. The OD of the bacterial solution was measured every 1 hour. 600nm Calculate the OD of each concentration 600nm The value is 0 (v / v), which is the OD of the control 600nm The ratio of the values ​​was plotted over time, and the sub-inhibitory concentrations of the six organic volatiles on Enterobacter DH5α were determined, namely, the sub-inhibitory concentrations of 1-methyl-1H-pyrrole, 2,3-butanediol, ethyl propionate, ethyl acetate, acetic acid and ethanol were 1 / 1000, 1 / 200, 1 / 5000, 1 / 5000, 1 / 5000 and 1 / 500 (v / v), respectively. Figure 3 shown.

[0062] 3.2 Determination of the induced luminescence effects of 14 whole-cell biosensors on 6 sub-inhibitory concentrations of organic volatiles

[0063] After culturing 14 whole-cell biosensors, i.e., recombinant luminescent bacteria, in LB culture medium supplemented with 100 μg / mL ampicillin overnight, 100 μL of the bacterial solution was added to the newly prepared LB culture medium and the culture was continued until the OD 600nmis about 0.2. Then, 6 kinds of organic volatiles at sub-inhibitory concentrations were added respectively, and the luminescence intensity of the whole-cell biosensor was measured every 1 hour for 6 consecutive hours. At the same time, the luminescence intensity of the whole-cell biosensor bacterial solution without organic volatiles was used as a control, and the ratio of the luminescence intensity of the whole-cell biosensor after adding organic volatiles to the luminescence intensity of the control was calculated, that is, the induction factor. The results showed that the induction coefficients of the 14 whole-cell biosensors to the 6 organic volatiles had different induction responses at different times, some of which were significantly up-regulated and some were significantly down-regulated. In summary, each whole-cell biosensor has a certain degree of response to the headspace organic volatiles of peanuts and corn infected with Aspergillus flavus, that is, they can all be used as characteristic variables for monitoring the process of peanuts and corn infected with Aspergillus flavus.

[0064] 4. Whole-cell biosensor array combined with machine learning prediction model to monitor early mildew of peanuts with high accuracy

[0065] 4.1 Preparation of whole-cell biosensor arrays

[0066] Scrape the cells from the LB plates containing 14 recombinant bacteria into 5 mL of liquid LB medium and a shaker tube containing ampicillin at a final concentration of 100 μg / mL. Incubate the cells in a 37°C incubator overnight with shaking. Then, dilute the cultured cells with fresh LB medium to an OD of 600nm The concentration of calcium alginate is about 0.2. After 1 mL of diluted bacterial solution and 1 mL of 2% (m / v) sodium alginate solution are evenly mixed, the mixture is drawn into the syringe and installed on the syringe pump. The injection speed is set to 2 mL / min and the solution is uniformly injected into the 0.5 M CaCl2 solution placed on the magnetic stirrer. After solidification for 30 minutes, the formed calcium alginate microspheres containing the recombinant bacteria are taken out, i.e., a whole-cell biosensor. Figure 2 As shown in A. 14 whole-cell biosensors were arranged on a 96-well plate, with 4 replicates for each, to form a whole-cell biosensor array. Figure 2 As shown in B.

[0067] 4.2 Preparation of peanut seeds infected with Aspergillus flavus

[0068] Weigh 5g of shelled peanut seeds and disinfect their surfaces with 80% (v / v) ethanol for 1 minute. Rinse twice with sterile distilled water, then soak in 5% (v / v) sodium hypochlorite for 20 minutes and rinse three times with sterile distilled water. Add 1mL of sterile water to the peanuts and incubate in a 4°C refrigerator for 48 hours. Measure the water activity every two hours using a water activity meter until it reaches approximately 0.90.

[0069] Inoculate the Aspergillus flavus spores onto potato dextrose agar medium and, after culturing for 6 days, rinse the spores with 1× PBS solution and dilute to a concentration of 10 6 Spore suspension of spores / mL. Pipette 1mL of spore liquid and mix it with 5g peanuts with adjusted water activity, and place it in an incubator at 26℃. Figure 4 As shown in A, the period 0, 1, and 2 days after inoculation with the Aspergillus flavus spore liquid is the early mildew stage, i.e., there is no obvious mycelial growth; while the period 3, 4, and 5 days after inoculation is the mildew stage, i.e., there is obvious mycelial growth.

[0070] 4.3 Determination of headspace organic volatiles in peanuts during the early stages of aflatoxin mildew by whole-cell biosensor array

[0071] The prepared whole-cell biosensor array was inoculated with 1 mL of 10 6 The peanut samples on the 1st and 2nd day after the aflatoxin spore liquid were placed together in a relatively closed aluminum box without direct contact. Figure 4 As shown in Figure B. As a control, the prepared whole-cell biosensor array was placed in a relatively closed aluminum box with peanut samples 1 and 2 days after inoculation with 1×PBS solution, without direct contact. After every 1 hour of incubation, the luminescence response of the whole-cell biosensor array was measured using a multifunctional microplate reader and recorded continuously for 5 hours. This experiment was repeated 30 times, resulting in 30 groups of response values ​​1 day after aflatoxin inoculation, 30 groups of response values ​​2 days after aflatoxin inoculation, 30 groups of response values ​​1 day after control treatment, and 30 groups of response values ​​2 days after control treatment. By merging the experimental data observed each day, a total of 120 groups of observed response data were obtained. Each group of observation data included the luminescence response values ​​of each recombinant bacterial strain at 5 time points. For the 14 recombinant bacterial strains, a total of 70 luminescence response values ​​were obtained, resulting in a data matrix of 120 rows × 70 columns.

[0072] 4.4 Construction of peanut mold prediction model

[0073] A prediction model was constructed using each row of data as an observation and each column as an eigenvalue, for a total of 120 observations and 70 eigenvalues. Six machine learning algorithms within the R language Caret data processing framework were used to construct binary and three-class machine learning models, respectively: random forest, support vector machine (linear kernel), support vector machine (radial basis kernel), sparse partial least squares discriminant analysis, artificial neural network, and high-dimensional discriminant analysis. The binary classification model was based on the induction coefficient (i.e., the ratio of the luminescence value of the recombinant bacteria in response to the moldy period to the luminescence value of the control treatment at the same time). This model was able to distinguish between headspace organic volatiles in peanuts that had been moldy for one day ("1 dpi_peanut") and those that had been moldy for two days ("2 dpi_peanut"). The three-class model is a model that reconstructs the luminescence value of bacteria in response to mildew for 1 day, the luminescence value in response to mildew for 2 days, and the luminescence value in response to the control treatment. This model can distinguish the differences in headspace organic volatiles between the control, peanuts that have been mildewed for 1 day ("1 dpi_peanut"), and peanuts that have been mildewed for 2 days ("2 dpi_peanut"), as shown in the following example: Figure 5 shown.

[0074] By comparing the two-class and three-class models established by 6 machine learning algorithms, it was found that the three-class model has a higher classification accuracy than the two-class model, reaching a prediction accuracy of 90%, especially the three-class model based on random forest, with a classification accuracy of 97.5%. In order to reduce redundant eigenvalues ​​and increase prediction accuracy, a further recursive feature elimination method was used to eliminate eigenvalues ​​with low contribution, reducing the original 70 eigenvalues ​​to the current 30 eigenvalues. Although the prediction accuracy remains the original high accuracy of 97.5%, the additional cost caused by collecting eigenvalues ​​is greatly reduced. It was finally determined that the three-class random forest prediction model based on 30 eigenvalues ​​is the best model for predicting the early stage of peanut aflatoxin mildew, with a prediction accuracy of 97.5%. It can predict the mildew status of peanuts as early as one day after the infection of peanut aflatoxin. Figure 5 shown.

[0075] 5. Whole-cell biosensor array combined with machine learning prediction model to monitor early mildew of corn with high accuracy

[0076] 5.1 Preparation of corn seeds infected with Aspergillus flavus

[0077] Weigh 5g of corn seeds and disinfect their surfaces with 80% (v / v) ethanol for 1 minute. Rinse twice with sterile distilled water, then soak in 5% (v / v) sodium hypochlorite for 20 minutes and rinse three times with sterile distilled water. Add 1mL of sterile water to the corn and incubate in a 4°C refrigerator for 48 hours. Measure the water activity every two hours using a water activity meter until it reaches approximately 0.90.

[0078] Inoculate the Aspergillus flavus spores onto potato dextrose agar medium and, after culturing for 6 days, rinse the spores with 1× PBS solution and dilute to a concentration of 10 6 Spore suspension of spores / mL. Pipette 1mL of spore liquid and mix it with 5g corn with adjusted water activity, and place it in an incubator at 26℃. Figure 2 As shown in A, the period 0, 1, and 2 days after inoculation with the Aspergillus flavus spore liquid is the early mildew stage, i.e., there is no obvious mycelial growth; while the period 3, 4, and 5 days after inoculation is the mildew stage, i.e., there is obvious mycelial growth.

[0079] 5.2 Determination of headspace organic volatiles in the early stages of aflatoxin mildew in corn using a whole-cell biosensor array

[0080] The prepared whole-cell biosensor array was inoculated with 1 mL of 10 6 The corn samples 1 and 2 days after the aflatoxin spore solution were placed together in a relatively closed aluminum box without direct contact. Figure 4 As shown in Figure B. As a control, the prepared whole-cell biosensor array was placed in a relatively closed aluminum box with no direct contact with corn samples 1 and 2 days after inoculation with 1×PBS solution. After incubation for 1 hour, the response luminescence value of the whole-cell biosensor array was measured using a multifunctional microplate reader and recorded continuously for 5 hours. This experiment was repeated 30 times, and 30 groups of response values ​​were obtained for 1 day after Aspergillus flavus inoculation, 30 groups of response values ​​for 2 days after Aspergillus flavus inoculation, 30 groups of response values ​​for 1 day after control treatment, and 30 groups of response values ​​for 2 days after control treatment. By merging the experimental data observed every day, a total of 120 groups of observed response data were obtained. Each group of observation data included the luminescence response values ​​of each recombinant bacterial strain at 5 time points. For the 14 recombinant bacterial strains, a total of 70 luminescence response values ​​were obtained, and a data matrix of 120 rows × 70 columns was finally obtained.

[0081] 5.3 Construction of corn mold prediction model

[0082] A prediction model was constructed using each row of data as an observation and each column as an eigenvalue, for a total of 120 observations and 70 eigenvalues. Six machine learning algorithms within the R language Caret data processing framework were used to construct binary and ternary classification models, respectively: random forest (RF), support vector machine (linear kernel) (svmLinear), support vector machine (radial basis kernel) (svmRadial), sparse partial least squares discriminant analysis (sPLSDA), artificial neural network (NN), and high-dimensional discriminant analysis (HDDA). The binary classification model was based on the induction coefficient (i.e., the ratio of the luminescence value of the recombinant bacteria in response to the moldy period to the luminescence value of the control treatment at the same time). This model was able to distinguish between headspace organic volatiles (VOCs) from corn 1 day after moldy ("1 dpi_maize") and 2 days after moldy ("2 dpi_maize"). The three-class model is a model that reconstructs the luminescence value of bacteria in response to moldy 1 day, the luminescence value in response to moldy 2 days, and the luminescence value in response to the control treatment. This model can distinguish the differences in headspace organic volatiles between the control, corn moldy 1 day ("1 dpi_maize"), and peanut moldy 2 days ("2 dpi_maize"), as shown in the following example: Figure 6 shown.

[0083] By comparing the two-classification and three-classification models established by 6 machine learning algorithms, it was found that the three-classification model has a higher classification accuracy than the two-classification model, reaching a prediction accuracy of more than 90%, especially the three-classification model based on random forest (RF), with a classification accuracy of 97.5%. In order to reduce redundant eigenvalues ​​and increase prediction accuracy, a further recursive feature elimination method was used to eliminate eigenvalues ​​with low contribution, reducing the original 70 eigenvalues ​​to the current 29 eigenvalues. Although the prediction accuracy remains the original high accuracy of 97.5%, the additional cost caused by collecting eigenvalues ​​is greatly reduced. It was finally determined that the three-classification random forest (RF) prediction model based on 29 eigenvalues ​​is the best model for predicting the early stage of corn aflatoxin mildew, with a prediction accuracy of 97.5%. It can predict the mildew status of corn as early as one day after corn aflatoxin infection. Figure 6 shown.

[0084] 6. Whole-cell biosensor arrays combined with machine learning prediction models can accurately distinguish early mildew in peanuts and corn.

[0085] 6.1 Preparation of Whole-Cell Biosensor Arrays and Peanut and Corn Samples Infected with Aspergillus flavus Whole-cell biosensor arrays and peanut and corn samples infected with Aspergillus flavus were prepared as described in Examples 4 and 5;

[0086] 6.2 Whole-cell biosensor array determination of headspace organic volatiles in peanuts and corn at the early stage of aflatoxin mildew The process of whole-cell biosensor array determination of headspace organic volatiles in peanuts and corn at the early stage of aflatoxin mildew is as described in Examples 4 and 5.

[0087] 6.3 Construction of a prediction model for distinguishing peanut and corn mold

[0088] Using the daily induction coefficients of peanuts and corn (i.e., the ratio of the luminescence value of the recombinant bacteria responding to the moldy samples to the luminescence value of the recombinant bacteria responding to the control treatment at the same time), a 120-row × 70-column data matrix was established, i.e., a total of 60 groups of peanut mold response values ​​(1 day and 2 days after moldy) and 60 groups of corn mold response values ​​(1 day and 2 days after moldy). Six machine learning-based binary classification models were established. The binary classification models were divided into two categories: moldy peanuts ("Infected_peanut") and moldy corn ("Infected_maize"). The best performing model was the one based on an artificial neural network (NN), with an accuracy of over 90%. Figure 7 shown.

[0089] In order to further improve the accuracy of the prediction model and reduce the number of eigenvalues, the recursive feature elimination method was used to eliminate eigenvalues ​​with low contribution, reducing the original 70 eigenvalues ​​to the current 44 eigenvalues. Although the prediction accuracy was slightly reduced, the additional cost of collecting eigenvalues ​​was greatly reduced. Finally, it was determined that the binary classification artificial neural network (NN) prediction model based on 44 eigenvalues ​​was the best model for distinguishing the early stages of aflatoxin mold in peanuts and corn, with a prediction accuracy of 92.8%. Figure 7 shown.

[0090] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method for constructing a whole-cell biosensor array, characterized in that: The following steps are involved: S1. Obtain host cells and recombinant vectors, and transform the recombinant vectors into the host cells using a transformation method to obtain 14 whole-cell biosensors: the promoters in the 14 whole-cell biosensors are recAp, dnaKp, grpEp, katGp, fabAp, yibTp, soxSp, glgSp, oxyRp, uvrAp, ompFp, spyp, pspAp, and leuAp, According to the promoter sequence in the E. coli promoter database, 14 amplification primers of the promoter region were designed respectively, and the DNA fragment containing the promoter was amplified from the genomic DNA of E. coli DH5α. The amplified fragment containing the promoter and the fragment containing Photorhabdus luminescens were double-digested with restriction endonucleases BamHI and PmeI. Photorhabdus luminescens The backbone plasmid pGEN-luxCDABE containing the luciferase operon and ampicillin resistance was recovered and purified by agarose gel excision. The double-digested plasmid and promoter fragment were ligated with T4-DNA ligase to construct a promoter-bacterial luciferase recombinant plasmid. Finally, the recombinant plasmid was transformed into DH5α competent Escherichia coli using the heat shock method to construct 14 whole-cell biosensors based on the promoter-bacterial luciferase recombinant plasmid. The promoter dnaKp is amplified using the base sequences shown in SEQ ID NOs: 1 and 2 as primers. The promoter fabAp was amplified using the base sequences shown in SEQ ID NOs: 3 and 4 as primers. The promoter glgSp was amplified using the base sequences shown in SEQ ID NOs: 5 and 6 as primers. The promoter grpEp was amplified using the base sequences shown in SEQ ID NOs: 7 and 8 as primers. The promoter katGp was amplified using the base sequences shown in SEQ ID NOs: 9 and 10. The promoter leuAp was amplified using the base sequences shown in SEQ ID NOs: 11 and 12 as primers. The promoter ompFp was amplified using the base sequences shown in SEQ ID NOs: 13 and 14 as primers. The promoter oxyRp was amplified using the base sequences shown in SEQ ID NOs: 15 and 16 as primers. The promoter pspAp was amplified using the base sequences shown in SEQ ID NOs: 17 and 18 as primers. The promoter recAp was amplified using the base sequences shown in SEQ ID NOs: 19 and 20. The promoter soxSp was amplified using the base sequences shown in SEQ ID NOs: 21 and 22 as primers. The promoter spyp was amplified using the base sequences shown in SEQ ID NOs: 23 and 24 as primers. The promoter uvrAp was amplified using the base sequences shown in SEQ ID NOs: 25 and 26. The promoter yibTp was amplified using the base sequences shown in SEQ ID NOs: 27 and 28 as primers; S2. Each whole-cell sensor is made into calcium alginate gel microspheres, and then placed on a 96-well plate, with the number of each whole-cell sensor being 3-4, to obtain a whole-cell sensor array.

2. A whole-cell biosensor array constructed by the construction method according to claim 1.

3. The detection method of the whole-cell biosensor array according to claim 2, wherein: The following steps are involved: Sa. Place the test sample and the whole-cell sensor array in a sealed aluminum box, and monitor the headspace organic volatiles of the test sample without contact at room temperature; Sb. Measure the response data of the whole-cell biosensor array every 1 h using a multi-wavelength microplate reader, then reseal the aluminum box and continue to monitor and analyze headspace organic volatiles for a total of 6 h. Sc. Repeat steps Sa-Sb at least 30 times to organize the data into a data matrix of 120 rows × 70 columns; Sd, using the caret machine learning analysis package in R language, respectively used random forest, support vector machine linear kernel function, support vector machine radial basis kernel function, sparse partial least squares discriminant analysis, and high-dimensional data discriminant analysis to build a machine learning model, screened out the model with the highest accuracy for feature value screening and optimization, and finally established a high-accuracy classification model.

4. Use of the whole-cell biosensor array according to claim 2 in detecting early mildew of peanuts.

5. Use of the whole-cell biosensor array according to claim 2 in detecting early mildew of corn.

Images