Multi-dimensional tobacco mildew marker screening method

Abstract

The invention discloses a multi-dimensional tobacco mildew marker screening method, which comprises the following steps: carrying out standardized sample collection and grouping, and constructing a sample system; obtaining multi-dimensional feature data in the tobacco leaf mildewing process through multi-dimensional detection; preprocessing and fusing the multi-dimensional feature data to obtain integrated data; on the basis of the integrated data, utilizing a machine learning model to screen features related to the tobacco leaf mildewing state; the screened features are verified, and the high-specificity marker is determined. According to the method, technologies such as metabonomics, microbiomics, volatile organic compound analysis and spectral imaging are combined, the high-specificity biomarkers are screened through multi-source data integration and machine learning, and a scientific basis is provided for intelligent and accurate detection, quality monitoring and early warning of tobacco leaf mildewing.

Description

Technical Field

[0001] This invention relates to the field of tobacco detection technology, and more specifically to a method for screening multidimensional tobacco leaf mold markers. Background Technology

[0002] Currently, tobacco leaves are susceptible to mold growth during storage due to temperature and humidity fluctuations, leading to quality degradation and even the production of harmful toxins, severely impacting the safety and economic value of tobacco products. The current limitations of tobacco leaf mold detection technologies mainly include:

[0003] 1) Limitations of characteristic markers: Existing methods mostly rely on single or small amounts of volatile compounds (such as 1-octen-3-ol, pentanal) as mold markers, which may miss other key metabolites, resulting in insufficient detection sensitivity and comprehensiveness, especially poor adaptability in complex environments or different tobacco varieties.

[0004] 2) High dependence on subjective evaluation: Some technologies require the use of human sensory evaluation such as vision and smell (e.g., grading of moldy area and scoring of moldy smell), which can lead to individual differences in judgment and affect the objectivity and repeatability of the results.

[0005] 3) Cumbersome operation process: Existing methods require multiple steps (such as mold isolation, re-inoculation, colony counting, GC-IMS detection), which results in long experimental cycles and complex operations, easily introducing human error and making it difficult to achieve rapid on-site detection.

[0006] 4) Insufficient biomarker validation: Some methods determine biomarkers only by comparing naturally moldy and artificially moldy samples, raising questions about the specificity and reliability of the biomarkers.

[0007] In summary, existing technologies generally suffer from problems such as one-sided marker selection, insufficient detection sensitivity, and strong subjectivity of results, making it difficult to meet the needs of intelligent and precise management of modern tobacco warehousing.

[0008] Therefore, there is an urgent need for a technical solution that can comprehensively, objectively, and efficiently screen for highly specific markers of tobacco mold. Summary of the Invention

[0009] In view of this, the present invention provides a multi-dimensional method for screening tobacco leaf mold biomarkers that at least solves some of the above-mentioned technical problems. The present invention integrates metabolomics, microbiome, volatile organic compound (VOCs) analysis and spectral imaging technology, combined with data fusion and machine learning, to achieve a comprehensive analysis of the tobacco leaf mold process, screen out highly specific tobacco leaf mold biomarkers, and provide a scientific basis for intelligent and accurate detection, quality monitoring and early warning of tobacco leaf mold.

[0010] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0011] This invention provides a method for screening multidimensional tobacco leaf mold markers, the method comprising the following steps:

[0012] S1. Conduct standardized sample collection and grouping to construct a sample system;

[0013] S2. Obtain multi-dimensional characteristic data of tobacco leaf mold growth process through multi-dimensional detection; the multi-dimensional characteristic data includes: metabolomics data, microbiome data, volatile organic compound (VOC) data and spectral imaging data of tobacco leaf samples;

[0014] S3. Preprocess and fuse the multi-dimensional feature data to obtain integrated data;

[0015] S4. Based on the integrated data, use a machine learning model to screen features related to the moldy state of tobacco leaves; verify the screened features to determine highly specific biomarkers.

[0016] Furthermore, in step S1, tobacco leaf samples with different storage conditions, varieties, and degrees of mold are selected and divided into:

[0017] Control group: No moldy tobacco leaves. No moldy tobacco leaves were stored for 0-30 days with a mold count <10. 3 CFU / g, no visible mold spots;

[0018] Mold group: classified according to the degree of mold, including:

[0019] Initial stage: Localized minor mold spots, mold count 10 3 -10 5 CFU / g;

[0020] Mid-term: Mold spots account for 10%-30%, mold count 10 5 -10 7 CFU / g;

[0021] Later stage: Mold spots account for >30%, mold count >10 7 CFU / g;

[0022] Negative control: Fresh tobacco leaves from the same batch that have no risk of mold growth.

[0023] Furthermore, in S2, the multi-dimensional detection includes:

[0024] (1) Metabolomics detection: to capture changes in endogenous metabolites in tobacco leaves;

[0025] (2) Microbiome detection: analyzing changes in mold-related microbial communities:

[0026] (3) VOCs analysis: Identify the characteristic volatile organic compounds released by mold growth;

[0027] (4) Spectral imaging: Obtain the physicochemical characteristic spectrum of mold growth.

[0028] Furthermore, in S3, the preprocessing of multi-dimensional feature data includes: standardization, dimensionality reduction, and missing value processing;

[0029] The fusion of multi-dimensional feature data includes:

[0030] The preprocessed data is concatenated into a sample-multi-feature matrix;

[0031] Extract features from individual source data, then integrate the feature sets;

[0032] Weights are assigned to different data sources through correlation analysis.

[0033] Furthermore, in step S4, the use of a machine learning model to screen features related to the moldy state of tobacco leaves specifically includes:

[0034] Primary screening: Partial least squares discriminant analysis was used to identify features significantly associated with mold growth;

[0035] Fine-grained selection: Use random forests, support vector machines, or gradient boosting trees to sort by feature importance scores;

[0036] Deep learning: using convolutional neural networks to extract deep features from high-dimensional data, and combining attention mechanisms to locate key markers.

[0037] Furthermore, in step S4, the selected features are verified, specifically including:

[0038] Internal validation: Model accuracy was evaluated using 5-fold cross-validation;

[0039] External validation: Validating the specificity and sensitivity of biomarkers using independent sample sets.

[0040] As can be seen from the above technical solution, the present invention provides a multi-dimensional method for screening tobacco leaf mold markers. Compared with the prior art, the present invention has at least the following beneficial effects:

[0041] This invention achieves highly specific screening of mold markers through a process of "parallel detection of multiple technologies - fusion of multiple data sources - intelligent screening". It can comprehensively, objectively and efficiently screen highly specific markers of mold in tobacco leaves, providing a scientific basis for intelligent and accurate detection, quality monitoring and early warning of mold in tobacco leaves.

[0042] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0043] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0046] Figure 1 This is a flowchart illustrating a multi-dimensional method for screening tobacco mold markers provided by the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0048] In the description of this invention, it should be noted that some processes described in this application specification and drawings include multiple operations that appear in a specific order. However, it should be clearly understood that these operations may be performed in any order or in parallel. Furthermore, various numbers are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0049] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0050] See Figure 1As shown, this invention provides a multi-dimensional method for screening biomarkers of tobacco mold. This method mainly combines metabolomics, microbiome, volatile organic compound (VOCs) analysis, and spectral imaging technologies. Through multi-source data integration and machine learning, it screens highly specific biomarkers, including the following steps:

[0051] S1. Conduct standardized sample collection and grouping to construct a sample system;

[0052] S2. Obtain multi-dimensional feature data of tobacco leaf mold growth process through multi-dimensional detection;

[0053] S3. Preprocess and fuse multi-dimensional feature data to obtain integrated data;

[0054] S4. Based on the integrated data, use machine learning models to screen features related to the mold state of tobacco leaves; verify the screened features and identify highly specific biomarkers.

[0055] The specific embodiments, technical processes, and working principles of the present invention will be described in detail below:

[0056] This invention achieves highly specific screening of mold biomarkers through a closed-loop process of "parallel detection of multiple technologies - fusion of multiple data sources - intelligent screening". Specifically, it includes four core parts: sample system construction, application of multi-dimensional detection technologies, data integration framework and machine learning screening model.

[0057] 1. Sample system construction: standardized sample collection and grouping.

[0058] To ensure the reliability of the biomarkers, a sample library covering the entire mold growth cycle needs to be established, and the grouping and processing standards for the samples need to be clearly defined.

[0059] ①Sample source:

[0060] In one specific implementation, tobacco leaf samples with different storage conditions (such as temperature and humidity, storage time), different varieties, and different degrees of mold were selected, including:

[0061] Control group: Unmolded tobacco leaves (stored for 0-30 days, mold count <10) 3 CFU / g, with no visible mold spots);

[0062] Mold group: graded according to the degree of mold (early stage: localized minor mold spots, mold count 10) 3 -10 5 CFU / g; Intermediate stage: mold coverage 10%-30%, mold count 10 5 -10 7 CFU / g; Later stage: Mold coverage >30%, mold count >10 7 CFU / g);

[0063] Negative control: Fresh tobacco leaves from the same batch with no risk of mold (as baseline reference).

[0064] ② Sample processing:

[0065] In one specific implementation, each sample is divided into four parallel subsamples, which are used for metabolomics, microbiome, VOCs analysis, and spectral imaging detection, respectively.

[0066] Microbiome samples should be immediately rinsed with sterile PBS (to remove environmental contaminants) and frozen at -80°C; metabolomics samples should be ground with liquid nitrogen and stored at -80°C; VOCs samples should be sealed in headspace vials (to avoid loss of volatiles); spectral imaging samples should be kept in their original form and tested immediately.

[0067] 2. Specific applications of multi-dimensional detection technology:

[0068] In one specific implementation, four types of technologies are used in parallel detection to obtain multi-dimensional characteristic data of the "biological-chemical-physical" process of tobacco leaf mold growth:

[0069] (1) Metabolomics detection: to capture changes in endogenous metabolites in tobacco leaves;

[0070] Detection objective: To screen for small molecule metabolites (such as amino acids, organic acids, sugars, lipids, etc.) that show significant differences during the mold growth process, reflecting abnormalities in the physiological state of tobacco leaves.

[0071] Technical process:

[0072] Metabolite extraction: Take 50 mg of tobacco powder, extract with 80% methanol (containing internal standard) by ultrasonic extraction for 30 min, centrifuge at 12000 rpm for 10 min, and filter the supernatant (0.22 μm filter membrane);

[0073] Detection platform: Ultra-high performance liquid chromatography-tandem mass spectrometry (UHPLC-MS / MS) with positive / negative ion mode detection; or gas chromatography-mass spectrometry (GC-MS) (for volatile / polar metabolites); Data output: Obtain a metabolite peak area matrix (rows: sample, columns: metabolites, values: relative concentrations), and match metabolite identities using databases (such as HMDB, Metlin).

[0074] (2) Microbiome detection: analyzing changes in mold-related microbial communities;

[0075] Detection objective: To screen for microorganisms (fungi / bacteria) that are significantly associated with mold growth and to clarify the "pathogen-mold growth" association (e.g., Aspergillus, Penicillium, etc. may be key pathogens).

[0076] Technical process:

[0077] Microbial DNA extraction: Total microbial DNA was extracted from the surface and interior of tobacco leaves using the CTAB method, and then amplified by PCR using primers for the V4-V5 region (bacteria) and ITS1 region (fungi) of the 16S rRNA gene.

[0078] Sequencing platform: Illumina MiSeq high-throughput sequencing, obtaining OTU (operational taxonomic unit) data;

[0079] Data output: Species abundance matrix (rows: sample, columns: species / OTU, values: relative abundance) is obtained, and community α / β diversity and differential species are analyzed using QIIME2.

[0080] (3) VOCs analysis: Identify the characteristic volatile organic compounds released by mold growth;

[0081] Detection objective: To screen for volatile organic compounds (such as aldehydes, ketones, and terpenes) specific to mold growth as "non-invasive" mold growth markers (which can be used for rapid detection).

[0082] Technical process:

[0083] VOCs capture: headspace solid-phase microextraction (HS-SPME, such as 50 / 30μm DVB / CAR / PDMS fiber head) was used, with headspace equilibration at 60℃ for 30 min and adsorption for 15 min;

[0084] Detection platform: Gas chromatography-mass spectrometry (GC-MS), DB-5MS column (30m×0.25mm×0.25μm), programmed temperature rise (40℃ for 3 min, then 5℃ / min to 250℃);

[0085] Data output: VOCs type and relative content matrix (compounds matched by NIST library), such as "3-methylbutyraldehyde, benzaldehyde" which may be enriched by mold.

[0086] (4) Spectral imaging: to obtain physicochemical characteristic spectra of mold growth;

[0087] Detection objective: To screen for spectral features related to mold growth (such as absorption peaks at specific wavelengths) to achieve non-destructive and rapid mold growth identification (which can be correlated with spatial distribution).

[0088] Technical process:

[0089] Detection platform: Hyperspectral imaging system (such as GaiaField-V10, 400-1000nm band) or Raman spectrometer;

[0090] Spectral acquisition: Hyperspectral imaging requires line scanning of tobacco leaves (5nm resolution) to obtain the spectral curve of each pixel; Raman spectroscopy requires focusing on mold spots and normal areas (laser wavelength 785nm) to collect characteristic peaks;

[0091] Data output: spectral matrix (rows: sample / pixel, columns: wavelength, values: absorbance / intensity), extracting mold-related spectral features (e.g., 1650 cm⁻¹) using chemometrics. -1 (Changes in protein peaks).

[0092] 3. Integrate multi-source data to build a unified analysis framework.

[0093] The data types from different technologies vary greatly (metabolic genomics: continuous variables; microbiome: count data; spectroscopy: high-dimensional waveforms), requiring preprocessing and fusion for integration.

[0094] ① Data preprocessing:

[0095] Standardization: Metabolomics / VOCs data were normalized using internal standard; microbiome data were normalized using relative abundance conversion; spectral data were normalized using multivariate scattering correction (MSC) to remove baseline drift.

[0096] Dimensionality reduction: Principal component analysis (PCA) is used to retain principal components with more than 85% variance for high-dimensional data (such as 1000+ wavelengths of spectrum);

[0097] Missing value handling: Metabolomics / microbiome is filled with "minimum values", and spectra are filled with "interpolation".

[0098] ② Data fusion:

[0099] Early fusion: preprocessed multi-source data are stitched together into a "sample-multi-feature" matrix (e.g., metabolites + species + VOCs + spectral principal components);

[0100] Intermediate fusion: Extract features from single-source data (such as differential metabolites in the metabolome and differential species in the microbiome) and then integrate the feature sets;

[0101] Weighting: Different data sources are assigned weights through correlation analysis (such as Pearson coefficient) (e.g., VOCs with a high correlation to mold growth have higher weights).

[0102] 4. Machine learning for screening highly specific biomarkers:

[0103] In one specific implementation, based on integrated data, a supervised learning model is used to screen for markers that have a strong ability to distinguish between "moldy / normal" and "degree of mold":

[0104] ① Model selection:

[0105] Primary screening: Partial least squares discriminant analysis (PLS-DA) was used to identify features significantly associated with mold growth (VIP>1);

[0106] Fine-grained selection: Use Random Forest (RF), Support Vector Machine (SVM), or Gradient Boosting Tree (XGBoost) to sort by "feature importance score" (such as the Gini index of RF);

[0107] Deep learning: Convolutional neural networks (CNNs) are used to extract deep features from high-dimensional data (such as spectral imaging) and attention mechanisms are combined to locate key markers.

[0108] ②Marker verification:

[0109] Internal validation: 5-fold cross-validation was used to evaluate the model accuracy (target >90%).

[0110] External validation: Validate the specificity (true negative rate >95%) and sensitivity (true positive rate >90%) of the biomarker using an independent sample set (20% of the total sample).

[0111] Final selection: Features that were identified in at least three models and had an area under the ROC curve (AUC > 0.9) on the validation set were retained as core biomarkers (e.g., Aspergillus abundance + 3-methylbutyraldehyde concentration + 1650 cm⁻¹). -1 (Spectral peak combination).

[0112] In summary, this approach utilizes a multi-dimensional complementary technology approach (microbiome reveals the inducing factors, metabolome reflects the host response, and VOCs / spectroscopy provides convenient detection). The screened biomarkers possess both "high specificity" (distinguishing mold from other deteriorations) and "practicality" (can be used for the development of rapid detection equipment). It can be applied to tobacco storage monitoring (such as online VOCs sensors), quality grading (spectral imaging screening), and mold early warning systems.

[0113] As described in the above embodiments, those skilled in the art will understand that the present invention proposes a multi-dimensional method for screening tobacco mold biomarkers. This method mainly combines metabolomics, microbiome, volatile organic compound (VOCs) analysis, and spectral imaging technologies. Through multi-source data integration and machine learning, it screens highly specific biomarkers. The present invention achieves highly specific screening of mold biomarkers through a closed-loop process of "parallel detection of multiple technologies - multi-source data fusion - intelligent screening," providing a scientific basis for intelligent and accurate detection and early warning of tobacco mold. Compared with existing technologies, the present invention has the following advantages:

[0114] High comprehensiveness: It integrates four types of technologies: metabolomics, microbiome, VOCs and spectral imaging, which can comprehensively capture the biological, chemical and physical changes in the process of tobacco mold growth, avoiding the risk of missing a single marker;

[0115] High specificity: Through multi-model cross-validation and independent sample testing, a combination of biomarkers with high AUC, high sensitivity and specificity can be screened out, significantly improving the accuracy of discrimination;

[0116] High objectivity: It abandons human sensory evaluation, realizes full-process data-driven, and the results are repeatable and traceable;

[0117] High practicality: The screened markers can be directly used to develop portable sensors (such as electrochemical VOCs sensors) and hyperspectral online detection equipment, supporting the construction of intelligent warehousing systems;

[0118] Highly scalable: The method is applicable to the detection of mold in other agricultural products (such as tea and Chinese medicinal materials), and has broad application value.

[0119] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, devices, or computer program products, etc. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0120] It should be noted that the word "comprising" does not exclude the presence of components or steps not listed in the claims. The words "a" or "an" preceding a component do not exclude the presence of a plurality of such components. This invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer.

[0121] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for screening multidimensional tobacco leaf mold markers, characterized in that, The method includes the following steps: S1. Conduct standardized sample collection and grouping to construct a sample system; S2. Obtain multi-dimensional feature data of tobacco leaf mold growth process through multi-dimensional detection; S3. Preprocess and fuse the multi-dimensional feature data to obtain integrated data; S4. Based on the integrated data, use a machine learning model to screen features related to the moldy state of tobacco leaves; verify the screened features to determine highly specific biomarkers.

2. The method for screening multidimensional tobacco leaf mold markers according to claim 1, characterized in that, In step S1, tobacco leaf samples with different storage conditions, varieties, and degrees of mold are selected and divided into: Control group: No moldy tobacco leaves. No moldy tobacco leaves were stored for 0-30 days with a mold count <10. 3 CFU / g, no visible mold spots; Mold group: classified according to the degree of mold, including: Initial stage: Localized minor mold spots, mold count 10 3 -10 5 CFU / g; Mid-term: Mold spots account for 10%-30%, mold count 10 5 -10 7 CFU / g; Later stage: Mold spots account for >30%, mold count >10 7 CFU / g; Negative control: Fresh tobacco leaves from the same batch that have no risk of mold growth.

3. The method for screening multidimensional tobacco leaf mold markers according to claim 1, characterized in that, In S2, the multi-dimensional detection includes: (1) Metabolomics detection: to capture changes in endogenous metabolites in tobacco leaves; (2) Microbiome detection: analyzing changes in mold-related microbial communities: (3) VOCs analysis: Identify the characteristic volatile organic compounds released by mold growth; (4) Spectral imaging: Obtain the physicochemical characteristic spectrum of mold growth.

4. The method for screening multidimensional tobacco leaf mold markers according to claim 1, characterized in that, In step S3, the preprocessing of multi-dimensional feature data includes: standardization, dimensionality reduction, and missing value processing; The fusion of multi-dimensional feature data includes: The preprocessed data is concatenated into a sample-multi-feature matrix; Extract features from individual source data, then integrate the feature sets; Weights are assigned to different data sources through correlation analysis.

5. The method for screening multidimensional tobacco leaf mold markers according to claim 1, characterized in that, In step S4, the machine learning model is used to screen features related to the moldy state of tobacco leaves, specifically including: Primary screening: Partial least squares discriminant analysis was used to identify features significantly associated with mold growth; Fine-grained selection: Use random forests, support vector machines, or gradient boosting trees to sort by feature importance scores; Deep learning: using convolutional neural networks to extract deep features from high-dimensional data, and combining attention mechanisms to locate key markers.

6. The method for screening multidimensional tobacco leaf mold markers according to claim 1, characterized in that, In step S4, the selected features are verified, specifically including: Internal validation: Model accuracy was evaluated using 5-fold cross-validation; External validation: Validating the specificity and sensitivity of biomarkers using independent sample sets.

Images