Metagenome sequencing-based method for analyzing pathogens carried by vector organisms

Abstract

The invention discloses a method for analyzing pathogens carried by a vector organism based on metagenome sequencing, belongs to the technical field of vector organism analysis, and realizes the crossing of pathogen identification from single sequence alignment to system biological analysis by constructing a multi-dimensional data fusion framework. The method has the core advantages that ecological background data and sequencing data in a sample collection stage are deeply coupled, the limitation that traditional species recognition depends on a database is broken through a dynamic weight network model, and association recognition of unknown pathogens can be achieved by means of cross-species conservative property of a gene function module under the condition that sequence homology is insufficient. According to the design, the breadth of metagenome high-throughput detection is reserved, and the internal relation of species-gene-function is strengthened through the three-layer network structure, so that an identification result not only contains pathogen classification information, but also can be synchronously associated with metabolic activity and propagation potential of the pathogen, and a more stereoscopic basis is provided for pathogen traceability and risk assessment.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of vector biological analysis, and particularly relates to a vector biological pathogen analysis method based on metagenomic sequencing. BACKGROUND

[0002] Vector biological pathogen analysis refers to a process of systematically researching and evaluating pathogenic agents that are transmitted through vectors such as mosquitoes, flies, rats, and cockroaches, in order to identify their species, transmission mechanisms, pathogenicity, and potential threats to human health. The core goal of vector biological pathogen analysis is to understand the interaction between pathogenic agents and vectors, and to reveal how pathogenic agents are transmitted through vectors, thereby providing scientific basis for disease prevention, control, and the formulation of public health policies. Vector biological pathogen analysis generally includes the isolation and identification of pathogenic agents, the life cycle of pathogenic agents in vectors, the transmission efficiency of pathogenic agents, and the adaptability of vectors to pathogenic agents. For example, mosquitoes are the main transmission vectors of malaria, dengue fever, and yellow fever, and the Plasmodium, dengue virus, and yellow fever virus carried in mosquitoes will undergo specific developmental stages during transmission, which determines whether the pathogenic agents can successfully infect new hosts. In addition, vector biological pathogen analysis also involves the study of vector population dynamics, ecological environment, and their interaction with human activities, in order to assess the risk and trend of pathogenic agent transmission. Through in-depth analysis of vector biological pathogens, key information can be provided for the formulation of effective prevention and control strategies, such as the selection of appropriate insecticides, the optimization of vector control measures, and the improvement of vaccine development direction. Vector biological pathogen analysis not only helps to cope with current infectious disease threats, but also provides scientific support for coping with future emerging infectious diseases, thereby protecting human health and social stability.

[0003] However, existing technologies are often limited to single sequencing data alignment analysis, and the use of sample collection background and functional association information is insufficient. Moreover, the reliance on static databases limits the identification ability of unknown pathogens. At the same time, the data lacks dynamic feedback mechanism, the adaptability to sample quality difference is poor, the result interpretation is isolated, and the visualization degree is low, which is difficult to directly support on-site prevention and control decision-making. SUMMARY

[0004] The purpose of the present application is to solve the above-mentioned problems, and to provide a vector biological pathogen analysis method based on metagenomic sequencing.

[0005] The technical solution adopted by the present application is as follows: a vector biological pathogen analysis method based on metagenomic sequencing, the method comprising the following steps:

[0006] S1: Collect target vector samples, and record the collection region, climate characteristics, host contact history, and sample preservation conditions at the same time. These information will be used as basic parameters for subsequent data analysis;

[0007] S2: Extract total DNA from the samples using the modified CTAB method. Evaluate the DNA integrity by qubit quantification and agarose gel electrophoresis. The classification results will directly affect the selection of subsequent library construction strategies;

[0008] S3: According to the DNA integrity classification results of S2, adjust the probe hybridization time and PCR cycle number, construct pathogen target enrichment library, and reserve 10% non-enrichment control group for evaluation of enrichment efficiency;

[0009] S4: Use Illumina NovaSeq platform for double-end sequencing. After quality control by FastQC, the raw data is automatically associated with the library construction parameters of S3 to form a quality traceability matrix;

[0010] S5: Integrate the sequencing data of S4, sample annotation information of S1, and pathogen genome data from public databases. Use the self-developed base quality correction algorithm to eliminate batch effects;

[0011] S6: Based on graph neural network, construct pathogen recognition model, convert S5 preprocessed data into species-gene-function three-layer network structure, and realize cross-species correlation identification of unknown pathogens by dynamically adjusting node weights;

[0012] S7: Combine the pathogen sequences identified by S6 for KEGG pathway annotation, and simultaneously call the regional climate data of S1 to construct a transmission risk prediction model, and generate a spatiotemporal transmission heat map;

[0013] S8: According to the prediction results of S7, design specific qPCR verification experiments, and use metabolomics methods to detect characteristic metabolites in samples to verify the activity state of pathogens;

[0014] S9: Integrate the key data nodes of S2-S8, automatically generate dynamic reports containing sample traceability path, methodology parameters, identification results, and risk prediction, and support interactive secondary analysis by adjusting the network weight parameters of S6.

[0015] In a preferred embodiment, in step S1, hierarchical sampling method is used to set up collection points in the target area, at least 3 monitoring traps per square kilometer, CO2 traps are used to collect mosquitoes continuously for 48 hours from 18:00 to 22:00 every day, white cloth drag net method is used to collect ticks, each plot is dragged for a distance of not less than 500 meters. After sample collection, immediately use sterile PBS buffer to rinse 3 times to remove surface contaminants, transfer to 2 mL RNase-free centrifuge tube, add RNAlater preservative solution with a volume ratio of 1:1. Label information needs to include the latitude and longitude of the collection point to six decimal places, record the average temperature, humidity, air pressure and host animal activity frequency of the day. Samples suspected of carrying pathogens are labeled separately and transported to the laboratory within 2 hours through the cold chain, maintaining a 4°C low temperature environment throughout the process.

[0016] In a preferred embodiment, in step S2, 200 μL of sample homogenate is added to 600 μL of modified CTAB extraction buffer, which contains 2% cetyltrimethylammonium bromide, 100 mmol / L Tris-HCl pH 8.0, 20 mmol / L EDTA, 1.4 mol / L NaCl, and 0.2% β-mercaptoethanol. During 65°C water bath heating for 30 minutes, gently invert and mix every 10 minutes. Add an equal volume of chloroform-isoamyl alcohol mixture with a volume ratio of 24:1, centrifuge at 2000 rpm for 10 minutes, take the supernatant and repeat the extraction 2 times. Use 0.8 times the volume of isopropanol to precipitate DNA at 20°C for 2 hours, wash twice with 75% ethanol, dry, and dissolve in 50 μL TE buffer. Use Qubit 4.0 fluorometer to quantitate DNA concentration, which needs to reach 50 ng / μL or more, detect DNA integrity index by Agilent 4200 TapeStation, when DIN value ≥ 8.0, it is determined as an intact sample, 6.0-8.0 is moderately degraded, and < 6.0 is severely degraded.

[0017] In a preferred embodiment, in step S3, the amount of capture probe is adjusted according to the DNA integrity classification results, the intact sample uses the regular probe amount, and the degraded sample is appropriately increased. The probe hybridization system includes hybridization buffer SDS Cot-1 DNA, and the reaction is carried out at a suitable temperature, the intact sample has a shorter reaction time, the degraded sample is appropriately prolonged, and the severely degraded sample uses single-stranded DNA library construction strategy. After hybridization product is captured by magnetic beads, PCR amplification reaction system is carried out, which includes HiFi enzyme mixture and upstream and downstream primers. The amplification program is set to pre-denaturation, annealing and extension steps, and the cycle number is adjusted according to the sample quality. Library quality inspection confirms that the fragment distribution and concentration meet the sequencing requirements.

[0018] In a preferred embodiment, in step S4, Illumina NovaSeq platform is used for double-end sequencing to ensure sufficient data quantity for each sample. After the raw data is converted into fastq format, FastQC is used for quality control to detect base quality value, GC content, adapter contamination, etc. Unqualified data is trimmed to retain high-quality sequences. The processed data is associated with the library construction parameters to form a quality traceability table, ensuring that the sequencing data can be traced back to the previous processing parameters.

[0019] In a preferred embodiment, in step S5, Kraken2 v2.1.2 is used for preliminary species annotation when integrating sequencing data, and the RefSeq bacteria, archaea, and virus joint database is selected. Sample annotation information is converted into a numerical matrix by Python pandas library, and regional climate data is standardized to the 0-1 interval. Host contact history is processed using one-hot encoding. Public database data is downloaded from NCBI GenBank, updated in the third quarter of 2024, containing 120,000 complete pathogen genome sequences. The self-developed base quality correction algorithm DeepCorrect is based on the Transformer architecture, and the training set contains 1000 clinical sample Illumina sequencing data. By comparing PacBio HiFi data, the base error rate correction model is optimized, and the systematic error is reduced by 42%. In the feature fusion stage, the weighted average method is used, with sequencing data weight 0.6, sample annotation information weight 0.3, and public database data weight 0.1. Principal component analysis is used to reduce the feature dimension to 512.

[0020] In a preferred embodiment, in step S6, the multi-modal data pre-processed in S5 is first converted into a "species-gene-function" three-layer heterogeneous graph structure: the species layer takes OTU (operational taxonomic unit) as the node;

[0021] The node features include regional host information annotated in S1 and sequencing depth in S4. The gene layer takes ORF prediction results as nodes, and the features include sequence conservation domain and GC content. The function layer takes KO number as node, and the feature is metabolic pathway correlation strength. The edge weight between layers is constructed by bidirectional mapping: the species-gene layer is based on CRISPR spacer matching degree, the gene-function layer uses BLASTP alignment score, and the species-function layer is indirectly connected by weighting the gene nodes.

[0022] The model training stage adopts a dynamic weight propagation mechanism: the initial weight matrix is initialized by the library enrichment efficiency of S3 and the batch effect correction coefficient of S5. In each iteration, the node weight is adjusted in real time through the attention mechanism: for known pathogens, the direct connection weight between species and genes is strengthened; for potential unknown pathogens, the weight of the cross-species conserved functional module (such as the virulence factor family) in the gene-function layer is automatically increased. In the identification process, the algorithm embeds the k-mer features of the input sequence into the graph structure, aggregates the three-layer neighborhood features through GraphSAGE, and outputs the species classification probability and functional association confidence. When a low-confidence species node is detected, a cross-layer backtracking verification is automatically triggered: if its gene node has a strong connection with a high-confidence functional node (such as an antibiotic resistance gene), the potential host species is inferred through the phylogenetic tree of the functional module, realizing the cross-species association recognition of unknown pathogens.

[0023] wherein the dynamic weight update formula is:

[0024]

[0025] wherein:

[0026] W i (t) represents the weight value of node i in the tth iteration;

[0027] α represents the historical weight retention coefficient (value range 0.3-0.5, dynamically adjusted by the DNA integrity classification results of S2);

[0028] β represents the neighborhood information contribution coefficient (value range 0.5-0.7, positively correlated with the propagation risk prediction confidence of S7);

[0029] N(i) represents the three-order neighborhood node set of node i (including cross-layer neighbors);

[0030] A ij represents the initial adjacency matrix value of nodes i and j (based on the feature fusion results of S5);

[0031] Hi represents the high-dimensional feature vector of node i (concatenating species / gene / function layer features);

[0032] cos(H i ,H j ) represents the cosine similarity of feature vectors (measuring sequence / function similarity);

[0033] F i represents the functional importance score of node i (calculated by KEGG pathway topological centrality);

[0034] D i represents the degree centrality of node i (controls the upper limit of weight contribution of high-connectivity nodes).

[0035] In a preferred embodiment, in step S7, KEGG pathway annotation is performed using the KAAS online tool with e-value threshold 1e-10, prokaryotes as the database for comparison. The functional annotation results are used to construct a pathway interaction network in Cytoscape v3.9.1, with the node size being positively correlated with the gene expression amount, and the edge width representing the co-expression coefficient between pathways. The propagation risk prediction model adopts a spatiotemporal LSTM architecture, and the input variables include the daily average temperature, humidity, host density in S1, and the population flow data of the predicted area. The time series length is set to 14 days, and the spatial resolution is 5km x 5km grid. The model training uses the Adam optimizer with a learning rate of 0.001, a batch size of 32, and 100 iterations. The hyperparameters are adjusted through 5-fold cross-validation. The risk thermodynamic map is drawn using the Matplotlib basemap toolbox, with the color gradient using the RdYlBu color scheme. The risk value is divided into 5 levels, and the areas at levels 4-5 need to be marked as key prevention and control areas.

[0036] In a preferred embodiment, in step S8, qPCR verification is performed using specific primers designed for the conserved gene region of the pathogen, with the product length being controlled. The primers are designed using professional software, and the annealing temperature is set appropriately. The reaction system contains SYBR Green mixture, primers, and template DNA. The amplification program is set to perform pre-denaturation, annealing, and extension steps with appropriate cycle numbers, and the melting curve analysis is performed to ensure the specificity of amplification. Metabolomics detection is performed using a mass spectrometry system, and appropriate chromatographic columns are selected. The mobile phase is set to gradient elution, and the flow rate and column temperature are controlled. The mass spectrometer uses an electrospray ion source, selects the positive ion mode, sets the scanning range and resolution, and ensures the detection sensitivity.

[0037] In a preferred embodiment, in step S9, the report framework is developed based on Shiny, and includes data overview, methodology parameters, result visualization, and risk prediction modules. The data overview page displays the sample collection distribution thermodynamic map and the DNA integrity distribution histogram, supporting data filtering according to collection conditions. The methodology parameter module presents the key steps in the form of a flowchart, and clicking can expand the detailed parameters. The result visualization module integrates the species abundance graph, functional pathway graph, and risk prediction curve, supporting the adjustment of the confidence interval threshold. The risk prediction module provides an interactive weight adjustment slider, which can be dragged to update the identification results and thermodynamic map in real time. The report supports exporting PDF and HTML formats, and retains the original data and interactive functions.

[0038] In summary, by adopting the above technical solutions, the present application has the following advantages:

[0039] 1、In the present application, by constructing a multi-dimensional data fusion framework, the identification of pathogen is realized from single sequence alignment to system biology analysis. Its core advantage lies in the deep coupling of ecological background data and sequencing data in the sample collection stage, breaking the limitations of traditional species identification on database through dynamic weight network model, which can realize the associated identification of unknown pathogens in the case of insufficient sequence homology by means of cross-species conservation of gene function modules. This design not only retains the breadth of metagenomic high-throughput detection, but also strengthens the internal relationship of species-gene-function through three-layer network structure, so that the identification result not only contains the classification information of pathogen, but also synchronously correlates its metabolic activity and transmission potential, providing more stereoscopic basis for pathogen tracing and risk assessment.

[0040] 2、In the present application, from multi-dimensional annotation in the sample collection stage to the final interactive report generation, each step forms an organic feedback through data nodes, for example, the weight parameters of the pathogen identification model can directly affect the accuracy of the transmission risk prediction, and the verification experiment results can also feed back to the model optimization. This dynamic adjustment mechanism makes the analysis process have good adaptability, which can not only deal with samples with different degradation degrees, but also realize the key tracking of specific pathogens by adjusting network weights. In addition, the combination of multi-omics cross-validation and visualization report tool reduces the threshold of complex data interpretation, provides intuitive and operable scientific basis for the disease control department to develop targeted prevention and control strategies, and effectively shortens the transformation cycle from laboratory detection to field application. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 The figure is a schematic diagram of the flow principle of the present application. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical scheme and advantages of the present application more clear and explicit, the present application is further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application, and do not limit the present application.

[0043] Example:

[0044] Reference Figure 1 A vector-borne pathogen analysis method based on metagenomic sequencing, the method comprising the following steps:

[0045] S1: Collecting target vector biological samples, and synchronously recording the collection region, climate characteristics, host contact history and sample preservation conditions, which will be used as basic parameters for subsequent data analysis;

[0046] S2: Extracting total DNA of the sample by using improved CTAB method, evaluating DNA integrity by qubit quantification and agarose gel electrophoresis, and the classification results directly affect the selection of subsequent library construction strategy;

[0047] S3: According to the DNA integrity grading results of S2, adjust the probe hybridization time and the number of PCR cycles, construct a pathogen target enrichment library, and reserve 10% non-enrichment control group for evaluating the enrichment efficiency;

[0048] S4: Use Illumina NovaSeq platform for double-end sequencing, and after the original data are quality controlled by FastQC, automatically associate the library construction parameters of S3 to form a quality traceability matrix;

[0049] S5: Integrate the sequencing data of S4, the sample annotation information of S1 and the pathogen genome data of public database, and eliminate batch effects by using a self-developed base quality correction algorithm;

[0050] S6: Based on the graph neural network, a pathogen recognition model is constructed, the preprocessed data of S5 are converted into a three-layer network structure of species-gene-function, and through dynamic adjustment of node weights, cross-species correlation recognition of unknown pathogens is realized;

[0051] S7: Combine the pathogen sequences recognized by S6 to perform KEGG pathway annotation, and simultaneously call the regional climate data of S1 to construct a transmission risk prediction model, and generate a spatiotemporal transmission heat map;

[0052] S8: According to the prediction results of S7, design specific qPCR verification experiments, and simultaneously use metabolomics methods to detect characteristic metabolites in samples to verify the activity state of pathogens;

[0053] S9: Integrate the key data nodes of S2-S8, automatically generate a dynamic report containing sample traceability path, methodological parameters, recognition results and risk prediction, and support interactive secondary analysis by adjusting the network weight parameters of S6.

[0054] In step S1, hierarchical sampling method is used to set collection points in the target area, at least 3 monitoring traps per square kilometer, CO2 traps are used for mosquito collection, continuous collection for 48 hours from 18:00 to 22:00 every day, white cloth drag net method is used for tick collection, each sample site is dragged for a distance of not less than 500 meters; after sample collection, the samples are immediately washed with sterile PBS buffer for 3 times to remove surface contaminants, and then transferred into 2 mL RNase-free centrifuge tubes, and 1:1 volume ratio of RNAlater preservative is added; the label information needs to include the collection point latitude and longitude accurate to six decimal places, the average temperature, humidity, air pressure and host animal activity frequency of the day.

[0055] In step S2, 200 μL of sample homogenate was added to 600 μL of modified CTAB extraction buffer containing 2% cetyltrimethylammonium bromide, 100 mmol / L Tris-HCl pH 8.0, 20 mmol / L EDTA, 1.4 mol / L NaCl, and 0.2% β-mercaptoethanol; 65°C water bath heating for 30 minutes, gently inverted and mixed every 10 minutes; add an equal volume of chloroform-isoamyl alcohol mixture (24:1, 2,000 rpm centrifuge for 10 minutes, take the supernatant and repeat extraction twice; use 0.8 times the volume of isopropanol-20°C to precipitate DNA for 2 hours, wash twice with 75% ethanol, and then dissolve in 50 μL TE buffer.

[0056] In step S3, the amount of capture probe was adjusted according to the DNA integrity classification results; the regular probe amount was used for intact samples, and the appropriate amount was increased for degraded samples; the probe hybridization system included hybridization buffer SDS Cot-1 DNA; the reaction time was shorter for intact samples and appropriately longer for degraded samples at an appropriate temperature; single-stranded DNA library construction strategy was used for severely degraded samples; after hybridization products were captured by magnetic beads, PCR amplification reaction system included HiFi enzyme mixture and upstream and downstream primers; the amplification program was set to adjust the cycle number according to the sample quality; library quality inspection confirmed that the fragment distribution and concentration met the sequencing requirements.

[0057] In step S4, Illumina NovaSeq platform was used for double-end sequencing to ensure that each sample obtained sufficient data volume; after the original data was converted to fastq format, FastQC was used for quality control to detect base quality value, GC content, and adapter contamination; unqualified data was trimmed to retain high-quality sequences; the processed data was associated with library construction parameters to form a quality traceability table, recording PCR cycle number, hybridization time, and probe concentration, to ensure that the sequencing data could be traced back to the previous processing parameters.

[0058] In step S5, Kraken2 v2.1.2 was used for preliminary species annotation when integrating sequencing data; RefSeq bacterial, archaeal, and viral joint library was selected; sample annotation information was converted to a numerical matrix through Python pandas library, regional climate data was standardized to 0-1 interval, and host contact history was processed using one-hot encoding; public database data was downloaded from NCBI GenBank, updated in the third quarter of 2024, containing 120,000 complete pathogen genome sequences; the self-developed base quality correction algorithm DeepCorrect was based on Transformer architecture, the training set contained 1000 clinical sample Illumina sequencing data, the base error rate was optimized by comparing PacBio HiFi data, and the system error was reduced by 42%.

[0059] In step S6, the multi-modal data pre-processed in S5 is first converted into a "species-gene-function" three-layer heterogeneous graph structure: the species layer takes the operational taxonomic unit as the node;

[0060] The node features include the regional host information annotated in S1 and the sequencing depth in S4; the gene layer takes the ORF prediction result as the node, and the features include the sequence conserved domain and GC content; the function layer takes the KO number as the node, and the feature is the metabolic pathway correlation strength; the edge weight between layers is constructed through bidirectional mapping: the species-gene layer is based on the CRISPR spacer matching degree, the gene-function layer uses the BLASTP alignment score, and the species-function layer is indirectly connected by weighting the gene nodes;

[0061] The model training stage adopts a dynamic weight propagation mechanism: the initial weight matrix is initialized by the library enrichment efficiency in S3 and the batch effect correction coefficient in S5, and in each iteration, the node weight is adjusted in real time through the attention mechanism - for known pathogens, the direct connection weight between species and genes is strengthened; for potential unknown pathogens, the weight of the conserved function module across species in the gene-function layer is automatically increased; in the identification process, the algorithm embeds the k-mer features of the input sequence into the graph structure, aggregates the three-layer neighborhood features through GraphSAGE, and outputs the species classification probability and the function correlation confidence; when a low-confidence species node is detected, automatic cross-layer backtracking verification is triggered: if its gene node has a strong connection with a high-confidence function node, the potential host species is inferred through the phylogenetic tree of the function module, realizing the cross-species association identification of unknown pathogens;

[0062] wherein the dynamic weight update formula is:

[0063]

[0064] wherein:

[0065] W i (t) represents the weight value of node i in the tth iteration;

[0066] α represents the historical weight retention coefficient;

[0067] β represents the neighborhood information contribution coefficient;

[0068] N(i) represents the three-order neighborhood node set of node i;

[0069] A ij represents the initial adjacency matrix value of nodes i and j;

[0070] Hi represents the high-dimensional feature vector of node i;

[0071] cos(H i ,H j ) represents the feature vector cosine similarity;

[0072] F i denotes the functional importance score of node i;

[0073] D i denotes the degree centrality of node i.

[0074] In step S7, the KEGG pathway annotation parameter setting e-value threshold 1e-10 is performed using the KAAS online tool, the database selection prokaryotes is performed, the functional annotation result is constructed into a pathway interaction network by Cytoscape v3.9.1, the node size is positively correlated with the gene expression amount, the edge width represents the co-expression coefficient between pathways, the propagation risk prediction model adopts a space-time LSTM architecture, the input variables include the daily average temperature, humidity, host density of S1 and the population flow data of the predicted area, the time series length is set to 14 days, the spatial resolution is 5km*5km grid, the model training adopts the Adam optimizer, the learning rate is 0.001, the batch size is 32, and the iteration is 100 rounds, the hyperparameters are adjusted through 5-fold cross-validation, and the risk thermodynamic map is drawn using the Matplotlib basemap toolbox, the color gradient adopts the RdYlBu color scheme, and the risk value is divided into 5 levels, and the level 4-5 area needs to be marked as a key prevention and control area.

[0075] In step S8, the qPCR verification experiment is designed, specific primers are designed for the conserved gene region of the pathogen, the product length is controlled, the primer design uses professional software to set the appropriate annealing temperature, the reaction system contains SYBRGreen mixed solution, primer and template DNA, the amplification program is set to pre-denaturation, annealing and extension steps, appropriate cycle number is performed, and the solubility curve analysis is performed to ensure the specificity of amplification, the metabolomics detection adopts a mass spectrometry system, a suitable chromatographic column is selected, the gradient elution program is set, the flow rate and column temperature are controlled, the mass spectrometry adopts an electrospray ion source, the positive ion mode is selected, the scanning range and resolution are set to ensure the detection sensitivity.

[0076] In step S9, the report framework is developed based on Shiny and contains data overview, methodology parameters, result visualization and risk prediction modules; the data overview page shows the sample collection distribution thermodynamic map and the DNA integrity distribution histogram to support data filtering according to collection conditions; the methodology parameter module presents the key steps in the form of a flowchart, and the detailed parameters can be expanded by clicking; the result visualization module integrates the species abundance graph, the functional pathway graph and the risk prediction curve to support the adjustment of the confidence interval threshold; the risk prediction module provides an interactive weight adjustment slider, which can be dragged to update the identification result and the thermodynamic map in real time; the report supports exporting PDF and HTML formats, retains the original data and interactive functions.

[0077] It can be known from the above that:

[0078] In the present application, by constructing a multi-dimensional data fusion framework, the identification of pathogens is realized from single sequence alignment to system biology analysis. The core advantage is to deeply couple the ecological background data and sequencing data in the sample collection stage, break the limitations of traditional species identification on database through dynamic weight network model, and realize the associated identification of unknown pathogens with the help of cross-species conservation of gene function modules in the case of insufficient sequence homology. This design not only retains the breadth of metagenomic high-throughput detection, but also strengthens the internal relationship of species-genes-function through three-layer network structure, so that the identification result not only contains the classification information of pathogens, but also synchronously correlates the metabolic activity and transmission potential, providing more stereoscopic basis for pathogen tracing and risk assessment.

[0079] In the present application, from multi-dimensional annotation in the sample collection stage to the final interactive report generation, each step forms an organic feedback through data nodes, for example, the weight parameters of the pathogen identification model can directly affect the accuracy of the transmission risk prediction, and the verification experiment results can in turn optimize the model. This dynamic adjustment mechanism makes the analysis process have good adaptability, which can not only cope with samples with different degrees of degradation, but also realize the key tracking of specific pathogens by adjusting the network weight. In addition, the combination of multi-omics cross-validation and visual report tools reduces the threshold of interpreting complex data, provides intuitive and operable scientific basis for the disease control department to formulate targeted prevention and control strategies, and effectively shortens the conversion cycle from laboratory detection to field application.

[0080] It should be noted that, in this text, relationship terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "includes" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment. Without more limitations, the element defined by the statement "includes a" does not exclude the presence of additional identical elements in the process, method, article or equipment including the element.

[0081] The above description enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for analysis of pathogen-carrying vectors based on metagenomic sequencing, characterized by: The method comprises the following steps: S1: Collect target vector samples, and record the collection region, climate characteristics, host contact history, and sample preservation conditions synchronously, which will be used as basic parameters for subsequent data analysis; S2: Extract total DNA from the samples using a modified CTAB method, evaluate the DNA integrity through qubit quantification and agarose gel electrophoresis, and the classification results directly affect the selection of subsequent library construction strategies; S3: According to the DNA integrity classification results of S2, adjust the probe hybridization time and the number of PCR cycles, construct a pathogen target enrichment library, and reserve 10% non-enrichment control group for evaluating the enrichment efficiency; S4: Perform double-end sequencing using the Illumina NovaSeq platform, and after the raw data are subjected to FastQC quality control, automatically associate the library construction parameters of S3 to form a quality traceability matrix; S5: Integrate the sequencing data of S4, the sample annotation information of S1, and the pathogen genome data in the public database, eliminate batch effects through a base quality correction algorithm developed independently; S6: Construct a pathogen recognition model based on a graph neural network, convert the preprocessed data of S5 into a three-layer network structure of species-gene-function, and realize cross-species correlation recognition of unknown pathogens by dynamically adjusting node weights; S7: Perform KEGG pathway annotation on the pathogen sequences recognized in S6, simultaneously call the regional climate data of S1 to construct a transmission risk prediction model, and generate a spatiotemporal transmission heat map; S8: According to the prediction results of S7, design specific qPCR verification experiments, and detect characteristic metabolites in the samples using metabolomics methods to verify the activity state of the pathogens; S9: Integrate the key data nodes of S2-S8, automatically generate a dynamic report containing sample traceability path, methodological parameters, recognition results, and risk prediction, and support interactive secondary analysis by adjusting the network weight parameters of S6.

2. The method of vector-borne pathogen analysis based on metagenomic sequencing as claimed in claim 1, wherein: In the step S1, hierarchical sampling method is used to set collection points in the target area, at least 3 monitoring traps per square kilometer, CO2 traps are used for mosquito collection, continuous collection for 48 hours from 18:00 to 22:00 every day, white cloth drag net method is used for tick collection, each sample site is dragged for a distance of not less than 500 meters; after sample collection, the samples are immediately washed with sterile PBS buffer for 3 times to remove surface contaminants, then transferred into 2mL RNase-free centrifuge tubes, and RNAlater preservative solution is added in a volume ratio of 1:1; the label information needs to include the collection point latitude and longitude accurate to six decimal places, the average temperature, humidity, air pressure and host animal activity frequency of the day.

3. The method of claim 1, wherein the method is a method for analyzing a pathogen carried by a vector organism based on metagenomic sequencing, characterized by: In step S2, 200 μL of sample homogenate is added to 600 μL of modified CTAB extraction buffer containing 2% cetyltrimethylammonium bromide, 100 mmol / L Tris-HCl pH 8.0, 20 mmol / L EDTA, 1.4 mol / L NaCl, and 0.2% β-mercaptoethanol; 65°C water bath heating for 30 minutes, gently invert and mix every 10 minutes; add an equal volume of chloroform-isoamyl alcohol mixture (24:1, 2,000 rpm centrifuge for 10 minutes, take the supernatant and repeat extraction twice; use 0.8 times the volume of isopropanol-20°C to precipitate DNA for 2 hours, wash twice with 75% ethanol, and then dissolve in 50 μL TE buffer.

4. The method of vector-borne pathogen analysis based on metagenomic sequencing as claimed in claim 1, wherein: In step S3, the amount of capture probe is adjusted according to the DNA integrity classification results; the probe hybridization system includes hybridization buffer SDS Cot-1 DNA; the reaction temperature is appropriate; the reaction time is shorter for intact samples and longer for degraded samples; single-stranded DNA library construction strategy is used for severely degraded samples; the hybridization products are captured by magnetic beads, and then PCR amplification reaction system is performed, which includes HiFi enzyme mixture and upstream and downstream primers; the amplification program is set, and the cycle number is adjusted according to the sample quality; library quality inspection confirms that the fragment distribution and concentration meet the sequencing requirements.

5. The method of analyzing pathogen-carrying vectors based on metagenomic sequencing according to claim 1, wherein: In step S4, Illumina NovaSeq platform is used for double-end sequencing to ensure that each sample obtains sufficient data volume; after the original data is converted into fastq format, FastQC is used for quality control to detect base quality value, GC content, and adapter contamination; unqualified data is trimmed to retain high-quality sequences; the processed data is associated with the library construction parameters to form a quality traceability table, ensuring that the sequencing data can be traced back to the previous processing parameters.

6. The method of vector-borne pathogen analysis based on metagenomic sequencing as claimed in claim 1, wherein: In step S5, Kraken2 v2.1.2 is used for preliminary species annotation when integrating sequencing data, and the database RefSeq bacteria archaea virus joint library is selected; sample annotation information is converted into a numerical matrix through Python pandas library, regional climate data is standardized to 0-1 interval, and host contact history is processed by one-hot encoding; public database data is downloaded from NCBI GenBank, the third quarter of 2024 update version, containing 120,000 complete pathogen genome sequences; the self-developed base quality correction algorithm DeepCorrect is based on Transformer architecture, the training set contains 1000 clinical sample Illumina sequencing data, the base error rate correction model is optimized by comparing PacBio HiFi data, and the systematic error is reduced by 42%.

7. The method of analyzing pathogen-carrying vectors based on metagenomic sequencing according to claim 1, wherein: In step S6, first, the multi-modal data after preprocessing in S5 is converted into a "species-gene-function" three-layer heterogeneous graph structure: the species layer takes the operational classification unit as the node; The node features include: S1 labeled regional host information and S4 sequencing depth; the gene layer takes the ORF prediction result as the node, and the features cover the sequence conserved domain and GC content; the function layer takes the KO number as the node, and the feature is the metabolic pathway correlation strength; the interlayer edge weight is constructed through bidirectional mapping: the species-gene layer is based on the CRISPR spacer sequence matching degree, the gene-function layer adopts the BLASTP alignment score, and the species-function layer is indirectly connected by weighting the gene nodes; The model training stage adopts a dynamic weight propagation mechanism: the initial weight matrix is initialized by the library enrichment efficiency of S3 and the batch effect correction coefficient of S5, and in each iteration, the node weight is adjusted in real time through the attention mechanism - for known pathogens, the direct connection weight of the species-gene is strengthened; for potential unknown pathogens, the weight of the conserved functional module across species in the gene-function layer is automatically increased; in the identification process, the algorithm embeds the k-mer features of the input sequence into the graph structure, aggregates the three-layer neighborhood features through GraphSAGE, and outputs the species classification probability and functional association confidence; when a low-confidence species node is detected, automatic cross-layer backtracking verification is triggered: if its gene node has a strong connection with a high-confidence function node, the potential host species is inferred through the phylogenetic tree of the functional module, realizing the cross-species association identification of unknown pathogens; Wherein: the dynamic weight update formula is: In the formula: W i (t) denotes the weight value of node i in the tth round of iteration; Alpha represents the historical weight retention coefficient; Beta represents the neighborhood information contribution coefficient; N(i) represents the three-order neighborhood node set of node i; A ij represents the initial adjacency matrix value for nodes i and j; Hi represents the high-dimensional feature vector of node i; cos(H i ,H j ) represents the eigenvector cosine similarity; F i denotes the functional importance score of node i; D i denotes the degree centrality of node i.

8. The method of analyzing pathogen-carrying vectors based on metagenomic sequencing according to claim 1, wherein: In step S7, the KEGG pathway annotation parameter setting e-value threshold is 1e-10, the comparison database is selected as prokaryotes using the KAAS online tool; the functional annotation result is used to construct the pathway interaction network by Cytoscapev3.9.1, the node size is positively correlated with the gene expression amount, and the edge width represents the co-expression coefficient between pathways; the propagation risk prediction model adopts a spatiotemporal LSTM architecture, the input variables include the daily average temperature, humidity, host density of S1 and the population flow data of the predicted area, the time series length is set to 14 days, and the spatial resolution is 5km*5km grid; the model training adopts the Adam optimizer, the learning rate is 0.001, the batch size is 32, and the iteration is 100 rounds; the risk heat map is drawn using the Matplotlib basemap toolbox, the color gradient adopts the RdYlBu color scheme, and the risk value is divided into five levels, and the level 4-5 area needs to be marked as the key prevention and control area.

9. The method of analyzing pathogen-carrying vectors based on metagenomic sequencing according to claim 1, wherein: In step S8, the qPCR verification experiment is designed, the specific primers are designed for the conserved gene region of the pathogen, the product length is controlled, the primer design uses professional software to set the appropriate annealing temperature; the reaction system contains SYBRGreen mixture, primers and template DNA. The amplification procedure is set to perform a proper number of cycles of denaturation, annealing and extension steps to ensure amplification specificity; the metabolomics detection uses a mass spectrometry system, selects a suitable chromatographic column, sets a gradient elution program, controls the flow rate and column temperature; the mass spectrometry uses an electrospray ion source, selects a positive ion mode, sets a scanning range and resolution to ensure detection sensitivity.

10. The method of analyzing pathogen-carrying vectors based on metagenomic sequencing according to claim 1, wherein: In the step S9, the report framework is developed based on Shiny to include data overview, methodology parameter, result visualization, and risk prediction modules; The data overview page displays a sample collection distribution heat map and a DNA integrity distribution histogram to support data filtering according to collection conditions; the methodology parameter module presents key steps in the form of a flowchart, and detailed parameters can be expanded by clicking; the result visualization module integrates species abundance chart, pathway map and risk prediction curve to support confidence interval threshold adjustment; The risk prediction module provides an interactive weight adjustment slider to update the identification result and heat map in real time; the report supports PDF and HTML format export to retain original data and interactive functions.

Images