Method and system for analyzing breeding strategy of peanut against bacterial wilt based on growth cycle

Abstract

The application relates to the technical field of data processing, and discloses a peanut anti-rhizoctonia solani breeding strategy analysis method and system based on a growth cycle. The method comprises the following steps: collecting and analyzing peanut growth period characteristic data to obtain a growth cycle characteristic database; comparing phenotypes of healthy plants and diseased plants to establish an anti-rhizoctonia solani early warning model; performing real-time monitoring according to the early warning model to obtain a risk early warning result; tracking the growth period based on the early warning result to establish an anti-rhizoctonia solani resistance scoring system; obtaining strain stability parameters through environmental response analysis; and finally determining excellent strains and forming an anti-rhizoctonia solani breeding scheme. Through the establishment of a data monitoring and analysis system for the whole growth period, dynamic evaluation and early warning of the anti-rhizoctonia solani performance of plants are realized, and the accuracy and efficiency of the selection and breeding of anti-rhizoctonia solani varieties are improved.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to a method and system for analyzing peanut bacterial wilt resistance breeding strategies based on the growth cycle. Background Technology

[0002] Peanuts are an important oilseed and economic crop. Cultivated peanuts are strictly self-pollinating, resulting in a rather narrow genetic base. Existing germplasm resources and traditional breeding techniques are unlikely to further improve breeding efficiency. Bacterial wilt, a bacterial disease, is currently a major disease affecting peanuts, primarily concentrated in the central and southern regions, and showing a gradual northward spread.

[0003] Existing peanut bacterial wilt resistance breeding methods mainly rely on field observation and experience-based judgment, lacking systematic analysis and utilization of growth cycle data. In traditional breeding processes, the evaluation of bacterial wilt resistance is often limited to phenotypic observation at a specific stage, failing to comprehensively reflect the plant's resistance throughout its entire growth period. Furthermore, environmental factors have a significant impact on bacterial wilt resistance evaluation, leading to insufficient stability and reliability of screening results. In addition, existing technologies suffer from problems such as single evaluation indicators, insufficient data analysis, and a lack of early warning mechanisms, all of which hinder the improvement of bacterial wilt resistance breeding efficiency. Summary of the Invention

[0004] This application provides a method and system for analyzing peanut bacterial wilt resistance breeding strategies based on the growth cycle. By establishing a data monitoring and analysis system for the entire growth period, it enables dynamic evaluation and early warning of plant resistance to bacterial wilt, thereby improving the accuracy and efficiency of bacterial wilt resistance breeding.

[0005] Firstly, this application provides a method for analyzing peanut bacterial wilt resistance breeding strategies based on the growth cycle. This method includes: continuously collecting morphological characteristics, physiological indicators, and disease symptoms of peanuts during their growth period using a phenotypic monitoring system; dividing the data into key growth stages through time series segmentation to construct a time-series feature vector, thus obtaining a growth cycle feature library; comparing and analyzing the phenotypic trajectories of healthy and diseased plants based on the growth cycle feature library; establishing early warning indicators through differential feature extraction to obtain a bacterial wilt resistance early warning model; and, based on the bacterial wilt resistance early warning model, analyzing the growth patterns of breeding materials. Real-time monitoring and analysis of phenotypic data are performed to identify potentially susceptible individuals through early warning thresholds, resulting in risk warnings. Based on these warnings, plants with different resistance levels are tracked throughout their growth period. A resistance scoring system is established by comparing phenotypic characteristics at key stages, yielding a dynamic resistance evaluation scheme. According to this scheme, multi-point experimental data are collected for environmental response analysis. Adaptability indicators are established through plant-environment interaction assessment, resulting in strain stability parameters. Based on these stability parameters, breeding materials are comprehensively evaluated and graded. Superior strains are identified through a grading and screening system, resulting in a growth cycle-based bacterial wilt resistance breeding scheme.

[0006] Secondly, this application provides a peanut bacterial wilt resistance breeding strategy analysis system based on the growth cycle, the peanut bacterial wilt resistance breeding strategy analysis system based on the growth cycle includes:

[0007] The data acquisition module is used to continuously collect morphological characteristics, physiological indicators and disease symptoms of peanuts during their growth period through a phenotypic monitoring system. The data is then divided into key growth periods through time series segmentation to construct a time series feature vector and obtain a growth cycle feature library.

[0008] The comparison module is used to compare and analyze the phenotypic trajectories of healthy plants and diseased plants based on the growth cycle feature library, and to establish early warning indicators through differential feature extraction to obtain a bacterial wilt resistance early warning model.

[0009] The monitoring module is used to monitor and analyze the growth phenotypic data of the breeding materials in real time based on the bacterial wilt resistance early warning model, identify potentially susceptible individuals by judging the early warning threshold, and obtain risk early warning results.

[0010] The tracking module is used to track plants with different resistance levels throughout their growth period based on the risk warning results, establish a resistance scoring system by comparing phenotypic characteristics at key periods, and obtain a dynamic resistance evaluation scheme.

[0011] The analysis module is used to collect multi-point test data for environmental response analysis according to the dynamic resistance evaluation scheme, establish adaptability indicators through plant-environment interaction assessment, and obtain strain stability parameters.

[0012] The grading module is used to comprehensively evaluate and grade breeding materials based on the stability parameters of the varieties. A grading screening system is used to identify superior varieties, resulting in a bacterial wilt resistance breeding program based on the growth cycle. In the technical solution provided in this application, a complete growth cycle feature library is constructed by continuously collecting and segmenting the morphological characteristics, physiological indicators, and disease symptoms of peanuts during their growth period, enabling comprehensive monitoring and recording of the plant's growth and development process. By comparing and analyzing the phenotypic trajectories of healthy and diseased plants, an early warning indicator system is established, providing data support for timely detection of potentially susceptible individuals. A deep learning algorithm is introduced into the bacterial wilt resistance early warning model, utilizing the time-series analysis capabilities of recurrent neural networks to monitor and analyze growth phenotypic data in real time, improving the accuracy of early warning. A dynamic resistance evaluation scheme is established by tracking the entire growth period of plants with different resistance levels and comparing phenotypic characteristics at key stages, making the bacterial wilt resistance evaluation more objective and comprehensive. Through environmental response analysis of multi-point experimental data and plant-environment interaction assessment, an adaptability indicator system is constructed, enhancing the stability of variety selection. Finally, a tiered screening system was used to comprehensively evaluate the breeding materials, achieving precise selection of superior lines. This method applies artificial intelligence algorithms to the field of peanut bacterial wilt resistance breeding. Through long short-term memory networks, feature extraction and time-series analysis of plant growth data were performed, fully mining the bacterial wilt resistance information contained in the growth cycle data, providing a data-driven scientific decision-making basis for the breeding of bacterial wilt-resistant varieties. Simultaneously, this method established a complete data processing and analysis workflow, realizing intelligent processing from data collection, feature extraction, early warning analysis to line selection, greatly improving the efficiency and accuracy of bacterial wilt resistance breeding work. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of an embodiment of the peanut bacterial wilt resistance breeding strategy analysis method based on the growth cycle in this application;

[0015] Figure 2 This is a schematic diagram of the hybridization and mating process in the embodiments of this application;

[0016] Figure 3 This is a flowchart illustrating the process of dividing data into key reproductive periods through time series segmentation in this embodiment of the application, constructing time series feature vectors, and obtaining a growth cycle feature library.

[0017] Figure 4 This is a schematic diagram of an embodiment of the peanut bacterial wilt resistance breeding strategy analysis system based on the growth cycle in this application. Detailed Implementation

[0018] This application provides a method and system for analyzing peanut bacterial wilt resistance breeding strategies based on the growth cycle. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the peanut bacterial wilt resistance breeding strategy analysis method based on the growth cycle in this application includes:

[0020] Step S101: Continuously collect morphological characteristics, physiological indicators and disease symptoms of peanuts during the growth period through a phenotypic monitoring system. Divide the data into key growth periods through time series segmentation, construct time series feature vectors, and obtain a growth cycle feature library.

[0021] Step S102: Based on the growth cycle feature library, compare and analyze the phenotypic trajectories of healthy plants and diseased plants, and establish early warning indicators through differential feature extraction to obtain a bacterial wilt resistance early warning model.

[0022] Step S103: Based on the bacterial wilt resistance early warning model, the growth phenotypic data of the breeding materials are monitored and analyzed in real time. Potentially susceptible individuals are identified by the early warning threshold, and the risk warning results are obtained.

[0023] Step S104: Based on the risk warning results, track the plants with different resistance levels throughout their growth period, establish a resistance scoring system by comparing phenotypic characteristics at key stages, and obtain a dynamic resistance evaluation scheme.

[0024] Step S105: According to the dynamic resistance evaluation scheme, collect multi-point test data for environmental response analysis, establish adaptability indicators through plant-environment interaction assessment, and obtain strain stability parameters.

[0025] Step S106: Based on the stability parameters of the strains, the breeding materials are comprehensively evaluated and graded. The superior strains are determined through the grading and screening system to obtain a bacterial wilt resistance breeding program based on the growth cycle.

[0026] It is understood that the executing entity of this application can be a peanut bacterial wilt resistance breeding strategy analysis system based on the growth cycle, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiment uses a server as an example for illustration.

[0027] Specifically, such as Figure 2 The diagram shown illustrates the hybridization process in this embodiment of the application. The female parent is a mutagenic material with strong overall resistance to stress and bacterial wilt, while the male parent is a high-yielding, high-quality variety. F1 generation individuals are obtained through hybridization. After collecting and analyzing growth data from the F1 generation individuals, self-pollination is performed to obtain the F2 generation. The F2 generation is planted in a bacterial wilt nursery for resistance assessment, and individual plants with excellent performance are selected for the F3 generation in southern breeding. The F3 generation materials undergo line comparison trials, variety comparison trials, and then multi-location trials to evaluate variety stability. Finally, superior varieties enter regional trials, and after comprehensive evaluation, stable bacterial wilt-resistant superior varieties are selected. In the breeding material selection stage, peanut varieties with strong overall resistance to stress and bacterial wilt are selected as the female parent, and high-yielding, high-quality peanut varieties are selected as the male parent for hybridization. After obtaining the F1 generation hybrid individuals, growth data is collected for the F1 generation individuals. Data collection for peanut growth was conducted across three key indicator dimensions: morphological characteristics (plant height, number of branches, and leaf area), physiological indicators (photosynthetic intensity, respiration intensity, and chlorophyll content), and disease symptoms (lesion area, disease progression, and incidence rate). Data collection spanned the seedling, vegetative growth, and reproductive growth stages. The raw data was segmented into time series according to different growth stages, and a feature matrix was constructed for each stage. Rows in the matrix represented individual plants, and columns represented various monitoring indicators. The feature matrix underwent dimensionality reduction, and key features were extracted to form temporal feature vectors. The elements in these vectors were arranged chronologically, reflecting the plant's growth, development, and resistance to bacterial wilt throughout the entire growth cycle. These temporal feature vectors were then compiled into a growth cycle feature database.

[0028] Based on the growth cycle characteristic database of F1 generation individuals, plants were divided into a healthy group and a diseased group according to their health status. A time-series comparative analysis was performed on the morphological characteristics, physiological indicators, and disease symptoms of the two groups, calculating the characteristic differences between the two groups at each time point. The difference values ​​were calculated using standardized numerical differences to eliminate the influence of different dimensions. The difference values ​​were arranged according to time series to construct a phenotypic trajectory difference matrix, where each row represents a time point and each column represents a characteristic indicator. Statistical analysis was performed based on the difference matrix to calculate the significance of differences in various characteristics at different growth stages, selecting key indicator combinations for early warning and establishing a bacterial wilt resistance early warning model.

[0029] During the F2 generation breeding process, the population obtained from self-pollination of the F1 generation was planted in a bacterial wilt nursery for resistance identification. Simultaneously, the F2 generation population was monitored in real time based on a bacterial wilt resistance early warning model. The growth phenotypic data of each plant was input into the early warning model, and risk assessment was conducted using a set early warning threshold. The early warning threshold was set based on statistical differences in characteristics between healthy and diseased plants in the early stages, selecting the threshold with the highest statistical significance. When a plant's phenotypic characteristics exceeded the early warning threshold, it was marked as a potentially susceptible individual. This method provided risk warnings for the F2 generation population, promptly identifying individuals with poor resistance performance. For plants with different resistance levels selected from the F2 generation, targeted breeding of the F3 generation continued. During the F3 generation expansion in southern China, plants with different resistance levels were tracked throughout their entire growth period, recording phenotypic characteristic data at each key growth stage. A phenotypic characteristic scoring standard was established, quantifying different characteristic indicators into scores to form a resistance scoring system. The scoring standard was based on the numerical range of characteristic values, transforming continuous data into discrete scores. This method quantifies the resistance level of plants, resulting in a dynamic resistance evaluation scheme.

[0030] Superior strains were introduced into the strain comparison trial phase, planted at multiple experimental sites, and environmental factor data such as soil physicochemical properties and climatic conditions were collected at each site. Correlation analysis was performed between environmental factor data and plant phenotypic data to establish plant-environment interactions. Genotype effects, environmental effects, and interaction effects were separated using variance decomposition to assess the environmental adaptability of the strains and calculate stability parameters. Materials exhibiting excellent stability parameters were selected for the regional trial phase. A comprehensive evaluation system was established, including stability indicators, bacterial wilt resistance indicators, and yield indicators, and the experimental materials were scored across multiple dimensions. The scoring results were graded, and a grading and screening standard was established. Based on the grading standard, strains with excellent overall performance were selected, and a bacterial wilt resistance breeding program was developed.

[0031] For example, in the morphological characteristic collection of F1 generation hybrid materials, plant height was measured on individual plants, and the initial data included continuous measurements at 30 time points during the growth period. These time points were segmented according to growth stages, with 10 measurement points each for the seedling stage, vegetative growth stage, and reproductive growth stage. The mean and coefficient of variation were calculated for the data of each stage to form period characteristic values. Simultaneously, physiological indicators such as photosynthetic intensity and chlorophyll content, as well as disease indicators such as lesion size, were recorded. These indicators were arranged chronologically to construct a time-series data matrix containing multiple feature dimensions. The data matrix was then dimensionality-reduced, and key features were extracted to form feature vectors. In subsequent early warning analysis, the similarity between these feature vectors and the standard feature vectors in the early warning model was calculated to determine the plant's health status. Based on this data analysis method, a batch of plants with excellent resistance were successfully screened in the F2 generation population, and this was further validated in the F3 generation. Through analysis of multi-point experimental data, the stability of these excellent plants was evaluated, and a stable and resistant superior strain to bacterial wilt was bred.

[0032] In this embodiment, a complete growth cycle feature library was constructed by continuously collecting and segmenting the morphological characteristics, physiological indicators, and disease symptoms of peanuts during their growth period, enabling comprehensive monitoring and recording of the plant's growth and development process. By comparing and analyzing the phenotypic trajectories of healthy and diseased plants, an early warning indicator system was established, providing data support for timely detection of potentially susceptible individuals. Deep learning algorithms were introduced into the bacterial wilt resistance early warning model, utilizing the time-series analysis capabilities of recurrent neural networks to monitor and analyze growth phenotypic data in real time, improving the accuracy of early warning. A dynamic resistance evaluation scheme was established by tracking the entire growth period of plants with different resistance levels and comparing phenotypic characteristics at key stages, making the evaluation of bacterial wilt resistance more objective and comprehensive. An adaptability indicator system was constructed through environmental response analysis of multi-point experimental data and plant-environment interaction assessment, enhancing the stability of strain selection. Finally, a tiered screening system was used to comprehensively evaluate breeding materials, achieving precise selection of superior strains. This method applies artificial intelligence algorithms to the field of peanut bacterial wilt resistance breeding. By using long short-term memory networks to extract features and perform time-series analysis on plant growth data, it fully mines the bacterial wilt resistance information contained in the growth cycle data, providing a data-driven scientific decision-making basis for the breeding of bacterial wilt-resistant varieties. Simultaneously, this method establishes a complete data processing and analysis workflow, realizing intelligent processing from data collection, feature extraction, early warning analysis to strain selection, greatly improving the efficiency and accuracy of bacterial wilt resistance breeding work.

[0033] In one specific embodiment, the process of performing step S101 may specifically include the following steps:

[0034] (1) Periodic segmentation of morphological characteristics were collected, and the morphological characteristics were divided into seedling stage, vegetative growth stage and reproductive growth stage. The plant height, number of branches and leaf area of ​​each growth stage were quantitatively measured.

[0035] (2) Extract time-series variation data of photosynthetic intensity, respiration intensity and chlorophyll content from physiological indicators, and perform data standardization and outlier removal;

[0036] (3) Rate and score the disease symptoms and establish a symptom index dataset that includes lesion size, disease index and incidence rate;

[0037] (4) Integrate the segmented data of morphological features, the data of temporal changes and the data of symptom indicators to construct a multidimensional dataset that reflects the characteristics of different reproductive stages;

[0038] (5) Extract features from the multidimensional dataset through data dimensionality reduction processing and establish a feature vector containing key reproductive period indicators;

[0039] (6) Sort and organize the feature vectors according to the time series, construct a data association network covering the entire reproductive period, and obtain the growth cycle feature library.

[0040] Specifically, such as Figure 3 The diagram shows the process of dividing data into key growth periods through time series segmentation in this embodiment of the application, constructing time series feature vectors, and obtaining a growth cycle feature library. Morphological feature data, physiological indicator data, and disease symptom data are collected during specified time periods. After standardization and outlier removal, the collected data are aligned with the time axis to form a multidimensional dataset. Then, principal component analysis is used for feature dimensionality reduction. Finally, the data are arranged in time series to construct a data association network and form a complete growth cycle feature library.

[0041] After obtaining F1 generation individuals by hybridizing mutagenic materials as the female parent and superior varieties as the male parent, the entire growth period was divided into three stages: seedling stage (15-30 days after sowing), vegetative growth stage (31-60 days after sowing), and reproductive growth stage (61-120 days after sowing). During each growth stage, plant height was measured using calipers, and plant height data for each plant was collected twice weekly using a plant height measuring instrument. The number of branches was counted using the leaf counting method, recording the number of branches on the main stem and lateral branches. Leaf area index was measured using a leaf area meter, and functional leaves of each plant were selected for measurement. An independent data table was created for each growth stage, containing three dimensions: plant number, measurement time, and measurement value.

[0042] Physiological indicators collected focused on changes in photosynthetic capacity, respiratory metabolism, and chlorophyll content. Photosynthetic intensity was measured using a photosynthesis meter between 9:00 AM and 11:00 AM, recording the net photosynthetic rate. Respiratory intensity was obtained by measuring oxygen consumption rate using the Clark oxygen electrode method. Chlorophyll content was measured in situ using a chlorophyll meter. The collected raw data underwent standardization, employing z-score standardization to unify data of different dimensions to the same scale. Outlier detection was performed on the standardized data, using box plots to identify outliers. Data points exceeding 1.5 times the interquartile range were marked as outliers and removed. Disease symptom rating was based on a 0-9 scale, with 0 indicating no lesions and 9 indicating complete plant death. Lesion size was determined by measuring the length and width of lesions with a ruler, and the lesion area was calculated. The disease index was calculated based on the severity of the disease: Disease Index = (∑(Number of diseased plants at each level × Corresponding level) / (Total number of plants surveyed × Highest level)) × 100. The incidence rate was obtained by calculating the ratio of diseased plants to the total number of plants. These data were stored in a symptom index dataset, which included fields such as time, plant number, symptom level, and measurement value.

[0043] The data integration phase correlates segmented morphological feature data, time-series physiological indicator data, and disease symptom data. A unified data format is established, with each record including a timestamp, plant identifier, and values ​​for various indicators. Data alignment methods are used to map data collected at different time points onto a unified timeline, forming a complete multidimensional dataset. Each row in the multidimensional dataset represents an observation point, and columns contain all collected feature indicators. Feature extraction employs principal component analysis (PCA) to reduce high-dimensional data to a lower dimension while retaining key information. The feature covariance matrix is ​​calculated, eigenvalues ​​and eigenvectors are solved, and principal components with higher contribution rates are selected as key features. Principal components are selected based on a cumulative contribution rate of 85% or higher. The selected principal components are then recombine with their corresponding eigenvectors to obtain the dimensionality-reduced feature vectors.

[0044] By arranging feature vectors in chronological order, a data association network reflecting the entire process of plant growth and development is constructed. Nodes in the network structure represent feature states at different stages, and edges represent the temporal relationships between features. By calculating the correlation coefficients between features, feature connection weights are established, forming a complete growth cycle feature database.

[0045] For example, F1 generation populations were obtained by crossing a bacterial wilt-resistant maternal parent variety A with a high-yielding paternal parent variety B, and data collection began at the seedling stage. The plant height data for each seedling included six measurement points, and the obtained plant height data were standardized using z-scores. Simultaneously collected photosynthetic intensity data showed diurnal variation patterns, and short-term fluctuations were eliminated using moving averages to obtain trend data. Disease symptom ratings showed that some plants developed lesions during the vegetative growth period, and the size and expansion rate of the lesions were recorded. These data were aligned by time to construct a data matrix containing 15 feature dimensions. Principal component analysis reduced the feature dimensions to five principal components, which reflected the main characteristics of plant growth, metabolism, and bacterial wilt resistance. These features were then organized into a time series to form a growth cycle data record.

[0046] In one specific embodiment, the process of performing step S102 may specifically include the following steps:

[0047] (1) Extract morphological characteristics, physiological indicators and disease symptoms of healthy plants from the growth cycle feature database and establish a phenotypic dataset of healthy plants.

[0048] (2) Extract morphological characteristics, physiological indicators and disease symptoms of diseased plants from the growth cycle feature database and establish a phenotypic dataset of diseased plants.

[0049] (3) Compare the phenotypic datasets of healthy plants and diseased plants over time, calculate the feature difference values ​​at each time point, and construct the phenotypic trajectory difference matrix.

[0050] (4) The phenotypic trajectory difference matrix is ​​segmented according to the reproductive period, and the distribution of difference characteristics in each reproductive period is statistically analyzed to form characteristic difference statistics;

[0051] (5) Based on the characteristic difference statistics, set the difference threshold, screen the significant difference characteristics of the key reproductive period, and construct an early warning indicator set;

[0052] (6) Input the early warning indicator set into the early warning evaluation network constructed by the recurrent neural network. The early warning evaluation network consists of an input layer, a feature extraction layer, a time series analysis layer and an early warning output layer.

[0053] The input layer contains N input nodes, where N is the feature dimension of the early warning indicator, and each node corresponds to a phenotypic feature data. The feature extraction layer consists of long short-term memory units, each containing a forget gate, an input gate, and an output gate, used to capture the long-term dependencies of phenotypic features. The time series analysis layer adopts a bidirectional structure, processing the feature sequence from the current time to the future time in the forward direction and processing the feature sequence from the historical time to the current time in the reverse direction, realizing the analysis of data throughout the entire reproductive period. The early warning output layer calculates the probability distribution of different risk levels through the Softmax function, compares the probability values ​​with preset thresholds to generate early warning judgment rules, and obtains the bacterial wilt resistance early warning model.

[0054] Specifically, when extracting data from the growth cycle feature library, plants are classified according to the disease symptom rating results. Plants with a disease symptom rating of 0-2 are classified as healthy plants, and plants with a disease symptom rating of 3 or higher are classified as diseased plants. Morphological characteristics (plant height, number of branches, leaf area), physiological indicators (photosynthetic intensity, respiration intensity, chlorophyll content), and disease symptoms (lesion size, disease index, incidence rate) are extracted from the healthy plant group to establish a healthy plant phenotypic dataset. The same method is used to extract the above indicator data from the diseased plant group to establish a diseased plant phenotypic dataset. When comparing the two phenotypic datasets over time, the data are aligned along the time axis to ensure comparability at the same time points. The characteristic difference value between the healthy group and the diseased group at each time point is calculated using a relative change method, i.e., (disease group value - healthy group value) / healthy group value. The calculated difference values ​​are arranged in chronological order to construct a phenotypic trajectory difference matrix. The rows of the matrix represent time series data, and the columns represent different feature indicators.

[0055] When segmenting the phenotypic trajectory difference matrix, the data was divided into three growth stages: seedling stage, vegetative growth stage, and reproductive growth stage. Statistical analysis was performed on the differences within each growth stage, calculating the mean, standard deviation, coefficient of variation, and other statistical measures for each characteristic indicator within that period. In this way, characteristic difference statistics reflecting the distribution patterns of differences across different growth stages were obtained.

[0056] When setting difference thresholds based on feature difference statistics, statistical significance testing is employed. A normality test is performed on the feature difference statistics to determine the data distribution characteristics. Based on the test results, either parametric or non-parametric testing methods are selected to calculate the significance level. Features with a significance level of 0.05 are marked as significantly different features. For significantly different features, their performance characteristics at different reproductive stages are further analyzed, and features exhibiting significant differences in the early stages are selected as early warning indicators to construct an early warning indicator set. When the early warning indicator set is input into the early warning assessment network, the network's input layer is constructed. The number of nodes N in the input layer is equal to the number of dimensions of the early warning indicators, with each node corresponding to one feature indicator. The feature extraction layer uses a Long Short-Term Memory (LSTM) structure, with each LSTM unit containing three gating units: a forget gate, an input gate, and an output gate. The forget gate determines the information to be discarded from the cell state, the input gate determines the new information to be updated, and the output gate determines the output cell state. The temporal analysis layer adopts a bidirectional structure, containing two LSTM layers: a forward layer and a backward layer. The forward LSTM layer processes sequence information from the current time to future time, while the backward LSTM layer processes sequence information from historical time to the current time. The warning output layer uses the Softmax activation function to convert the network output into a probability distribution of different risk levels. The probability values ​​are compared with preset warning thresholds to generate a warning judgment result.

[0057] For example, in the F2 generation of peanut plants resistant to bacterial wilt, 75 healthy plants and 25 diseased plants were identified from 100 plants. Growth period data, including phenotypic characteristic records at 15 time points, were extracted from both groups. Time-series comparative analysis revealed a significant difference in chlorophyll content between the healthy and diseased groups in the early vegetative growth stage, a difference that gradually widened in subsequent periods. This feature was incorporated into the early warning indicator set. When constructing the early warning assessment network, the input layer had 12 nodes, corresponding to 12 early warning indicators. The feature extraction layer used 64 LSTM units, and the time-series analysis layer employed a bidirectional structure with 32 LSTM units. After network training, early warning assessments were performed on new plant samples. Based on the probability distribution of the Softmax output layer, the early warning threshold was set to 0.7. When the probability of a plant's disease risk exceeded this threshold, it was marked as a potentially susceptible individual.

[0058] In one specific embodiment, the process of performing step S103 may specifically include the following steps:

[0059] (1) Input the morphological characteristics, physiological indicators and disease symptoms of the breeding materials into N nodes of the input layer, and standardize the phenotypic characteristic data of each node to obtain standardized growth phenotypic data;

[0060] (2) Standardized growth phenotypic data are fed into the long short-term memory unit of the feature extraction layer. Historical information is filtered through the forget gate, current information is received through the input gate, and information is integrated through the output gate to obtain temporal correlation data of phenotypic features.

[0061] (3) Input the phenotypic feature time-series correlation data into the time-series analysis layer, and process the feature sequences from the current time to the future time and from the historical time to the current time through a bidirectional structure to obtain the full-process analysis data of the reproductive period;

[0062] (4) The probability distribution values ​​of different risk levels are obtained by calculating the data of the entire reproductive period through the Softmax function of the early warning output layer;

[0063] (5) Compare the probability distribution values ​​with the warning threshold, mark potentially susceptible individuals who exceed the threshold, and construct risk monitoring results;

[0064] (6) Perform characteristic analysis and risk level ranking on susceptible individuals in the risk monitoring results to obtain risk warning results.

[0065] Specifically, the input data processing for the early warning assessment network involves extracting morphological characteristics (plant height, number of branches, leaf area), physiological indicators (photosynthetic intensity, respiration intensity, chlorophyll content), and disease symptoms (lesion size, disease index, incidence rate) from breeding materials. This data is input into N input nodes of the early warning network, where N equals the total number of feature dimensions. The input data is then standardized using the z-score standardization method to convert all indicators to a uniform scale.

[0066] The standardized growth phenotypic data is fed into the Long Short-Term Memory (LSTM) units of the feature extraction layer. Each LSM unit contains three gating units: a forget gate, an input gate, and an output gate. The forget gate selectively forgets the cell state from the previous time step, primarily filtering out historical information that contributes little to the current warning judgment. The input gate receives new information from the current time step, focusing on the changes in phenotypic features. The output gate integrates the filtered historical information with the current information to generate temporal correlation data of phenotypic features. This temporal correlation data reflects the changing patterns of plant phenotypic features over time.

[0067] Phenotypic feature time-series correlation data are input into the time-series analysis layer, which employs a bidirectional structure for data processing. The forward processing analyzes the feature change trends from the current time to the predicted future time, while the backward processing analyzes the feature evolution patterns from historical time to the current time. Through bidirectional analysis, complete full-cycle analysis data is obtained, which includes information on the plant's growth and development characteristics and resistance to bacterial wilt throughout the entire growth cycle.

[0068] Risk level probabilities were calculated using the Softmax function on data analyzed throughout the reproductive period. The calculation formula is as follows:

[0069]

[0070] in: This represents the probability value of the i-th risk level; Indicates the number of characteristics throughout the reproductive period; Represents the temporal weight of the j-th feature; This represents the contribution of the j-th feature to resistance to bacterial wilt; Represents the growth cycle coefficient of the j-th feature; This indicates the total number of risk levels.

[0071] The calculated probability distribution values ​​are compared with preset warning thresholds. These thresholds are critical values ​​determined based on historical data statistical analysis; when the probability value for a certain risk level exceeds the corresponding threshold, the plant is marked as a potentially susceptible individual. A risk monitoring results dataset is created for all marked susceptible individuals. Further analysis is performed on the susceptible individuals in the risk monitoring results, statistically analyzing their characteristics, calculating the degree of deviation of various characteristic indicators from healthy plants, and ranking the susceptible individuals by risk level based on the degree of deviation, thus forming risk warning results.

[0072] For example, in the F2 generation selection process for peanut bacterial wilt resistance breeding, phenotypic data of hybrid offspring from mutagenic materials with strong bacterial wilt resistance and superior varieties are input into the early warning assessment network. Data on nine characteristic indicators, including plant height and photosynthetic intensity, are standardized. The standardized data are then fed into the LSTM unit of the feature extraction layer, where a gating mechanism processes the temporal features of the data. For example, for chlorophyll content, the forgetting gate filters out short-term diurnal fluctuations while retaining long-term trends; the input gate focuses on mutation points in chlorophyll content; and the output gate integrates this information to form a temporal feature reflecting the changing patterns of chlorophyll content. The bidirectional structure of the temporal analysis layer processes the historical performance of plants from emergence to the present moment, as well as predicts future growth and development trends. Based on the processed data, the probability of plants belonging to different risk levels is calculated using the Softmax function. By comparing the risk probability with the early warning threshold, potentially susceptible individuals are identified, and the risk level is determined based on the degree of abnormality of each characteristic indicator, selecting individuals suitable for continued breeding.

[0073] In one specific embodiment, the process of executing step S104 may specifically include the following steps:

[0074] (1) Based on the risk warning results, the resistance level of the plants was divided into four levels: high resistance, medium resistance, mild susceptibility and high susceptibility, and a graded dataset was established.

[0075] (2) Track and monitor plants with different resistance levels in the graded dataset throughout their entire growth period, collect morphological characteristics, physiological indicators and disease symptoms data at each key growth stage, and construct a growth period tracking data matrix;

[0076] (3) Extract the characteristic data of each key period from the growth period tracking data matrix, compare the characteristics of plants in the same resistance level group, and obtain the characteristic consistency index within the group.

[0077] (4) Compare and analyze the consistency index of characteristics within the group with the characteristics between groups with different resistance levels, calculate the significance of characteristic differences, and screen out the key period phenotypic characteristics;

[0078] (5) Establish scoring criteria for phenotypic characteristics during critical periods, divide the characteristic values ​​into different scores according to the numerical range, and form resistance scoring rules;

[0079] (6) Based on the resistance scoring rules, the plant phenotypic characteristics are scored and calculated, and the scores of each characteristic are combined to obtain a dynamic resistance evaluation scheme.

[0080] Specifically, a four-level classification standard was used to classify the resistance levels of plants based on risk warning results. High resistance corresponds to a risk warning value less than 0.3, exhibiting no obvious disease symptoms throughout the entire growth period; medium resistance corresponds to a risk warning value between 0.3 and 0.5, showing only mild disease symptoms in the later stages of growth; mild susceptibility corresponds to a risk warning value between 0.5 and 0.7, showing disease symptoms starting in the middle of growth; and high susceptibility corresponds to a risk warning value greater than 0.7, showing obvious disease symptoms in the early stages of growth. Independent datasets were established for each of the four levels, recording plant number, resistance level, and risk warning value.

[0081] The graded plant populations were monitored throughout their entire growth cycle, with key monitoring points established at the seedling, vegetative growth, and reproductive growth stages. At each monitoring point, data on morphological characteristics (plant height, number of branches, leaf area), physiological indicators (photosynthetic intensity, respiration intensity, chlorophyll content), and disease symptoms (lesion size, disease index, incidence rate) were collected. The collected data were organized into a growth cycle tracking data matrix according to time sequence and characteristic type, with rows representing different time points and columns representing different characteristic indicators. Characteristic data for each key growth stage were extracted from the growth cycle tracking data matrix. For plants within the same resistance level group, the mean and standard deviation of each characteristic indicator were calculated to assess the consistency of characteristics within the group. The coefficient of variation was used to evaluate consistency; a smaller coefficient of variation indicates more consistent performance among plants within the group. For each characteristic indicator, the significance of characteristic differences among the four resistance level groups was calculated using one-way ANOVA, calculating the F-statistic and P-value. Characteristics with P-values ​​less than 0.05 were marked as significantly different.

[0082] The scoring criteria for phenotypic characteristics during critical periods were established using the following formula:

[0083]

[0084] in: This represents the score of the k-th plant at the j-th resistance level during the i-th growth stage; Indicates the number of feature indicators; This represents the weight coefficient of the m-th feature; Indicates the actual measured value; This represents the minimum value of the feature; This represents the maximum value of the feature; The effectiveness coefficient represents the feature.

[0085] Formula for calculating the comprehensive resistance score based on the scoring rules:

[0086]

[0087] in: This represents the overall resistance score of the k-th plant; Indicates the number of reproductive years; Indicates the quantity of resistance levels; This represents the temporal weight of the i-th reproductive period; This represents the level coefficient for the j-th resistance level; This represents the interaction coefficient between reproductive period and resistance level.

[0088] The calculated comprehensive score results were compiled into a dynamic resistance evaluation scheme. For example, in the F2 generation population selection, individuals with different performance were selected from plants grown under the bacterial wilt spectrum and graded. When tracking the growth period of the highly resistant plants, it was recorded that the chlorophyll content of these plants remained at a high level during the vegetative growth stage, and the coefficient of variation was small, indicating that this characteristic had good consistency within the highly resistant group. Further comparison of chlorophyll content data from different resistance levels revealed significant differences in this indicator among different resistance levels. The chlorophyll content indicator was incorporated into the scoring criteria, and corresponding weighting coefficients were set. Through comprehensive scoring of multiple periods and multiple characteristic indicators, a group of individuals with stable bacterial wilt resistance were selected. These individuals entered the F3 generation southern breeding stage for further resistance verification.

[0089] In one specific embodiment, the process of executing step S105 may specifically include the following steps:

[0090] (1) Collect plant growth data in dynamic resistance evaluation schemes at different geographical locations, record growth conditions, soil characteristics and climate parameters, and construct multi-point test datasets;

[0091] (2) Perform site-specific statistics on the multi-point experimental dataset, calculate the correlation between plant phenotypic data and environmental factors at each experimental site, and form an environmental response data matrix;

[0092] (3) Extract the temporal variation patterns of plant growth and environmental factors from the environmental response data matrix and establish a plant-environment interaction data table;

[0093] (4) Perform variance decomposition on the data in the plant-environment interaction data table to separate genotype effects, environmental effects and interaction effects to obtain effect component data;

[0094] (5) Calculate the performance stability of each genotype based on the effect component data, and construct an adaptive index system that includes the interaction between genotype and environment;

[0095] (6) Perform numerical analysis on the adaptability index system, screen out genotype combinations with high stability, and obtain strain stability parameters.

[0096] Specifically, data collection during the regional trial phase involved setting up experimental sites in representative ecological regions, including cold, temperate, and subtropical zones. Detailed environmental parameters were recorded at each site, including growth conditions (light intensity, sunshine duration, precipitation), soil characteristics (pH, organic matter content, trace elements), and climate parameters (temperature, humidity, wind speed). Simultaneously, phenotypic data were collected from the planted peanut varieties, recording growth and development indicators and resistance to bacterial wilt. This data was compiled into a multi-site experimental dataset. When analyzing the multi-site experimental dataset, data were processed in groups according to the experimental sites. The Pearson correlation coefficient was used to calculate the correlation between plant phenotypic data and environmental factors, establishing a correlation matrix. Each element in the correlation matrix represents the correlation strength between a phenotypic indicator and an environmental factor; the correlation coefficient ranges from -1 to 1, with positive values ​​indicating a positive correlation and negative values ​​indicating a negative correlation.

[0097] When extracting temporal variation patterns from the environmental response data matrix, the data is divided according to the growth stage. For each stage, the dynamic correlation between plant phenotype and environmental factors is calculated, and the impact of environmental factor changes on plant growth and development is analyzed. The analysis results are compiled into a plant-environment interaction data table, which includes the interaction relationships at different growth stages.

[0098] The variance of the plant-environment interaction data was decomposed using a two-way ANOVA. The total variance was decomposed into three components: genotype effect (differences between different strains), environmental effect (differences between different experimental sites), and interaction effect (differences in the response of strains to the environment). By calculating the variance components of each component, the influence of different factors on plant performance was assessed.

[0099] The formula for calculating the stability index based on effect component data is:

[0100]

[0101] in: It represents the stability index of the k-th trait of the i-th genotype in the j-th environment; Indicates the number of environmental points; Indicates the number of test periods; This represents the environmental weighting coefficient; Indicates the period weighting coefficient; Indicates the measured value; This represents the mean; It represents the standard deviation.

[0102] The formula for calculating the comprehensive stability parameter is:

[0103]

[0104] in: Indicates the overall stability parameter; Indicates the number of genotypes; Indicates the quantity of a trait; Indicates genotype weight; Indicates trait weights; This represents the coefficient of variation.

[0105] For example, superior strains obtained from the F3 generation of southern breeding entered the multi-site trial phase. The same batch of materials was planted at test sites in three different climatic regions, and temperature variation curves, precipitation distribution, and soil nutrient status were recorded at each test site during the trial. Phenotypic data collection revealed differences in morphological characteristics such as plant height and number of branches among different test sites for a particular strain. Environmental correlation analysis determined that these differences were significantly correlated with environmental factors such as daily average temperature and precipitation. Further variance decomposition revealed that the phenotypic variation of this strain mainly stemmed from environmental effects, with relatively small genotypic and interaction effects. Based on stability indicators, superior strains exhibiting stable performance were selected; these strains demonstrated strong environmental adaptability and resistance to bacterial wilt.

[0106] In one specific embodiment, the process of executing step S106 may specifically include the following steps:

[0107] (1) Numerical analysis of the stability parameters of the strains was conducted to establish an evaluation matrix that includes stability indicators, resistance to bacterial wilt indicators and yield indicators, and to construct a comprehensive evaluation system for breeding materials;

[0108] (2) Quantify and calculate the various indicators in the comprehensive evaluation system of breeding materials according to their weights to form a multi-dimensional evaluation score and establish a comprehensive scoring dataset;

[0109] (3) Cluster the comprehensive score dataset and divide the breeding materials into different grades according to the score distribution to form a graded evaluation result;

[0110] (4) Set a screening threshold for the graded evaluation results, and include materials with values ​​higher than the threshold into the candidate set of superior strains to construct a database of superior materials;

[0111] (5) Analyze the growth cycle characteristics of the lines in the database of superior materials, evaluate their resistance to bacterial wilt and growth status throughout the entire growth period, and form a characteristic profile of the lines;

[0112] (6) Based on the characteristic profile of the strain, the superior strains are screened. Taking into account the growth cycle performance and resistance to bacterial wilt, a bacterial wilt resistance breeding scheme based on the growth cycle is obtained.

[0113] Specifically, numerical analysis of strain stability parameters requires the establishment of a complete evaluation index system. Stability indicators include genotype effect values, environmental effect values, and interaction effect values, reflecting the stability of plant performance under different environments. Bacterial wilt resistance indicators include disease incidence rate, disease index, lesion size, and disease progression rate, reflecting the plant's resistance to bacterial wilt. Yield indicators include yield per plant, number of pods, seed setting rate, and weight per 100 pods, reflecting the plant's yield performance. These indicators are organized into an evaluation matrix, where rows represent different breeding materials and columns represent different evaluation indicators. During the quantitative calculation stage, different weights are assigned to each indicator in the evaluation matrix. The stability indicator weight is set to 0.4, considering the core objective of bacterial wilt resistance breeding, the bacterial wilt resistance indicator weight is set to 0.4, and the yield indicator weight is set to 0.2. Corresponding weight coefficients are also set for specific indicators within each indicator category. After standardizing the raw data, the weights are multiplied by the weight coefficients to obtain a weighted score. The weighted scores of all indicators are summarized to obtain a comprehensive evaluation score for each breeding material, establishing a comprehensive scoring dataset.

[0114] When performing cluster analysis on the comprehensive score dataset, the K-means clustering algorithm was used to classify breeding materials into four levels: excellent, good, average, and poor. During clustering, the comprehensive score was used as the feature dimension, and the Euclidean distance between samples was calculated. Cluster centers were determined through iterative optimization to obtain the graded evaluation results.

[0115] A screening threshold was set based on the grading evaluation results, and materials at the excellent level were selected as candidate superior lines. Materials with a comprehensive score higher than 90 points were directly included in the candidate set of superior lines, while materials with scores between 80 and 90 points required further evaluation to determine if their key indicators met the requirements. The information of the selected materials was stored in the superior material database, which recorded information such as material number, comprehensive score, and values ​​of various indicators. A full growth period analysis was performed on the lines in the superior material database, focusing on the performance of the three key stages: seedling stage, vegetative growth stage, and reproductive growth stage. Morphological characteristics (plant height, number of branches, leaf area), physiological indicators (photosynthetic intensity, respiration intensity, chlorophyll content), and resistance to bacterial wilt (lesion size, disease index, incidence rate) were recorded for each stage. These data were then compiled into a time series to construct a characteristic profile reflecting the entire process of plant growth and development.

[0116] When screening based on strain characteristics, three key aspects are considered: growth vigor, resistance to bacterial wilt, and stability. Growth vigor mainly assesses the robustness of plant morphology; resistance to bacterial wilt focuses on the occurrence and development of the disease; and stability evaluates the consistency of various traits. By comprehensively analyzing these characteristics, superior strains are selected, and bacterial wilt-resistant breeding programs are developed.

[0117] For example, after obtaining strain stability parameters through multi-point trials, a batch of F3 generation materials were comprehensively evaluated. An evaluation matrix was established, containing 15 evaluation indicators. The weight coefficients of each indicator were determined through preliminary data analysis; for example, the disease index was weighted at 0.4 in the bacterial wilt resistance indicator because this indicator has the highest correlation with bacterial wilt resistance. After standardization and weighted calculation of the raw data, a comprehensive score was obtained for each material. K-means clustering was used to classify the materials into four levels, and a comprehensive score of 90 was set as the screening threshold for superior strains. The selected superior strains were tracked throughout their growth period, and key growth indicators were recorded. Based on the growth characteristic profiles of the materials, superior strains possessing both stable bacterial wilt resistance and good agronomic traits were selected, and these strains entered the regional trial stage.

[0118] In one specific embodiment, the process of extracting stability index data, including genotype effect values ​​in strain stability parameters, from the evaluation system can specifically include the following steps:

[0119] (1) Extract stability index data from the comprehensive evaluation system of breeding materials, including genotype effect value, environmental effect value and interaction effect value in the stability parameters of the line, and construct a stability scoring matrix;

[0120] (2) Extract bacterial wilt resistance index data from the comprehensive evaluation system of breeding materials, including disease incidence, disease index, lesion size and disease progression rate, and construct a bacterial wilt resistance scoring matrix;

[0121] (3) Extract yield index data from the comprehensive evaluation system of breeding materials, including yield per plant, number of pods, seed setting rate and weight of 100 pods, and construct a yield scoring matrix;

[0122] (4) Set weight coefficients for the stability score matrix, bacterial wilt resistance score matrix and yield score matrix, and construct a weight allocation scheme;

[0123] (5) The scoring matrix is ​​weighted based on the weight allocation scheme, and the scores of each indicator are integrated to form a comprehensive score table;

[0124] (6) Normalize the comprehensive score table to unify all indicators to the same scale and obtain the comprehensive score dataset.

[0125] Specifically, the comprehensive evaluation of breeding materials begins with the extraction of stability index data. Genotype effect values, environmental effect values, and interaction effect values ​​are extracted from the strain stability parameters. These effect values ​​are obtained through variance decomposition. The genotype effect value reflects the contribution of the plant's own genetic characteristics, the environmental effect value represents the degree of influence of external environmental conditions, and the interaction effect value reflects the plant's response to the environment. These three effect values ​​are organized into a stability scoring matrix according to the material number, with rows representing different breeding materials and columns corresponding to the three effect values. The extraction of bacterial wilt resistance index data focuses on four key indicators: disease incidence rate is obtained by statistically analyzing the ratio of diseased plants to the total number of plants; the disease severity index is calculated based on the severity level of the disease; lesion size is measured and recorded using a ruler; and the disease progression rate is determined by continuously observing the speed of lesion expansion. These indicators need to be recorded during the seedling stage, vegetative growth stage, and reproductive growth stage, and the dynamic process of disease development is reflected by comparing data from different periods. These indicator data are organized into a bacterial wilt resistance scoring matrix, with each row representing one material and four columns corresponding to the four bacterial wilt resistance indicators.

[0126] The extraction of yield index data mainly includes: yield per plant obtained by weighing; number of pods counted manually; seed setting rate calculated based on the ratio of pods to total pods; and weight of 100 pods obtained by weighing 100 seeds. These index data are primarily collected during the harvest period and organized into a yield scoring matrix according to the material number. The matrix structure is similar to the previous two matrices, with rows representing materials and columns corresponding to the four yield indicators. When setting weight coefficients for the three scoring matrices, a balance between the goals of bacterial wilt resistance breeding and production needs was considered. The stability scoring matrix has a weight of 0.3 because stability is the foundation for variety promotion; the bacterial wilt resistance scoring matrix has a weight of 0.4, serving as a core indicator for bacterial wilt resistance breeding; and the yield scoring matrix has a weight of 0.3 to ensure that the selected materials have good yield performance. Within each matrix, different indicators also have corresponding weights. For example, in the bacterial wilt resistance indicators, the disease severity index has a higher weight because it comprehensively reflects the severity of the disease.

[0127] When performing weighted calculations based on a weighting scheme, the original data in each matrix is ​​multiplied by the corresponding indicator weight to obtain the weighted indicator score. Then, the weighted scores of the same material in different matrices are summed to obtain the material's overall score. All material scores are then compiled into an overall score table, which includes the material number and its corresponding overall score.

[0128] When normalizing the comprehensive score table, the maximum-minimum normalization method is used. The score of each indicator is converted to the range of 0-1, eliminating the influence of different dimensions between indicators. This normalization process yields a standardized comprehensive score dataset, facilitating subsequent material grading and screening.

[0129] For example, during the F3 generation breeding process, a comprehensive evaluation was conducted on a batch of materials that had undergone multi-location trials. Effect values ​​were extracted from the line stability parameters, revealing that the genotype effect value of a particular material was dominant, indicating that its phenotypic expression was primarily determined by genetic factors. In the evaluation of resistance to bacterial wilt, the disease data of this material was continuously tracked, recording a slow lesion expansion rate and a low disease index. Yield indicators were measured at harvest, showing good yield per plant and seed setting rate. These data were input into three scoring matrices, and calculated according to set weighting coefficients to obtain the material's comprehensive score. This evaluation method allows for the selection of superior materials that possess both stability and good resistance to bacterial wilt and yield performance.

[0130] In one specific embodiment, the process of extracting bacterial wilt resistance index data from the comprehensive evaluation system of breeding materials may specifically include the following steps:

[0131] (1) The disease incidence data were divided into seedling disease incidence, vegetative growth period disease incidence and reproductive growth period disease incidence according to different growth stages, and a time series matrix of disease incidence was constructed.

[0132] (2) Classify and statistically analyze the disease index data according to the lesion area, degree of wilting and affected parts, and establish a disease grading data table;

[0133] (3) Extract the time series changes of lesion length, width and area from the lesion size data to form a lesion development dataset;

[0134] (4) Track the disease progression rate data over time, calculate the lesion expansion rate and disease severity per unit time, and obtain the dynamic parameters of the disease;

[0135] (5) Integrate the incidence rate time series matrix, disease severity grading data table, lesion development dataset and disease dynamic parameters to form a database of bacterial wilt resistance characteristics;

[0136] (6) Based on the various indicators in the bacterial wilt resistance characteristic database, weighted processing is performed, and the comprehensive score is calculated by combining the weight coefficients to construct the bacterial wilt resistance scoring matrix.

[0137] Specifically, the disease data processing categorized disease incidence rates according to growth stages. Seedling stage incidence rates were statistically analyzed 15-30 days after sowing, obtained by investigating the ratio of diseased plants to the total number of plants. Vegetative growth stage incidence rates were statistically analyzed 31-60 days after sowing, focusing on the increase in newly infected plants. Reproductive growth stage incidence rates were statistically analyzed 61-120 days after sowing, recording the spread of the disease within the population. The incidence rate data from these three stages were compiled into a time-series matrix, where rows represent different time points and columns represent incidence rates at different growth stages. Disease index data were processed using a categorical statistical method, rating the disease based on three dimensions: lesion area, wilting severity, and affected body part. The lesion area is classified into five levels: Level 1 indicates lesions covering less than 10% of the leaf area, Level 2 10-25%, Level 3 26-50%, Level 4 51-75%, and Level 5 over 76%. The degree of wilting is also classified into five levels: Level 1 slight wilting, Level 2 obvious wilting, Level 3 partial wilting, Level 4 mostly wilting, and Level 5 complete death. Affected parts are categorized into roots, stems, and leaves. This classification data is compiled into a disease severity grading table, which records the disease severity level of each plant at different stages.

[0138] The processing of lesion size data involves three dimensions: dynamic changes in lesion length, width, and area. The length and width of the lesions are measured using a ruler, and the area is calculated based on the lesion shape. Measurements are taken every three days to record the expansion of the lesions. These measurement data are then organized chronologically to form a dataset reflecting the lesion development process. This dataset contains a complete record of lesion size changes over time. The processing of disease progression rate data focuses on two aspects: the speed of lesion expansion and the severity of disease progression. The lesion expansion speed is obtained by calculating the increase in lesion area per unit time, while the severity of disease progression is represented by the rate of change in disease severity. These data are continuously tracked, recording changes within each time interval to obtain parameters reflecting the dynamics of disease development.

[0139] The data integration phase involved correlation analysis of the incidence rate time-series matrix, disease severity grading data table, lesion development dataset, and disease dynamic parameters. A unified data format was established, aligning data from different sources to the same timeline. Correlation analysis determined the relationships between different indicators, constructing a complete database of bacterial wilt resistance characteristics. When performing comprehensive scoring based on this database, weighting coefficients were assigned to each indicator. The incidence rate indicator had a weight of 0.3, reflecting population-level resistance; the disease severity index had a weight of 0.3, reflecting the severity of disease on a single plant; the lesion development indicator had a weight of 0.2, indicating the disease's expansion trend; and the disease progression rate had a weight of 0.2, reflecting the dynamic characteristics of bacterial wilt resistance. By combining these weighting coefficients, a comprehensive bacterial wilt resistance score for each plant was calculated, constructing a bacterial wilt resistance scoring matrix.

[0140] For example, in evaluating bacterial wilt resistance in the F2 generation population, continuous observation was conducted from sowing. During the seedling stage, a small number of plants showed disease symptoms, and the initial incidence rate was recorded. After entering the vegetative growth stage, newly infected plants were counted, and the disease began to spread throughout the population. During the reproductive growth stage, the disease development was continued to be tracked. Diseased plants were graded according to their severity, and changes in lesion size were recorded. If a plant's lesions initially appeared on the leaves, and continuous measurements showed a slow rate of lesion expansion and minimal change in disease severity, this indicated that the plant possessed a certain degree of bacterial wilt resistance. These observational data were input into a scoring system to calculate the plant's overall bacterial wilt resistance score, which serves as an important reference indicator in the bacterial wilt resistance breeding process.

[0141] The above describes the peanut bacterial wilt resistance breeding strategy analysis method based on the growth cycle in the embodiments of this application. The following describes the peanut bacterial wilt resistance breeding strategy analysis system based on the growth cycle in the embodiments of this application. Please refer to [link / reference]. Figure 4 One embodiment of the peanut bacterial wilt resistance breeding strategy analysis system based on the growth cycle in this application includes:

[0142] The data acquisition module 201 is used to continuously collect morphological characteristics, physiological indicators and disease symptoms of peanuts during the growth period through the phenotypic monitoring system. The data is divided into key growth periods through time series segmentation, and a time series feature vector is constructed to obtain a growth cycle feature library.

[0143] The comparison module 202 is used to compare and analyze the phenotypic trajectories of healthy plants and diseased plants based on the growth cycle feature library, and to establish early warning indicators through differential feature extraction to obtain a bacterial wilt resistance early warning model.

[0144] Monitoring module 203 is used to monitor and analyze the growth phenotypic data of breeding materials in real time based on the bacterial wilt resistance early warning model, identify potentially susceptible individuals by early warning threshold, and obtain risk early warning results;

[0145] The tracking module 204 is used to track plants with different resistance levels throughout their growth period based on the risk warning results, establish a resistance scoring system by comparing phenotypic characteristics at key periods, and obtain a dynamic resistance evaluation scheme.

[0146] Analysis module 205 is used to collect multi-point test data for environmental response analysis according to the dynamic resistance evaluation scheme, establish adaptability indicators through plant-environment interaction assessment, and obtain strain stability parameters.

[0147] The grading module 206 is used to comprehensively evaluate and grade the breeding materials based on the stability parameters of the strains, determine the superior strains through the grading and screening system, and obtain a bacterial wilt resistance breeding program based on the growth cycle.

[0148] Through the collaborative efforts of the aforementioned components, a complete growth cycle feature database was constructed by continuously collecting and segmenting morphological characteristics, physiological indicators, and disease symptoms during the peanut growth period, enabling comprehensive monitoring and recording of the plant's growth and development. By comparing and analyzing the phenotypic trajectories of healthy and diseased plants, an early warning indicator system was established, providing data support for the timely identification of potentially susceptible individuals. Deep learning algorithms were introduced into the bacterial wilt resistance early warning model, utilizing the time-series analysis capabilities of recurrent neural networks to monitor and analyze growth phenotypic data in real time, improving the accuracy of early warnings. A dynamic resistance evaluation scheme was established by tracking the entire growth period of plants with different resistance levels and comparing phenotypic characteristics at key stages, making the evaluation of bacterial wilt resistance more objective and comprehensive. An adaptability indicator system was constructed through environmental response analysis of multi-point experimental data and plant-environment interaction assessment, enhancing the stability of strain selection. Finally, a tiered screening system was used to comprehensively evaluate breeding materials, achieving precise selection of superior strains. This method applies artificial intelligence algorithms to the field of peanut bacterial wilt resistance breeding. By using long short-term memory networks to extract features and perform time-series analysis on plant growth data, it fully mines the bacterial wilt resistance information contained in the growth cycle data, providing a data-driven scientific decision-making basis for the breeding of bacterial wilt-resistant varieties. Simultaneously, this method establishes a complete data processing and analysis workflow, realizing intelligent processing from data collection, feature extraction, early warning analysis to strain selection, greatly improving the efficiency and accuracy of bacterial wilt resistance breeding work.

[0149] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system and units described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0150] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for analyzing a breeding strategy for a peanut resistant to a bacterial wilt disease based on a growth cycle, characterized by, The peanut resistance to bacterial wilt breeding strategy analysis method based on the growth cycle comprises: The morphological characteristics, physiological indexes and disease symptoms of the peanut growth period are continuously collected through a phenotype monitoring system, the data is divided according to key growth periods through time series segmentation, a time sequence feature vector is constructed, and a growth cycle feature library is obtained; According to the growth cycle feature library, the phenotype trajectories of healthy plants and diseased plants are compared and analyzed, early warning indexes are established through difference feature extraction, and a bacterial wilt resistance early warning model is obtained, including: extracting morphological characteristics, physiological indexes and disease symptom data of healthy plants from the growth cycle feature library to establish a healthy plant phenotype data set; extracting morphological characteristics, physiological indexes and disease symptom data of diseased plants from the growth cycle feature library to establish a diseased plant phenotype data set; comparing the healthy plant phenotype data set and the diseased plant phenotype data set in time sequence, calculating the feature difference value at each time point, and constructing a phenotype trajectory difference matrix; segmenting the phenotype trajectory difference matrix according to growth periods, counting the difference feature distribution of each growth period, and forming a feature difference statistic; setting a difference threshold based on the feature difference statistic, screening significant difference features in key growth periods, and constructing an early warning index set; inputting the early warning index set into a warning evaluation network constructed by a recurrent neural network, the warning evaluation network being composed of an input layer, a feature extraction layer, a time sequence analysis layer and a warning output layer; the input layer includes N input nodes, N being the feature dimension of the early warning index, and each node corresponding to a phenotype feature data; the feature extraction layer is composed of long short-term memory units, each unit including a forgetting gate, an input gate and an output gate, for capturing long-term dependencies of phenotype features; the time sequence analysis layer adopts a bidirectional structure, processing feature sequences from the current time to the future time in a forward direction and processing feature sequences from the history time to the current time in a reverse direction, to realize analysis of growth period data; the warning output layer calculates the probability distribution of different risk levels through a Softmax function, compares the probability value with a preset threshold to generate a warning judgment rule, and obtains a bacterial wilt resistance early warning model; According to the bacterial wilt resistance early warning model, the growth phenotype data of breeding materials are monitored and analyzed in real time, potential susceptible individuals are determined through a warning threshold judgment, and a risk warning result is obtained; Based on the risk warning result, plants with different resistance levels are tracked throughout the growth period, a resistance scoring system is established through key period phenotype feature comparison, and a dynamic resistance evaluation scheme is obtained. According to the dynamic resistance evaluation scheme, multi-point test data is collected for environmental response analysis, an adaptability index is established through plant-environment interaction evaluation, and line stability parameters are obtained, including: collecting plant growth data in the dynamic resistance evaluation scheme at different geographical sites, recording growth conditions, soil properties, and climate parameters, and constructing a multi-point test data set; performing point-by-point statistics on the multi-point test data set, calculating the correlation between plant phenotype data and environmental factors at each test point, and forming an environmental response data matrix; extracting the time sequence variation law of plant growth and environmental factors from the environmental response data matrix, and establishing a plant-environment interaction data table; performing variance decomposition on the data in the plant-environment interaction data table, separating genotypic effects, environmental effects, and interaction effects, and obtaining effect component data; calculating the performance stability of each genotype based on the effect component data, and constructing an adaptability index system containing genotype and environment interaction; performing numerical analysis on the adaptability index system, and screening out genotype combinations with high stability to obtain line stability parameters; According to the line stability parameters, the breeding materials are comprehensively evaluated and classified, and excellent lines are determined through a classification screening system to obtain a growth cycle-based resistance to bacterial wilt breeding scheme.

2. The method for analyzing the growth cycle-based peanut anti-rhizoctonia solani breeding strategy according to claim 1, wherein, The morphological characteristics, physiological indicators, and disease symptoms of peanuts during the growth period are continuously collected by the phenotype monitoring system, the data is divided according to key growth stages through time sequence segmentation, a time sequence feature vector is constructed, and a growth cycle feature library is obtained, including: The morphological characteristics are collected by period segmentation, and the morphological characteristics are divided into seedling stage, vegetative growth stage, and reproductive growth stage. The plant height, branch number, and leaf area of each growth stage are quantitatively determined. The time sequence variation data of photosynthesis intensity, respiration intensity, and chlorophyll content are extracted from the physiological indicators, and the data is standardized and outliers are removed. The disease symptoms are rated and scored to establish a symptom index data set containing lesion size, disease index, and incidence rate. The segmented morphological characteristic data, time sequence variation data, and symptom index data are integrated to construct a multi-dimensional data set reflecting the characteristics of different growth stages. Feature extraction is performed on the multi-dimensional data set through data dimension reduction processing, a feature vector containing key growth stage indicators is established, the feature vector is sorted and organized according to the time sequence, a data correlation network covering the entire growth period is constructed, and the growth cycle feature library is obtained.

3. The method for analyzing the growth cycle-based peanut anti-rhizoctonia solani breeding strategy according to claim 1, wherein, According to the bacterial wilt resistance early warning model, the growth phenotype data of the breeding materials are monitored and analyzed in real time, potential susceptible individuals are determined through early warning threshold determination, and risk early warning results are obtained, including: The morphological characteristics, physiological indicators, and disease symptom data of the breeding materials are input into the N nodes of the input layer, and the phenotype characteristic data of each node is standardized to obtain standardized growth phenotype data; The standardized growth phenotype data is transmitted to the long short-term memory unit of the feature extraction layer, historical information is filtered through the forget gate, current information is received through the input gate, and information is integrated through the output gate to obtain phenotype feature time sequence correlation data; The phenotype characteristic time sequence association data is input into the time sequence analysis layer, and the feature sequence from the current time to the future time and from the historical time to the current time is processed through a bidirectional structure to obtain whole growth period analysis data of the growth period; The whole growth period analysis data is calculated through a Softmax function of the early warning output layer to obtain probability distribution values of different risk levels; The probability distribution values are compared with early warning threshold values, potential susceptible individuals exceeding the threshold values are marked, and a risk monitoring result is constructed; the susceptible individuals in the risk monitoring result are subjected to feature analysis and risk degree sorting to obtain a risk early warning result.

4. The method for analyzing the growth cycle-based peanut anti-rhizoctonia solani breeding strategy according to claim 1, wherein, Based on the risk early warning result, plants with different resistance levels are subjected to whole growth period tracking, a resistance scoring system is established through key period phenotype characteristic comparison, and a dynamic resistance evaluation scheme is obtained, including: Plants are divided into high resistance, medium resistance, light susceptibility and high susceptibility according to the risk early warning result, and the plants are divided into four levels to establish a hierarchical data set; The plants with different resistance levels in the hierarchical data set are subjected to whole growth period tracking and monitoring, and morphological characteristics, physiological indexes and disease symptom data at each key growth period are collected to construct a growth period tracking data matrix; Feature data of each key period is extracted from the growth period tracking data matrix, and the plants in the same resistance level group are subjected to feature comparison to obtain intra-group feature consistency indexes; The intra-group feature consistency indexes are compared and analyzed with different resistance level groups, feature difference significance is calculated, and key period phenotype characteristics are screened out; Scoring standards are established for the key period phenotype characteristics, feature values are divided into different scores according to numerical intervals, and a resistance scoring rule is formed; The plant phenotype characteristics are scored and calculated based on the resistance scoring rule, and the dynamic resistance evaluation scheme is obtained by comprehensively considering the feature scores.

5. The method for analyzing the growth cycle-based peanut anti-rhizoctonia solani breeding strategy according to claim 1, wherein, According to the strain stability parameters, the breeding materials are comprehensively evaluated and classified, the excellent strains are determined through a classification screening system, and a Ralstonia solanacearum resistance breeding scheme based on the growth cycle is obtained, including: The strain stability parameters are subjected to numerical analysis, an evaluation matrix containing stability indexes, Ralstonia solanacearum resistance indexes and yield indexes is established, and a breeding material comprehensive evaluation system is constructed; The indexes in the breeding material comprehensive evaluation system are quantitatively calculated according to weights to form multi-dimensional evaluation scores, and a comprehensive scoring data set is established; The comprehensive scoring data set is subjected to clustering processing, the breeding materials are divided into different grades according to the score distribution, and a classification evaluation result is formed; A screening threshold value is set for the classification evaluation result, the materials higher than the threshold value are included in an excellent strain candidate set, and an excellent material database is constructed; The strains in the excellent material database are subjected to growth cycle feature analysis to evaluate the Ralstonia solanacearum resistance performance and growth conditions in the whole growth period, and a strain characteristic portrait is formed; Based on the strain characteristic portrait, excellent strains are screened, the growth cycle performance and Ralstonia solanacearum resistance performance are comprehensively considered, and a Ralstonia solanacearum resistance breeding scheme based on the growth cycle is obtained.

6. The method for analyzing the growth cycle-based peanut anti-rhizoctonia solani breeding strategy according to claim 5, wherein, The indexes in the breeding material comprehensive evaluation system are quantitatively calculated according to weights, multi-dimensional evaluation scores are formed, and a comprehensive score data set is established, including: Stability index data is extracted from the breeding material comprehensive evaluation system, including genotype effect value, environment effect value and interaction effect value in line stability parameters, and a stability score matrix is constructed; Rice bacterial wilt resistance index data is extracted from the breeding material comprehensive evaluation system, including disease incidence, disease index, lesion size and disease progression rate, and a rice bacterial wilt resistance score matrix is constructed; Yield index data is extracted from the breeding material comprehensive evaluation system, including single plant yield, pod number, seed setting rate and hundred-pod weight, and a yield score matrix is constructed; The stability score matrix, rice bacterial wilt resistance score matrix and yield score matrix are set with weight coefficients, and a weight distribution scheme is constructed; Based on the weight distribution scheme, the score matrix is weighted and calculated, the scores of various indexes are integrated, and a comprehensive score table is formed; The comprehensive score table is normalized to unify the indexes to the same scale, and a comprehensive score data set is obtained.

7. The method for analyzing growth cycle based peanut anti-rhizoctonia solani breeding strategy of claim 6, wherein, The rice bacterial wilt resistance index data is extracted from the breeding material comprehensive evaluation system, including disease incidence, disease index, lesion size and disease progression rate, and a rice bacterial wilt resistance score matrix is constructed, including: The disease incidence data is divided into seedling stage incidence, vegetative growth stage incidence and reproductive growth stage incidence according to different growth periods, and a time sequence matrix of incidence is constructed; The disease index data is classified and counted according to lesion area, wilting degree and affected part to establish a disease classification data table; The time sequence changes of lesion length, width and area are extracted from the lesion size data to form a lesion development data set; The disease progression rate data is tracked in time sequence, the lesion expansion speed and disease severity in unit time are calculated, and disease dynamic parameters are obtained; The time sequence matrix of incidence, disease classification data table, lesion development data set and disease dynamic parameters are integrated to form a rice bacterial wilt resistance characteristic database; Based on the indexes in the rice bacterial wilt resistance characteristic database, weighted processing is performed, the comprehensive score is calculated through weight combination, and a rice bacterial wilt resistance score matrix is constructed.

8. A growth cycle-based peanut resistance to Ralstonia solanacearum breeding strategy analysis system for implementing the growth cycle-based peanut resistance to Ralstonia solanacearum breeding strategy analysis method according to any one of claims 1 to 7, characterized in that, The peanut bacterial wilt resistance breeding strategy analysis system based on growth cycle includes: A collection module is used to continuously collect morphological characteristics, physiological indexes and disease symptoms of peanuts during the growth period through a phenotype monitoring system, divide the data according to key growth periods through time sequence segmentation, construct a time sequence feature vector, and obtain a growth cycle feature library. The comparison module is configured to compare the phenotypic trajectories of the healthy plants and the diseased plants according to the growth cycle feature library, extract early warning indicators through difference features, and obtain a bacterial wilt resistance early warning model, including: extracting morphological features, physiological indicators, and disease symptom data of the healthy plants from the growth cycle feature library to establish a healthy plant phenotypic dataset; extracting morphological features, physiological indicators, and disease symptom data of the diseased plants from the growth cycle feature library to establish a diseased plant phenotypic dataset; performing time series comparison on the healthy plant phenotypic dataset and the diseased plant phenotypic dataset, calculating feature difference values at each time point, and constructing a phenotypic trajectory difference matrix; segmenting the phenotypic trajectory difference matrix according to growth periods, counting difference feature distributions of the growth periods, and forming feature difference statistics; setting a difference threshold based on the feature difference statistics, screening significant difference features of key growth periods, and constructing an early warning indicator set; inputting the early warning indicator set into an early warning evaluation network constructed by a recurrent neural network, the early warning evaluation network including an input layer, a feature extraction layer, a time series analysis layer, and an early warning output layer; the input layer includes N input nodes, N is a feature dimension of the early warning indicators, and each node corresponds to a phenotypic feature data; the feature extraction layer is composed of long short-term memory units, each unit includes a forgetting gate, an input gate, and an output gate, and is configured to capture long-term dependencies of the phenotypic features; the time series analysis layer adopts a bidirectional structure, processes feature sequences from the current time to the future time in a forward direction, and processes feature sequences from the historical time to the current time in a reverse direction, to analyze data of the whole growth period; the early warning output layer calculates probability distributions of different risk levels by using a Softmax function, compares the probability values with a preset threshold to generate an early warning judgment rule, and obtains the bacterial wilt resistance early warning model; The monitoring module is configured to monitor and analyze growth phenotypic data of breeding materials in real time according to the bacterial wilt resistance early warning model, determine potential susceptible individuals through an early warning threshold judgment, and obtain a risk early warning result; The tracking module is configured to track plants with different resistance levels throughout the growth period based on the risk early warning result, establish a resistance scoring system through comparison of key period phenotypic features, and obtain a dynamic resistance evaluation scheme. An analysis module is configured to collect multi-point test data for environmental response analysis according to the dynamic resistance evaluation scheme, establish an adaptability index through plant-environment interaction evaluation, and obtain a line stability parameter, including: collecting plant growth data in the dynamic resistance evaluation scheme at different geographical sites, recording growth conditions, soil characteristics, and climate parameters, and constructing a multi-point test data set; performing site-by-site statistics on the multi-point test data set, calculating the correlation between plant phenotype data and environmental factors at each test site, and forming an environmental response data matrix; extracting the time sequence variation law of plant growth and environmental factors from the environmental response data matrix, and establishing a plant-environment interaction data table; performing variance decomposition on the data in the plant-environment interaction data table, separating genotypic effects, environmental effects, and interaction effects, and obtaining effect component data; calculating the performance stability of each genotype based on the effect component data, and constructing an adaptability index system including genotype and environment interaction; performing numerical analysis on the adaptability index system, screening out a genotype combination with high stability, and obtaining a line stability parameter; A grading module is configured to comprehensively evaluate and grade breeding materials according to the line stability parameter, determine excellent lines through a grading screening system, and obtain a growth cycle-based resistance to bacterial wilt breeding scheme.

Images