Database construction method for integrally identifying resistance of rice blast and sheath blight

Abstract

The invention discloses a method for constructing a database for integrally identifying resistance of rice blast and sheath blight. The method comprises the following steps: step 1, acquiring multi-source data; step 2, carrying out standardized preprocessing on the multi-source data; step 3, extracting multi-dimensional features and carrying out association modeling; according to the method, phenotype, genotype and other multi-source data of rice blast and sheath blight are integrated to construct a database, the problem of traditional data islands is solved, meanwhile, through multi-dimensional feature extraction and correlation modeling, the resistance mechanism of double diseases can be synchronously analyzed, an accurate breeding decision is provided, and the database is dynamically updated, so that the breeding efficiency is improved. Disease source variation or environmental change can be quickly responded, the timeliness of a resistance database is ensured, and efficient and accurate data support is provided for breeding of disease-resistant rice varieties.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural bioinformatics and plant pathology, and particularly relates to a database construction method for integrated identification of rice blast and sheath blight resistance. BACKGROUND

[0002] Rice blast and sheath blight are two major fungal diseases in global rice production, causing more than 10% of global rice yield loss annually, and breeding rice varieties with double resistance is the key to prevention and control.

[0003] However, the traditional resistance identification method is a separate identification method, which is time-consuming and costly. Meanwhile, the existing database lacks multi-dimensional correlation data such as phenotype, genotype, environment and experimental process, and it is difficult to reveal the double-resistance synergistic mechanism. Moreover, the current identification database has poor dynamic adaptability and cannot update the resistance correlation rules of variant pathogenic strains in time, which limits the actual value of guiding breeding. Therefore, the present application provides a database construction method for integrated identification of rice blast and sheath blight resistance to solve the problems in the prior art. SUMMARY

[0004] In view of the above problems, the present application aims to provide a database construction method for integrated identification of rice blast and sheath blight resistance. The database construction method for integrated identification of rice blast and sheath blight resistance realizes the integrated database construction of double-disease resistance synchronous identification, resistance gene mining and environmental adaptability prediction through multi-dimensional data fusion, intelligent correlation analysis and dynamic updating mechanism.

[0005] To achieve the object of the present application, the present application realizes the following technical scheme: a database construction method for integrated identification of rice blast and sheath blight resistance, comprising the following steps:

[0006] Step 1: Multi-source data acquisition, synchronously acquiring phenotype, genotype, environment and test process data for resistance identification of rice blast and sheath blight;

[0007] Step 2: Multi-source data standardization preprocessing, standardizing and preprocessing the phenotype data, genotype data, environment data and test process data to generate standardized parameters;

[0008] The standardization of the phenotype data specifically includes segmenting the lesion area by a convolutional neural network model, extracting lesion area and color features, and combining artificial scoring calibration to generate standardized phenotype parameters; the text data is processed by a natural language processing tool to extract descriptive phenotype terms and map them to standardized grades;

[0009] The genotype data standardization specifically refers to removing low-quality SNP sites with a missing rate of >10% and a minor allele frequency of <0.05, uniformly encoding genotypes according to the rule of "0" homozygous reference type, "1" heterozygous type and "2" homozygous variant type, aligning the positions of disease resistance genes / QTLs and labeling the linkage disequilibrium interval;

[0010] The environmental data standardization specifically refers to calculating the cumulative environmental index and clustering rice blast races according to pathogenic spectrum similarity and classifying the strain lines of rice sheath blight according to pathogenicity;

[0011] The test process data standardization specifically refers to converting the test design type, the number of repetitions and the plot layout into a machine-readable format, labeling "to be reviewed" for abnormal data and removing unreliable data without recording the inoculation time;

[0012] Step three, multi-dimensional feature extraction and correlation modeling, based on the standardized data, key features of phenotype-genotype-environment are extracted, and a collaborative correlation model is constructed;

[0013] Step four, integrated database construction, based on the correlation model, a hierarchical architecture database system including a meta-database, a core database, a correlation database and an application interface layer is designed to store core data and correlation rule information;

[0014] Step five, dynamic database updating, through real-time data collection, model iteration, resistance marker updating and user feedback correction, the database is dynamically updated to ensure the timeliness of the database.

[0015] Further improvement lies in that: the phenotype data in step one includes plant height, leaf color and biomass, and the lesion morphology, occurrence period, severity, latent period, resistance classification of rice blast, and the lesion height, expansion mode, severity and stem decay degree of sheath blight;

[0016] The genotype data includes whole genome SNP, SSR marker, genotype of known rice blast resistance genes and sheath blight resistance QTLs, and whole genome resequencing data;

[0017] The environmental data includes meteorological data, cultivation data and pathogen data; the meteorological data includes daily temperature, relative humidity, rainfall and sunshine duration, the cultivation data includes nitrogen, phosphorus and potassium application amount, transplanting density and irrigation method, and the pathogen data includes rice blast physiological race and sheath blight pathogenic strain line;

[0018] The test process data includes experimental design type, number of repetitions, plot area, rice blast inoculum, sheath blight inoculum, inoculation time, field management record and data verification record.

[0019] Further improvement lies in that the lesion shape of rice blast in the phenotype data includes fusiform, oval and edge blurring degree, the incubation period is the days from inoculation to onset, and the resistance grading is divided into susceptible, intermediate resistance and high resistance; the number of repetitions in the test process data is ≥3 times, and the plot area is ≥10m 2 , and the data verification record adopts double-entry verification.

[0020] Further improvement lies in that when the environmental data is standardized to calculate the cumulative environmental index, the rice blast stress index = (average temperature × 0.3 + continuous rainy days × 0.5 + relative humidity × 0.2), and the sheath blight threat index = (the number of days with daily average humidity > 80% × 0.4 + nitrogen fertilizer application amount × 0.3 + planting density × 0.3).

[0021] Further improvement lies in that the multi-dimensional feature extraction and correlation modeling in step three are specifically:

[0022] S1, phenotype-resistance correlation feature extraction, the phenotype parameters significantly related to the severity of rice blast / sheath blight are selected through statistical analysis;

[0023] S2, genotype-resistance correlation feature extraction, the SNP sites significantly related to double resistance are selected through single marker analysis or whole genome association analysis, and then new resistance candidate genes are mined by combining gene function annotation;

[0024] S3, environment-resistance correlation feature construction, the contribution degree of environmental factors to resistance is quantified through a random forest model, and resistance stability zones are divided based on the cumulative environmental index;

[0025] S4, multi-factor synergistic correlation model construction, phenotype, genotype and environmental data are integrated to construct a resistance prediction model and a resistance synergistic optimization module.

[0026] Further improvement lies in that the resistance stability zones divided in S3 include a double resistance dominant zone, a rice blast susceptible zone and a sheath blight susceptible zone.

[0027] Further improvement lies in that in step four, the meta database records data sources, collection time, quality control tags and data format information; the core database stores structured data tables including phenotype tables, genotype tables, environment tables and test tables; the correlation database stores multi-dimensional correlation rules; and the application interface layer provides RESTful API interfaces.

[0028] Further improvement lies in that the dynamic updating mechanism in step five includes data automatic collection, model iterative optimization, resistance marker updating and user feedback correction.

[0029] Further improvements include: automatic data acquisition via field temperature and humidity sensors and sequencing platforms to access new data in real time, with a preprocessing process triggered daily at midnight; model iteration and optimization by collecting new experimental data quarterly, retraining the association model, and retaining model version records; resistance marker updates specifically involve updating the association rule base and labeling it "new discovery" after review when new resistance genes or pathogen mutations are discovered; and user feedback corrections are made through an open user interaction interface, with verified data corrections performed monthly.

[0030] The beneficial effects of this invention are as follows: This invention is the first to integrate multi-source data such as phenotypes and genotypes of rice blast and sheath blight to construct a database, which solves the problem of traditional data silos. At the same time, through multi-dimensional feature extraction and correlation modeling, the resistance mechanisms of the two diseases can be analyzed simultaneously, providing accurate breeding decisions. Moreover, the dynamic updating of the database can quickly respond to pathogen mutations or environmental changes, ensuring the timeliness of the resistance database and providing efficient and accurate data support for the breeding of disease-resistant rice varieties. Attached Figure Description

[0031] Figure 1 This is a flowchart of the construction method of the present invention. Detailed Implementation

[0032] To enhance understanding of the present invention, the present invention will be further described in detail below with reference to embodiments. These embodiments are only used to explain the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0033] Rice blast and sheath blight are the two major fungal diseases affecting rice production globally, causing an average annual reduction of over 10% in global rice yields. Breeding rice varieties with dual resistance is crucial for control, but traditional resistance identification methods have the following significant drawbacks:

[0034] 1. Separate identification leads to resource waste: Rice blast and sheath blight require separate field / indoor trials. For example, rice blast requires inoculation with different physiological races, and sheath blight requires setting different humidity gradients. The identification cycle for a single disease is about 2-3 months, and the identification of two diseases requires repeating the experiment twice, with a total cycle of ≥4 months. This is time-consuming, and repeated experiments consume manpower, land, reagents and other resources, which greatly increases the cost.

[0035] 2. Analysis of data isolation and limitation mechanisms: Existing databases such as Rice blast DB and RSDB for rice blast resistance focus on single diseases and store only fragmented data such as phenotypic severity or genotype markers. They lack multi-dimensional correlation data of phenotype, genotype, environment and experimental process, making it difficult to reveal the synergistic genetic mechanism or environmental regulation of dual resistance.

[0036] 3. Insufficient dynamic adaptability: For example, the physiological races of the variant pathogens of rice blast ZB1-ZG4 or the pathogenic strains of R0-R5 strains of rice sheath blight are prone to variation with the number of years of planting. Environmental factors such as temperature and humidity fluctuations caused by global warming are constantly changing. Traditional static databases cannot update resistance association rules in a timely manner, which leads to a decline in their actual value in guiding breeding over time.

[0037] Based on the above issues, according to Figure 1 As shown in the figure, this embodiment provides a method for constructing a database for integrated identification of resistance to rice blast and sheath blight, including the following steps:

[0038] Step 1: Multi-source data acquisition, simultaneously collecting phenotypic, genotypic, environmental, and experimental process data for resistance identification of rice blast and sheath blight;

[0039] The phenotypic data include plant height, leaf color (SPAD value), and biomass (dry weight of aboveground / belowground parts), as well as agronomic traits such as lesion morphology (including spindle-shaped, elliptical, and blurred edges) of rice blast, occurrence time (tillering stage / heading stage), severity, latent period, resistance grading, and lesion height (leaf sheath / leaf), expansion pattern (horizontal / vertical direction), severity, and stem rot degree (1-5) of rice sheath blight; the latent period is the number of days from inoculation to the onset of disease, and the resistance grading is divided into susceptible, intermediate resistant, and highly resistant;

[0040] Genotypic data include whole-genome SNPs (density ≥100K), SSR markers, genotypes of known rice blast resistance genes (such as Pi-ta, Pi9) and sheath blight resistance QTLs (such as qRsg1, qRsg5), such as homozygous, heterozygous and deletion, as well as whole-genome resequencing data;

[0041] Environmental data includes meteorological data, cultivation data, and pathogen data; meteorological data includes daily temperature, relative humidity, rainfall, and sunshine duration; cultivation data includes nitrogen, phosphorus, and potassium application rates, transplanting density, and irrigation methods (moistening or flooding); pathogen data includes physiological races of rice blast fungus and pathogenic lines of rice sheath blight fungus, such as Chinese differential types ZB1-ZG4, and pathogenic lines of rice sheath blight fungus such as R0-R5.

[0042] Experimental data include experimental design type (e.g., randomized block or split-plot design), number of replicates ≥ 3, and plot area ≥ 10m². 2 Rice blast inoculum, rice sheath blight inoculum, inoculation time (divided into tillering stage or booting stage), field management records, and data verification records; data verification records are verified by two people entering and verifying data twice. The spore suspension concentration used for rice blast inoculum is 10. 5 The inoculum size was 10 cells / mL, and the sclerotium inoculum concentration was 10 when inoculating with sheath blight inoculum. 3Each hole contains one plant, and field management records include information on the use of disease prevention agents.

[0043] Step 2: Standardization and preprocessing of multi-source data. Phenotypic data, genotypic data, environmental data, and experimental process data are standardized and preprocessed to generate standardized parameters.

[0044] Specifically, phenotypic data standardization involves segmenting lesion regions in image data using convolutional neural network models such as ResNet-50, extracting lesion area and color features (i.e., HSV spatial mean), and combining this with manual scoring for calibration to generate standardized phenotypic parameters. For example, the severity of rice blast is calculated as "lesion area percentage × 0.7 + latent period × 0.3". Text data uses natural language processing tools such as BERT to extract descriptive phenotypic terms, such as "brown spindle-shaped lesions appear on leaf sheaths", which are then mapped to standardized grading, such as a level 3 severity for rice sheath blight.

[0045] Genotype data standardization specifically involves removing low-quality SNP sites with a deletion rate >10% and a minor allele frequency <0.05, unifying genotype coding according to the pattern of "0" homozygous reference type, "1" heterozygous type, and "2" homozygous variant type, and aligning the disease resistance gene / QTL positions based on the rice reference genome IRGSP-1.0 and marking linkage disequilibrium (LD) intervals.

[0046] The standardization of environmental data specifically involves calculating the cumulative environmental index (CEI). The rice blast stress index is calculated as (average temperature × 0.3 + number of consecutive rainy days × 0.5 + relative humidity × 0.2), and the sheath blight threat index is calculated as (number of days with average daily humidity > 80% × 0.4 + nitrogen fertilizer application rate × 0.3 + planting density × 0.3). Rice blast races are clustered according to the similarity of their pathogenicity spectrum, such as ZB1-ZB3 being highly pathogenic. Sheath blight strains are classified according to their pathogenicity, with R0-R2 being weakly pathogenic and R3-R5 being highly pathogenic.

[0047] The standardization of trial data specifically involves converting the trial design type, number of repetitions, and cell layout into a machine-readable format, such as JSON or XML. Abnormal data is marked as "requires verification," such as when the severity difference between repetitions in the same cell is greater than 50%, and unreliable data that does not record the vaccination time is removed.

[0048] Step 3: Multidimensional feature extraction and association modeling. Based on standardized data, extract key phenotypic, genotypic, and environmental features to construct a collaborative association model.

[0049] Specifically:

[0050] S1. Phenotypic-resistance association feature extraction: phenotypic parameters that are significantly associated with the severity of rice blast / sheath blight are screened through statistical analysis such as Pearson correlation coefficient, such as the incubation period of rice blast at the tillering stage ≤5 days and the vertical expansion and transfer time of sheath blight lesions.

[0051] S2. Genotype-resistance association feature extraction: First, use single-marker analysis or genome-wide association analysis (GWAS) of GLM model to screen for SNP sites that are significantly associated with dual resistance, such as sites near the Pi9 gene. Then, combine gene function annotation to mine new resistance candidate genes, such as the CHI gene encoding chitinase, and gene function annotation such as the NLR disease resistance gene family or transcription factor family.

[0052] S3. Construction of environment-resistance association features: The contribution of environmental factors to resistance is quantified by random forest model. For example, the weight of the number of consecutive rainy days during the heading period accounts for 40% of the resistance to rice blast. The resistance stable zone is divided based on the cumulative environmental index.

[0053] The resistance-stable zone includes the dual-resistance-advantage zone, the rice blast-sensitive zone, and the sheath blight-sensitive zone;

[0054] S4. Construction of a multi-factor synergistic association model: Integrating phenotypic, genotypic, and environmental data to construct a resistance prediction model and a resistance synergistic optimization module;

[0055] The resistance prediction model takes genotype and environmental parameters as input and outputs dual resistance levels with an accuracy of ≥85%. The resistance co-optimization module is used to identify key gene combinations or environmental regulation strategies to enhance dual resistance.

[0056] Step 4: Construct an integrated database. Based on the association model, design a layered database system including a metadata database, a core database, an association database, and an application interface layer to store core data and association rule information.

[0057] The metadata database records information such as data source, collection time, quality control labels, and data format. For example, the data source is obtained from the National Rice Industry Technology System Experimental Station. The quality control labels include "verified" or "pending verification". The data format includes CSV, FASTA, or JSON.

[0058] The core database stores structured data tables including phenotypic tables, genotypic tables, environmental tables, and experimental tables. The phenotypic table includes information such as variety ID, rice blast severity, and sheath blight lesion height. The genotypic table includes information such as variety ID, SNP locus genotype, and QTLqRsg5 genotype. The environmental table includes information such as experimental ID, year, average temperature, and rice blast race. The experimental table includes information such as experimental ID, design type, number of replicates, and inoculum concentration.

[0059] The association database stores multi-dimensional association rules, such as "SNP site with AA genotype → high probability of rice blast resistance ≥90%" and "variety carrying qRsg1 gene has a 50% increased risk of sheath blight in areas with CEI>0.6".

[0060] The application interface layer provides a RESTful API interface, which supports breeding software in querying resistance prediction results and researchers in downloading association rule data.

[0061] Step 5: Dynamic database updates. The database is dynamically updated through real-time data collection, model iteration, resistance marker updates, and user feedback corrections to ensure its timeliness.

[0062] The dynamic update mechanism includes automatic data collection, iterative model optimization, resistance marker updates, and user feedback corrections.

[0063] The data is automatically collected in real time by field temperature and humidity sensors (IoT sensors) and sequencing platforms (automated SNP detection equipment), and a preprocessing process is set to be triggered every morning.

[0064] Model iteration optimization involves collecting new experimental data every quarter, with a data volume of ≥100 varieties, and retraining the association model using methods such as updating the GWAS significance threshold or adjusting the random forest feature weights, while retaining model version records.

[0065] The specific update of resistance markers involves updating the association rule base after reviewing the discovery of new resistance genes (such as the RGA1 gene identified in pan-genome analysis) or pathogen variations (such as the new rice blast race ZG5) and labeling them with the "new discovery" tag.

[0066] User feedback corrections are made through an open user interaction interface. After verification, erroneous data is corrected once a month. The user interaction interface is used to receive feedback from breeding experts, such as information that the predicted resistance of a certain variety does not match the actual resistance.

[0067] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A database construction method for integrated identification of rice blast and sheath blight resistance, characterized by, The method comprises the following steps: Step one, multi-source data collection, synchronously collecting phenotype, genotype, environment and test process data of resistance identification of rice blast and sheath blight; Step two, multi-source data standardization preprocessing, standardizing and preprocessing the phenotype data, genotype data, environment data and test process data to generate standardized parameters; The standardization of the phenotype data specifically comprises segmenting the lesion area through a convolutional neural network model, extracting lesion area and color features, and combining artificial scoring calibration to generate standardized phenotype parameters; the text data is extracted through natural language processing tools to extract descriptive phenotype terms and map them to standardized grades; The standardization of the genotype data specifically comprises removing low-quality SNP sites with a missing rate of more than 10% and a minor allele frequency of less than 0.05, uniformly encoding the genotypes according to the rule of "0" homozygous reference type, "1" heterozygous type and "2" homozygous variant type, and aligning the linkage disequilibrium interval of the resistance genes / QTL positions; The standardization of the environment data specifically comprises calculating the cumulative environment index, and clustering the rice blast races according to the pathogenicity spectrum similarity and the sheath blight strains according to the pathogenicity classification; The standardization of the test process data specifically comprises converting the test design type, the number of repetitions, and the plot layout into a machine-readable format, marking the abnormal data as "to be reviewed", and removing unreliable data without recording the inoculation time; Step three, multi-dimensional feature extraction and correlation modeling, extracting phenotype-genotype-environment key features based on standardized data, and constructing a collaborative correlation model; Step four, integrated database construction, designing a hierarchical architecture database system including a meta-database, a core database, a correlation database and an application interface layer based on the correlation model, and storing core data and correlation rule information; Step five, dynamic database updating, dynamically updating the database through real-time data collection, model iteration, resistance marker updating and user feedback correction to ensure the timeliness of the database.

2. The database construction method for identifying rice blast and sheath blight resistance in one step according to claim 1, wherein: The phenotype data in step one includes plant height, leaf color and biomass, as well as lesion morphology, occurrence period, severity, latent period, resistance classification of rice blast, and lesion height, expansion mode, severity and stem decay degree of sheath blight; The genotype data includes whole-genome SNP, SSR marker, genotype of known rice blast resistance genes and sheath blight resistance QTL, and whole-genome resequencing data; The environment data includes meteorological data, cultivation data and pathogen data; the meteorological data includes daily temperature, relative humidity, rainfall and sunshine duration, the cultivation data includes nitrogen, phosphorus and potassium application amount, transplanting density and irrigation method, and the pathogen data includes rice blast physiological race and sheath blight pathogenic strain; The test process data includes experimental design type, number of repetitions, plot area, rice blast inoculum, sheath blight inoculum, inoculation time, field management record and data verification record.

3. The database construction method for integrated identification of rice blast and sheath blight resistance according to claim 2, characterized in that: The lesion shape of rice blast in the phenotype data includes fusiform, oval and edge blur, the incubation period is the days from inoculation to disease onset, and the resistance classification is divided into susceptible, intermediate resistance and high resistance; the number of repetitions in the process data is ≥3 times, the plot area is ≥10 m 2 , and the data verification record adopts double entry verification by two people.

4. The database construction method for integrated identification of rice blast and sheath blight resistance according to claim 2, characterized in that: When the cumulative environment index is calculated, the rice blast stress index is (average temperature x 0.3 + consecutive rainy days x 0.5 + relative humidity x 0.2); the sheath blight threat index is (number of days with average humidity > 80% x 0.4 + nitrogen fertilizer application amount x 0.3 + planting density x 0.3).

5. The database construction method for integrated identification of rice blast and sheath blight resistance according to claim 1, characterized in that: The multi-dimensional feature extraction and correlation modeling in the third step is specifically as follows: S1, phenotype-resistance correlation feature extraction, significant phenotype parameters related to rice blast / dryback disease severity are screened through statistical analysis; S2, genotype-resistance correlation feature extraction, SNP sites significantly associated with double resistance are screened through single marker analysis or whole genome association analysis, and then new resistance candidate genes are mined by combining gene function annotation; S3, environment-resistance correlation feature construction, the contribution of environmental factors to resistance is quantified through a random forest model, and resistance stability zones are divided based on cumulative environmental index; S4, multi-factor synergistic correlation model construction, phenotype, genotype and environment data are integrated to construct resistance prediction model and resistance synergistic optimization module.

6. The database construction method for integrated identification of rice blast and sheath blight resistance according to claim 5, characterized in that: The resistance stability zones divided in S3 include double resistance dominant zone, rice blast sensitive zone and dryback disease sensitive zone.

7. The database construction method for integrated identification of rice blast and sheath blight resistance according to claim 1, wherein: In the fourth step, the metadata database records data sources, collection time, quality control labels and data format information; the core database stores structured data tables including phenotype table, genotype table, environment table and test table; the correlation database stores multi-dimensional correlation rules; and the application interface layer provides RESTful API interface.

8. The database construction method for integrated identification of rice blast and sheath blight resistance according to claim 1, wherein: The dynamic updating mechanism in the fifth step includes data automatic collection, model iterative optimization, resistance marker updating and user feedback correction.

9. The database construction method for identifying rice blast and sheath blight resistance in one step according to claim 8, wherein: The data automatic collection accesses new data in real time through field temperature and humidity sensors and sequencing platforms, and sets the pre-processing process to be triggered at dawn every day; The model iterative optimization re-trains the correlation model by collecting new experimental data every quarter, and retains model version records; the resistance marker updating specifically updates the correlation rule library and labels "newly discovered” tag after the new resistance genes or pathogen variations are found and audited; and the user feedback correction corrects the error data after verification through the open user interaction interface, and the correction frequency is once a month.

Images