A biological-environmental multidimensional information data model for crop regional experiments

By constructing a multidimensional information data model for crop regional experiments, combining bioinformatics and regional experiment data, the problem of incomplete analysis of crop genotype, environment and interaction effects in the existing technology is solved, and the efficiency of regional experiments is improved. The accuracy is recommended, and the breeding and planting decisions of rice varieties are optimized.

CN118262786BActive Publication Date: 2024-12-03HUAZHONG AGRI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410358688.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-03-27
Publication Date
2024-12-03
Estimated Expiration
2044-03-27

AI Technical Summary

Technical Problem

The prior art is difficult to effectively combine bioinformatics methods with regional experiment data, resulting in insufficient analysis of the impact of crop genotype, environment and its interaction effects on crop traits, affecting the efficiency and accuracy of regional experiments.

Method used

Construct a multidimensional information data model for crop regional experiments, load and process regional experiment data, combine the analysis of whole genome biological information data, perform data interpolation and prediction, generate complete phenotypic data, and use environmental index to predict new environments The crop phenotype data below is recommended.

Benefits of technology

It improves the efficiency and accuracy of regional experiments, provides data support for multiple verifications, and has multiple applicability and scalability at the theoretical and application levels, helps to select suitable candidate variety test points, and optimizes the breeding and planting decisions of rice varieties.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118262786B_ABST
    Figure CN118262786B_ABST
Patent Text Reader

Abstract

The present invention discloses a crop regional test biological-environmental multidimensional information data model, which integrates whole genome biological information and environmental-phenotypic data to focus on the ideal test points of candidate varieties in crop regional tests, thereby combining bioinformatics methods with regional test data to characterize the effects of genotype, environment and their interaction effects on crop traits, thereby optimizing and improving the efficiency and accuracy of crop tests.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of computational biology technology, and in particular to a crop regional test biological-environmental multidimensional information data model and a research framework thereof that can promote the selection of ideal test sites for candidate varieties in regional tests. Background Art

[0002] Breeding high-quality and high-yield rice varieties is one of the main ways to increase rice yield. Multi-environment trials are designed to evaluate the adaptability and stability of untested candidate varieties under multiple environmental conditions, and to provide farmers and growers with a basis for planting decisions. Regional trials often need to consider multiple factors, such as varieties, fertilizers, pesticides, irrigation, etc. In order to reduce errors caused by interference factors, multiple repeated experiments are often required. However, due to the inconsistency of the environments adapted to different candidate varieties, the phenotypic data matrix (variety × environment) generated by regional trials may be incomplete. Therefore, it is necessary to obtain complete regional trial data, further analyze the data reasonably, select appropriate test sites for the cultivation of rice varieties in different planting areas, improve the efficiency and accuracy of regional trials, and save resources and time costs. Selecting a suitable statistical model is an important factor in comprehensively evaluating the quality of regional trials. The regional trials of crop varieties were comprehensively evaluated by using methods such as variance analysis model, linear regression model, additive main effect and multiplicative interaction model, cluster analysis and grey correlation analysis. At present, regional trial data are mainly used to compare and evaluate the advantages and disadvantages of participating varieties. These data include different crop types, different trial levels, and the diversity and richness of trial locations, trial times, and trial varieties. The rational application of these trial data to the study of genotype-environment interactions will help understand the response of genotypes to natural environmental conditions and provide insights for evaluating the pilot selection of candidate varieties in different environments.

[0003] On the other hand, existing technologies use bioinformatics tools and techniques, such as genomics, transcriptomics, and proteomics, to analyze crop genotype and phenotypic data, and combine data on environmental factors (such as rainfall, temperature, light, etc.) to understand the genetic characteristics, growth and development processes, and responses to the environment of crops. This comprehensive analysis helps to discover key genes and traits that affect crop yield, quality, and adaptability, and provide a scientific basis and technical support for crop breeding and agricultural production.

[0004] In summary, there is an urgent need to build a research framework and data model that can combine bioinformatics methods and regional trial data to characterize the impact of genotype, environment and their interaction effects on crop traits. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a crop regional test biological-environmental multidimensional information data model, thereby combining bioinformatics methods with regional test data to characterize the effects of genotype, environment and their interaction effects on crop traits, thereby optimizing and improving the efficiency and accuracy of regional tests.

[0006] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows.

[0007] A biological-environmental multidimensional information data model for crop regional trials, which integrates whole genome biological information and environmental-phenotypic data to focus on the ideal test sites of candidate varieties in crop regional trials.

[0008] As a preferred technical solution of the present invention, at the front end of the data model, after loading the acquired regional test data set and performing underlying data processing, a whole genome bioinformatics data analysis framework is embedded and the regional test data with missing phenotypes in the existing regional test data set are interpolated to obtain complete phenotypic data.

[0009] As a preferred technical solution of the present invention, the regional test data set is loaded with a phenotypic data file as a carrier, and the following data process is set: the kinship matrix is ​​read and converted into a symmetric matrix, the row and column names in the symmetric matrix are all genotype data, the environment, genotype, yield and other related data are sorted and converted into a wide format data frame spread_data so that each column represents the yield data under one environment; the underlying data processing includes generating a training set and a test set, and the training set and the test set are generated by repeating the loop n times, and in each loop, a certain proportion of the data of each environment is randomly selected as the test set and the rest is used as the training set; at the same time, the position and true value in each generated test set are recorded for standby use.

[0010] As a preferred technical solution of the present invention, based on a complete crop phenotypic data set, the data model constructs real potency data pointing to crop phenotypes from key environmental information parameters, and provides qualitative or quantitative data attention to ideal test points for candidate varieties in crop regional trials.

[0011] As a preferred technical solution of the present invention, at the back end of the data model, existing crop planting data and / or newly built field trials in new environments are combined and test data is collected to verify the accuracy of phenotypic prediction of tested genotypes in untested environments in combination with environmental information.

[0012] As a preferred technical solution of the present invention, in the data link of phenotypic prediction accuracy verification, the data model performs multi-dimensional data comparison and verification based on single phenotype or multi-phenotype data within a single time period or multiple time periods of the crop.

[0013] As a preferred technical solution of the present invention, at the application end of the data model, the crop phenotypic data under the new environment is output using the environmental index as a bridge, and the ecological types of the crop varieties are divided based on the obtained data output, thereby making data recommendations for ideal test points for candidate varieties in crop regional trials.

[0014] As a preferred technical solution of the present invention, in the ecological type classification link of crop varieties, the data model also includes a visualization component, which generates a network diagram based on the ecological zone division results, uses the size of the points to represent the frequency of the pilot sites being divided into ecological zones, and uses the thickness of the lines to represent the similarity of the ecological types of the pilot sites.

[0015] As a preferred technical solution of the present invention, a data structure example of the data model includes the following specific data processes:

[0016] A. Data loading and its underlying normalization: Load the phenotypic data file based on the existing experimental data set, read the kinship matrix and convert it into a symmetric matrix A_matrix, with the row and column names being genotypes; then select and process the environment Environment, genotype Genotype, and yield GY related data from the County_note_abb data frame, and filter out specific environmental data; organize the data and convert it into a wide-format data frame spread_data, with each column representing the yield data in one environment; then generate training sets and test sets, and repeat the loop 10 times to generate training sets and test sets; in each loop, randomly select 50% of the data for each environment as the test set, set this part of the data to NA, and use the remaining 50% as the training set; record the NA position and true value in each generated test set for the estimation of the subsequent model prediction accuracy;

[0017] B. Embedding of the multi-module whole genome bioinformatics data analysis framework: specifically including: loading the BGGE package and related data; setting the working directory and the intermediate variable middle_par; reading the genotype data file GAPIT.Genotype.Numerical.txt; using the BGGE method to predict the yield data in the test set; saving the prediction results; also including: loading the MegaLMM package and related data; setting the working directory and the intermediate variable middle_par_6; reading the kinship matrix; using the MegaLMM method to predict the yield data in the test set; saving the prediction results; also including: loading the rrBLUP package and related data; reading the kinship matrix; using the rrBLUP method to predict the yield data in the test set; saving the prediction results; also including: for each generated test set, calculating the average yield of each environment and adding it to the data frame; using the impute_lm function in the simputation package to interpolate the yield data through a linear model; saving the interpolation results;

[0018] C. Data interpolation: For each test set generated, calculate the average yield of each environment and add it to the data frame; use the impute_lm function in the simulation package to interpolate the yield data through a linear model; save the interpolation results; evaluate the prediction accuracy of the multi-module whole genome bioinformatics data analysis framework, and calculate the average of ten prediction accuracies; select the whole genome bioinformatics data analysis framework module with high prediction ability in predicting phenotypic data with severe missing, and use it to interpolate the phenotypic data of yield, plant height and flowering period under multiple environments to obtain a complete regional test data set;

[0019] D. Identify key environmental factors and conduct accuracy assessment: specify the working directory, planting information file planting.txt, phenotypic data file phenotype_GYPHFT.txt, phenotype GY to be analyzed, number of environmental data days t to download, and multiple target sub-functions; embed CERIS.r to download information on environmental factors from 1 to t days after rice planting; use the Perl script mergeEnvironmentalIndexRank.pl to merge the ranking results of the correlation between environmental index and phenotypic mean output by multiple sub-functions; then read the environmental index file bestEn with an absolute correlation value greater than 0.5 vIndex-FT.txt; load a total of k environmental data from multiple environments and newly created environmental data, and merge them to obtain environmental data envk_weatherdata; filter environmental factor data with correlation greater than a specific value according to the name of the environmental factor; calculate the average value of each environmental factor in each environment to form the env_factors_means_df data frame; read the flowering period phenotype data of k environments and merge them into lm_input; use a linear model to fit each environmental index with the flowering period phenotype, and predict the specified crop parameters of the newly created environment, such as flowering period; calculate the correlation between the predicted value and the observed value, and generate a correlation chart;

[0020] E. Predict the rice yield data for x years and conduct ecological type analysis to realize the data focus of the test sites: Download the meteorological data of x years based on CERIS, use the key environmental indices related to yield and the phenotypic information of the tested genotypes in multiple environments to predict the yield data in untested environments; then use the R language GGEBiplotGUI package to analyze the yield data of multiple rice varieties in different provinces and regions for x years to obtain the ecological zone division results for each year; and build or load visualization components to visualize the ecological zone division results for x years and generate a network diagram. The size of the points indicates the frequency with which the pilot is divided into ecological zones, and the thickness of the lines indicates the similarity of the ecological types of the pilots.

[0021] As a preferred technical solution of the present invention, the crop is rice.

[0022] The beneficial effects of adopting the above technical solution are:

[0023] The present invention develops a data model based on environmental information and complete regional test data and performs multiple verifications, providing data support for the selection of ideal test sites for candidate varieties in different environments, and has multiple applicability and scalability at the theoretical and application levels.

[0024] The present invention combines bioinformatics methods with regional test data to construct a research framework and data model that can combine bioinformatics methods with regional test data to characterize the effects of genotype, environment and their interaction effects on crop traits, which can effectively optimize and improve the efficiency and accuracy of regional tests. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is the overall flow chart of the analysis method recommended for the rice regional trial pilot based on environmental information.

[0026] Figure 2 This is the prediction accuracy evaluation process and result diagram of the whole genome selection model.

[0027] Figure 3 It is a prediction accuracy framework diagram of the model built with environmental index as the bridge.

[0028] Figure 4 This is a macro-environment classification result chart based on 39 years of rice yield data. DETAILED DESCRIPTION

[0029] The present invention is described in detail in the following examples. All raw materials and equipment used in the present invention are conventional commercial products and can be directly obtained through market purchase.

[0030] In the description of the following embodiments, specific details such as specific system structures, technologies, etc. are provided for illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0031] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or their collections. It should also be understood that the term "and / or" used in the present specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0032] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0033] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0034] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0035] Example 1: Technical model framework

[0036] In general, the present invention constructs a new data model research framework to promote the selection of ideal test sites for candidate varieties in regional trials. In summary, the present invention first evaluates the prediction accuracy of the whole genome selection method, selects the method with the highest accuracy to interpolate the regional test data with missing phenotypes to obtain complete phenotypic data; then predicts the phenotype of rice based on the identified key environmental indexes; finally, analyzes the predicted phenotypic data to qualitatively or quantitatively promote the selection of ideal test sites for candidate varieties in regional trials. In the prediction accuracy evaluation of the whole genome selection method, the prediction accuracy of BGGE for the complex trait of rice yield was 0.73, and the prediction accuracy for plant height and flowering period was 0.84 and 0.87 respectively; in the variety evaluation, high-yield and stable varieties were planted every year, and the average annual gains in 2014, 2015 and 2016 were 2.77%, 3.72% and 2.12% respectively; for the selection of candidate variety test sites, the ecological type of Guangdong Province in the rice-growing area of ​​the middle and lower reaches of the Yangtze River is specific, and the ecological types of Fujian Province and Anhui Province are similar to those of surrounding provinces.

[0037] The research of this invention provides ideas for the genetic improvement of agronomic traits of rice varieties in different rice-growing areas, provides insights for the ideal pilot selection of candidate varieties in different environments, and has certain universality and scalability.

[0038] Example 2: Leading Data Steps of Data Model

[0039] The dominant data steps of the data model of the present invention are summarized as follows.

[0040] In terms of initial data set acquisition, regional test data from the rice-growing areas in the middle and lower reaches of the Yangtze River were collected. It mainly includes yield-related data of late indica rice varieties in 36 pilots in 8 provinces, including Guangdong, Hubei, Anhui, Hunan, Fujian, Guangxi Zhuang Autonomous Region and Zhejiang. The test data covers a variety of different environmental types and is representative, which can truly reflect the environmental conditions and planting methods in the region. Taking this as basic data, the invention results can be more applicable when analyzing regional test data of different crops. Multiple late indica rice varieties have been tested in the rice-growing areas in the middle and lower reaches of the Yangtze River for many years. Phenotypic data of rice yield, plant height, flowering period, growth period, panicle length, thousand-grain weight, grain length and other traits are collected every year. The province and year are combined as one environment, with a total of 19 environments. However, due to the limitation of the test area and the limitation of the test period, the regional test data has a small sample size. The observation data of each environment have more than 50% missing values, and the highest missing rate reaches 90%.

[0041] The phenotypic data were interpolated using biological algorithms to obtain a complete phenotypic data set. The prediction accuracy of the whole genome selection model was evaluated, mainly including four methods: BGGE (Bayesian Genomic Genotype × Environment Interaction), MegaLMM (Mega-scale linear mixed models), rrBLUP (Ridge regression best linear unbiased prediction) and JRM (Joint regression method). These four methods have different characteristics in theory and practice, and corresponding open source software packages are available. They have been widely used and studied in whole genome selection and related fields. The yield, plant height and flowering period data were divided into training sets and test sets in a ratio of 5:5 to meet the subsequent accuracy requirements for interpolation of phenotypic data with a missing rate of more than 50%.

[0042] Based on the Pearson correlation coefficient between the observed and predicted values, the method with the highest prediction accuracy in the three traits was selected to interpolate the regional test data with missing phenotypes to obtain a complete test data set. The prediction accuracy of BGGE for the complex trait of rice yield was 0.73, and the prediction accuracy for plant height and flowering period was 0.84 and 0.87 respectively. Next, based on the BGGE method, complete phenotypic data of late indica rice yield, plant height and flowering period were obtained in 19 environments.

[0043] The meteorological data of the past 39 years (1990-2016) were downloaded from the NASA POWER (https: / / power.larc.nasa.gov / ) database using the R package nasa power. The meteorological data covers five major environmental information categories: daily light, temperature, rainfall, air, and soil conditions during the growth and development of rice, including a total of 25 environmental factors (maximum temperature, minimum temperature, relative humidity, rainfall, day length, light and effective radiation, net photosynthetically active radiation, ultraviolet A, ultraviolet B, surface pressure, wind speed, wind direction, soil humidity, soil moisture, and some composite factors based on the above environmental factors, DTR = maximum temperature-minimum temperature, TSR = maximum temperature 2 -lowest temperature 2 , MMR = minimum temperature / maximum temperature, GDD = (maximum temperature + minimum temperature) / 2-basal temperature of growth, dGDD = |GDD(day_n)-GDD(day_n-1)|, PTT = GDD×day length, dPTT = |PTT (day_n)-PTT(day_n-1)|, PTR = GDD / day length, PTD1 = DTR×day length, PTD2 = DTR / day length, PRDTR = PR / DTR).

[0044] An environmental index (a combination of environmental factors and growth windows) was constructed using 25 environmental factors. Growth windows are formed at different stages during plant development, and windows of 10 to 50 were selected as the growth days that are meaningful to rice growth. The average values ​​of different environmental factors with different starting days and a span of at least 10 days were calculated for each environment. The phenotypic mean under each environment was fitted with the mean of the environmental index, and the size of the correlation value was compared to identify the environmental index that can represent the phenotypic mean; based on the key environmental indices that have a significant impact on the phenotype, the phenotype of yield-related traits of tested rice varieties in untested environments was predicted in combination with joint regression analysis.

[0045] Furthermore, in order to evaluate the accuracy of phenotypic prediction of tested genotypes in untested environments by combining environmental information, we conducted a field experiment in Hubei in 2018 (HB_2018) in the new environment to collect rice flowering date data. The flowering date of rice in 2018 was predicted using the complete flowering date data of 2014, 2015 and 2016 combined with the identified key environmental indices, and the correlation between the predicted and observed values ​​was compared to evaluate the prediction accuracy of the model. The predicted value of flowering date had a moderate correlation of 0.55 with the observed value, which can be explained by the correlation of flowering date in HB_2018 with flowering date in other 19 environments ranging from 0.43 to 0.64. Furthermore, after evaluating the prediction accuracy of the model, the rice yield in the Yangtze River production area in the past 30 years was predicted by combining the meteorological data of the rice-growing areas in the middle and lower reaches of the Yangtze River for the past 39 years and the complete phenotypic data of 2014, 2015 and 2016.

[0046] The yield data of each year were analyzed using the genotype plus genotype-by-environment interaction biplot. The "Which wins where" functional map in the GGE biplot analysis makes a comprehensive evaluation of the characteristics of the test site while considering the interaction between genotype and environment, providing a basis for the classification of varieties into corresponding ecological zones. This method can divide multiple planting environments into different ecological types. The macro-environment classification results based on 39 years of rice yield data were obtained. The classification results were visualized using Cytoscape software. We can see that the ecological type of Guangdong Province in the rice-growing area of ​​the middle and lower reaches of the Yangtze River is specific, and the ecological types of Fujian Province and Anhui Province are similar to those of surrounding provinces. In other words, for the selection of variety testing pilots, Guangdong should be selected as a mandatory pilot under limited budget; for variety breeding, varieties bred in Fujian and Anhui can be promoted in surrounding pilots.

[0047] Embodiment 3: A concrete example of a data model

[0048] Based on the above embodiment, a specific example of the data model of the present invention includes:

[0049] A. Data loading and its underlying normalization: Load the phenotypic data file based on the existing experimental data set, read the kinship matrix and convert it into a symmetric matrix A_matrix, with the row and column names being genotypes; then select and process the environment Environment, genotype Genotype, and yield GY related data from the County_note_abb data frame, and filter out specific environmental data; organize the data and convert it into a wide-format data frame spread_data, with each column representing the yield data in one environment; then generate training sets and test sets, and repeat the loop 10 times to generate training sets and test sets; in each loop, randomly select 50% of the data for each environment as the test set, set this part of the data to NA, and use the remaining 50% as the training set; record the NA position and true value in each generated test set for the estimation of the subsequent model prediction accuracy;

[0050] B. Embedding of the multi-module whole genome bioinformatics data analysis framework: specifically including: loading the BGGE package and related data; setting the working directory and the intermediate variable middle_par; reading the genotype data file GAPIT.Genotype.Numerical.txt; using the BGGE method to predict the yield data in the test set; saving the prediction results; also including: loading the MegaLMM package and related data; setting the working directory and the intermediate variable middle_par_6; reading the kinship matrix; using the MegaLMM method to predict the yield data in the test set; saving the prediction results; also including: loading the rrBLUP package and related data; reading the kinship matrix; using the rrBLUP method to predict the yield data in the test set; saving the prediction results; also including: for each generated test set, calculating the average yield of each environment and adding it to the data frame; using the impute_lm function in the simputation package to interpolate the yield data through a linear model; saving the interpolation results;

[0051] C. Data interpolation: For each test set generated, calculate the average yield of each environment and add it to the data frame; use the impute_lm function in the simulation package to interpolate the yield data through a linear model; save the interpolation results; evaluate the prediction accuracy of the multi-module whole genome bioinformatics data analysis framework, and calculate the average of ten prediction accuracies; select the whole genome bioinformatics data analysis framework module with high prediction ability in predicting phenotypic data with severe missing, and use it to interpolate the phenotypic data of yield, plant height and flowering period under multiple environments to obtain a complete regional test data set;

[0052] D. Identify key environmental factors and conduct accuracy assessment: specify the working directory, planting information file planting.txt, phenotypic data file phenotype_GYPHFT.txt, phenotype GY to be analyzed, number of environmental data days t to download, and multiple target sub-functions; embed CERIS.r to download information on environmental factors from 1 to t days after rice planting; use the Perl script mergeEnvironmentalIndexRank.pl to merge the ranking results of the correlation between environmental index and phenotypic mean output by multiple sub-functions; then read the environmental index file bestEn with an absolute correlation value greater than 0.5 vIndex-FT.txt; load a total of k environmental data from multiple environments and newly created environmental data, and merge them to obtain environmental data envk_weatherdata; filter environmental factor data with correlation greater than a specific value according to the name of the environmental factor; calculate the average value of each environmental factor in each environment to form the env_factors_means_df data frame; read the flowering period phenotype data of k environments and merge them into lm_input; use a linear model to fit each environmental index with the flowering period phenotype, and predict the specified crop parameters of the newly created environment, such as flowering period; calculate the correlation between the predicted value and the observed value, and generate a correlation chart;

[0053] E. Predict the rice yield data for x years and conduct ecological type analysis to realize the data focus of the test sites: Download the meteorological data of x years based on CERIS, use the key environmental indices related to yield and the phenotypic information of the tested genotypes in multiple environments to predict the yield data in untested environments; then use the R language GGEBiplotGUI package to analyze the yield data of multiple rice varieties in different provinces and regions for x years to obtain the ecological zone division results for each year; and build or load visualization components to visualize the ecological zone division results for x years and generate a network diagram. The size of the points indicates the frequency with which the pilot is divided into ecological zones, and the thickness of the lines indicates the similarity of the ecological types of the pilots.

[0054] Example 4: Procedural construction of data model instances and their code implementation

[0055] Based on Example 3, see the attached Figure 1 For a specific example of the data model of the present invention, the data construction and code implementation process includes:

[0056] 1. Collecting regional test data in the rice-growing areas of the middle and lower reaches of the Yangtze River

[0057] In order to obtain phenotypic data of rice varieties under multiple environments, it is necessary to collect rice regional trial data. The regional trial data of rice in the middle and lower reaches of the Yangtze River in 2014, 2015 and 2016 include 191 late indica rice varieties, which were planted in 36 pilot sites in 8 provinces including Guangdong, Hubei, Anhui, Hunan, Fujian, Guangxi Zhuang Autonomous Region and Zhejiang. The trial data include rich information such as growth status, pest and disease occurrence, yield and quality during the growth period. The province and year are combined as one environment, with a total of 19 environments; not all varieties were tested in every environment, and less than 50% of the phenotypic data were observed in a single environment.

[0058] 2. Evaluation of the prediction accuracy of whole genome selection methods

[0059] In order to obtain high-quality and complete phenotypic data under different environments, the prediction accuracy of four methods, including JRM, rrBLUP ("rrBLUP" package), BGGE ("BGGE" package) and MegaLMM ("MegaLMM" package), was evaluated using R language.

[0060] For each trait, 50% of the observations in each environment are randomly selected and masked as the test set, and the remaining 50% are used as the training set. When dividing the training set and the test set, this step is repeated ten times, and the mean of the ten prediction accuracies is used as the final evaluation indicator of the model. By repeatedly randomly dividing the training set and the test set for model evaluation, the performance of the model can be better evaluated and more reliable results can be provided, thereby helping to select the most suitable prediction method. The specific details are (taking rice yield data as an example):

[0061] #Generate a test set train_data_gather_2env_list that masks 50% of the phenotypic data. This variable is used in the following four methods.

[0062] load("D: / desktop / recent / yield220808.RData")#Load phenotypic data

[0063] A_matrix<-

[0064] read.table(

[0065] "D: / desktop / Rhistory files / kinship_191.txt",

[0066] header = FALSE,

[0067] sep = "\t",

[0068] skip = 3,

[0069] row.names = 1 )

[0071] genotype<- rownames(A_matrix)

[0072] colnames(A_matrix)<- rownames(A_matrix)

[0073] yield_data<- County_note_abb %>%

[0074] mutate(Environment = paste(Prov_abb, TrialYear, sep = '_')) %>%

[0075] select(Environment, Genotype, GY) %>%

[0076] filter(Environment != "ZJ_2015")

[0077] spread_data<- yield_data %>%

[0078] group_by(Environment, Genotype) %>%

[0079] summarise(GY = mean(GY, na.rm = T)) %>%

[0080] spread(key = Environment, value = GY) %>%

[0081] as.data.frame()

[0082] pheno_names<- colnames(spread_data)[-1]

[0083] GY_df<- spread_data %>%

[0084] select(-Genotype) %>%

[0085] 'rownames<-'(genotype)

[0086] train_df_list<- list()

[0087] NA_position_list_10<- list()

[0088] true_value_list_10<- list()

[0089] NA_position_df_list10<- list()

[0090] for (j in 1:10) {

[0091] train_df<- as.data.frame(matrix(nrow = 191, ncol = 19))

[0092] rownames(train_df)<- genotype

[0093] colnames(train_df)<- pheno_names

[0094] NA_position_list<- list()

[0095] true_value_list<- list()

[0096] NA_position_df_list<- list()

[0097] for (i in 1:19) {

[0098] middle_df<- GY_df %>%

[0099] select(pheno_names[i])

[0100] True_position<- apply(middle_df, 2, function(x)

[0101] which(!is.na(x))) %>%

[0102] as.data.frame()

[0103] NA_scale<- floor(length(True_position[, 1]) / 2)

[0104] set.seed(j)

[0105] NA_position<- sample(True_position[, 1], NA_scale)

[0106] Genotype<- genotype[NA_position]

[0107] NA_position_df<- data.frame(Genotype, NA_position)

[0108] NA_position_df_list[[i]]<- NA_position_df

[0109] NA_position_list[[pheno_names[i]]]<- NA_position

[0110] true_value<- middle_df[, 1][NA_position]

[0111] true_value_list[[pheno_names[i]]]<- true_value

[0112] train_data<- middle_df

[0113] train_data[, 1][NA_position]<- NA

[0114] train_df[, i]<- train_data

[0115] }

[0116] NA_position_df_list10[[j]]<- NA_position_df_list

[0117] NA_position_list_10[[j]]<- NA_position_list

[0118] true_value_list_10[[j]]<- true_value_list

[0119] train_df_list[[j]]<- train_df

[0120] }

[0121] train_data_gather_2env_list<- list()

[0122] geno_1_env_list<- list()

[0123] for (i in 1:10) {

[0124] train_data<- train_df_list[[i]] %>%

[0125] 'rownames<-'(1:191)

[0126] train_data$Genotype<- genotype

[0127] train_data<- train_data %>%

[0128] select(Genotype, everything())

[0129] train_data_gather<- train_data %>%

[0130] gather(Environment, GY, AH_2015:ZJ_2016)

[0131] multi_env_data<- train_data_gather

[0132] geno_no_NA<- na.omit(train_data_gather)

[0133] geno_no_NA_count<- geno_no_NA %>%

[0134] group_by(Genotype) %>%

[0135] count(sort = TRUE)

[0136] geno_1_env<- geno_no_NA_count %>%

[0137] filter(n == 1) %>%

[0138] as.data.frame()

[0139] geno_1_env<- geno_1_env[, 1]

[0140] geno_1_env_list[[i]]<- geno_1_env

[0141] NA_position_vec<- c()

[0142] for (j in 1:length(geno_1_env)) {

[0143] geno_1_env_position<-

[0144] which(multi_env_data$Genotype == geno_1_env[j]&

[0145] !is.na(multi_env_data$GY))

[0146] NA_position_vec<- c(NA_position_vec, geno_1_env_position)

[0147] }

[0148] multi_env_data$GY[NA_position_vec]<- NA

[0149] multi_env_data<- multi_env_data %>%

[0150] select(Environment, Genotype, GY)

[0151] train_data_gather_2env_list[[i]]<- multi_env_data

[0152] }

[0153] Use the BGGE method to predict the phenotypic value of rice yield in the test set (taking the division into training set and test set as an example).

[0154] load(" / public / home / hpei / BGGE / BGGE_08_24 / yield220808.RData")

[0155] setwd("D: / Documents")

[0156] library(tidyverse)

[0157] library(BGGE)

[0158] middle_par<- 1 #1:10

[0159] input_file<- file(" / public / home / hpei / BGGE / BGGE_08_24 / GAPIT.Genotype.Numerical.txt", "r")

[0160] line<- readLines(input_file, n = 1)

[0161] line_vector<- strsplit(line, split = '\t')[[1]]

[0162] df<- data.frame(matrix(nrow = 1,ncol =length(line_vector) ))

[0163] for (i in 1:length(line_vector)) {

[0164] df[1, i]<- line_vector[i]

[0165] }

[0166] line<- readLines(input_file, n = 1)

[0167] while(length(line)==1){

[0168] line_vector<- strsplit(line, split = '\t')[[1]]

[0169] df<- rbind(df,line_vector)

[0170] line<- readLines(input_file, n = 1)

[0171] }

[0172] close(input_file)

[0173] rownames(df)<-df[,1]

[0174] colnames(df)<- df[1,]

[0175] geno<- df[-1,-1]

[0176] train_data<- train_data_gather_2env_list[[middle_par]]

[0177] ne<- as.vector(table(train_data$Environment))

[0178] print(ne)

[0179] K<- getK(Y = train_data, X = geno, kernel = "GK", bandwidth = c(0.25,1,2.5), model = "MDs")

[0180] print(K)

[0181] fit<- BGGE(y = train_data$GY, K = K, ne = ne)

[0182] print(fit)

[0183] save(fit,file=paste(" / public / home / hpei / BGGE / BGGE_megalmm_rep10_220909_env2 / BGGE_megalmm_rep10_0.5NA_220913_env2 / BGGE_pre_env2_rep10_0.5NA",middle_par,".RData",sep=""))

[0184] Use the MegaLMM method to predict the phenotypic value of rice yield in the test set (taking the division into training set and test set as an example). Download the source code of MegaLMM at https: / / github.com / deruncie / MegaLMM / tree / v0.9.3.

[0185] # Load feature values ​​(combination of province phenotype and year)

[0186] load("D: / desktop / Rhistory files / 220531-megalmm-zhang / avg_data.RData")

[0187] E_number<- phenotype_pivot_wide_filtered %>%

[0188] Gather(key = Trait, value = Value, -Genotype) %>%

[0189] na.omit() %>%

[0190] Group_by(Trait) %>%

[0191] Summarise(total = n()) %>%

[0192] Arrange(-total)

[0193] E_number_trait <- E_number %>%

[0194] Mutate(Multi_rait = sapply(strsplit(Trait, split='_', fixed=TRUE), function(x)(x[2])))

[0195] pheno_name_middle <- E_number_trait %>%

[0196] Filter(Multi_rait == "GY") %>% as.data.frame()

[0197] pheno_name_middle <- pheno_name_middle[,1]

[0198] pheno_name_middle <- sort(pheno_name_middle)

[0199] #Megalmm input data format

[0200] megalmm_df_list <- list()

[0201] For(i in 1:10){

[0202] MgeaLMM_data <- train_data_gather_2env_list[[i]]

[0203] MgeaLMM_data <- MgeaLMM_data %>%

[0204] spread(key = Environment,value=GY)%>%

[0205] set_names(c("Genotype",pheno_name_middle))

[0206] megalmm_df<- phenotype_pivot_wide_filtered

[0207] MgeaLMM_data<- MgeaLMM_data%>%

[0208] select(-Genotype)

[0209] megalmm_df[,colnames(MgeaLMM_data)]<- MgeaLMM_data

[0210] megalmm_df<-megalmm_df %>%

[0211] select(-Genotype)%>%

[0212] 'rownames<-'(genotype)

[0213] megalmm_df_list[[i]]<- megalmm_df

[0214] }

[0215] middle_par_6<- 1 #1:10

[0216] setwd(" / public / home / hpei / BGGE / megalmm_rep10_220922_env2_half_NA_fromshao / megalmm_input_73trait_0.5NA_shao / megalmm1")

[0217] library(tidyverse)

[0218] library(MegaLMM)

[0219] set_MegaLMM_nthreads(RcppParallel::defaultNumThreads()*0.1)

[0220] A_matrix<-

[0221] read.table(

[0222] " / public / home / hpei / BGGE / BGGE_08_24 / MegaLMM_220901 / kinship_191.txt",

[0223] header = FALSE,

[0224] sep = "\t",

[0225] skip = 3,

[0226] row.names = 1 )

[0228] genotype<- rownames(A_matrix)

[0229] colnames(A_matrix)<- rownames(A_matrix)

[0230] train_data<- megalmm_df_list[[middle_par_6]]

[0231] h2_d<- 20

[0232] K_value<- round(dim(train_data)[2] / 2)

[0233] run_parameters = MegaLMM_control(

[0234] drop0_tol = 1e-10,

[0235] scale_Y = T,

[0236] h2_divisions = h2_d,

[0237] h2_step_size = NULL,

[0238] burn = 00,

[0239] thin = 3,

[0240] K = K_value )

[0242] priors = MegaLMM_priors(

[0243] tot_Y_var = list(V = 0.5, nu = 10),

[0244] tot_F_var = list(V = 18 / 20, nu = 20),

[0245] Lambda_prior = list(

[0246] sampler = sample_Lambda_prec_horseshoe,

[0247] prop_0 = 0.1,

[0248] delta = list(shape = 3, scale = 1),

[0249] delta_iterations_factor = 300

[0250] ),

[0251] h2_priors_resids_fun = function(h2s, n)

[0252] 1,

[0253] h2_priors_factors_fun = function(h2s, n) 1 )

[0256] MegaLMM_state = setup_model_MegaLMM(

[0257] train_data,

[0258] ~ 1 + (1 | Genotype),

[0259] data = data.frame(Genotype = genotype),

[0260] relmat = list(Genotype = A_matrix),

[0261] run_parameters = run_parameters, )

[0263] maps = make_Missing_data_map(MegaLMM_state,

[0264] max_NA_groups = 50,

[0265] verbose = F)

[0266] MegaLMM_state = set_Missing_data_map(MegaLMM_state, maps$Missing_data_map_list[[4]])

[0267] MegaLMM_state = set_priors_MegaLMM(MegaLMM_state, priors)

[0268] MegaLMM_state = initialize_variables_MegaLMM(MegaLMM_state)

[0269] MegaLMM_state = initialize_MegaLMM(MegaLMM_state)

[0270] MegaLMM_state = clear_Posterior(MegaLMM_state)

[0271] MegaLMM_state$Posterior$posteriorSample_params = c('Lambda')

[0272] MegaLMM_state$Posterior$posteriorFunctions = list(pred = 'U_R[,1] +U_F %*% Lambda[,1]')

[0273] MegaLMM_state = clear_Posterior(MegaLMM_state)

[0274] n_iter = 20

[0275] for (i in 1:20) {

[0276] print(sprintf('Run %d', i))

[0277] MegaLMM_state = sample_MegaLMM(MegaLMM_state, n_iter)

[0278] MegaLMM_state = save_posterior_chunk(MegaLMM_state)

[0279] print(MegaLMM_state)

[0280] plot(MegaLMM_state)

[0281] if (MegaLMM_state$current_state$nrun<MegaLMM_state$run_parameters$burn ||

[0282] i<= 20) {

[0283] MegaLMM_state = reorder_factors(MegaLMM_state, drop_cor_threshold= 0.6) # Factor order doesn't "mix" well in the MCMC. We can help it bymanually re-ordering from biggest to smallest

[0284] print(MegaLMM_state$run_parameters$burn)

[0285] }

[0286] }

[0287] estimate<- MegaLMM_state$current_state$Eta %>%

[0288] as.data.frame()

[0289] parameters<- train_data %>%

[0290] as.data.frame() %>%

[0291] gather(key = 'trait', value = 'value') %>%

[0292] group_by(trait) %>%

[0293] summarise(mean = mean(value, na.rm = T),

[0294] sd = sd(value, na.rm = T))

[0295] estimate_test<- mapply(`*`,estimate %>% as.data.frame(), parameters$sd) %>%

[0296] as.data.frame() %>%

[0297] mapply(`+`,.,parameters$mean) %>%

[0298] as.data.frame()

[0299] save(estimate_test,file=paste(" / public / home / hpei / BGGE / megalmm_rep10_220922_env2_half_NA_fromshao / megalmm_input_73trait_0.5NA_shao / megalmm",middle_par_6," / megalmm_half_NA_pred_73_trait_shaocode_",middle_par_6,".RData",sep=""))

[0300] Use the rrBLUP method to predict the phenotypic values of rice yield in the test set (taking the division of the training set and the test set as an example).

[0301] ibrary(rrBLUP)

[0302] library(tidyverse)

[0303] load("D: / desktop / recent / yield220808.RData")# Load phenotypic data

[0304] kinship<- read.table("D: / desktop / R historical file / kinship_191.txt",header=FALSE,sep="\t",

[0305] skip = 3,row.names = 1)

[0306] yield_data<- County_note_abb %>%

[0307] mutate(Environment=paste(Prov_abb,TrialYear, sep='_'))%>%

[0308] select(Environment,Genotype,GY)%>%

[0309] filter(Environment!="ZJ_2015")

[0310] spread_data<- yield_data %>%

[0311] group_by(Environment,Genotype)%>%

[0312] summarise(GY=mean(GY,na.rm=T))%>%

[0313] spread(key = Environment,value=GY)%>%

[0314] as.data.frame()

[0315] pheno_names<- colnames(spread_data)[-1]

[0316] rrblup_df_list<-list()

[0317] for(i in 1:10){

[0318] rrblup_data<- train_data_gather_2env_list[[i]]

[0319] rrblup_data<- rrblup_data %>%

[0320] spread(key = Environment,value=GY)%>%

[0321] set_names(c("Genotype",pheno_names))

[0322] rrblup_df_list[[i]]<- rrblup_data

[0323] }

[0324] colnames(kinship)<-rownames(kinship)

[0325] kinship<- kinship%>%

[0326] as.matrix()

[0327] Geno<- colnames(kinship)

[0328] pred_df_list<- list()

[0329] for(k in 1:10) {

[0330] test_data<- rrblup_df_list[[k]]

[0331] pred_df<- data.frame(matrix(nrow = 191, ncol = 19))

[0332] rownames(pred_df)<- Geno

[0333] colnames(pred_df)<- colnames(test_data)[-1]

[0334] trait_name<- colnames(test_data)[-1]

[0335] for (i in 2:(length(trait_name) + 1)) {

[0336] sub_data<- test_data[, c(1, i)]

[0337] A<-

[0338] kin.blup(sub_data,

[0339] geno = 'Genotype',

[0340] pheno = trait_name[i - 1],

[0341] K = kinship)

[0342] pred<- A$pred %>%

[0343] as.data.frame()

[0344] pred_df[, i - 1]<- pred

[0345] }

[0346] pred_df_list[[k]]<- pred_df

[0347] }

[0348] Use regression analysis to predict the phenotypic value of rice yield in the test set (taking the division into training set and test set as an example).

[0349] im_list<-list()

[0350] all_NA_geno_vec_list<-list()

[0351] impute_lm_input_data<-list()

[0352] for(q in 1:10){

[0353] GY_data<- train_data_gather_2env_list[[q]]

[0354] Env_mean_df<- GY_data %>%

[0355] group_by(Environment)%>%

[0356] summarize(Env_mean = mean(GY,na.rm =T))%>%

[0357] as.data.frame()

[0358] Env_mean_vec<- Env_mean_df[,2]

[0359] GY_data$Env_mean<- rep(Env_mean_vec,each=191)

[0360] impute_lm_data<- GY_data%>%

[0361] select(Genotype,GY,Env_mean)

[0362] impute_lm_data$position<- rownames(impute_lm_data)

[0363] library(simputation)

[0364] all_NA_geno<- impute_lm_data%>%

[0365] filter(is.na(GY))%>%

[0366] group_by(Genotype)%>%

[0367] count(Genotype,name = "NA_count")%>%

[0368] as.data.frame()%>%

[0369] filter(NA_count == 19)

[0370] all_NA_geno_vec<- all_NA_geno[,1]

[0371] all_NA_geno_vec_list[[q]]<- all_NA_geno_vec

[0372] genotype_vec<- setdiff(impute_lm_data$Genotype,all_NA_geno_vec)

[0373] impute_lm_noallNA_data<- impute_lm_data %>%

[0374] filter(impute_lm_data$Genotype %in% genotype_vec)

[0375] impute_lm_input_data[[q]]<- impute_lm_noallNA_data

[0376] im<- impute_lm(impute_lm_noallNA_data,GY~Env_mean+Genotype)

[0377] im_list[[q]] <- im

[0378] }

[0379] The prediction accuracy evaluation results of the four methods for yield, plant height and flowering period (average of ten Pearson correlation coefficients) showed that BGGE had a higher prediction ability in predicting phenotypic data with severe missing data, which were 0.73, 0.84 and 0.87, respectively ( Figure 2 ).

[0380] 3. Obtaining complete phenotypic data based on whole genome selection methods

[0381] The phenotypic data of yield, plant height and flowering period under 19 environments were interpolated using the BGGE method in combination with the genotype data to obtain a complete regional trial data set.

[0382] 4. Identification of key environmental factors of response phenotypes based on environmental information

[0383] The Critical Environmental Regressor through InformedSearch (CERIS) method was used to identify key environmental indicators. The source code of CERIS can be downloaded at https: / / github.com / jmyu / CERIS_JGRA.

[0384] The CERIS method mainly includes three R functions (CERIS.r, CERIS_functions_1.r and CERIS_functions_2.r). CERIS.r is the main function, which contains the input setting values ​​of important parameters. CERIS_functions_1.r and CERIS_functions_2.r are sub-functions used to download meteorological data. They can download data of 25 environmental factors throughout the growth period of rice. The specific details are (taking rice yield data as an example):

[0385] Rscript CERIS.r -w E: / work_report / 1205 / CERIS-BGGE-outdata -gplanting.txt -e phenotype_GYPHFT.txt -a GY -d 150 -c CERIS_functions_1.r. Download the information of 14 environmental factors from 1 to 150 days after rice planting. The -w parameter is used to set the working directory, the -g parameter inputs the rice planting information, mainly including the longitude and latitude of the planting site, the planting date and harvest date of rice under each environment, the -e parameter is the phenotypic data to be input, the -a parameter specifies the phenotype to be analyzed, the -d parameter specifies the number of days of environmental data to be downloaded, and the -c parameter selects the target sub-function.

[0386] Rscript CERIS.r -w E: / work_report / 1205 / CERIS-BGGE-outdata -gplanting.txt -e phenotype_GYPHFT.txt -a GY -d 150 -c CERIS_functions_2.r Download the information of the remaining 11 environmental factors from 1 to 150 days after rice planting.

[0387] Each run will obtain the ranking results of the correlation between the target environmental index and the phenotypic mean, which is contained in the file GY_19Envs_environmental_index_ranking.txt. The output results of the two sub-functions are merged using the perl language to obtain the environmental index that is most significantly correlated with the phenotypic data (perlmergeEnvironmentalIndexRank.pl E:\work_report\1205\CERIS-BGGE-outdata\CERIS_output.1\GY\GY_19Envs_environmental_index_ranking.txt E:\work_report\1205\CERIS-BGGE-outdata\CERIS_output.2\GY\GY_19Envs_environmental_index_ranking.txt GY_19Envs_environmental_index_ranking-merge.txt).

[0388] The above steps were repeated to obtain key environmental indices related to rice yield, plant height and flowering period.

[0389] 5. Evaluation of the prediction accuracy of new environmental phenotype data using environmental index as a bridge (taking flowering period as an example)

[0390] The absolute value of the correlation coefficient between the environmental index and the phenotypic mean is greater than 0.5, and the one with the largest correlation is selected for each of the 25 environmental factors.

[0391] Based on the complete flowering period data of 19 environments obtained previously, combined with the identified key environmental indices, the flowering period of HB_2018 was predicted using linear fitting ( Figure 3 Here, a single environmental index is used to predict the flowering period phenotype, and then the phenotypic data is averaged to obtain the final predicted value. The specific details of the phenotypic data prediction based on key environmental indices are as follows (taking rice flowering period data as an example):

[0392] setwd("E: / work_report / 1205 / CERIS-BGGE-outdata")

[0393] #Environmental index with absolute value of correlation of flowering period > 0.5

[0394] best_Indedx<- read.table("bestEnvIndex-FT.txt",header = T)

[0395] best_Indedx<- best_Indedx[abs(best_Indedx$cor)>= 0.5,]

[0396] #19 Environment data

[0397] weather_data1<- read.table(

[0398] "E: / work_report / 1205 / CERIS-BGGE-outdata / CERIS_output.1 / Envs_envParas_DAP150",

[0399] header = T ,

[0400] sep = " " )

[0402] weather_data2<- read.table(

[0403] "E: / work_report / 1205 / CERIS-BGGE-outdata / CERIS_output.2 / Envs_envParas_DAP150",

[0404] header = T ,

[0405] sep = " " )

[0407] weather_data2<- weather_data2[,6:16]

[0408] weather_data<- cbind(weather_data1,weather_data2)

[0409] Environmental data for #HB_2018

[0410] wheather_HB2018_1<- read.table("E: / work_report / 20230203 / FT_valadition / CERIS_output.1 / Envs_envParas_DAP150",header=T)

[0411] wheather_HB2018_2<- read.table("E: / work_report / 20230203 / FT_valadition / CERIS_output.2 / Envs_envParas_DAP150",header=T)

[0412] wheather_HB2018_2<- wheather_HB2018_2[,6:16]

[0413] wheather_HB2018<- cbind(wheather_HB2018_1,wheather_HB2018_2)

[0414] #Environmental data for 20 environments

[0415] env20_weatherdata<- rbind(weather_data,wheather_HB2018)

[0416] env20_weatherdata<- env20_weatherdata[order(env20_weatherdata$env_code),]

[0417] #Name of the environmental factor whose correlation is greater than a certain value

[0418] env_factor_names<- best_Indedx$envName

[0419] colnames(env20_weatherdata)

[0420] target_env_weatherData<- env20_weatherdata[,c("env_code","Date",env_factor_names)]

[0421] env20_names<- unique(target_env_weatherData$env_code)

[0422] start <- best_Indedx$start

[0423] end <- best_Indedx$end

[0424] # Group by environment

[0425] env_factors_means_df <- as.data.frame(matrix(nrow = length(env20_names), ncol = length(env_factor_names)))

[0426] for (i in 1:length(env20_names)) {

[0427] env_factors_means <- c()

[0428] for (j in 1:length(env_factor_names)){

[0429] mid_df <- target_env_weatherData[target_env_weatherData$env_code == env20_names[i], ]

[0430] mid_df <- mid_df[,c("env_code","Date",env_factor_names[j])]

[0431] mid_df <- mid_df[start[j]:end[j], ]

[0432] Eindex_mean <- mean(mid_df[,3], na.rm = T)

[0433] names(Eindex_mean) <- paste(env20_names[i], env_factor_names[j], sep = "_")

[0434] env_factors_means <- c(env_factors_means, Eindex_mean)

[0435] }

[0436] env_factors_means_df[i,]<- env_factors_means

[0437] }

[0438] colnames(env_factors_means_df)<- env_factor_names

[0439] rownames(env_factors_means_df)<- env20_names

[0440] #20 Phenotypic data for complete environments

[0441] FT_pheno_2018<- read.table("E: / work_report / 20230203 / FT_valadition / FT_phenotype191_2NA.txt",header = T)

[0442] train_data<-FT_pheno_2018

[0443] #19 Phenotypic data for the environment

[0444] FT_pheno<-

[0445] read.table(

[0446] "E: / work_report / 1205 / CERIS-BGGE-outdata / reaction_norm_cor / phenotype_GYPHFT.txt",

[0447] header = T )

[0449] FT_pheno<- FT_pheno %>%

[0450] select(1,2,5)

[0451] #20 Flowering period phenotypes in different environments

[0452] lm_input<- rbind(FT_pheno,train_data)

[0453] lm_input<- lm_input[order(lm_input$env_code),]

[0454] # Add the means of multiple environmental indices to lm_input

[0455] lm_input_fac_mean_df <- lm_input

[0456] lm_input_fac_mean_df_spread <- lm_input_fac_mean_df %>%

[0457] spread(key = env_code, value = FT)

[0458] no_l <- dim(lm_input_fac_mean_df_spread)[1]

[0459] no_e <- dim(lm_input_fac_mean_df_spread)[2] - 1

[0460] pheno_hat_list <- list()

[0461] for (index in 1:length(env_factor_names)) {

[0462] env1_factors_means_df <- env_factors_means_df[, index]

[0463] pheno_hat = matrix(0, no_l, no_e);

[0464] for (i in 1:no_l) {

[0465] mid_pheno <- as.numeric(lm_input_fac_mean_df_spread[i, -1])

[0466] for (j in 1:no_e) {

[0467] target_mid_pheno <- mid_pheno[-j]

[0468] target_mid_factor <- env1_factors_means_df[-j]

[0469] mid_df <- rbind(mid_pheno, env1_factors_means_df)

[0470] model<- lm(y ~ x, data = data.frame(x = target_mid_factor, y =target_mid_pheno))

[0471] new<- data.frame(x = c(mid_df[2, j]))

[0472] y_hat<- predict(model, newdata = new)

[0473] pheno_hat[i, j] = y_hat

[0474] }

[0475] }

[0476] pheno_hat_list[[index]]<- pheno_hat

[0477] }

[0478] names(pheno_hat_list)<- env_factor_names

[0479] FT_imputed_vec<- c(rep(0,191))

[0480] #HB_2018

[0481] for (factor in 1:length(pheno_hat_list)){

[0482] mid_Df<- pheno_hat_list[[factor]]

[0483] FT_imputed<- mid_Df[,12]

[0484] FT_imputed_vec<- FT_imputed_vec + FT_imputed

[0485] }

[0486] FT_imputed<- FT_imputed_vec / length(pheno_hat_list)

[0487] cor(FT_imputed,FT_pheno_2018$FT,use = "complete.obs")#0.55

[0488] plot_Df<- data.frame(imputed = FT_imputed, origin = FT_pheno_2018$FT)

[0489] The predicted values ​​of flowering date in HB_2018 have a moderate correlation of 0.55 with the observed values, which can be explained by the correlations of flowering date in HB_2018 with the other 19 environmental flowering dates ranging from 0.43 to 0.64.

[0490] 6. Forecast of rice production data for the past 39 years

[0491] Using environmental index as a bridge and combining phenotypic data from 19 tested environments, the yield data of rice producing areas in the middle and lower reaches of the Yangtze River for the past 39 years were predicted.

[0492] Download meteorological data from rice sowing to 150 days after planting from 1984 to 2022 based on CERIS. Note that only 22 types of environmental factors are used here, because the meteorological data of three environmental factors, net photosynthetically active radiation, ultraviolet B and ultraviolet A from 1984 to 2000, do not exist in the NASA POWER database.

[0493] Using the key environmental indices related to yield that meet the conditions obtained above and the phenotypic information of tested genotypes (191 rice varieties) in 19 environments, the yield data in untested environments (8 provinces from 1984 to 2022) were predicted and used in the subsequent ecotype analysis of different environments.

[0494] 7. Combine regional test evaluation models to classify ecological types and make pilot recommendations

[0495] After obtaining the yield data of 191 rice varieties in 8 provinces in 39 years, the yield data of each year was analyzed using the R language GGEBiplotGUI package, and the ecological zone division results of each year were obtained based on the "Which wins where" map. Finally, a classification result covering the overall environment of 8 provinces in 39 years was obtained.

[0496] The results of the 39-year ecological division were visualized using Cytoscape software, as shown in the network diagram ( Figure 4 ). The size of the dots reflects the frequency with which each pilot site is classified into an ecological zone. Since Guangdong has a unique ecological type, it is a reasonable choice to include it in the variety testing pilot under a limited budget. The thickness of the line indicates the similarity of the ecological types of the pilot sites. Fujian and Anhui are similar to the ecological types of other surrounding pilot sites. Therefore, based on the consideration of variety breeding, varieties bred in Fujian and Anhui can be considered for promotion in surrounding pilot sites.

[0497] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0498] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. The biological-environmental multidimensional information data model for crop regional experiments is characterized by: The data model integrates whole genome bioinformatics and environmental-phenotypic data to focus on the ideal test sites of candidate varieties in crop regional trials; the data model includes the following data processes: A. Data loading and its underlying normalization: Load the phenotypic data file based on the existing experimental data set, read the kinship matrix and convert it into a symmetric matrix A_matrix, with the row and column names being genotypes; then select and process the environment Environment, genotype Genotype, and yield GY related data from the County_note_abb data frame, and filter out specific environmental data; organize the data and convert it into a wide format data frame spread_data, with each column representing the yield data under one environment; then generate training sets and test sets, and repeat the loop 10 times to generate training sets and test sets; in each loop, randomly select 50% of the data for each environment as the test set, set this part of the data to NA, and use the remaining 50% as the training set; Record the NA position and true value in each generated test set for estimation of subsequent model prediction accuracy; B. Embedding of the multi-module whole genome bioinformatics data analysis framework: specifically including: loading the BGGE package and related data; setting the working directory and the intermediate variable middle_par; reading the genotype data file GAPIT.Genotype.Numerical.txt; using the BGGE method to predict the yield data in the test set; saving the prediction results; also including: loading the MegaLMM package and related data; setting the working directory and the intermediate variable middle_par_6; reading the kinship matrix; using the MegaLMM method to predict the yield data in the test set; saving the prediction results; also including: loading the rrBLUP package and related data; reading the kinship matrix; using the rrBLUP method to predict the yield data in the test set; saving the prediction results; also including: for each generated test set, calculating the average yield of each environment and adding it to the data frame; using the impute_lm function in the simputation package to interpolate the yield data through a linear model; saving the interpolation results; C. Data interpolation: For each test set generated, calculate the average yield of each environment and add it to the data frame; use the impute_lm function in the simulation package to interpolate the yield data through a linear model; save the interpolation results; evaluate the prediction accuracy of the multi-module whole genome bioinformatics data analysis framework, and calculate the average of ten prediction accuracies; select the whole genome bioinformatics data analysis framework module with high prediction ability in predicting phenotypic data with severe missing, and use it to interpolate the phenotypic data of yield, plant height and flowering period under multiple environments to obtain a complete regional test data set; D. Identify key environmental factors and conduct accuracy assessment: specify the working directory, planting information file planting.txt, phenotypic data file phenotype_GYPHFT.txt, phenotype to be analyzed GY, number of days t of environmental data to download, and multiple target sub-functions; embed CERIS.r to download information on environmental factors from 1 to t days after rice planting; use the Perl script mergeEnvironmentalIndexRank.pl to merge the ranking results of the correlation between environmental index and phenotypic mean output by multiple sub-functions; then read the environmental index file bestE with an absolute correlation value greater than 0.5 nvIndex-FT.txt; load a total of k environmental data from multiple environments and newly created environmental data, and merge them to obtain the environmental data envk_weatherdata; filter the environmental factor data with a correlation greater than a specific value according to the name of the environmental factor; calculate the average value of each environmental factor in each environment to form the env_factors_means_df data frame; read the flowering period phenotype data of k environments and merge them into lm_input; use a linear model to fit each environmental index with the flowering period phenotype, and predict the flowering period of the specified crop parameter of the newly created environment; calculate the correlation between the predicted value and the observed value, and generate a correlation chart; E. Predict the rice yield data for x years and conduct ecological type analysis to realize the data focus of the test sites: Download the meteorological data of x years based on CERIS, use the key environmental indices related to yield and the phenotypic information of the tested genotypes in multiple environments to predict the yield data in untested environments; then use the R language GGEBiplotGUI package to analyze the yield data of multiple rice varieties in different provinces and regions for x years to obtain the ecological zone division results for each year; and build or load visualization components to visualize the ecological zone division results for x years and generate a network diagram. The size of the points indicates the frequency with which the pilot is divided into ecological zones, and the thickness of the lines indicates the similarity of the ecological types of the pilots.

2. The crop regional test biological-environmental multidimensional information data model according to claim 1, characterized in that: The crop is rice.