Method for identifying wheat drought-resistant candidate gene loci based on deterministic regulatory path topic modeling

Abstract

The application discloses a kind of wheat drought-resistant candidate gene site identification methods based on deterministic regulation path theme modeling, comprising the following steps: obtaining the gene expression under the condition of wheat drought stress, candidate gene site, drought-resistant phenotype, environmental factor, functional annotation and gene regulation network data;Expression support regulation diagram is constructed;In expression support regulation diagram, deterministic regulation path and transcription factor central module are generated, and they are used as regulation word;Obtain regulation path theme feature;Generate regulation path theme evidence;Dynamic regulation network is constructed and dynamic network importance evidence is calculated;Further, drought-resistant phenotype association, environmental response, functional annotation and dynamic network importance evidence are fused, and the output result is obtained.The application can convert complex gene regulation network into computable, traceable and interpretable regulation path theme feature, improve the accuracy, stability and biological interpretability of wheat drought-resistant candidate gene site identification.

Description

Technical Field

[0001] This invention relates to the field of plant gene recognition technology, and in particular to a method for identifying candidate gene loci for drought resistance in wheat based on deterministic regulatory pathway theme modeling. Background Technology

[0002] Currently, wheat is one of the world's most important food crops, and its drought resistance trait is influenced by multiple genes, multiple environmental factors, and a complex regulatory network. Under drought stress, wheat gene expression exhibits sparsity, dynamics, and sample heterogeneity; some key drought-resistant genes may be activated only in specific tissues, at specific time points, under specific treatment conditions, or under specific environmental gradients. With the development of high-throughput sequencing, phenomics, and environmental monitoring technologies, researchers can simultaneously obtain multimodal data on gene expression, candidate variant sites, drought-resistant phenotypes, environmental factors, and functional annotations, providing a data foundation for identifying drought-resistant candidate gene loci.

[0003] In existing technologies, differential expression analysis, genome-wide association analysis, co-expression network analysis, machine learning classification, and graph neural networks have been used to screen candidate genes for drought resistance in wheat. These methods can discover potential drought-related genes from different perspectives, but they still have the following limitations.

[0004] First, traditional methods often use single-gene expression differences, single association statistics, or the importance of a single machine learning feature as screening criteria, which makes it difficult to fully utilize the pathway structure, transcription factor central modules, and multi-gene synergistic regulatory information in gene regulatory networks.

[0005] Second, existing co-expression network or regulatory network analyses mostly remain at the node or edge level, making it difficult to transform complex regulatory networks into computable, traceable, and interpretable regulatory path features, resulting in insufficient explanation of the regulatory mechanisms of candidate gene loci.

[0006] Third, although some deep learning or graph neural network methods can extract nonlinear features, their results are usually expressed as latent space vectors or black box feature importance, making it difficult to directly output the regulatory pathways, transcription factor central modules, and validation priorities corresponding to candidate gene sites.

[0007] Fourth, the drought stress response of wheat has obvious temporal dynamics and environmental dependence. Traditional static network modeling methods are difficult to capture the dynamic changes of the regulation path at different time points, under different treatment conditions or different environmental gradients, thus limiting the ability to identify the core drought resistance regulation module.

[0008] Therefore, there is an urgent need for a method that can transform the wheat gene regulatory network into deterministic regulatory pathways and transcription factor central modules, and obtain interpretable regulatory pathway theme features through topic modeling, thereby improving the accuracy and interpretability of drought-resistant candidate gene loci identification. Summary of the Invention

[0009] The purpose of this invention is to provide a method for identifying wheat drought-resistant candidate gene loci based on deterministic regulatory pathway theme modeling, which can solve the problem that traditional candidate gene loci screening methods cannot fully utilize the pathway structure and transcription factor central modules in the gene regulatory network;

[0010] Furthermore, it can address the problem that existing machine learning or deep learning methods lack traceable evidence of regulatory pathways and biological explanations for their output results;

[0011] Furthermore, it can solve the problem of difficulty in characterizing the dynamic changes of the regulatory network under multiple time points, conditions, and environmental factors in wheat drought stress response;

[0012] Furthermore, it can solve the problem that the identification results of candidate drought-resistant gene loci are difficult to form a verification priority and explanation of regulatory mechanisms.

[0013] The technical solution adopted in this invention is as follows:

[0014] A method for identifying candidate gene loci for drought resistance in wheat based on deterministic regulatory pathway theme modeling includes the following steps:

[0015] Step 1: Obtain multimodal data of wheat under drought stress conditions. The multimodal data includes gene expression matrices to characterize gene expression levels in wheat samples, genotypes, tissues, treatment conditions, or time points; candidate gene locus matrices to characterize SNPs, InDel, structural variations, allele frequencies, locus coverage, or genotype encoding; drought resistance phenotype data to characterize leaf water content, biomass, root length, root volume, stomatal conductance, water use efficiency, relative survival rate, chlorophyll content, or drought stress index; environmental factor data to characterize soil moisture, air humidity, temperature, light intensity, drought treatment time, field water gradient, or transpiration intensity; and gene function annotation data and gene regulatory networks to characterize the relationship between candidate genes and stress resistance, ABA signaling, ROS scavenging, root development, stomatal regulation, osmotic regulation, or known drought resistance pathways.

[0016] Step 2: Based on the observable gene set in the gene expression matrix, filter the gene regulatory network to obtain an expression support regulation map;

[0017] Step 3: Construct deterministic regulatory pathways and transcription factor central modules based on the expression support regulation graph, and use the deterministic regulatory pathways and transcription factor central modules as regulatory terms;

[0018] Step 4: Calculate the regulatory activity of the regulatory words in wheat samples, treatment conditions, or time points to obtain the regulatory word activity matrix;

[0019] Step 5: Based on the regulatory word activity matrix obtained in Step 4, and combined with the drought resistance phenotype data and environmental factor data obtained in Step 1, regulatory words are screened according to non-zero support ratio, activity variance, drought resistance phenotype correlation or environmental response intensity to obtain the screened regulatory words and their activity characteristics.

[0020] Step 6: Based on the gene expression matrix obtained in Step 1 and the activity features of the selected regulatory words obtained in Step 5, assemble a non-negative document matrix and obtain the regulatory path theme features through topic modeling.

[0021] Step 7: Based on the candidate gene locus matrix obtained in Step 1, the screened regulatory words obtained in Step 5, and the topic-regulatory word payload matrix obtained in Step 6, the candidate gene loci are mapped to the corresponding genes, regulatory words, and regulatory pathway topics to generate regulatory pathway topic evidence for the candidate gene loci.

[0022] Step 8: Based on the regulatory word activity matrix obtained in Step 4 and the corresponding time points, treatment conditions, or environmental gradient information in Step 1, construct a dynamic regulatory network and calculate the evidence of the importance of the dynamic network; based on the candidate gene locus matrix, drought resistance phenotype data, environmental factor data, and gene function annotation data obtained in Step 1, calculate the evidence of drought resistance phenotype association, environmental response, and functional annotation, respectively; then, based on the regulatory pathway theme evidence obtained in Step 7 and the above-mentioned evidence of drought resistance phenotype association, environmental response, functional annotation, and dynamic network importance, obtain the comprehensive identification score of the candidate gene locus.

[0023] Step 9: Based on the comprehensive identification score of candidate gene loci obtained in Step 8, determine the set of highly reliable wheat drought-resistant candidate gene loci according to the preset score threshold or sorting ratio, and output the set of highly reliable wheat drought-resistant candidate gene loci and its corresponding regulatory pathway evidence, transcription factor central module evidence, regulatory pathway theme evidence and verification priority.

[0024] The support regulation graph in step 2 is constructed through the following steps:

[0025] Step 2.1: Extract the set of observable genes from the gene expression matrix;

[0026] Step 2.2: Retain regulatory edges from the original gene regulatory network where both ends of the regulatory edge belong to the observable gene set;

[0027] Step 2.3: The retained nodes and control edges form an expression support control graph;

[0028] Step 2.4, the expression support regulation map is used to ensure that the deterministic regulatory pathways, transcription factor central modules and regulatory word activities constructed in the subsequent process are all calculated from the measurable genes in the current gene expression matrix.

[0029] The deterministic control path in step 3 is constructed through the following steps:

[0030] Step 3.1: Using the transcription factors in the regulatory graph as anchor nodes, extend downstream along the regulatory edge to target genes to generate an ordered regulatory pathway starting from the transcription factors.

[0031] Step 3.2, limit the maximum length of the regulatory pathway and the maximum number of pathways retained by each transcription factor to avoid excessive domination of the regulatory word set by highly connected transcription factors;

[0032] Step 3.3: Screen regulatory pathways based on evidence from gene expression of pathway members, drought resistance phenotype, environmental response intensity, or functional annotation.

[0033] The transcription factor central module in step 3 is constructed through the following steps:

[0034] Step 3.4: Determine the set of outgoing neighborhood target genes for each transcription factor in the expression support regulation graph;

[0035] Step 3.5: Sort the target genes according to their expression variance, drought resistance phenotype correlation, environmental response intensity, and the strength of functional annotation evidence or regulatory evidence.

[0036] Step 3.6: Select target genes that meet the preset quantity range or evidence threshold, and together with the corresponding transcription factors, form the transcription factor central module.

[0037] In step 4, the activity of the regulatory word is calculated through the following steps:

[0038] Step 4.1: Define each deterministic regulatory pathway or each central module of a transcription factor as a regulatory term;

[0039] Step 4.2: Calculate sample-level regulatory activity based on the expression values ​​of regulatory term member genes in wheat samples;

[0040] Step 4.3: Calculate the non-zero support ratio and activity variance based on the distribution of regulatory word activity in different samples, treatment conditions, or time points;

[0041] Step 4.4 outputs the non-zero support ratio and activity variance of the regulatory words, which serve as the basis for screening regulatory words in step 5.

[0042] The control path theme features in step 6 are obtained through the following steps:

[0043] Step 6.1: Scale and splice the sample-level gene expression features formed by the gene expression matrix obtained in Step 1 and the screened regulatory word activity features obtained in Step 5 to form a non-negative document matrix;

[0044] Step 6.2: Perform topic modeling on the non-negative document matrix to obtain the sample topic representation matrix and the topic-regulation word load matrix;

[0045] Step 6.3: From the sample topic representation matrix and topic-regulatory word load matrix, obtain the regulatory pathway topic features used for candidate gene site mapping and regulatory pathway topic evidence calculation in Step 7.

[0046] The candidate gene locus mapping in step 7 is performed through the following steps:

[0047] Step 7.1: Based on the candidate gene locus matrix obtained in Step 1, and combined with wheat genome public annotation or linkage interval information, map the candidate gene loci to genes that overlap with, are adjacent to, are located in the same functional region, are located in the promoter region, are located in the enhancer region, or are located within the linkage interval.

[0048] Step 7.2: Based on whether the mapped gene belongs to the deterministic regulatory pathway or transcription factor central module retained after screening in Step 5, establish the correspondence between candidate gene sites and screened regulatory words.

[0049] Step 7.3: Based on the correspondence between the candidate gene loci and the selected regulatory terms, and the contribution of the selected regulatory terms to the regulatory pathway theme, establish the correspondence between the candidate gene loci and the regulatory pathway theme.

[0050] The comprehensive identification score in step 8 is obtained by fusing the following evidence:

[0051] Evidence on regulatory pathway themes is used to characterize the contribution of candidate gene loci to regulatory pathway themes;

[0052] Evidence of drought resistance phenotype association is used to characterize the association between candidate gene loci or mapping genes and drought resistance phenotype;

[0053] Environmental response evidence is used to characterize the association between candidate gene loci or mapped genes and drought environmental factors;

[0054] Functional annotation evidence is used to characterize the relationship between candidate gene loci or mapped genes and drought-related functional pathways.

[0055] Evidence of the importance of dynamic networks is used to characterize the topological importance of candidate gene loci or mapped genes in deterministic regulatory pathways, transcription factor central modules, or dynamic regulatory networks.

[0056] The dynamic control network parsing step in step 8 includes:

[0057] A dynamic regulatory network was constructed based on data on the activity of regulatory terms and drought resistance phenotypes at different time points, under different treatment conditions or environmental gradients.

[0058] Identify core control paths and core control modules based on time dependencies, causal relationships, edge weight changes, node importance, or module stability;

[0059] The core regulatory pathways and core regulatory modules are used as the basis for prioritizing candidate gene loci for validation.

[0060] This invention transforms the transcription factor-target gene relationships in the wheat gene regulatory network into deterministic regulatory pathways and transcription factor central modules, further defining them as regulatory terms. This allows complex regulatory networks to be transformed into computationally computable, traceable, and interpretable regulatory pathway thematic features. The invention employs a deterministic regulatory pathway construction method, independent of random walks or random graph sampling, achieving a stable and consistent set of regulatory terms under the same input data and parameter conditions, thus improving the reproducibility of candidate gene locus identification results. Through regulatory term activity matrices and thematic modeling, this invention integrates multiple regulatory pathways and transcription factor central modules into a regulatory pathway thematic representation, revealing multi-gene synergistic regulatory patterns in wheat drought stress response. Mapping candidate gene loci to regulatory pathways, transcription factor central modules, and regulatory pathway themes allows for the simultaneous output of candidate gene loci and corresponding regulatory mechanism evidence, improving the biological interpretability of candidate gene locus screening results. Furthermore, this invention combines drought resistance phenotype, environmental response, functional annotation, and the importance of dynamic regulatory networks for comprehensive scoring, improving the accuracy, stability, and validation priority ranking capabilities of candidate drought resistance gene locus identification. This invention not only outputs a list of candidate gene loci, but also provides evidence of regulatory pathways, evidence of transcription factor central modules, evidence of regulatory pathway themes, and validation priorities. This helps in the selection of targets for subsequent quantitative polymerase chain reaction (qPCR), CRISPR-Cas9 gene editing system (clustered regularly interspaced shortpalindromic repeats and CRISPR-associated protein 9, CRISPR-Cas9), chromatin immunoprecipitation sequencing (ChIP-seq), or molecular breeding validation experiments. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is a flowchart of the present invention.

[0063] Figure 2 is a schematic diagram of the construction of the expression support regulation graph into a set of deterministic regulation words according to the present invention.

[0064] Figure 3 is a schematic diagram of the activity matrix of regulatory words and the topic feature learning of regulatory paths described in this invention.

[0065] Figure 4 is a schematic diagram of the candidate gene site mapping, dynamic regulatory network parsing and verification priority output described in this invention. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] like Figure 1 , 2 As shown in Figures 3 and 4, the present invention includes the following steps:

[0068] Step 1: Obtain multimodal data of wheat under drought stress conditions. The multimodal data includes gene expression matrices to characterize gene expression levels in wheat samples, genotypes, tissues, treatment conditions, or time points; candidate gene locus matrices to characterize single nucleotide polymorphism (SNP), insertion / deletion (InDel) variants, structural variations, allele frequencies, locus coverage, or genotype encoding; drought resistance phenotype data to characterize leaf water content, biomass, root length, root volume, stomatal conductance, water use efficiency, relative survival rate, chlorophyll content, or drought stress index; environmental factor data to characterize soil moisture, air humidity, temperature, light intensity, drought treatment time, field water gradient, or transpiration intensity; and gene function annotation data and gene regulatory networks to characterize the relationship between candidate genes and stress resistance, abscisic acid (ABA) signaling, reactive oxygen species (ROS) scavenging, root development, stomatal regulation, osmotic regulation, or known drought resistance pathways.

[0069] Step 2: Based on the observable gene set in the gene expression matrix, the gene regulatory network is filtered to obtain an expression support regulation map; the expression support regulation map in Step 2 is constructed through the following steps:

[0070] Step 2.1: Extract the set of observable genes from the gene expression matrix;

[0071] Step 2.2: Retain regulatory edges from the original gene regulatory network where both ends of the regulatory edge belong to the observable gene set;

[0072] Step 2.3: The retained nodes and control edges form an expression support control graph;

[0073] Step 2.4, the expression support regulation map is used to ensure that the deterministic regulatory pathways, transcription factor central modules and regulatory word activities constructed in the subsequent process are all calculated from the measurable genes in the current gene expression matrix.

[0074] Step 3: Construct deterministic regulatory pathways and transcription factor central modules based on the expression support regulation graph, and use the deterministic regulatory pathways and transcription factor central modules as regulatory terms; the deterministic regulatory pathways in Step 3 are constructed through the following steps:

[0075] Step 3.1: Using the transcription factors in the regulatory graph as anchor nodes, extend downstream along the regulatory edge to target genes to generate an ordered regulatory pathway starting from the transcription factors.

[0076] Step 3.2, limit the maximum length of the regulatory pathway and the maximum number of pathways retained by each transcription factor to avoid excessive domination of the regulatory word set by highly connected transcription factors;

[0077] Step 3.3: Screen regulatory pathways based on evidence from gene expression of pathway members, drought resistance phenotype, environmental response intensity, or functional annotation.

[0078] The transcription factor central module in step 3 is constructed through the following steps:

[0079] Step 3.4: Determine the set of outgoing neighborhood target genes for each transcription factor in the expression support regulation graph;

[0080] Step 3.5: Sort the target genes according to their expression variance, drought resistance phenotype correlation, environmental response intensity, and the strength of functional annotation evidence or regulatory evidence.

[0081] Step 3.6: Select target genes that meet the preset quantity range or evidence threshold, and together with the corresponding transcription factors, form the transcription factor central module.

[0082] Step 4: Calculate the regulatory activity of the regulatory term in wheat samples, treatment conditions, or time points to obtain the regulatory term activity matrix; the regulatory term activity in Step 4 is calculated through the following steps:

[0083] Step 4.1: Define each deterministic regulatory pathway or each central module of a transcription factor as a regulatory term;

[0084] Step 4.2: Calculate sample-level regulatory activity based on the expression values ​​of regulatory term member genes in wheat samples;

[0085] Step 4.3: Calculate the non-zero support ratio and activity variance based on the distribution of regulatory word activity in different samples, treatment conditions, or time points;

[0086] Step 4.4 outputs the non-zero support ratio and activity variance of the regulatory words, which serve as the basis for screening regulatory words in step 5.

[0087] Step 5: Based on the regulatory word activity matrix obtained in Step 4, and combined with the drought resistance phenotype data and environmental factor data obtained in Step 1, regulatory words are screened according to non-zero support ratio, activity variance, drought resistance phenotype correlation or environmental response intensity to obtain the screened regulatory words and their activity characteristics.

[0088] Step 6: Based on the gene expression matrix obtained in Step 1 and the regulatory word activity features obtained in Step 5, the sample-level gene expression features and regulatory word activity features are scaled and spliced ​​to form a non-negative document matrix, and the regulatory path theme features are obtained through topic modeling.

[0089] The control path theme features in step 6 are obtained through the following steps:

[0090] Step 6.1: The sample-level gene expression features obtained from the gene expression matrix and the screened regulatory word activity features are scaled and spliced ​​to form a non-negative document matrix;

[0091] Step 6.2: Perform topic modeling on the non-negative document matrix to obtain the sample topic representation matrix and the topic-regulation word load matrix;

[0092] Step 6.3: From the sample topic representation matrix and topic-regulatory word load matrix, obtain the regulatory pathway topic features used for candidate gene site mapping and regulatory pathway topic evidence calculation in Step 7.

[0093] Step 7: Based on the candidate gene locus matrix obtained in Step 1, the screened regulatory terms obtained in Step 5, and the topic-regulatory term payload matrix obtained in Step 6, the candidate gene loci are mapped to corresponding genes, regulatory terms, and regulatory pathway topics to generate regulatory pathway topic evidence for the candidate gene loci; the candidate gene locus mapping in Step 7 is performed through the following steps:

[0094] Step 7.1: Based on the candidate gene locus matrix obtained in Step 1, and combined with wheat genome public annotation or linkage interval information, map the candidate gene loci to genes that overlap with, are adjacent to, are located in the same functional region, are located in the promoter region, are located in the enhancer region, or are located within the linkage interval.

[0095] Step 7.2: Based on whether the mapped gene belongs to the deterministic regulatory pathway or transcription factor central module retained after screening in Step 5, establish the correspondence between candidate gene sites and screened regulatory words.

[0096] Step 7.3: Based on the correspondence between the candidate gene loci and regulatory terms and the contribution of the regulatory terms to the regulatory pathway theme, establish the correspondence between the candidate gene loci and the regulatory pathway theme.

[0097] Step 8: Based on the regulatory word activity matrix obtained in Step 4 and the corresponding time points, treatment conditions, or environmental gradient information in Step 1, construct a dynamic regulatory network and calculate the evidence of the importance of the dynamic network; based on the candidate gene locus matrix, drought resistance phenotype data, environmental factor data, and gene function annotation data obtained in Step 1, calculate the evidence of drought resistance phenotype association, environmental response, and functional annotation, respectively; then, based on the regulatory pathway theme evidence obtained in Step 7 and the above-mentioned evidence of drought resistance phenotype association, environmental response, functional annotation, and dynamic network importance, obtain the comprehensive identification score of the candidate gene locus.

[0098] Step 9: Based on the comprehensive identification score of candidate gene loci obtained in Step 8, determine the set of highly reliable wheat drought-resistant candidate gene loci according to the preset score threshold or sorting ratio, and output the set of highly reliable wheat drought-resistant candidate gene loci and its corresponding regulatory pathway evidence, transcription factor central module evidence, regulatory pathway theme evidence and verification priority.

[0099] This invention transforms the complex regulatory network of wheat into computable "regulatory terms" thematic features, ensuring complete determinism in path construction and stable, reproducible results. Through regulatory term activity matrices and thematic modeling, it reveals multi-gene synergistic regulatory patterns. The candidate gene loci output simultaneously includes three levels of evidence: path, module, and theme, significantly enhancing biological interpretability. A comprehensive scoring system is applied, considering drought resistance phenotype, environmental response, functional annotation, and network dynamic importance, improving identification accuracy and prioritizing validation. The final output not only includes a list of candidate genes but also provides supporting regulatory evidence and experimental validation target suggestions, directly supporting subsequent experiments such as qPCR, CRISPR-Cas9, ChIP-seq, and molecular breeding, achieving an efficient closed loop from computational prediction to experimental validation.

[0100] To provide a detailed explanation and illustration of the present invention, the following specific examples will be used for further elaboration:

[0101] Step 1: Obtain multimodal data on wheat drought stress

[0102] Multimodal data of wheat under normal water supply and drought stress conditions were acquired, including gene expression matrices. Candidate gene locus matrix Drought resistance phenotypic matrix Environmental factor matrix Function annotation matrix and gene regulatory networks .

[0103] The gene expression matrix is ​​represented as follows:

[0104]

[0105] in, This indicates the number of wheat samples, the number of treatment conditions, the number of genotypes, or the number of time points. Indicates the number of genes detected. Indicates the first In the nth sample The expression value of each gene, Represents the real number field.

[0106] The candidate gene locus matrix is ​​represented as follows:

[0107]

[0108] in, Indicates the number of candidate variant sites. Indicates the first The sample at the th Genotype coding, allele frequency, locus coverage, or locus characteristic value at each candidate gene locus.

[0109] The drought resistance phenotypic matrix is ​​represented as follows:

[0110]

[0111] in, This indicates the number of drought resistance phenotypic indicators. Indicates the first The first sample Several drought resistance phenotypic indicators. The environmental factor matrix is ​​represented as follows:

[0112]

[0113] in, Indicates the quantity of environmental factors. Indicates the first The first sample Environmental factors.

[0114] Standardize the input data:

[0115]

[0116] in, and They represent the first The mean and standard deviation of each gene across all samples. To prevent positive numbers with a denominator of zero.

[0117] In one implementation, the gene expression matrix is ​​first processed by a zero-inflation model, a graph-enhanced variational autoencoder, or a denoised expression modeling method to obtain a denoised expression matrix or sparse expression latent features, which are then used as inputs for subsequent regulatory word activity calculations and regulatory path topic modeling.

[0118] Step 2: Construct an expression support regulation graph

[0119] Let the original gene regulatory network be:

[0120]

[0121] in, For the original set of control network nodes, This is the original set of control edges.

[0122] An initial set of observable genes is extracted from the gene expression matrix, denoted as the set shown in the following equation:

[0123]

[0124] Both endpoints belong to Regulation edge:

[0125]

[0126] The initial set of observable genes is further filtered by preserving the endpoints of regulatory edges. The filtered set of expression support nodes is still denoted as:

[0127]

[0128] This results in the expression support regulation diagram:

[0129]

[0130] This step ensures that the deterministic regulatory pathways, transcription factor central modules, and regulatory word activities constructed subsequently are all calculated from measurable genes in the current gene expression matrix.

[0131] Step 3: Constructing a deterministic regulatory path

[0132] In expressing support for the control diagram In this context, deterministic regulatory pathways are constructed using transcription factors as anchors. For transcription factors... The regulation path is represented as follows:

[0133]

[0134] in:

[0135]

[0136] The path length satisfies:

[0137]

[0138] And any adjacent nodes satisfy:

[0139]

[0140] The number of pathways for each anchored transcription factor satisfies:

[0141]

[0142] in, Indicates the maximum path length. This represents the maximum number of pathways retained for each anchored transcription factor.

[0143] The set of globally deterministic control paths is represented as:

[0144]

[0145] in, This represents the set of anchored transcription factors that are retained.

[0146] The above construction method does not rely on random walks or random graph sampling, making the set of control paths obtained under the same input data and the same parameter settings repeatable.

[0147] Step 4: Constructing the transcription factor center module

[0148] For transcription factors Its outgoing neighborhood is defined as:

[0149]

[0150] Neighborhood target genes were ranked based on expression variance, drought resistance phenotype relevance, environmental response intensity, and the strength of functional annotation or regulatory evidence, and a subset of target genes was selected.

[0151]

[0152] The number of selected target genes satisfies the following:

[0153]

[0154] Where s represents the number of selected target genes. and These represent the lower and upper limits of the number of target genes, respectively.

[0155] Constructing the transcription factor central module:

[0156]

[0157] The set of retained transcription factor central modules is represented as follows:

[0158]

[0159] The final set of regulatory terms is obtained:

[0160]

[0161] in, For a set of deterministic control paths, It is a collection of central modules of transcription factors. It is a set of regulatory terms composed of deterministic regulatory pathways and transcription factor central modules.

[0162] Step 5: Calculate the activity of regulatory terms and form the activity characteristics of the selected regulatory terms.

[0163] For regulatory terms Its member gene set is as follows:

[0164]

[0165] No. Regulatory words in each sample The activity is:

[0166]

[0167] The activity matrix of the regulatory words was obtained:

[0168]

[0169] in, Indicates the number of regulatory words. It represents the field of nonnegative real numbers.

[0170] Calculate the non-zero support ratio of regulatory terms:

[0171]

[0172] in, This indicates an indicator function; it takes the value 1 if the condition within the parentheses is true, and 0 otherwise.

[0173] Calculate the average activity of the regulatory term:

[0174]

[0175] Calculate the variance of the activity of regulatory terms:

[0176]

[0177] Retain the control terms that meet the following conditions:

[0178]

[0179]

[0180] in, , and These represent the minimum non-zero support ratio, the maximum non-zero support ratio, and the minimum activity variance threshold, respectively.

[0181] The regulatory terms retained under the above conditions and their corresponding activity matrix columns constitute the activity features of the selected regulatory terms, and serve as inputs for subsequent non-negative document matrix construction, candidate gene site mapping, and dynamic regulatory network construction.

[0182] Step 6: Learn the thematic characteristics of the regulatory path

[0183] The gene expression features formed from the gene expression matrix obtained in step 1, and the regulatory word activity features formed from the regulatory word activity matrix selected in step 5, are assembled into a non-negative document matrix:

[0184]

[0185] Among them, the The document vector for each sample is:

[0186]

[0187] in, Indicates the first Standardized gene expression feature vectors of each sample Indicates the first The feature vector of the regulatory word activity of each sample and These are scaling factors for gene expression characteristics and regulatory word activity characteristics, respectively. For the document matrix... Perform topic modeling to obtain the sample topic representation matrix:

[0188]

[0189] And the topic-regulatory word load matrix:

[0190]

[0191] in, The sample topic representation matrix, For the topic-modulation word load matrix, To regulate the number of topics along the path, Indicates the first The sample at the th Weighting of each regulatory path theme Indicating regulatory terms In the Loads in the theme of each regulation path.

[0192] In one implementation, the topic modeling employs latent Dirichlet assignment, nonnegative matrix factorization, latent semantic analysis, probabilistic topic modeling, or nonnegative low-rank decomposition.

[0193] Step 7: Mapping candidate gene loci to thematic evidence on regulatory pathways

[0194] Based on the candidate gene locus matrix obtained in step 1, and combined with wheat genome public annotations or linkage interval information, candidate gene loci... Mapping to one or more genes, the mapping set is represented as:

[0195]

[0196] like Any gene in the list belongs to the regulatory term Then the regulatory word is used as a candidate gene locus. Evidence for regulatory pathways. Candidate gene loci. The set of regulatory evidence is as follows:

[0197]

[0198] The thematic evidence for the regulatory pathways of candidate gene loci is represented as follows:

[0199]

[0200] in:

[0201]

[0202] here, Indicates candidate gene loci The set of genes mapped to. Indicating regulatory terms The set of member genes, Indicates the relationship with candidate gene loci A set of related regulatory terms, Indicates the first The average weight of each regulatory pathway theme in all samples Indicating regulatory terms In the Loads in the theme of each regulation path.

[0203] This step enables the connection of candidate SNPs, InDel sites, or structural variant sites with regulatory pathways, transcription factor central modules, and regulatory pathway theme representations.

[0204] Step 8: Construct a dynamic regulation network from the regulation word activity matrix, and calculate the importance evidence of the dynamic regulation network: combined recognition score:

[0205] A dynamic regulation network is constructed for the word activity matrix under different time points, processing conditions, or environmental gradients.

[0206] Let the first The activity matrix of regulatory terms under specific time points, treatment conditions, or environmental gradients is as follows:

[0207]

[0208] In step 5, the elements in the activity matrix represent the activity of a certain regulatory word in a certain sample at the corresponding time point, treatment conditions or environmental gradient. The number of wheat samples under this condition is determined by the corresponding sample set, and the number of regulatory words is the number of elements in the regulatory word set. Indicates the first At a given time point, under specific processing conditions or environmental gradient, the [number]th [time point]... Regulatory words in each sample The activity, This indicates the number of wheat samples under this condition. Indicates the number of regulatory terms.

[0209] The corresponding drought resistance phenotypic matrix is ​​represented as follows:

[0210]

[0211] in, Indicates the first At a given time point, under specific processing conditions or environmental gradient, the [number]th [time point]... The first sample One drought resistance phenotypic indicator This indicates the number of drought resistance phenotypic indicators.

[0212] Establish time-dependent regulatory relationships between adjacent time points, treatment conditions, or environmental gradients:

[0213]

[0214] in, Indicates the first Control words at a specific time point, under specific treatment conditions, or under environmental gradients , Indicates the first Control words at a specific time point, under specific treatment conditions, or under environmental gradients .

[0215] The time dependence strength between regulatory terms is expressed as follows:

[0216]

[0217] in, Indicating regulatory terms In the An activity vector at a given time point, under specific treatment conditions, or within an environmental gradient. Indicating regulatory terms In the An activity vector at a given time point, under specific treatment conditions, or within an environmental gradient. This represents the function for calculating time dependencies. The time dependence calculation function is determined based on the changes in the activity of the regulatory term, temporal correlation, Granger causality, dynamic Bayesian network, graph neural network edge weights, or the joint changes of the regulatory term and the drought resistance phenotype. Specifically, the input of the time dependence calculation function is the activity vector of the first regulatory term at the current time point, the activity vector of the second regulatory term at the next time point, and the corresponding drought resistance phenotype matrix. Its internal operations include the calculation of changes in the activity of the regulatory term, temporal correlation, causality, dynamic Bayesian network, graph neural network edge weights, or the joint changes of the regulatory term and the drought resistance phenotype. Its output is the time dependence strength from the first regulatory term to the second regulatory term.

[0218] When the time dependency strength meets the preset threshold, the corresponding dynamic adjustment edge is retained:

[0219]

[0220] in, This indicates the threshold for dynamically adjusting the edge retention.

[0221] The dynamic control network is composed of control word nodes and their dynamic control edges:

[0222]

[0223] in, This represents the set of control word nodes in a dynamic control network. This represents the set of time-dependent control edges in a dynamic control network.

[0224] For nodes in a dynamic control network Its network importance score is expressed as:

[0225]

[0226] in, Represents a node The degree or weighted degree, Represents a node betweenness centrality, Represents a node The PageRank value, a webpage ranking algorithm. Represents a node Stability at different time points, under different treatment conditions, or under different environmental gradients The weights are non-negative and satisfy:

[0227]

[0228] Node stability is represented as:

[0229]

[0230] in, Represents a node Sets of activity or importance scores at different time points, under different treatment conditions, or under different environmental gradients. express standard deviation express The average value, To prevent positive numbers with a denominator of zero.

[0231] Through the above steps, the system identifies core regulatory terms, core regulatory pathways, and core regulatory modules that exhibit temporal variation characteristics, high network centrality, and relatively stable cross-time points or environmental gradients during drought stress. These are then used as the basis for calculating the comprehensive identification score and verification priority of candidate gene loci.

[0232] Step 9: Construct a comprehensive identification score for candidate gene loci based on evidence of dynamic network importance.

[0233] For candidate gene loci Calculate the following evidence scores.

[0234] The evidence for the regulatory path theme is obtained in step 7. .

[0235] Evidence of the association between drought resistance phenotype and drought resistance is expressed as follows:

[0236]

[0237] in, Indicates candidate gene loci Genotype coding, allele frequencies, or locus characteristics in different samples This indicates the drought resistance phenotype index or the comprehensive drought resistance phenotype score.

[0238] Evidence of environmental response is expressed as follows:

[0239]

[0240] in, Indicates gene Expression vectors at different samples, processing conditions, or time points This indicates drought-related environmental factors or environmental stress intensity indicators.

[0241] The evidence for the functional annotation is represented as follows:

[0242]

[0243] in, Indicates gene Functional annotation scores are assigned based on whether the gene is involved in stress resistance, ABA signaling, ROS scavenging, root development, stomatal regulation, osmotic regulation, or known drought resistance pathways.

[0244] Evidence of the importance of dynamic networks is expressed as follows:

[0245]

[0246] in, This represents the dynamic importance score of the mapped gene, obtained by aggregating the importance scores of relevant regulatory term nodes containing the mapped gene g. This represents the node importance score of the regulatory term w associated with candidate gene locus j in the dynamic regulatory network. Candidate gene locus The overall recognition score is:

[0247]

[0248] in, The weights are non-negative and satisfy:

[0249]

[0250] Here, the summation constraint indicates that the sum of the weights of the five comprehensive identification scores is one, and u represents the evidence type index. This represents the weight corresponding to the u-th type of evidence; the weight is determined through cross-validation, expert priors, Bayesian factors, model interpretation results, or validation data.

[0251] Step 10: Output candidate gene loci and validation priority

[0252] Calculate the validation priority for each candidate gene locus:

[0253]

[0254] in, Indicates candidate gene loci The comprehensive recognition score, Evidence indicating the thematic approach to regulation. This demonstrates the importance of dynamic networks. This represents the consistency score, calculated based on the stability of the comprehensive identification score obtained in step 9 across different groups, taking into account the sample, time point, or environmental conditions grouped in step 1. The weights are non-negative.

[0255] in, satisfy:

[0256]

[0257] Here, the summation constraint indicates that the sum of the four verification priority weights is one, and r represents the index of the priority component. This represents the weight corresponding to the r-th component; the system follows... Output highly reliable wheat drought resistance candidate gene loci, corresponding candidate genes, deterministic regulatory pathways, transcription factor central modules, regulatory pathway themes, drought resistance phenotype association evidence, environmental response evidence, functional annotation evidence, and recommended validation priorities, from highest to lowest.

[0258] Example 1: Theme Modeling of Regulatory Pathways under Multimodal Data of Wheat Drought Stress

[0259] As shown in Figure 1, in one specific embodiment, the present invention is used to identify high-confidence drought-resistant candidate gene loci under wheat drought stress conditions.

[0260] First, samples from multiple wheat varieties were collected under normal water supply and drought stress conditions. For each sample, gene expression matrices, candidate SNP or InDel site matrices, drought resistance phenotype data, environmental factor data, functional annotation data, and wheat gene regulatory networks were obtained. The drought resistance phenotype data included leaf water content, biomass, root length, water use efficiency, and relative survival rate; environmental factors included soil moisture, temperature, air humidity, and drought treatment duration.

[0261] Then, the input data was standardized to obtain a standardized expression matrix, a standardized phenotype matrix, and a standardized environmental factor matrix. Subsequently, the wheat gene regulatory network was subjected to intersection filtering with the observable gene set in the expression matrix to obtain an expression support regulation map.

[0262] As shown in Figure 2, the original wheat transcription factor-target (TF-target) regulatory network or co-expression regulatory network is first filtered by an expression support network to remove unobservable gene nodes and their related regulatory edges in the current expression matrix. Subsequently, the expression support regulatory graph is transformed into deterministic regulatory pathways and transcription factor central modules, respectively. Together, they constitute a set of regulatory terms, providing a foundation for subsequent construction of the regulatory term activity matrix and thematic modeling of regulatory pathways.

[0263] In the expression-supported regulatory map, deterministic regulatory pathways are constructed using dehydration-responsive element-binding protein (DREB), the WRKY transcription factor family, the MYB transcription factor family, the NAC transcription factor family, and the basic leucine zipper (bZIP) transcription factor family as anchor nodes. Each pathway begins with an anchor transcription factor and connects to downstream target genes along the regulatory edge. The maximum pathway length is set to... Each transcription factor retains at most There are several pathways. Simultaneously, for each transcription factor, target genes with high expression variance, strong phenotypic correlation, or significant environmental response in its neighborhood are selected to construct a transcription factor central module.

[0264] As shown in Figure 3, the system uses deterministic regulatory pathways and transcription factor central modules as regulatory terms, calculates their activities in different samples, treatment conditions, or time points, and forms a regulatory term activity matrix. Subsequently, the system assembles the regulatory term activity features and gene expression features into a non-negative document matrix, and obtains the sample topic representation matrix and topic-regulatory term payload matrix through topic modeling. The regulatory pathway topic represents a weighted combination of multiple regulatory pathways and modules, rather than a single regulatory pathway or a single module.

[0265] Furthermore, the system maps candidate SNPs, InDel sites, or structural variant sites to adjacent genes, overlapping genes, promoter region genes, enhancer region genes, or genes within the same functional region. If the gene mapped to a candidate gene site belongs to a certain regulatory term, then that regulatory term is used as evidence of the regulatory pathway of that candidate gene site.

[0266] As shown in Figure 4, candidate SNPs, InDel sites, or structural variant sites are first mapped to adjacent genes, overlapping genes, promoter region genes, enhancer region genes, or genes in the same functional region. Subsequently, the mapped genes are further associated with regulatory terms and regulatory pathway themes. The system constructs a dynamic regulatory network by combining the changes in regulatory term activity at different time points or under environmental gradients, and outputs the validation priority of candidate gene sites based on the comprehensive recognition score and the importance of the dynamic network.

[0267] Finally, the system integrates evidence on regulatory pathways, evidence on drought resistance phenotype associations, evidence on environmental responses, evidence on functional annotations, and evidence on the importance of dynamic networks to calculate a comprehensive identification score. It then outputs highly reliable wheat drought resistance candidate gene loci and their corresponding regulatory pathways, transcription factor central modules, and validation priorities according to preset score thresholds or ranking ratios.

[0268] Example 2: Calculation of Regulatory Word Activity Based on Sparse Representation Modeling

[0269] Example 2 further defines the steps of “inputting wheat multimodal data” in Figure 1 and “constructing the regulatory word activity matrix” in Figure 3.

[0270] In another embodiment, the gene expression matrix in step 1 is first processed using a zero-inflation model to distinguish between true zero expression and noisy zero expression. For genes... In the sample expression values ​​in Modeling is performed using a zero-inflation distribution:

[0271]

[0272] in, Represents probability. Indicates the first In the nth sample The expression value of each gene, Represents positive integers to express counting. Indicates the first The probability of zero inflation for each gene. Indicates the first In the nth sample The expression intensity parameter of each gene, This represents the natural exponential function.

[0273] After zero-inflation modeling, the system obtains a denoised expression matrix or sparse expression latent features, which are then used as input for calculating regulatory word activity. This approach reduces the impact of technical zeros or noise zeros on regulatory pathway activity estimation and improves the identification ability of low-abundance drought-resistant response genes or weakly effective candidate gene loci.

[0274] Example 3: Tracing the Thematic Evidence from Candidate Gene Loci to Regulatory Pathways

[0275] Example 3 further defines the mapping relationship of "candidate gene locus - gene - regulatory word - regulatory pathway theme" in Figures 2, 3 and 4.

[0276] In another embodiment, candidate gene loci Mapping to one or more genes, the mapping set is represented as:

[0277]

[0278] like Any gene in the list belongs to the regulatory term Then the regulatory word is used as a candidate gene locus. Evidence for regulatory pathways. Candidate gene loci. The set of regulatory evidence is as follows:

[0279]

[0280] The supporting evidence for the regulatory pathways of candidate gene loci is further described as follows:

[0281]

[0282] This embodiment can link candidate SNPs, InDel sites, or structural variant sites with regulatory pathways, transcription factor central modules, and regulatory pathway theme representations, thereby providing traceable regulatory evidence for subsequent qPCR, CRISPR-Cas9, or ChIP-seq validation.

[0283] Example 4: Dynamically Adjusting the Priority Output of Network Parsing and Verification

[0284] Example 4 is a further limitation on the "dynamically adjusting network parsing and verification priority output" step in Figure 4.

[0285] In another embodiment, the system constructs a dynamic regulation network based on the activity matrix of the regulating words at different time points, under different processing conditions, or with different environmental gradients. This dynamic regulation network is determined based on time correlation, Granger causality, dynamic Bayesian networks, graph neural network edge weights, or the trend of changes in the activity of the regulating words.

[0286] The system further calculates the network importance score of the mapped gene or regulatory term corresponding to the candidate gene locus based on node degree, betweenness centrality, PageRank value, and cross-environment stability in the dynamic regulatory network. Subsequently, the system integrates the comprehensive identification score, regulatory pathway theme evidence, dynamic network importance evidence, and cross-environment consistency to calculate the validation priority.

[0287]

[0288] The system according to Output a list of candidate gene loci from high to low, and label each candidate gene locus with its corresponding regulatory pathway, transcription factor central module, regulatory pathway theme, drought resistance phenotype association evidence, environmental response evidence, functional annotation evidence, and recommended validation methods. The recommended validation methods include qPCR, CRISPR-Cas9, ChIP-seq, or molecular breeding validation experiments.

[0289] Compared with traditional methods for screening candidate gene loci for drought resistance in wheat, this invention has at least the following differences.

[0290] Traditional methods typically use single-gene expression differences, single genome-wide association study (GWAS) statistics, or single machine learning feature importance as screening criteria. This invention transforms gene regulatory networks into deterministic regulatory pathways and transcription factor central modules, and further obtains regulatory pathway theme features through topic modeling, realizing a traceable mapping from network structure to candidate gene locus evidence.

[0291] Traditional co-expression network analysis typically focuses on nodes or edges, making it difficult to explain multi-gene synergistic regulatory patterns. This invention integrates multiple regulatory pathways and transcription factor central modules into regulatory themes through regulatory word activity matrices and topic modeling, thereby enabling the identification of higher-level drought resistance regulatory programs.

[0292] Traditional graph models or deep learning methods often output black-box representations or candidate gene lists; this invention can simultaneously output candidate gene loci, regulatory pathway evidence, transcription factor central module evidence, regulatory theme evidence, and validation priorities, making the results more suitable for subsequent experimental validation.

[0293] Traditional static network methods are difficult to characterize the temporal dynamic changes under drought stress; this invention can combine the activity of regulatory words under different time points, treatment conditions and environmental gradients to construct a dynamic regulatory network and identify the core regulatory paths and core regulatory modules.

[0294] A computer-readable storage medium storing a computer program thereon, wherein when executed by a processor, the computer program causes the device containing the computer-readable storage medium to perform the wheat drought-resistant gene locus screening method described above, which integrates regulatory pathway theme features and controls for false discovery rates without p-values. The computer program includes computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory, and other memories.

[0295] An electronic device includes a memory and a processor, wherein the memory stores a program executable on the processor, and the processor executes the program to implement the wheat drought resistance gene locus screening method described above, which combines fusion regulatory pathway theme features with no P-value false discovery rate control.

[0296] If the modules / units integrated in the electronic device described in this application are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can also be implemented by a computer program instructing related hardware devices. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above.

[0297] Furthermore, the computer-readable storage medium may primarily include a stored program area and a stored data area, wherein the stored program area may store the operating system, an application program required for at least one function, etc.; and the stored data area may store data created based on the use of blockchain nodes, etc.

[0298] The computer-readable storage medium stores computer-readable instructions, which are executed by a processor in an electronic device to implement the wheat drought resistance gene locus screening method described in any of the above embodiments, which combines fusion regulatory pathway theme features with no P-value false discovery rate control.

[0299] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0300] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0301] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

[0302] It should be noted that the terms "comprising" and "having" and any variations thereof in the specification and claims of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or device.

[0303] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Those skilled in the art can modify, replace, or combine the data types, regulatory network sources, path length thresholds, module size thresholds, topic model types, dynamic network construction methods, candidate gene locus mapping methods, comprehensive identification score calculation methods, and verification priority ranking methods in the above embodiments without departing from the technical concept of the present invention; such modifications, replacements, or combinations should still fall within the scope of protection of the present invention.

[0304] Note that the above description is merely a preferred embodiment and application of the technical principles of the present invention. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the specific embodiments described herein, and may include many other effective embodiments without departing from the concept of the present invention. The scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for identifying candidate gene loci for drought resistance in wheat based on deterministic regulatory pathway theme modeling, characterized in that: Includes the following steps, Step 1: Obtain multimodal data of wheat under drought stress conditions. The multimodal data includes gene expression matrices to characterize gene expression levels in wheat samples, genotypes, tissues, treatment conditions, or time points; candidate gene locus matrices to characterize SNPs, InDel, structural variations, allele frequencies, locus coverage, or genotype encoding; drought resistance phenotype data to characterize leaf water content, biomass, root length, root volume, stomatal conductance, water use efficiency, relative survival rate, chlorophyll content, or drought stress index; environmental factor data to characterize soil moisture, air humidity, temperature, light intensity, drought treatment time, field water gradient, or transpiration intensity; and gene function annotation data and gene regulatory networks to characterize the relationship between candidate genes and stress resistance, ABA signaling, ROS scavenging, root development, stomatal regulation, osmotic regulation, or known drought resistance pathways. Step 2: Based on the observable gene set in the gene expression matrix, filter the gene regulatory network to obtain an expression support regulation map; Step 3: Construct deterministic regulatory pathways and transcription factor central modules based on the expression support regulation graph, and use the deterministic regulatory pathways and transcription factor central modules as regulatory terms; Step 4: Calculate the regulatory activity of the regulatory words in wheat samples, treatment conditions, or time points to obtain the regulatory word activity matrix; Step 5: Based on the regulatory word activity matrix obtained in Step 4, and combined with the drought resistance phenotype data and environmental factor data obtained in Step 1, regulatory words are screened according to non-zero support ratio, activity variance, drought resistance phenotype correlation or environmental response intensity to obtain the screened regulatory words and their activity characteristics. Step 6: Based on the gene expression matrix obtained in Step 1 and the activity features of the selected regulatory words obtained in Step 5, assemble a non-negative document matrix and obtain the regulatory path theme features through topic modeling. Step 7: Based on the candidate gene locus matrix obtained in Step 1, the screened regulatory words obtained in Step 5, and the topic-regulatory word payload matrix obtained in Step 6, the candidate gene loci are mapped to the corresponding genes, regulatory words, and regulatory pathway topics to generate regulatory pathway topic evidence for the candidate gene loci. Step 8: Based on the regulatory word activity matrix obtained in Step 4 and the corresponding time point, processing conditions or environmental gradient information in Step 1, construct a dynamic regulatory network and calculate the evidence of the importance of the dynamic network. Based on the candidate gene locus matrix, drought resistance phenotype data, environmental factor data, and gene function annotation data obtained in step 1, drought resistance phenotype association evidence, environmental response evidence, and functional annotation evidence are calculated respectively. Then, based on the regulatory pathway theme evidence obtained in step 7, as well as the above-mentioned drought resistance phenotype association evidence, environmental response evidence, functional annotation evidence, and dynamic network importance evidence, a comprehensive identification score for candidate gene loci is obtained. Step 9: Based on the comprehensive identification score of candidate gene loci obtained in Step 8, determine the set of highly reliable wheat drought-resistant candidate gene loci according to the preset score threshold or sorting ratio, and output the set of highly reliable wheat drought-resistant candidate gene loci and its corresponding regulatory pathway evidence, transcription factor central module evidence, regulatory pathway theme evidence and verification priority.

2. The method for identifying wheat drought-resistant candidate gene loci based on deterministic regulatory pathway theme modeling according to claim 1, characterized in that, The support regulation graph in step 2 is constructed through the following steps: Step 2.1: Extract the set of observable genes from the gene expression matrix; Step 2.2: Retain regulatory edges from the original gene regulatory network where both ends of the regulatory edge belong to the observable gene set; Step 2.3: The retained nodes and control edges form an expression support control graph; Step 2.4, the expression support regulation map is used to ensure that the deterministic regulatory pathways, transcription factor central modules and regulatory word activities constructed in the subsequent process are all calculated from the measurable genes in the current gene expression matrix.

3. The method for identifying wheat drought-resistant candidate gene loci based on deterministic regulatory pathway theme modeling according to claim 1, characterized in that, The deterministic control path in step 3 is constructed through the following steps: Step 3.1: Using the transcription factors in the regulatory graph as anchor nodes, extend downstream along the regulatory edge to target genes to generate an ordered regulatory pathway starting from the transcription factors. Step 3.2, limit the maximum length of the regulatory pathway and the maximum number of pathways retained by each transcription factor to avoid excessive domination of the regulatory word set by highly connected transcription factors; Step 3.3: Screen regulatory pathways based on evidence from gene expression of pathway members, drought resistance phenotype, environmental response intensity, or functional annotation.

4. The method for identifying wheat drought-resistant candidate gene loci based on deterministic regulatory pathway theme modeling according to claim 1, characterized in that, The transcription factor central module in step 3 is constructed through the following steps: Step 3.4: Determine the set of outgoing neighborhood target genes for each transcription factor in the expression support regulation graph; Step 3.5: Sort the target genes according to their expression variance, drought resistance phenotype correlation, environmental response intensity, and the strength of functional annotation evidence or regulatory evidence. Step 3.6: Select target genes that meet the preset quantity range or evidence threshold, and together with the corresponding transcription factors, form the transcription factor central module.

5. The method for identifying wheat drought-resistant candidate gene loci based on deterministic regulatory pathway theme modeling according to claim 1, characterized in that, In step 4, the activity of the regulatory word is calculated through the following steps: Step 4.1: Define each deterministic regulatory pathway or each central module of a transcription factor as a regulatory term; Step 4.2: Calculate sample-level regulatory activity based on the expression values ​​of regulatory term member genes in wheat samples; Step 4.3: Calculate the non-zero support ratio and activity variance based on the distribution of regulatory word activity in different samples, treatment conditions, or time points; Step 4.4 outputs the non-zero support ratio and activity variance of the regulatory words, which serve as the basis for screening regulatory words in step 5.

6. The method for identifying wheat drought-resistant candidate gene loci based on deterministic regulatory pathway theme modeling according to claim 1, characterized in that, The control path theme features in step 6 are obtained through the following steps: Step 6.1: Scale and splice the sample-level gene expression features formed by the gene expression matrix obtained in Step 1 and the screened regulatory word activity features obtained in Step 5 to form a non-negative document matrix; Step 6.2: Perform topic modeling on the non-negative document matrix to obtain the sample topic representation matrix and the topic-regulation word load matrix; Step 6.3: From the sample topic representation matrix and topic-regulatory word load matrix, obtain the regulatory pathway topic features used for candidate gene site mapping and regulatory pathway topic evidence calculation in Step 7.

7. The method for identifying wheat drought-resistant candidate gene loci based on deterministic regulatory pathway theme modeling according to claim 1, characterized in that, The candidate gene locus mapping in step 7 is performed through the following steps: Step 7.1: Based on the candidate gene locus matrix obtained in Step 1, and combined with wheat genome public annotation or linkage interval information, map the candidate gene loci to genes that overlap with, are adjacent to, are located in the same functional region, are located in the promoter region, are located in the enhancer region, or are located within the linkage interval. Step 7.2: Based on whether the mapped gene belongs to the deterministic regulatory pathway or transcription factor central module retained after screening in Step 5, establish the correspondence between candidate gene sites and screened regulatory words. Step 7.3: Based on the correspondence between the candidate gene loci and the selected regulatory terms, and the contribution of the selected regulatory terms to the regulatory pathway theme, establish the correspondence between the candidate gene loci and the regulatory pathway theme.

8. The method for identifying wheat drought-resistant candidate gene loci based on deterministic regulatory pathway theme modeling according to claim 1, characterized in that, The comprehensive identification score in step 8 is obtained by fusing the following evidence: Evidence on regulatory pathway themes is used to characterize the contribution of candidate gene loci to regulatory pathway themes; Evidence of drought resistance phenotype association is used to characterize the association between candidate gene loci or mapping genes and drought resistance phenotype; Environmental response evidence is used to characterize the association between candidate gene loci or mapped genes and drought environmental factors; Functional annotation evidence is used to characterize the relationship between candidate gene loci or mapped genes and drought-related functional pathways. Evidence of the importance of dynamic networks is used to characterize the topological importance of candidate gene loci or mapped genes in deterministic regulatory pathways, transcription factor central modules, or dynamic regulatory networks.

9. The method for identifying wheat drought-resistant candidate gene loci based on deterministic regulatory pathway theme modeling according to claim 1, characterized in that, The dynamic control network parsing step in step 8 includes: A dynamic regulatory network was constructed based on data on the activity of regulatory terms and drought resistance phenotypes at different time points, under different treatment conditions or environmental gradients. Identify core control paths and core control modules based on time dependencies, causal relationships, edge weight changes, node importance, or module stability; The core regulatory pathways and core regulatory modules are used as the basis for prioritizing candidate gene loci for validation.

Images