Method for constructing and predicting plant intrinsic disordered region functional group based on sequence syntax

Abstract

The application discloses a plant intrinsic IDR functional group construction and prediction method based on sequence syntax, and the application constructs candidate GIN models and candidate GIN-Banks corresponding to various plant species through 90-dimensional truncated Z-score syntax feature vectors, and determines formal GIN models and formal GIN-Banks by combining low-differentiation mapping item proportions, the number of syntax clusters with significant functional annotations and functional annotation coverage degrees; the application takes intraspecific syntax clusters of different plant species as basic units, constructs cross-species syntax cluster similarity graphs based on the similarity of syntax cluster clustering centers, and combines graph community discovery and functional evidence to form formal cross-species plant IDR functional groups pIFC, so that even if IDR sequences in different plant species lack significant conservation, cross-species functional correspondence can be established according to sequence syntax similarity, and automatic division of plant IDR functional groups is realized.

Description

Technical Field

[0001] This invention belongs to the field of biotechnology, specifically relating to a method for constructing and predicting functional groups in the intrinsic disordered regions of plants based on sequence grammar. Background Technology

[0002] In recent years, with the development of proteomics, bioinformatics, and molecular biology, protein sequence and function databases have been continuously improved, and researchers have gradually moved from understanding to modifying functional motifs and conserved sites in proteins. Meanwhile, intrinsically disordered regions (IDRs), as a class of protein regions lacking stable three-dimensional structures, have increasingly attracted attention for their role in plant development, stress response, and signal transduction. However, unlike highly conserved functional domains, IDRs typically lack sequence conservation, making it difficult for existing functional annotation methods that rely on sequence conservation and protein structure to accurately classify and analyze them across species. Currently, IDR research is gradually expanding from disordered region identification to sequence characteristics, evolutionary constraints, and molecular grammar analysis. However, existing methods still lack a unified framework for cross-plant species analysis, making it difficult to systematically analyze the sequence grammar of IDRs, classify them functionally, construct cross-species functional maps, and predict the functions of new sequences. Summary of the Invention

[0003] The purpose of this invention is to address the aforementioned problems in the existing technology by providing a method for constructing and predicting functional groups in the intrinsic disordered region of plants based on sequence grammar.

[0004] The above-mentioned objectives of the present invention are achieved by the following technical means:

[0005] A method for constructing and predicting functional taxa in the intrinsic disordered region of plants based on sequence grammar includes the following steps:

[0006] Step 1: Obtain the complete proteome data file of the plant species;

[0007] Step 2: Identify the IDR sequences located in the intrinsically disordered region of protein sequences in the whole proteome data file of the plant species and clean the IDR sequences.

[0008] Step 3: Calculate the 90-dimensional truncated Z-score syntactic feature vector of the IDR sequence using the NARDINI+ algorithm;

[0009] Step 4: Set the range of K values. Based on the 90-dimensional truncated Z-score grammatical feature vector, use GIN models with different K values ​​to cluster all IDR sequences in each plant species. According to the clustering results of each plant species, select multiple candidate K values ​​for each plant species and use the GIN model with the candidate K values ​​as the candidate GIN model. K is the number of cluster centers of the GIN model.

[0010] Step 5: Cluster the IDR sequences within the corresponding plant species using the candidate GIN models for each plant species to obtain the candidate GIN-Bank for each plant species. The candidate GIN-Bank includes the IDR sequences and the grammatical clusters of the IDR sequences.

[0011] Step 6: Perform GO enrichment analysis and KEGG enrichment analysis on the grammar clusters and construct a functional annotation result table for the grammar clusters;

[0012] Step 7: Based on the functional annotation results table of each grammatical cluster of each plant species, determine the species-specific final cluster number K value of each plant species and the corresponding formal GIN model and formal GIN-Bank.

[0013] Step 8: Construct a cross-species grammar cluster similarity graph, which includes grammar cluster nodes, preserved edges, and edge weights; use the Leiden graph community detection algorithm to divide the cross-species grammar cluster similarity graph into cross-species grammar cluster communities; use the pvclust function to perform hierarchical clustering on all cross-species grammar cluster communities to obtain multiple candidate cross-species functional units; upgrade the candidate cross-species functional units to formal cross-species plant IDR functional groups pIFC;

[0014] Step 9: Construct a sequence-level plant IDR function predictor pIFCmap. The pIFCmap takes the IDR sequence to be tested as input, maps the IDR sequence to the syntax cluster of the formal GIN-Bank of the plant species, and outputs the pIFC function category prediction result of the IDR sequence to be tested according to the correspondence between the syntax cluster and pIFC.

[0015] As mentioned above, candidate GIN-Bank also includes the plant species associated with the IDR sequence, the grammatical cluster identification number, and the mapping distance from the IDR sequence to the nearest cluster center. Mapping distance from IDR sequence to the second nearest cluster center and mapping discrimination index Syntax clusters are constructed centered on each cluster center. The syntax cluster assignment result of the IDR sequence is the syntax cluster centered on the nearest cluster center, and the assignment number of the syntax cluster is the syntax cluster number centered on the nearest cluster center; the mapping discriminant index is... Map the discrimination index IDR sequences that are below a set threshold are marked as low-discrimination mapping entries;

[0016] The functional annotation results table of the grammatical clusters includes the plant species, the candidate K value corresponding to the candidate GIN-Bank, the grammatical cluster number, the significantly enriched GO annotation entries, the significantly enriched KEGG pathways, and the corrected p-value and enrichment fold of each significantly enriched GO annotation entry and each significantly enriched KEGG pathway.

[0017] As mentioned above, the cross-species grammatical cluster similarity graph is constructed through the following steps:

[0018] Step 8.1: Define each grammatical cluster in the formal GIN-Bank of each plant species as a grammatical cluster node. The sequence grammatical features of the grammatical cluster node are obtained by using the 90-dimensional truncated Z-score grammatical feature vector of the cluster center of the corresponding grammatical cluster.

[0019] Step 8.2: Define the connection between any two syntactic cluster nodes from different plant species as a candidate edge, and calculate the edge weight of the candidate edge between any two syntactic cluster nodes based on the following formula:

[0020] ;

[0021] In the formula, and These are the 90-dimensional truncated Z-score grammatical feature vectors representing the cluster centers of the two grammatical cluster nodes. and They are respectively and In the Dimensionally truncated coordinates on Z-score syntactic features, index Take values ​​from 1 to 90; Let the edge weights be the candidate edges between two grammar cluster nodes;

[0022] Step 8.3: Set the edge weight threshold, and select candidate edges whose edge weights are greater than the edge weight threshold as retained edges. All grammar cluster nodes, retained edges and their corresponding edge weights constitute a cross-species grammar cluster similarity graph.

[0023] The cross-species grammatical cluster community is divided through the following steps:

[0024] Step 8.4.1: Using the Leiden graph community detection algorithm, output a community number for each grammar cluster node in the cross-species grammar cluster similarity graph, and assign grammar cluster nodes with the same community number to a cross-species grammar cluster community. All cross-species grammar cluster communities constitute the community partitioning result.

[0025] Step 8.4.2: Pre-set a baseline random seed, use the Leiden graph community detection algorithm with the baseline random seed to partition the cross-species grammatical cluster similarity graph to obtain the baseline community partitioning result, and then use the Leiden graph community detection algorithm with multiple random seeds to partition the cross-species grammatical cluster similarity graph to obtain multiple sets of control community partitioning results. The baseline community partitioning result and the multiple sets of control community partitioning results constitute a total of N1 community partitioning results.

[0026] Step 8.4.3: First, in the cross-species grammar cluster community in the baseline community partitioning results, construct any two grammar cluster nodes within the cross-species grammar cluster community as an internal node pair. The set of internal node pairs of the cross-species grammar cluster community is composed of all internal node pairs of the cross-species grammar cluster community.

[0027] Secondly, the stability score is calculated based on the internal node pair set for cross-species grammatical cluster communities in the following manner:

[0028] The number of times two grammar cluster nodes in an internal node pair are assigned to the same community in N1 community partitioning results is counted, and the number is divided by N1 to obtain the co-community support rate of the internal node pair; the average co-community support rate of all internal node pairs in the set of internal node pairs of the cross-species grammar cluster community is calculated as the stability score of the cross-species grammar cluster community.

[0029] Finally, a stability threshold is set to select all cross-species grammar cluster communities with stability scores greater than the stability threshold from the baseline community partitioning results to enter hierarchical clustering; cross-species grammar cluster communities with stability scores less than the stability threshold are not included in hierarchical clustering.

[0030] As described above, the pvclust function is used to perform hierarchical clustering of all cross-species grammatical cluster communities to obtain multiple candidate cross-species functional units. The specific steps include:

[0031] Step 8.5.1: Group the grammar cluster nodes in the cross-species grammar cluster community according to plant species. If a group contains only one grammar cluster node, the 90-dimensional truncated Z-score grammar feature vector of the cluster center of the grammar cluster node is used as the species representative vector of the plant species in the cross-species grammar cluster community. If a group contains multiple grammar cluster nodes, the average of the 90-dimensional truncated Z-score grammar feature vectors of the cluster centers of all grammar cluster nodes in the group is used as the species representative vector of the plant species in the cross-species grammar cluster community.

[0032] Finally, the species representative vectors of all plant species within the cross-species grammar cluster community are averaged to obtain the 90-dimensional community average grammar vector of the cross-species grammar cluster community.

[0033] Construct a candidate community grammar matrix for cross-species grammar clusters. Each row of the candidate community grammar matrix corresponds to a cross-species grammar cluster community, and each column corresponds to a grammatical feature dimension in the 90-dimensional community average grammar vector.

[0034] Step 8.5.2: Using the candidate community grammar matrix as input, perform hierarchical clustering using the pvclust function, and output the hierarchical clustering tree between cross-species grammar cluster communities and the AU value of each clustering branch;

[0035] If multiple cross-species grammatical cluster communities are grouped into the same cluster branch in a hierarchical clustering tree, and the following conditions are met simultaneously, the cluster branch is considered a candidate for merging:

[0036] Condition 1: The clustering branch is not the root branch that includes all cross-species grammatical cluster communities; Condition 2: The clustering branch includes at least two cross-species grammatical cluster communities; Condition 3: The AU value of the clustering branch is not lower than the set AU value threshold.

[0037] Then, all candidate merge branches that do not contain each other are determined as a final merge branch; for multiple candidate merge branches that have parent-child inclusion relationships, only the candidate merge branch with the most cross-species grammar cluster communities is retained as the final merge branch.

[0038] Finally, all cross-species grammatical cluster communities within each final merge branch are merged into a candidate cross-species functional unit.

[0039] As described above, upgrading a candidate cross-species functional unit to a formal cross-species plant IDR functional group pIFC specifically includes the following steps:

[0040] Step 8.6.1: First, construct a summary table of functional evidence for candidate cross-species functional units. The summary table includes the plant species, taxonomic classes, grammatical cluster numbers, and significantly enriched GO annotation entries included in the candidate cross-species functional units. Then, count the number of grammatical clusters, plant species, and taxonomic classes corresponding to the same significantly enriched GO annotation entries in the summary table.

[0041] Step 8.6.2: When there are significantly enriched GO annotation entries in the functional evidence summary table that meet the requirements for cross-species functional consistency, the candidate cross-species functional unit is upgraded to a formal cross-species plant IDR functional group pIFC.

[0042] The requirement for cross-species functional consistency is as follows: if the proportion of plant species supporting a significantly enriched GO annotation entry and the proportion of plant taxonomic classes supporting a significantly enriched GO annotation entry in the candidate cross-species functional unit are both not lower than a preset threshold, then the significantly enriched GO annotation entry meets the requirement for cross-species functional consistency.

[0043] The plant species support ratio for a significantly enriched GO annotation entry is the proportion of the number of plant species supporting a significantly enriched GO annotation entry to the total number of plant species in the candidate cross-species functional unit; the plant taxonomic class support ratio is the proportion of the number of plant taxonomic classes supporting a significantly enriched GO annotation entry to the total number of plant taxonomic classes in the candidate cross-species functional unit.

[0044] Step 8.6.3: Sort and number all formal cross-species plant IDR functional groups pIFC:

[0045] Condition 1: There are more significantly enriched GO annotation entries that meet the requirement of cross-species functional consistency;

[0046] Condition 2: The formal cross-species plant IDR functional group pIFC covers more plant taxonomic classes;

[0047] Condition 3: The formal cross-species plant IDR functional group pIFC covers more plant species;

[0048] Condition 4: In the N1 community partitioning results, the total number of times that all two grammatical cluster nodes constituting internal node pairs in the formal cross-species plant IDR functional taxonomy pIFC are assigned to the same community is greater.

[0049] The comparison of the next condition only proceeds when the previous condition is the same; after sorting, all formal cross-species plant IDR functional groups pIFC are numbered sequentially according to the sorting order.

[0050] As described above, the sequence-level plant IDR function predictor pIFCmap obtains the pIFC function category prediction results of the IDR sequence to be tested through the following steps:

[0051] Step 9.1: Construct the pIFCmap reference package for the sequence-level plant IDR functional predictor pIFCmap. The pIFCmap reference package includes the numbered official cross-species plant IDR functional groups pIFC, as well as the official GIN model, official GIN-Bank, and functional annotation result table of each grammatical cluster in the official GIN-Bank for each plant species; the pIFCmap reference package also includes 54-dimensional compositional background statistical parameters for each plant species.

[0052] Step 9.2: The sequence-level plant IDR function predictor pIFCmap cleans the IDR sequence to be tested;

[0053] Step 9.3: The sequence-level plant IDR function predictor pIFCmap then calculates the 90-dimensional truncated Z-score grammatical feature vector of the cleaned IDR sequence to be tested relative to the 54-dimensional compositional background statistical parameters of different plant species.

[0054] Step 9.4: The sequence-level plant IDR function predictor pIFCmap uses the formal GIN model of each plant species in the pIFCmap reference package to calculate the mapping distance from the IDR sequence to be tested to the cluster center of each grammatical cluster in the formal GIN-Bank. Then, the mapping distance from the IDR sequence to be tested to the cluster center of all grammatical clusters in the formal GIN-Bank of all plant species is sorted from smallest to largest to obtain the sorting table of the grammatical clusters of the IDR sequence to be tested.

[0055] Step 9.5: The sequence-level plant IDR function predictor pIFCmap obtains the top 3 syntax clusters in the sorting table of the IDR sequence to be tested, and the corresponding function annotation result table are the function prediction reference syntax clusters and function prediction results of the IDR sequence to be tested.

[0056] Step 9.6: The sequence-level plant IDR function predictor pIFCmap determines whether the function prediction reference syntax cluster of the IDR sequence to be tested belongs to the formal cross-species plant IDR functional group pIFC. If one or more syntax clusters in the function prediction reference syntax cluster of the IDR sequence to be tested belong to the formal cross-species plant IDR functional group pIFC, then the pIFC function category prediction result of the IDR sequence to be tested is obtained by deduplicating all the formal cross-species plant IDR functional groups pIFC to which the function prediction reference syntax cluster of the IDR sequence to be tested belongs; if the top 3 syntax clusters in the ranking table of the syntax clusters of the IDR sequence to be tested do not belong to any formal cross-species plant IDR functional group pIFC, then there is no pIFC function category prediction result for the IDR sequence to be tested.

[0057] Step 9.7: The sequence-level plant IDR function predictor pIFCmap outputs the pIFCmap prediction result table of the IDR sequence to be tested. The pIFCmap prediction result table of the IDR sequence to be tested includes the sorting table of the grammar clusters of the IDR sequence to be tested, the function prediction reference grammar clusters and function prediction results of the IDR sequence to be tested, and the pIFC function category prediction results of the IDR sequence to be tested.

[0058] As described above, the 90-dimensional truncated Z-score grammatical feature vector of the IDR sequence includes the truncated Z-score value of the 90-dimensional Z-score grammatical feature vector of the IDR sequence, and the 90-dimensional Z-score grammatical feature vector of the IDR sequence includes the 54-dimensional composition class Z-score of the IDR sequence and the 36-dimensional permutation class Z-score of the IDR sequence.

[0059] The 54-dimensional component class Z-score of the IDR sequence is obtained in the following way:

[0060] First, the 54-dimensional compositional background statistical parameters of each plant species are pre-calculated: for each IDR sequence of a plant species, the original feature values ​​of each of the 54 compositional features are calculated; for each compositional feature, the set of original feature values ​​of all IDR sequences of the plant species on each compositional feature is taken as the background distribution of the plant species on the corresponding compositional feature; then the mean and standard deviation of the background distribution on each compositional feature are calculated, and a total of 54-dimensional compositional background statistical parameters of the plant species are obtained.

[0061] Secondly, calculate the 54-dimensional component class Z-score for each IDR sequence. The 54-dimensional component class Z-score for each IDR sequence includes the Z-score of each IDR sequence on each of the 54 component class features.

[0062] The Z-score of the IDR sequence on a single constituent class feature is obtained as follows: the original feature value of the IDR sequence on a single constituent class feature is subtracted from the mean of the background distribution of the corresponding constituent class feature of the plant species to which the IDR sequence belongs, and then divided by the standard deviation of the background distribution of the constituent class feature of the plant species to which the IDR sequence belongs.

[0063] The 36-dimensional permutation class Z-score of the IDR sequence is obtained through the following steps:

[0064] Step 3.3, Residue Type Identification: The amino acid composition of the IDR sequence is divided into 8 residue types according to physicochemical properties. The 8 residue types are polar residues, hydrophobic residues, positively charged residues, negatively charged residues, aromatic residues, alanine, proline and glycine.

[0065] Calculation of original residue arrangement features: The 8 residue types are paired to obtain 36 residue combinations; the arrangement parameters in the linear IDR sequence of each residue combination are calculated as the original residue arrangement features of the corresponding residue combination in the IDR sequence, resulting in a total of 36 original residue arrangement features of the IDR sequence.

[0066] Step 3.4: The residue type ratio is calculated as follows: Count the number of residues of each of the 8 residue types in the IDR sequence, and divide the number of residues of each residue type in the IDR sequence by the total length of the IDR sequence to obtain the residue ratio of each residue type in the IDR sequence. For any residue arrangement class original feature in the 36-dimensional residue arrangement class original features, if the residue type ratio of any one of the two residue types corresponding to the residue arrangement class original feature in the IDR sequence is lower than a set threshold, then the corresponding residue arrangement class original feature is marked as a low-frequency arrangement feature, and the residue combination corresponding to the low-frequency arrangement feature is a low-frequency residue combination.

[0067] Homogeneous random shuffling: While keeping the amino acid composition of the IDR sequence unchanged, the IDR sequence is randomly shuffled N2 times. Each random shuffling yields a random sequence, resulting in a total of N2 random sequences of the IDR sequence. These N2 random sequences together constitute the set of homogeneous random background sequences of the IDR sequence itself.

[0068] Calculation of 36-dimensional permutation-class random background statistical parameters: For any single-class combination of residues in the IDR sequence, calculate the permutation parameters of the single-class combination of residues in all random sequences in the random background sequence set of the same composition of the IDR sequence. Then calculate the mean and standard deviation of all permutation parameters of the single-class combination of residues to obtain the permutation-class random background statistical parameters of the single-class combination of residues. Calculate the permutation-class random background statistical parameters for each of the 36 classes of residues to obtain the 36-dimensional permutation-class random background statistical parameters of the IDR sequence.

[0069] Step 3.5, 36-dimensional permutation class Z-score calculation: Calculate the 36-dimensional permutation class Z-score for each IDR sequence: The 36-dimensional permutation class Z-score for each IDR sequence includes the Z-score for each IDR sequence on the 36 classes of residue combinations;

[0070] The Z-scores of the IDR sequence on single-class residue combinations are obtained as follows: First, the Z-score of the IDR sequence on low-frequency residue combinations is set to 0; then, the original characteristics of the residue arrangement class of the IDR sequence on the other single-class residue combinations (excluding low-frequency residue combinations) are subtracted from the mean of the random background statistical parameters of the IDR sequence on the corresponding residue combination, and then divided by the standard deviation of the IDR sequence on the random background statistical parameters of the corresponding residue combination.

[0071] When calculating the Z-score of the 36-dimensional permutation class, each amino acid in the IDR sequence is converted into an integer number sequence according to a preset encoding rule. Then, based on the integer number sequence corresponding to the IDR sequence, residue type identification, original feature calculation of residue permutation class, residue type ratio calculation, random shuffling of the same composition, calculation of random background statistical parameters of the 36-dimensional permutation class, and calculation of the Z-score of the 36-dimensional permutation class are performed. The calculation is accelerated by compilation, parallel computing, or vectorized computing.

[0072] The entire IDR sequence of all plant species is then divided into multiple computational blocks according to a preset number. Each computational block includes multiple IDR sequences. Within the same computational block, each IDR sequence independently constructs its own set of homologous random background sequences based on its own amino acid composition. Then, the 36-dimensional permutation random background statistical parameters and the 36-dimensional permutation Z-score of each IDR sequence are calculated. Finally, after the calculation of each computational block is completed, the corresponding block result file is output. After all computational blocks are completed, the block result files are merged to obtain the 36-dimensional permutation Z-score of the entire IDR sequence of all plant species.

[0073] As mentioned above, candidate GIN models are selected in the following way:

[0074] Step 4.1: Select all IDR sequences within the plant species whose length is within the set length range to construct a training subset of the GIN model within the species; extract the 90-dimensional truncated Z-score grammatical feature vector of each IDR sequence in the training subset of the GIN model within the species to form the training feature matrix of the GIN model of the plant species. Each row of the training feature matrix corresponds to an IDR sequence, and each column of the training feature matrix corresponds to a grammatical feature dimension in the 90-dimensional truncated Z-score grammatical feature vector.

[0075] Step 4.2: Use GIN models with different K values ​​to perform KMeans clustering on the training feature matrix, and take the clustering result with the smallest inertia value as the clustering result corresponding to the K value;

[0076] Step 4.3: Plot the inertia curve as the inertia value changes with the K value. The K values ​​corresponding to the inflection points of the inertia curve are K1 and K2, respectively, identified by the Kneedle algorithm and the tail-fitelbow algorithm.

[0077] When K1 and K2 are different, calculate the arithmetic mean of K1 and K2 and round it to the nearest integer to get K3. If K1 and K2 are not consecutive integers, take K1, K2 and K3 as three candidate K values. If K1 and K2 are consecutive integers, K3 is the larger of K1 and K2 plus 1. If K3 exceeds the preset range of K values, K3 is the smaller of K1 and K2 minus 1.

[0078] When K1 and K2 are the same, K0-2, K0 and K0+2 are taken as three candidate K values, and K0=K1=K2;

[0079] Step 4.4: The three candidate K values ​​of the plant species are used as the three candidate GIN models for the plant species.

[0080] As mentioned above, the significantly enriched GO annotation entries, significantly enriched KEGG pathways, and the corrected p-values ​​and enrichment folds of each significantly enriched GO annotation entry and each significantly enriched KEGG pathway are obtained through the following steps:

[0081] Step 6.1: Extract the genes to which all IDR sequences belong in the syntax cluster to form the gene set of the syntax cluster;

[0082] IDR sequences whose start position is less than 10 amino acid residues from the N-terminus of a protein or whose termination position is less than 10 amino acid residues from the C-terminus of a protein are marked as non-linked IDR sequences; otherwise, they are marked as linked IDR sequences.

[0083] exist IDR sequences with a length of ≥1.5 were selected, and non-linked IDR sequences with a length of 35 amino acid residues or linked IDR sequences with a length of 45 amino acid residues and their corresponding genes were constructed as functional enrichment analysis subsets. For candidate GIN-Bank, all non-redundant genes that entered the functional enrichment analysis subset were defined as the background gene set of the candidate GIN-Bank. For any grammar cluster in the candidate GIN-Bank, the non-redundant genes that entered the functional enrichment analysis subset within the gene set of the grammar cluster were defined as the gene set to be tested for the grammar cluster.

[0084] Step 6.2 First, obtain the original GO annotation table: using the gene set to be tested of the syntax cluster and the gene numbers involved in the background gene set of the candidate GIN-Bank to which the syntax cluster belongs as the search objects, obtain the correspondence between genes and GO annotation entries through Ensembl Plants BioMart to form the original GO annotation table;

[0085] Secondly, obtain the GO ontology file and the plant GO slim definition file;

[0086] Next, the GO annotation entries in the original GO annotation table are mapped to the GO annotation entries in the plant GO slim definition file to obtain the plant GO slim annotation table;

[0087] Finally, the plant GO slim annotation tables were organized into two types of files: GO-TERM2GENE and GO-TERM2NAME.

[0088] Step 6.3: Perform GO enrichment analysis on the set of genes to be tested for the grammatical clusters:

[0089] First, the non-redundant genes in the plant GO slim annotation table that are the genes to be tested of the grammar cluster are defined as the GO gene set of the grammar cluster, and the non-redundant genes in the plant GO slim annotation table that are the background genes of the candidate GIN-Bank to which the grammar cluster belongs are defined as the GO background gene set.

[0090] Secondly, using the GO background gene set as the statistical background, the enricher function was employed, with GO-TERM2GENE and GO-TERM2NAME as reference files, to perform enrichment analysis on the GO gene set to be tested for the grammar cluster. The original p-values ​​and enrichment folds of each GO annotation entry corresponding to the GO gene set to be tested were obtained. The original p-values ​​of each GO annotation entry were corrected using the Benjamini-Hochberg method with multiple tests to obtain the corrected p-values ​​of each GO annotation entry. GO annotation entries with corrected p-values ​​not greater than the first set p-value threshold were marked as significantly enriched GO annotation entries of the grammar cluster.

[0091] Step 6.4: For plant species with KEGG species IDs, obtain the KEGG pathway list of plant species, and obtain the correspondence between the genes involved in the gene set to be tested of the syntax cluster and the background gene set of the candidate GIN-Bank to which the syntax cluster belongs and the KEGG pathway, and organize them into local snapshots KEGG-TERM2GENE and KEGG-TERM2NAME.

[0092] Step 6.5: Perform KEGG enrichment analysis on the set of genes to be tested for the grammatical cluster:

[0093] First, non-redundant genes with KEGG pathway annotations in the gene set to be tested of the grammar cluster are defined as the KEGG gene set to be tested of the grammar cluster; non-redundant genes with KEGG pathway annotations in the background gene set of the candidate GIN-Bank to which the grammar cluster belongs are defined as the KEGG background gene set.

[0094] Secondly, using the KEGG background gene set as the statistical background, the enricher function was employed, with local snapshots KEGG-TERM2GENE and KEGG-TERM2NAME as reference files, to perform KEGG pathway enrichment analysis on the KEGG gene set to be tested in the syntax cluster. The original p-values ​​and enrichment folds of each KEGG pathway corresponding to the KEGG gene set to be tested in the syntax cluster were obtained. The original p-values ​​of each KEGG pathway were corrected using the Benjamini-Hochberg method with multiple tests to obtain the corrected p-values ​​of each KEGG pathway. KEGG pathways with corrected p-values ​​not greater than a second set p-value threshold were marked as significantly enriched KEGG pathways of the syntax cluster.

[0095] As mentioned above, the species-specific final cluster number K value of plant species, the corresponding formal GIN model, and the formal GIN-Bank are obtained through the following steps:

[0096] First, a candidate K comprehensive evaluation table is constructed, which includes candidate K value, proportion of low-discrimination mapping entries, number of syntactic clusters with significant functional annotations, and functional annotation coverage.

[0097] Secondly, based on the candidate K comprehensive evaluation table, the species-specific final cluster number K value is selected from the three candidate K values ​​of plant species according to the following conditions: Condition 1: The number of syntactic clusters with more significant functional annotations;

[0098] Condition 2: It has broader functional annotation coverage;

[0099] Condition 3: It has a lower proportion of low-discrimination mapping entries;

[0100] The comparison will only proceed to the next condition if the comparison results are the same as the previous condition.

[0101] Among them, the proportion of low-discrimination mapping entries in the candidate GIN-Bank is The proportion of IDR sequences with a value less than 1.5 to the total number of IDR sequences in the plant species; a grammatical cluster with significant functional annotation is a grammatical cluster that is significantly enriched with at least one GO annotation entry or significantly enriched with the KEGG pathway; functional annotation coverage is the proportion of grammatical clusters with significant functional annotation to the total number of grammatical clusters in the candidate GIN-Bank.

[0102] The candidate GIN models and candidate GIN-Banks corresponding to the species-specific final cluster number K are respectively the formal GIN model, the formal GIN-Bank, and the functional annotation results of each grammatical cluster in the formal GIN-Bank.

[0103] Compared with the prior art, the present invention has the following advantages:

[0104] (1) This invention uses intraspecific grammatical clusters of different plant species as basic units, constructs cross-species grammatical cluster similarity graphs based on the similarity of grammatical cluster cluster centers, and forms formal cross-species plant IDR functional groups pIFC by combining graph community discovery and functional evidence; thus, even if the IDR sequences in different plant species themselves lack significant conservation, cross-species functional correspondence can be established based on sequence grammatical similarity, realizing automatic division and graph organization of plant IDR functional groups.

[0105] (2) The present invention further integrates the formal GIN-Bank and the formal cross-species plant IDR functional group pIFC and its functional annotation results into the pIFCmap prediction reference system, and recalculates the grammatical features of the IDR sequence to be tested and maps them to the grammatical cluster and the formal cross-species plant IDR functional group pIFC; thus, the present invention can not only be used for IDR functional annotation of existing reference proteome plants, but also for functional prediction and candidate sequence screening of non-model plants, crops with incomplete annotation and artificially designed IDR sequences.

[0106] (3) This invention combines the proportion of low-discrimination mapping entries in the candidate GIN-Bank, the number of grammatical clusters with significant functional annotations, and the functional annotation coverage to determine the formal GIN model and the formal GIN-Bank; thus, the resulting intra-species grammatical clusters are not only clustering results in a mathematical sense, but also have corresponding functional tendencies, which can be used for entry-level functional annotation and candidate functional region screening of plant IDR.

[0107] (4) This invention identifies IDR sequences and further converts the identified and cleaned IDR sequences into 90-dimensional truncated Z-score grammatical feature vectors. Thus, plant IDRs that lack stable three-dimensional structures and have weak sequence conservation are transformed into quantifiable and comparable grammatical feature representations, thereby reducing the dependence on existing sequence homology alignment, conserved domains and stable protein structures.

[0108] (5) In view of the problem of large computational cost of 36-dimensional residue arrangement features, this invention accelerates the processing of residue type identification, original feature calculation of residue arrangement, residue type ratio calculation, random shuffling of the same composition, calculation of random background statistical parameters of 36-dimensional arrangement, and Z-score of 36-dimensional arrangement. Without reducing the number of random shuffling, without changing the residue category classification rules and grammatical feature definition, it significantly reduces the time cost of 90-dimensional grammatical feature calculation in the plant IDR sequence resource set pIDRome, making the construction of multi-species plant IDR grammatical maps practically feasible. Attached Figure Description

[0109] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0110] To facilitate understanding and implementation of the present invention by those skilled in the art, the present invention will be further described in detail below with reference to embodiments. The embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0111] Example 1:

[0112] like Figure 1 As shown, the method for constructing and predicting functional taxa in the intrinsic disordered region of plants based on sequence grammar includes the following steps:

[0113] Step 1: Obtain the whole proteome data files of plant species from the Ensembl database.

[0114] In this embodiment, the FASTA format whole proteome data files of each plant species in release-62 are downloaded as raw data for subsequent identification of plant intrinsic disordered regions and cross-species comparative analysis. The whole proteome data files of each plant species include all identifiable protein sequences (i.e., protein amino acid sequences) of the corresponding plant species stored in FASTA format, which serve as input for subsequent identification of intrinsic disordered regions.

[0115] For each plant species, a corresponding plant proteome input list is constructed for the whole proteome data file. The plant proteome input list records the plant species name, genome version, proteome file name, file source, download version, and protein sequence number corresponding to the whole proteome data file.

[0116] Step 2: Using the whole proteome data files of each plant species obtained in Step 1 as input, identify the intrinsically disordered regions (IDRs) in the protein sequences of each plant species' whole proteome data files, and construct the IDR sequence set IDRome for each plant species by combining the IDR sequences corresponding to the IDRs of all protein sequences identified for each plant species; finally, summarize the IDR sequence sets IDRome of each plant species to form the plant IDR sequence resource set pIDRome; specifically including the following steps:

[0117] Step 2.1: Identify intrinsic disordered regions using a unified IDR prediction tool: Input the whole proteome data files of each plant species into MobiDB-lite 4.0 (a consensus prediction tool for predicting intrinsic disordered regions in protein sequences), and predict the intrinsic disordered regions of each protein sequence in the whole proteome data files, obtaining the start and end positions of the intrinsic disordered regions in each protein sequence.

[0118] Step 2.2: Map the intrinsic disorder region back to the protein sequence and extract the IDR sequence located in the intrinsic disorder region. Specifically, based on the start and end positions of the intrinsic disorder region of each protein sequence obtained in Step 2.1, extract the amino acid fragments located in the intrinsic disorder region from the protein sequence. The extracted amino acid fragments constitute the actual protein amino acid sequence of the intrinsic disorder region of the corresponding protein sequence. The actual protein amino acid sequence of the intrinsic disorder region is denoted as the IDR sequence.

[0119] For each IDR sequence, information such as plant species, protein number, start and end positions of the intrinsic disorder region, and length of the intrinsic disorder region is recorded.

[0120] Step 2.3: Clean the IDR sequences and form the plant IDR sequence resource set pIDRome: Clean the IDR sequences obtained in Step 2.2 uniformly: retain only the IDR fragments with feature ="-" in the InterPro style output of MobiDB-lite 4.0; then delete the IDR sequences containing non-standard amino acid residues such as X, U, Z, J, B and O from the IDR fragments with feature ="-".

[0121] All cleaned IDR sequences for each plant species are packaged into a corresponding IDR sequence set IDRome. Then, all IDR sequence sets IDRome for all plant species are aggregated to obtain the plant IDR sequence resource set pIDRome, which serves as the basic dataset for subsequent sequence grammar representation, intra-species functional classification, and cross-species functional map construction.

[0122] Among them, X, U, Z, J, B, and O are extended IUPAC protein single-letter codes, representing non-standard or ambiguous amino acid residues in addition to the standard 20 amino acids.

[0123] Step 3: Calculate the 90-dimensional truncated Z-score grammatical feature vector for each IDR sequence in the plant IDR sequence resource set pIDRome using the NARDINI+ algorithm. The grammatical features include compositional features and residue arrangement features. Therefore, sequence grammatical characterization involves two types of feature calculations: the first type is a 54-dimensional compositional feature, used to characterize the amino acid composition, residue class ratios, physicochemical properties, and patch characteristics of the IDR sequence; the second type is a 36-dimensional residue arrangement feature, used to characterize the mixed, dispersed, or patchy distribution of different types of residues in the linear IDR sequence. Specifically, this includes the following steps:

[0124] Step 3.1: Pre-calculate the 54-dimensional compositional background statistical parameters for each plant species: For each IDR sequence in the IDR sequence set IDRome of a plant species, calculate the original feature value on each of the 54 compositional features; for each compositional feature, define the set of original feature values ​​of all IDR sequences of the plant species on each compositional feature as the background distribution of the plant species on the corresponding compositional feature; then calculate the mean and standard deviation of the background distribution on each compositional feature to obtain the 54-dimensional compositional background statistical parameters of the plant species. The 54-dimensional compositional background statistical parameters serve as the reference background for calculating the compositional class Z-score of each IDR sequence within the plant species.

[0125] Step 3.2: Calculate the 54-dimensional component class Z-score for each IDR sequence. The 54-dimensional component class Z-score for each IDR sequence includes the Z-score of each IDR sequence on each of the 54 component class features.

[0126] The Z-score of the IDR sequence on a single constituent class feature is obtained as follows: the original feature value of the IDR sequence on a single constituent class feature is subtracted from the mean of the background distribution of the corresponding constituent class feature of the plant species to which the IDR sequence belongs, and then divided by the standard deviation of the background distribution of the constituent class feature of the plant species to which the IDR sequence belongs.

[0127] Next, calculate the 36-dimensional permutation class Z-score of the IDR sequence, which includes the following steps:

[0128] Step 3.3: Calculate the 36-dimensional residue arrangement class original features of each IDR sequence, specifically including the following steps:

[0129] Step 3.3.1, Residue Type Identification: For any IDR sequence in the plant IDR sequence resource set pIDRome, the amino acids are classified into 8 residue types according to their physicochemical properties. The 8 residue types are polar residues, hydrophobic residues, positively charged residues, negatively charged residues, aromatic residues, alanine, proline, and glycine.

[0130] Step 3.3.2, Calculation of original features of residue arrangement: Combine the 8 residue types in pairs. Since the 8 residue types include combinations between the same category and combinations between different categories, a total of C(8,2) + 8 residue combinations are obtained, which is 36 residue combinations.

[0131] The arrangement parameters of each residue combination in the linear IDR sequence were calculated to characterize whether the two residue combinations tend to be uniformly mixed or clustered in the IDR sequence; a total of 36 residue combinations were obtained in the linear IDR sequence.

[0132] Then, the arrangement parameters in the linear IDR sequence for each type of residue combination are used as the original residue arrangement features of the corresponding type of residue combination, resulting in 36 original residue arrangement features corresponding to each type of residue combination. The original residue arrangement features of the 36 types of residue combinations together constitute the 36-dimensional original residue arrangement features of the IDR sequence.

[0133] Step 3.4: Calculate the 36-dimensional permutation-class random background statistical parameters for each IDR sequence, specifically including the following steps:

[0134] Step 3.4.1: The residue type ratio is calculated as follows: For any IDR sequence, the number of residues of each of the 8 residue types in the IDR sequence is counted, and the number of residues of each residue type in the IDR sequence is divided by the total length of the IDR sequence to obtain the residue ratio of each residue type in the IDR sequence. For any original residue arrangement feature in the 36-dimensional original residue arrangement features, if the residue type ratio of any one of the two residue types corresponding to the original residue arrangement feature is less than 10% in the IDR sequence, then the corresponding original residue arrangement feature is marked as a low-frequency arrangement feature, and the residue combination corresponding to the low-frequency arrangement feature is a low-frequency residue combination.

[0135] Step 3.4.2, Iso-compositional random shuffling: While keeping the amino acid composition of the IDR sequence unchanged, the IDR sequence is randomly shuffled 100,000 times. Each random shuffling yields a random sequence, resulting in a total of 100,000 random sequences with the same number of amino acids but different arrangements. These 100,000 random sequences together constitute the iso-compositional random background sequence set of the IDR sequence itself.

[0136] The original features of the 36-dimensional residue arrangement class do not use the compositional distribution of the entire species' IDRome as the background, but rather use the set of homologous random background sequences of each IDR sequence as the background.

[0137] Step 3.4.3, Calculation of 36-dimensional permutation-class random background statistical parameters: For any single category of residue combination in the IDR sequence, calculate the permutation parameters of the single category of residue combination in all random sequences in the same composition of the random background sequence set of the IDR sequence. Then calculate the mean and standard deviation of all permutation parameters of the single category of residue combination to obtain the permutation-class random background statistical parameters of the single category of residue combination. Calculate the permutation-class random background statistical parameters for each of the 36 categories of residue combinations to obtain the 36-dimensional permutation-class random background statistical parameters of the IDR sequence.

[0138] Step 3.5, 36-dimensional permutation class Z-score calculation: Calculate the 36-dimensional permutation class Z-score for each IDR sequence. The 36-dimensional permutation class Z-score for each IDR sequence includes the Z-score for each IDR sequence on the 36 types of residue combinations.

[0139] The Z-score of each IDR sequence on a single class of residue combinations is obtained as follows: First, the Z-score of the IDR sequence on low-frequency residue combinations is set to 0; then, the original characteristics of the residue arrangement class of the IDR sequence on the other single class of residue combinations (excluding low-frequency residue combinations) are subtracted from the mean of the random background statistical parameters of the IDR sequence on the corresponding residue combination, and then divided by the standard deviation of the IDR sequence on the random background statistical parameters of the corresponding residue combination.

[0140] To improve the computational efficiency of 36-dimensional residue arrangement features on the large-scale plant IDR sequence resource set pIDRome data, the calculation process of arrangement features is accelerated by compilation: First, each amino acid in the IDR sequence is converted into an integer number sequence according to a preset encoding rule. Then, based on the integer number sequence, residue type identification, original feature calculation of residue arrangement, residue type ratio calculation, random shuffling of the same composition, calculation of random background statistical parameters of 36-dimensional arrangement, and calculation of 36-dimensional arrangement Z-score are performed, thereby reducing the computational overhead caused by repeated string statistics and character sampling during random shuffling. Then, the calculation process is accelerated by compilation acceleration, parallel computing, or vectorized computing. In this embodiment, Numba is used to accelerate the compilation of residue type identification, original feature calculation of residue arrangement, residue type ratio calculation, random shuffling of the same composition, calculation of random background statistical parameters of 36-dimensional arrangement, and calculation of 36-dimensional arrangement Z-score.

[0141] Meanwhile, during the calculation of 36-dimensional residue arrangement features on the large-scale plant IDR sequence resource set pIDRome, the IDR sequences to be calculated in pIDRome are divided into multiple calculation blocks according to a preset number. Each calculation block contains several IDR sequences. In this embodiment, each calculation block contains 100 IDR sequences. Each calculation block runs as an independent calculation task. Within the same calculation block, each IDR sequence independently constructs its own set of random background sequences with the same composition according to its own amino acid composition, and performs 100,000 random shuffling, 36-dimensional residue arrangement parameter calculation, and 36-dimensional arrangement class Z-score calculation. After each calculation block is completed, the corresponding block result file is output. After all calculation blocks are completed, the results of each block are merged to form the 36-dimensional arrangement class Z-score result of all IDR sequences of all plant species in the plant IDR sequence resource set pIDRome.

[0142] Step 3.6: Merge the 54-dimensional composition class Z-score and the 36-dimensional permutation class Z-score sequentially to form a 90-dimensional Z-score grammatical feature vector, which includes the Z-score values ​​of the 90-dimensional grammatical features.

[0143] Step 3.7: To avoid a few extreme Z-scores from having an excessive impact on the overall grammatical space during subsequent distance calculations and clustering, the 90-dimensional Z-score grammatical feature vector is truncated: all Z-score values ​​greater than 3 are set to 3, all Z-score values ​​less than -3 are set to -3, and Z-score values ​​between -3 and 3 remain unchanged, resulting in the truncated Z-score values ​​of the 90-dimensional grammatical features; the 90-dimensional Z-score grammatical feature vector of each IDR sequence in each plant species is truncated, resulting in the 90-dimensional truncated Z-score grammatical feature vector of each IDR sequence in each plant species, which includes the truncated Z-score values ​​of the 90-dimensional grammatical features.

[0144] Step 4: Set the range of values ​​for K. Based on the 90-dimensional truncated Z-score grammatical feature vectors, use GIN (Grammars Inferred using NARDINI+) models with different K values ​​to perform unsupervised clustering of all IDR sequences within each plant species. Based on the clustering results for each plant species, select multiple candidate K values ​​for each species, and use the GIN models with these candidate K values ​​as candidate GIN models, thus establishing a set of candidate GIN models for each plant species. This specifically includes the following steps:

[0145] The GIN model is a grammatical feature organization model of intraspecific IDR sequences trained within the IDR sequence set IDRome of a single plant species. The GIN model is based on several cluster centers and is used to characterize the grammatical features of different types of IDR sequences in the IDR sequence set IDRome of a plant species. K is the number of cluster centers in the GIN model.

[0146] Step 4.1: Constructing the training subset of the intra-species GIN model: Select all IDR sequences with a length between 70 and 300 amino acids from the IDR sequence set IDRome of each plant species to form the training subset of the intra-species GIN model of the plant species; for each IDR sequence that enters the training subset of the intra-species GIN model, extract the 90-dimensional truncated Z-score grammatical feature vector obtained in step 3 to form the training feature matrix of the GIN model of the plant species. Each row of the training feature matrix corresponds to an IDR sequence, and each column of the training feature matrix corresponds to a grammatical feature dimension of the 90-dimensional truncated Z-score grammatical feature vector.

[0147] Step 4.2: Within the preset K value range (1-50), GIN models with different K values ​​are used to perform KMeans clustering on the training feature matrix multiple times with different random initial conditions. KMeans clustering uses each IDR sequence in the training feature matrix as a sample, and the 90-dimensional truncated Z-score grammatical feature vector of the IDR sequence as its coordinate value in the 90-dimensional grammatical feature space. For each K value, the clustering result with the smallest inertia value is taken as the clustering result corresponding to that K value. Multiple candidate K values ​​and corresponding candidate GIN models are selected based on the inertia curve showing the change of inertia value with K value. Specifically, this includes the following steps:

[0148] KMeans clustering was implemented using the official KMeans implementation in the scikit-learn package. The distance metric used was the centroid model corresponding to Euclidean distance. Clustering parameters were set to `init="random"`, `n_init=10`, and `random_state=1` to ensure consistency in the clustering process and reproducibility of results across different clusters. `init="random"` indicates that initial cluster centers are generated randomly; `n_init=10` means that KMeans clustering is performed 10 times under different random initial conditions for the same K value, and the cluster with the smallest inertia value is used as the clustering result for that K value; `random_state=1` means that the random number seed is fixed at 1 to ensure the repeatability of the random initialization process. Simultaneously, for each K value, the cluster centers output by KMeans clustering and the corresponding inertia value are recorded. The inertia value is the clustering evaluation metric in the official KMeans implementation, representing the sum of the squared Euclidean distances of each sample to its nearest cluster center.

[0149] Step 4.3: Select multiple candidate K values ​​within plant species based on the inertia curve: For each plant species, arrange the inertia values ​​corresponding to different K values ​​in Step 4.2 in ascending order of K value, and plot the inertia curve of the plant species as the inertia value changes with the K value with K value as the x-axis and inertia value as the y-axis. Then, use the Kneedle algorithm and the tail-fit elbow algorithm to identify the inertia curves. Record the K value corresponding to the inertia point identified by the Kneedle method as K1, and the K value corresponding to the inertia point identified by the tail-fit elbow algorithm as K2.

[0150] When K1 and K2 are different, calculate the arithmetic mean of K1 and K2 and round it to an integer, denoted as K3. If K1 and K2 are not consecutive integers, take K1, K2, and K3 as three candidate K values ​​for the plant species. If K1 and K2 are consecutive integers, the rounded K3 obtained from the arithmetic mean of K1 and K2 will overlap with one of K1 or K2. In this case, adjust K3 to the larger of K1 and K2 plus 1. If the adjusted K3 exceeds the preset K value range, adjust K3 to the smaller of K1 and K2 minus 1.

[0151] When K1 and K2 are the same, the common K value of K1 and K2 is recorded as K0 (i.e., K0=K1=K2). With K0 as the center, K0-2, K0 and K0+2 are taken as three candidate K values ​​for the plant species. Thus, each plant species can finally obtain three candidate K values, K1, K2 and K3, which are non-overlapping and within the preset K value range.

[0152] Step 4.4: Establish a set of candidate GIN models indexed by candidate K values: For each plant species, the three candidate K values ​​of the plant species represent three GIN models at three different clustering granularities, which are respectively the three candidate GIN models of the plant species; each plant species obtains three candidate GIN models corresponding to the three candidate K values, forming a set of candidate GIN models for the corresponding plant species.

[0153] Step 5: Using the 90-dimensional truncated Z-score grammatical feature vectors of each IDR sequence in the IDR sequence set IDRome obtained in Step 3 as input, cluster the IDR sequences within the corresponding plant species using each candidate GIN model for each plant species. This yields the candidate GIN-Bank corresponding to each candidate K-value and candidate GIN model for each plant species. A one-to-one correspondence is maintained between each candidate K-value, candidate GIN model, and candidate GIN-Bank. Only the candidate K-value needs to be determined to identify a unique candidate GIN model and candidate GIN-Bank. Specifically, this includes the following steps:

[0154] The candidate GIN model uses KMeans clustering to iteratively update K cluster centers and assigns each IDR sequence in the IDR sequence set IDRome of plant species to a grammatical cluster centered on each cluster center. GIN-Bank is an entry-level database of IDR sequences formed after the GIN model is mapped back to the IDR sequence set IDRome of plant species. GIN-Bank records each IDR sequence in the IDR sequence set IDRome of plant species, the assignment result of the grammatical cluster of the IDR sequence, the mapping distance of each IDR sequence to the cluster center of the grammatical cluster, and the mapping discrimination index, etc.

[0155] Step 5.1: For each plant species, read the candidate GIN model corresponding to the three candidate K values ​​output in Step 4, and the 90-dimensional truncated Z-score grammatical feature vector of each IDR sequence in the IDR sequence set IDRome obtained in Step 3. With each IDR sequence as a row and each grammatical feature dimension in the 90-dimensional truncated Z-score grammatical feature vector as a column, form the IDR sequence feature matrix to be mapped for the plant species.

[0156] Step 5.2: Calculate the mapping distance from each IDR sequence in the feature matrix of the IDR sequences to be mapped to the cluster center. In this embodiment, the mapping distance is the squared Euclidean distance. IDR sequence to the first Mapping distance of each cluster center for:

[0157] (1);

[0158] in, Indicates the first The IDR sequence in the first Truncation Z-score values ​​on dimensional grammatical features Indicates the first The cluster centers at the in Coordinate values ​​on dimensional grammatical features, This is the cluster center number.

[0159] Step 5.3: Determine the grammatical cluster assignment of each IDR sequence under the GIN model: Sort the squared Euclidean distances of each IDR sequence to all cluster centers in the current GIN model in ascending order; determine the cluster center with the smallest distance as the nearest cluster center of the IDR sequence, and the cluster center with the second smallest distance as the second nearest cluster center of the IDR sequence. The set of IDR sequences with the same nearest cluster center is defined as a grammatical cluster, and the grammatical cluster assignment of the IDR sequence is the grammatical cluster centered on the nearest cluster center.

[0160] Step 5.4, Evaluate mapping quality: Denote the mapping distance from an IDR sequence to the nearest cluster center as . The mapping distance to the second nearest cluster center is denoted as The mapping discrimination index of this IDR sequence is calculated using the following formula:

[0161] (2);

[0162] in, The degree of distinction between the nearest cluster center and the second nearest cluster center used to characterize the IDR sequence's affiliation. The larger the value, the more clearly the syntactic family affiliation of the IDR sequence under the current GIN model is determined; The smaller the value, the closer the IDR sequence is to the boundary between multiple syntax clusters. If the value is less than 1.5, the IDR sequence is marked as a low-discrimination mapping entry.

[0163] Step 5.5: Construct a candidate GIN-Bank. For each candidate GIN model of each plant species, summarize the grammatical cluster attribution results obtained in Step 5.3 and the mapping quality index obtained in Step 5.4 to form the candidate GIN-Bank corresponding to that candidate GIN model. The candidate GIN-Bank records the grammatical cluster attribution results of all IDR sequences in the IDR sequence set IDRome of the plant species under the corresponding candidate GIN model, the plant species of each IDR sequence, the start and end positions of the intrinsic disorder region, the length of the intrinsic disorder region, the IDR sequence, the grammatical cluster attribution number (the grammatical cluster attribution number is the grammatical cluster number centered on the nearest cluster center), and the mapping distance from the IDR sequence to the nearest cluster center. Mapping distance from IDR sequence to the second nearest cluster center and mapping discrimination index Ultimately, three candidate GIN-Banks were obtained for each plant species corresponding to three candidate GIN models. Before the formal GIN model is determined, the candidate GIN-Banks are only used as the data basis for evaluating the candidate GIN models and are not used as the annotation results of the formal IDR sequences.

[0164] Step 6: Perform GO enrichment analysis and KEGG enrichment analysis on the grammar clusters to construct a functional annotation result table for the grammar clusters. The functional annotation result table represents the functional tendency of the grammatical type of the IDR sequence represented by the grammar cluster, specifically including the following steps:

[0165] Step 6.1: Construct functional enrichment analysis subsets: For each candidate GIN-Bank of each plant species, extract the genes to which all IDR sequences belong in each grammatical cluster to form the gene set of each grammatical cluster.

[0166] Then, based on the position of the IDR sequence in the protein sequence, IDR sequences with a start position less than 10 amino acid residues from the N-terminus or an end position less than 10 amino acid residues from the C-terminus are marked as non-linked IDR sequences, otherwise they are marked as linked IDR sequences.

[0167] exist IDR sequences with a length of ≥1.5 were selected, and non-linked IDR sequences with a length of ≥35 amino acid residues or linked IDR sequences with a length of ≥45 amino acid residues, along with their corresponding genes, were constructed as functional enrichment analysis subsets. In the same candidate GIN-Bank of the same plant species, all non-redundant genes that entered the functional enrichment analysis subset were defined as the background gene set of the candidate GIN-Bank. For any grammar cluster in the candidate GIN-Bank, the non-redundant genes that entered the functional enrichment analysis subset within the gene set of the grammar cluster were defined as the gene set to be tested for the grammar cluster.

[0168] Among them, non-redundant genes are those obtained by removing duplicates from the IDR sequences in the functional enrichment analysis subset according to their source gene numbers; when the same gene corresponds to multiple IDR sequences, it is counted only once in the same gene set; the background gene set is used to limit the statistical background of functional enrichment analysis in candidate GIN-Bank, and the gene set to be tested is used to characterize the functional composition of the genes to which the IDR sequences belong within a single grammar cluster.

[0169] Step 6.2: Using the gene set to be tested of the grammar cluster and the gene numbers involved in the background gene set of the candidate GIN-Bank to which the grammar cluster belongs as the search objects, the correspondence between the genes of the plant species and GO (Gene Ontology) annotation entries is obtained through Ensembl Plants BioMart (Ensembl plant genome data mining tool), forming the original GO annotation table. For genes for which no GO annotation entries are found, it is considered that no GO annotation has been obtained under that annotation source, and no subsequent GO enrichment test is performed.

[0170] Then, the GO ontology file and the plant GO slim definition file are obtained. The GO ontology file provides the hierarchical relationship between GO annotation entries, and the plant GO slim definition file provides the set of high-level GO annotation entries in the plant general function annotation. Using the GOATOOLS package, the GO annotation entries in the original GO annotation table are mapped to the high-level GO annotation entries contained in the plant GO slim definition file according to the GO hierarchical relationship in the GO ontology file, thus obtaining the plant GO slim annotation table.

[0171] The Plant GO Slim annotation table is used to consolidate fine-grained GO annotation entries into higher-level functional entries suitable for multi-species plant comparisons. After mapping, the Plant GO Slim annotation table is organized into two types of files: GO-TERM2GENE and GO-TERM2NAME. GO-TERM2GENE records the correspondence between Plant GO Slim entries and gene numbers, while GO-TERM2NAME records the number, name, and GO category of each Plant GO Slim entry.

[0172] Step 6.3: Perform GO enrichment analysis on the set of genes to be tested for the grammatical clusters:

[0173] Step 6.3.1: For each candidate GIN-Bank of each plant species, define the non-redundant genes with plant GO slim annotations (genes in the plant GO slim annotation table with plant GO slim annotations) in the background gene set of the candidate GIN-Bank as the GO background gene set, and define the non-redundant genes with plant GO slim annotations in the gene set to be tested for each grammar cluster in the candidate GIN-Bank as the GO gene set to be tested for the corresponding grammar cluster.

[0174] Step 6.3.2: GO Enrichment Analysis. Enrichment analysis was performed on GO annotation entries in three categories: biological processes, molecular functions, and cellular components, according to their GO classification. Specifically, for each candidate GIN-Bank of each plant species, using the GO background gene set as the statistical background, the enricher function in the R language clusterProfiler package was used, with GO-TERM2GENE and GO-TERM2NAME obtained in Step 6.2 as reference files, to perform enrichment analysis on the GO gene sets to be tested for each grammatical cluster. The original p-values ​​and enrichment folds of each GO annotation entry corresponding to the GO gene sets to be tested were obtained. Subsequently, the original p-values ​​of each GO annotation entry in the same enrichment analysis were corrected using the Benjamini-Hochberg method with multiple tests to obtain the corrected p-values ​​of each GO annotation entry. GO annotation entries with corrected p-values ​​not greater than 0.05 were marked as significantly enriched GO annotation entries of the grammatical cluster, serving as evidence for GO functional annotation of the grammatical cluster.

[0175] Step 6.4: For plant species with KEGG species IDs, the KEGG pathway list of the plant species is obtained online using the R language KEGGREST package. The correspondence between the genes involved in the gene set to be tested for each syntax cluster and the background gene set of the candidate GIN-Bank to which the syntax cluster belongs, and the KEGG pathways is also obtained. This information is compiled into two types of files: KEGG-TERM2GENE and KEGG-TERM2NAME, and frozen as local snapshots. KEGG-TERM2GENE is used to record the correspondence between KEGG pathway IDs and gene IDs, and KEGG-TERM2NAME is used to record the correspondence between KEGG pathway IDs and pathway names. Subsequent KEGG enrichment analyses are performed based on the frozen local snapshots of KEGG-TERM2GENE and KEGG-TERM2NAME files to ensure the reproducibility of analysis results across different batches. For plant species without corresponding KEGG species IDs and whose gene IDs cannot be matched with KEGG IDs, they are marked as KEGG-free reference data.

[0176] Step 6.5: Perform KEGG enrichment analysis on the gene sets to be tested for each grammatical cluster obtained in Step 6.1:

[0177] Step 6.5.1: For each candidate GIN-Bank of a plant species with a KEGG species number, define the non-redundant genes with KEGG pathway annotations in the background gene set of the candidate GIN-Bank as the KEGG background gene set; define the non-redundant genes with KEGG pathway annotations in the gene set to be tested for each grammar cluster in the candidate GIN-Bank as the KEGG gene set to be tested for the grammar cluster.

[0178] Step 6.5.2: For each candidate GIN-Bank of each plant species, using the KEGG background gene set as the statistical background, the enricher function in the R language clusterProfiler package is used, with the locally frozen KEGG-TERM2GENE and KEGG-TERM2NAME obtained in Step 6.4 as reference files, to perform KEGG pathway enrichment analysis on the KEGG gene set to be tested for each syntax cluster. The original p-values ​​and enrichment folds of each KEGG pathway corresponding to the KEGG gene set to be tested for each syntax cluster are obtained. Subsequently, the original p-values ​​of each KEGG pathway in the same enrichment analysis are corrected using the Benjamini-Hochberg method with multiple tests to obtain the corrected p-values ​​of each KEGG pathway. KEGG pathways with corrected p-values ​​not greater than 0.05 are marked as significantly enriched KEGG pathways of the syntax cluster, serving as evidence for KEGG functional annotation of the syntax cluster. For plant species without KEGG reference data, no KEGG enrichment analysis is performed.

[0179] Step 6.6: Construct a functional annotation result table for grammatical clusters. Summarize the GO functional annotation evidence obtained in Step 6.3 and the KEGG functional annotation evidence obtained in Step 6.5. Construct a functional annotation result table for each grammatical cluster in each candidate GIN-Bank for each plant species. The functional annotation result table includes the plant species, the candidate K value corresponding to the candidate GIN-Bank, the grammatical cluster number, significantly enriched GO annotation entries, significantly enriched KEGG pathways, and the corrected p-value and enrichment fold of each significantly enriched GO annotation entry and each significantly enriched KEGG pathway.

[0180] Step 7: For each plant species, using the candidate GIN-Banks corresponding to the three candidate K values ​​obtained in Step 5 and the functional annotation results tables of each grammatical cluster in the candidate GIN-Banks of each plant species obtained in Step 6 as evaluation criteria, sort the three candidate K values ​​of the plant species to determine the species-specific final cluster number K value (final K). Since a one-to-one correspondence between candidate K values, candidate GIN models, and candidate GIN-Banks has been established in Steps 4 and 5, after determining the species-specific final cluster number K value (final K), the candidate GIN models and candidate GIN-Banks corresponding to the species-specific final cluster number K value (final K) can be directly extracted and upgraded to formal GIN models and formal GIN-Banks, respectively. This specifically includes the following steps:

[0181] Step 7.1: Construct a candidate K comprehensive evaluation table: For each plant species, integrate the functional annotation results of the candidate GIN-Bank corresponding to the three candidate K values ​​obtained in Step 5 and the candidate GIN-Bank grammar clusters obtained in Step 6 to construct a candidate K comprehensive evaluation table for the plant species. The candidate K comprehensive evaluation table includes each candidate K value, the proportion of low-discrimination mapping entries, the number of grammar clusters with significant functional annotations, and the functional annotation coverage.

[0182] The proportion of low-discrimination mapping entries is the candidate GIN-Bank obtained in step 5. The proportion of IDR sequences with a value less than 1.5 to the total number of IDR sequences in the plant species is defined as follows: a grammatical cluster with significant functional annotation is defined as a grammatical cluster that has obtained at least one significantly enriched GO annotation entry or a significantly enriched KEGG pathway in step 6; the functional annotation coverage is defined as the proportion of grammatical clusters with significant functional annotation to the total number of grammatical clusters in the candidate GIN-Bank.

[0183] Step 7.2: Based on the candidate K comprehensive evaluation table, select the species-specific final cluster number K value from the three candidate K values ​​for plant species according to the following conditions (the next condition is compared only if the comparison results under the previous condition are the same):

[0184] Condition 1: The number of syntactic clusters with more significant functional annotations;

[0185] Condition 2: It has broader functional annotation coverage;

[0186] Condition 3: It has a lower proportion of low-discrimination mapping entries.

[0187] Step 7.3: Based on the species-specific final cluster number K value (final K), form the formal GIN model and the formal GIN-Bank: The candidate GIN model and candidate GIN-Bank corresponding to the species-specific final cluster number K value are respectively the formal GIN model, the formal GIN-Bank, and the functional annotation result table of each grammatical cluster in the formal GIN-Bank; thus, each plant species obtains a formal GIN model, a formal GIN-Bank, and a functional annotation result table of each grammatical cluster in the formal GIN-Bank.

[0188] Step 8: Construct a cross-species grammatical cluster similarity graph, which includes grammatical cluster nodes, preserved edges, and edge weights; use the Leiden graph community detection algorithm to divide the cross-species grammatical cluster similarity graph into cross-species grammatical cluster communities; use the pvclust function to perform hierarchical clustering on all cross-species grammatical cluster communities to obtain multiple candidate cross-species functional units; upgrade the candidate cross-species functional units to formal cross-species plant IDR functional groups pIFC (Plant IDR Functional Class), where pIFC is the plant IDR cross-species functional category obtained after cross-species grammatical cluster similarity graph construction, graph community detection, functional evidence evaluation, and rule-based upgrading; specifically including the following steps:

[0189] Step 8.1: Read the official GIN model, official GIN-Bank, and functional annotation result table of each grammatical cluster in the official GIN-Bank of each plant species obtained in Step 7. Define each grammatical cluster in the official GIN-Bank of each plant species as a grammatical cluster node. The sequence grammatical features of the grammatical cluster node are obtained by using the 90-dimensional truncated Z-score grammatical feature vector of the cluster center of the corresponding grammatical cluster. The grammatical clusters in the official GIN-Bank of all plant species constitute a cross-species grammatical cluster node set.

[0190] Step 8.2: Define the connection between any two grammar cluster nodes from different plant species in the cross-species grammar cluster node set as a candidate edge. Read the 90-dimensional truncated Z-score grammar feature vectors of the cluster centers corresponding to these two grammar cluster nodes, and calculate the cosine similarity between the 90-dimensional truncated Z-score grammar feature vectors of the two cluster centers according to the following formula. Define the cosine similarity as the edge weight of the candidate edge. The cosine similarity is specifically based on the following formula:

[0191] (3);

[0192] in, The cosine similarity of the 90-dimensional truncated Z-score grammatical feature vectors of the cluster centers corresponding to two grammatical cluster nodes is calculated, which is also the edge weight of the candidate edges between the two grammatical cluster nodes. and These are 90-dimensional truncated Z-score grammatical feature vectors representing the cluster centers of two grammatical cluster nodes, respectively. and They represent and In the Dimensionally truncated coordinates on Z-score syntactic features, index Take values ​​from 1 to 90; The higher the value, the more consistent the direction of the cluster centers corresponding to the two grammar cluster nodes in the 90-dimensional truncated Z-score grammar feature vector, that is, the more similar the grammar features of the cluster centers of the two grammar clusters.

[0193] All nodes from different grammar clusters of plant species are paired up, and the edge weights of the candidate edges of each pair of paired grammar cluster nodes are calculated to form a candidate edge-weight table for cross-species grammar clusters, which is used to construct a cross-species grammar cluster similarity graph.

[0194] Step 8.3: Construct a cross-species grammar cluster similarity graph: Using the cross-species grammar cluster node set obtained in Step 8.1 and the candidate edge-weight table of the cross-species grammar cluster obtained in Step 8.2 as input, the candidate edges are filtered according to the preset edge weight threshold (0.90). If the edge weight is greater than the edge weight threshold, the two grammar cluster nodes connected by the candidate edge are considered to have high cross-species sequence grammar similarity. The candidate edges with edge weight greater than the edge weight threshold are retained as edges and are included in the cross-species grammar cluster similarity graph. If the edge weight of the candidate edge is lower than the edge weight threshold, the candidate edge is deleted and not included in the cross-species grammar cluster similarity graph.

[0195] All grammar cluster nodes, preserved edges, and corresponding edge weights constitute a cross-species grammar cluster similarity graph. The cross-species grammar cluster similarity graph is a weighted undirected graph. The edges in the cross-species grammar cluster similarity graph represent the grammar similarity relationship between grammar clusters of different plant species that meets the preset similarity requirements. The edge weights represent the cosine similarity between the cluster centers of two corresponding grammar clusters. The cross-species grammar cluster similarity graph only retains the connections between cross-species grammar clusters with high grammar similarity, which are used for subsequent graph community discovery.

[0196] Step 8.4: Divide cross-species grammatical cluster communities based on the graph community detection algorithm, specifically including the following steps:

[0197] Step 8.4.1: Using the cross-species grammatical cluster similarity graph obtained in Step 8.3 as input, a graph community detection algorithm is used to partition the grammatical cluster nodes into communities. The Leiden graph community detection algorithm is called using the igraph graph object and the leidenalg community detection package in Python. During the call, the cross-species grammatical cluster similarity graph obtained in Step 8.3 is converted into an igraph graph object. The grammatical cluster nodes are treated as graph nodes, the retained edges are treated as graph edges, and the edge weights are... The weights are used as the edges of the graph; then the leidenalg.find_partition function is used for community partitioning.

[0198] In this embodiment, the partition type for community division adopts RBConfigurationVertexPartition (a partition type name in the leidenalg software package), and the resolution parameter resolution_parameter is set to 1.50. After each run of the Leiden graph community discovery algorithm, a community number is output for each grammar cluster node in the cross-species grammar cluster similarity graph. In the same run result, two or more grammar cluster nodes with the same community number are assigned to the same cross-species grammar cluster community, and all cross-species grammar cluster communities constitute the community division result.

[0199] Step 8.4.2: To reduce the impact of random initialization on the community partitioning results in the graph community detection algorithm, the Leiden graph community detection algorithm is repeatedly run with 10 random seeds, and the stability of the community partitioning results is evaluated. Specifically:

[0200] A baseline random seed is pre-set (in this embodiment, the baseline random seed is set to 1). The Leiden graph community detection algorithm with the baseline random seed is used to partition the cross-species grammatical cluster similarity graph to obtain the baseline community partitioning result. Subsequently, the Leiden graph community detection algorithm with 9 random seeds is used to partition the cross-species grammatical cluster similarity graph to obtain 9 sets of control community partitioning results. The baseline community partitioning result and multiple sets of control community partitioning results constitute a total of 10 community partitioning results.

[0201] Step 8.4.3: First, in the cross-species grammar cluster community in the baseline community partitioning results, construct an internal node pair for any two grammar cluster nodes within the cross-species grammar cluster community. Traverse all pairwise combinations between different grammar cluster nodes within the cross-species grammar cluster community. The set of internal node pairs of the cross-species grammar cluster community is formed by all internal node pairs of the cross-species grammar cluster community.

[0202] For each internal node pair in the set of internal node pairs, count the number of times the two grammar cluster nodes in the internal node pair are assigned to the same community in 10 community partitioning results, and divide the number by 10 to obtain the co-community support rate of the internal node pair; calculate the average co-community support rate of all internal node pairs in the set of internal node pairs of the cross-species grammar cluster community as the stability score of the cross-species grammar cluster community.

[0203] Set a stability threshold (0.7) to filter out all cross-species grammar cluster communities with stability scores greater than the stability threshold from the baseline community partitioning results and enter the hierarchical clustering in step 8.5; cross-species grammar cluster communities with stability scores less than the stability threshold are not entered into the hierarchical clustering in step 8.5; since the cross-species grammar cluster communities that finally enter the hierarchical clustering come from the community partitioning results of the same baseline random seed, the cross-species grammar cluster communities that finally enter the hierarchical clustering are a set of non-overlapping communities, and each grammar cluster node that enters the hierarchical clustering belongs to at most one cross-species grammar cluster community.

[0204] Step 8.5: Merge and audit cross-species grammatical cluster communities: Using the cross-species grammatical cluster communities obtained in Step 8.4 as input, determine whether different cross-species grammatical cluster communities need to be merged into the same candidate cross-species functional unit; for each cross-species grammatical cluster community, read the 90-dimensional truncated Z-score grammatical feature vectors of the cluster centers corresponding to all grammatical cluster nodes within the cross-species grammatical cluster community; specifically including the following steps:

[0205] Step 8.5.1: To avoid excessive weighting of the 90-dimensional community average grammar vector when the same plant species contributes multiple grammar cluster nodes in the same cross-species grammar cluster community, this embodiment uses a species balancing method to calculate the 90-dimensional community average grammar vector of the cross-species grammar cluster community. Specifically, for a certain cross-species grammar cluster community, the grammar cluster nodes in the cross-species grammar cluster community are grouped according to plant species. If a group includes only one grammar cluster node, the 90-dimensional truncated Z-score grammar feature vector of the cluster center of the grammar cluster node is used as the species representative vector of the plant species in the cross-species grammar cluster community. If a group includes multiple grammar cluster nodes, the average value obtained by calculating the arithmetic mean of the 90-dimensional truncated Z-score grammar feature vectors of the cluster centers of all grammar cluster nodes in the group is used as the species representative vector of the plant species in the cross-species grammar cluster community. Subsequently, the coordinate values ​​of the species representative vectors of different plant species in the cross-species grammar cluster community are calculated in the same grammar feature dimension to obtain the 90-dimensional community average grammar vector of the cross-species grammar cluster community.

[0206] Construct a candidate community grammar matrix for cross-species grammar clusters. Each row of the candidate community grammar matrix corresponds to a cross-species grammar cluster community, and each column corresponds to a grammatical feature dimension in the 90-dimensional community average grammar vector.

[0207] Step 8.5.2: Take the candidate community grammar matrix as input, call the pvclust function of the pvclust package in R language to perform hierarchical clustering, and output the hierarchical clustering tree between cross-species grammar cluster communities and the AU value of each clustering branch. The AU value is used to characterize the degree of support for the cross-species grammar cluster communities contained in the corresponding clustering branch to be stably classified into the same group.

[0208] If multiple cross-species grammatical cluster communities are grouped into the same cluster branch in a hierarchical clustering tree, and simultaneously meet the following conditions, then that cluster branch is considered a candidate for merging:

[0209] Condition 1: The clustering branch is not the root branch that contains all cross-species grammatical cluster communities; Condition 2: The clustering branch contains at least two cross-species grammatical cluster communities; Condition 3: The AU value of the clustering branch is not less than 0.95.

[0210] Because the clustering branches in the hierarchical clustering tree have parent-child nesting relationships, the following principles are followed when determining the final merge branch: all candidate merge branches that do not contain each other are determined as independent final merge branches; for multiple candidate merge branches that have parent-child inclusion relationships, only the candidate merge branch with the most cross-species grammar cluster communities is retained as the final merge branch, and its subordinate candidate merge branches are not allowed to become the final merge branch.

[0211] Finally, all cross-species grammatical cluster communities within each final merge branch are merged into a candidate cross-species functional unit.

[0212] Step 8.6: Upgrade the candidate cross-species functional unit to a formal cross-species plant IDR functional group pIFC, which specifically includes the following steps:

[0213] Step 8.6.1: Construct a summary table of functional evidence for candidate cross-species functional units, summarizing the plant species, taxonomic classes, grammatical cluster numbers, and significantly enriched GO annotation entries included in the candidate cross-species functional units, to form a summary table of functional evidence for candidate cross-species functional units.

[0214] For each candidate cross-species functional unit, the number of grammatical clusters, plant species, and plant taxonomic classes corresponding to the same significantly enriched GO annotation entries in the summary table of functional evidence for the candidate cross-species functional units are counted.

[0215] Step 8.6.2: When there are significantly enriched GO annotation entries in the functional evidence summary table that meet the requirements for cross-species functional consistency, the candidate cross-species functional unit is upgraded to a formal cross-species plant IDR functional class (pIFC).

[0216] The requirement for cross-species functional consistency is as follows: if the proportion of plant species supporting a significantly enriched GO annotation entry and the proportion of plant taxonomic classes supporting a significantly enriched GO annotation entry in the candidate cross-species functional unit are both not lower than a preset threshold (80%), then the significantly enriched GO annotation entry meets the requirement for cross-species functional consistency.

[0217] The plant species support ratio for a significantly enriched GO annotation entry is the proportion of the number of plant species supporting a significantly enriched GO annotation entry to the total number of plant species in the candidate cross-species functional unit; the plant taxonomic class support ratio is the proportion of the number of plant taxonomic classes supporting a significantly enriched GO annotation entry to the total number of plant taxonomic classes in the candidate cross-species functional unit.

[0218] Step 8.6.3: Sort and number all formal cross-species plant IDR functional groups pIFC:

[0219] Condition 1: There are more significantly enriched GO annotation entries that meet the requirement of cross-species functional consistency;

[0220] Condition 2: The formal cross-species plant IDR functional group pIFC covers more plant taxonomic classes;

[0221] Condition 3: The formal cross-species plant IDR functional group pIFC covers more plant species;

[0222] Condition 4: In the N1 community partitioning results, the total number of times that all two grammatical cluster nodes constituting internal node pairs in the formal cross-species plant IDR functional taxonomy pIFC are assigned to the same community is greater.

[0223] The comparison proceeds only when the previous condition is the same; after sorting, all formal cross-species plant IDR functional groups pIFC are sequentially numbered as pIFC-001, pIFC-002, pIFC-003, etc., according to the sorting order.

[0224] Step 9: Using the formal cross-species plant IDR functional groups pIFC obtained in Step 8 and the formal GIN models and formal GIN-Banks of each plant species obtained in Step 7 as reference data, construct a sequence-level plant IDR functional predictor pIFCmap. pIFCmap takes the IDR sequence to be tested as input, maps it to the grammatical clusters of the formal GIN-Bank of the reference plant species, and outputs the pIFC category prediction results based on the grammatical cluster functions and the correspondence between the grammatical clusters and pIFCs. The IDR sequence to be tested can be obtained from a database or predicted by tools such as MobiDB-lite 4.0, or designed manually. Specifically, the following steps are included:

[0225] Step 9.1: Construct the pIFCmap reference package for the sequence-level plant IDR functional predictor pIFCmap. The pIFCmap reference package includes the numbered formal cross-species plant IDR functional groups pIFC obtained in Step 8. The pIFCmap reference package also includes the formal GIN model, formal GIN-Bank, and functional annotation result table of each grammatical cluster in the formal GIN-Bank for each plant species obtained in Step 7. The pIFCmap reference package also includes the 54-dimensional compositional background statistical parameters of each plant species obtained in Step 3.

[0226] Step 9.2: The sequence-level plant IDR function predictor pIFCmap cleans the IDR sequence to be tested: The IDR sequence to be tested is cleaned by deleting sequences containing non-standard amino acid residues such as X, U, Z, J, B, and O, which will be used as the objects for subsequent grammatical feature calculation and functional prediction.

[0227] Step 9.3: The sequence-level plant IDR function predictor pIFCmap calculates the 90-dimensional truncated Z-score grammatical feature vector of the cleaned IDR sequence relative to the 54-dimensional compositional background statistical parameters of different plant species. The 36-dimensional residue arrangement class feature is calculated according to the method in step 3.4, using a set of random background sequences with the same composition as the IDR sequence itself as the background. The 54-dimensional compositional class feature is calculated according to the methods in steps 3.1 and 3.2, calling the 54-dimensional compositional background statistical parameters of each plant species in the pIFCmap reference package to calculate the 54-dimensional compositional class Z-score of the IDR sequence relative to the background of different plant species. Then, the 36-dimensional residue arrangement class Z-score of the IDR sequence is merged with the 54-dimensional compositional class Z-score calculated relative to the background of each plant species to form the 90-dimensional Z-score grammatical feature vector of the IDR sequence relative to the 54-dimensional compositional background statistical parameters of different reference plant species. Finally, following the method described in step 3, the Z-score values ​​of each grammatical feature dimension in the 90-dimensional Z-score grammatical feature vector are truncated to [-3, 3] to obtain the 90-dimensional truncated Z-score grammatical feature vector of the IDR sequence to be tested relative to the 54-dimensional compositional background statistical parameters of different plant species.

[0228] Step 9.4: The sequence-level plant IDR function predictor pIFCmap uses the formal GIN model of each plant species in the pIFCmap reference package to calculate the mapping distance from the IDR sequence to be tested to the cluster centers of each grammatical cluster in the formal GIN-Bank. Then, the mapping distances from the IDR sequence to be tested to the cluster centers of all grammatical clusters in the formal GIN-Bank of all plant species are sorted from smallest to largest to obtain the sorting table of the grammatical clusters of the IDR sequence to be tested.

[0229] Step 9.5: The sequence-level plant IDR function predictor pIFCmap obtains the top 3 grammar clusters and their corresponding function annotation result tables from the grammar cluster attribution result ranking table of the IDR sequence to be tested. These are the function prediction reference grammar clusters and function prediction results of the IDR sequence to be tested, respectively. The higher the grammar cluster ranks in terms of mapping distance, the higher the priority of its corresponding function annotation result table as a candidate function of the IDR sequence to be tested.

[0230] Step 9.6: The sequence-level plant IDR function predictor pIFCmap generates pIFC function category prediction results for the IDR sequence to be tested: It is determined whether the function prediction reference syntax cluster of the IDR sequence to be tested belongs to the formal cross-species plant IDR function group pIFC. If one or more syntax clusters in the function prediction reference syntax cluster of the IDR sequence to be tested belong to the formal cross-species plant IDR function group pIFC, then all the formal cross-species plant IDR function groups pIFC to which the function prediction reference syntax cluster of the IDR sequence to be tested are deduplicated and used as the pIFC function category prediction result of the IDR sequence to be tested; if the top 3 syntax clusters in the ranking table of the syntax clusters of the IDR sequence to be tested do not belong to any formal cross-species plant IDR function group pIFC, then there is no pIFC function category prediction result for the IDR sequence to be tested.

[0231] Step 9.7: The sequence-level plant IDR function predictor pIFCmap outputs the pIFCmap prediction result table of the IDR sequence to be tested. The pIFCmap prediction result table of the IDR sequence to be tested includes the sorting table of the grammar clusters of the IDR sequence to be tested, the function prediction reference grammar clusters and function prediction results of the IDR sequence to be tested, and the pIFC function category prediction results of the IDR sequence to be tested.

[0232] It should be noted that the embodiments described in this invention are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains can make various modifications or additions to the described embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for constructing and predicting functional taxa in the intrinsic disordered region of plants based on sequence grammar, characterized in that, Includes the following steps: Step 1: Obtain the complete proteome data file of the plant species; Step 2: Identify the IDR sequences located in the intrinsically disordered region of protein sequences in the whole proteome data file of the plant species and clean the IDR sequences. Step 3: Calculate the 90-dimensional truncated Z-score syntactic feature vector of the IDR sequence using the NARDINI+ algorithm; Step 4: Set the range of K values. Based on the 90-dimensional truncated Z-score grammatical feature vector, use GIN models with different K values ​​to cluster all IDR sequences in each plant species. According to the clustering results of each plant species, select multiple candidate K values ​​for each plant species and use the GIN model with the candidate K values ​​as the candidate GIN model. K is the number of cluster centers of the GIN model. Step 5: Cluster the IDR sequences within the corresponding plant species using the candidate GIN models for each plant species to obtain the candidate GIN-Bank for each plant species. The candidate GIN-Bank includes the IDR sequences and the grammatical clusters of the IDR sequences. Step 6: Perform GO enrichment analysis and KEGG enrichment analysis on the grammar clusters and construct a functional annotation result table for the grammar clusters; Step 7: Based on the functional annotation results table of each grammatical cluster of each plant species, determine the species-specific final cluster number K value of each plant species and the corresponding formal GIN model and formal GIN-Bank. Step 8: Construct a cross-species grammar cluster similarity graph, which includes grammar cluster nodes, preserved edges, and edge weights; use the Leiden graph community detection algorithm to perform cross-species grammar cluster community partitioning on the cross-species grammar cluster similarity graph; The pvclust function was used to perform hierarchical clustering of all cross-species grammatical cluster communities to obtain multiple candidate cross-species functional units; these candidate cross-species functional units were then upgraded to formal cross-species plant IDR functional groups pIFC. Step 9: Construct a sequence-level plant IDR function predictor pIFCmap. The pIFCmap takes the IDR sequence to be tested as input, maps the IDR sequence to the syntax cluster of the formal GIN-Bank of the plant species, and outputs the pIFC function category prediction result of the IDR sequence to be tested according to the correspondence between the syntax cluster and pIFC.

2. The method for constructing and predicting functional taxa in plant intrinsic disordered regions based on sequence grammar according to claim 1, characterized in that, The candidate GIN-Bank also includes the plant species associated with the IDR sequence, the grammatical cluster index, and the mapping distance from the IDR sequence to the nearest cluster center. Mapping distance from IDR sequence to the second nearest cluster center and mapping discrimination index Syntax clusters are constructed centered on each cluster center. The syntax cluster assignment result of the IDR sequence is the syntax cluster centered on the nearest cluster center, and the assignment number of the syntax cluster is the syntax cluster number centered on the nearest cluster center; the mapping discriminant index is... ; Map the discrimination index IDR sequences that are below a set threshold are marked as low-discrimination mapping entries; The functional annotation results table of the grammatical clusters includes the plant species, the candidate K value corresponding to the candidate GIN-Bank, the grammatical cluster number, the significantly enriched GO annotation entries, the significantly enriched KEGG pathways, and the corrected p-value and enrichment fold of each significantly enriched GO annotation entry and each significantly enriched KEGG pathway.

3. The method for constructing and predicting functional taxa in plant intrinsic disordered regions based on sequence grammar according to claim 2, characterized in that, The cross-species grammatical cluster similarity graph is constructed through the following steps: Step 8.1: Define each grammatical cluster in the formal GIN-Bank of each plant species as a grammatical cluster node. The sequence grammatical features of the grammatical cluster node are obtained by using the 90-dimensional truncated Z-score grammatical feature vector of the cluster center of the corresponding grammatical cluster. Step 8.2: Define the connection between any two syntactic cluster nodes from different plant species as a candidate edge, and calculate the edge weight of the candidate edge between any two syntactic cluster nodes based on the following formula: ； In the formula, and These are the 90-dimensional truncated Z-score grammatical feature vectors representing the cluster centers of the two grammatical cluster nodes. and They are respectively and In the Dimensionally truncated coordinates on Z-score syntactic features, index Take values ​​from 1 to 90; Let the edge weights be the candidate edges between two grammar cluster nodes; Step 8.3: Set the edge weight threshold, and select candidate edges whose edge weights are greater than the edge weight threshold as retained edges. All grammar cluster nodes, retained edges and their corresponding edge weights constitute a cross-species grammar cluster similarity graph. The cross-species grammatical cluster community is divided through the following steps: Step 8.4.1: Using the Leiden graph community detection algorithm, output a community number for each grammar cluster node in the cross-species grammar cluster similarity graph, and assign grammar cluster nodes with the same community number to a cross-species grammar cluster community. All cross-species grammar cluster communities constitute the community partitioning result. Step 8.4.2: Pre-set a baseline random seed, use the Leiden graph community detection algorithm with the baseline random seed to partition the cross-species grammatical cluster similarity graph to obtain the baseline community partitioning result, and then use the Leiden graph community detection algorithm with multiple random seeds to partition the cross-species grammatical cluster similarity graph to obtain multiple sets of control community partitioning results. The baseline community partitioning result and the multiple sets of control community partitioning results constitute a total of N1 community partitioning results. Step 8.4.3: First, in the cross-species grammar cluster community in the baseline community partitioning results, construct any two grammar cluster nodes within the cross-species grammar cluster community as an internal node pair. The set of internal node pairs of the cross-species grammar cluster community is composed of all internal node pairs of the cross-species grammar cluster community. Secondly, the stability score is calculated based on the internal node pair set for cross-species grammatical cluster communities in the following manner: The number of times two grammar cluster nodes in an internal node pair are assigned to the same community in N1 community partitioning results is counted, and the number is divided by N1 to obtain the co-community support rate of the internal node pair; the average co-community support rate of all internal node pairs in the set of internal node pairs of the cross-species grammar cluster community is calculated as the stability score of the cross-species grammar cluster community. Finally, a stability threshold is set to filter out all cross-species grammatical cluster communities in the baseline community partitioning results whose stability scores are greater than the stability threshold and enter hierarchical clustering. Cross-species grammar cluster communities with stability scores below the stability threshold are not included in hierarchical clustering.

4. The method for constructing and predicting functional taxa in plant intrinsic disordered regions based on sequence grammar according to claim 3, characterized in that, The step of using the pvclust function to perform hierarchical clustering of all cross-species grammatical cluster communities to obtain multiple candidate cross-species functional units specifically includes the following steps: Step 8.5.1: Group the grammar cluster nodes in the cross-species grammar cluster community according to plant species. If a group contains only one grammar cluster node, the 90-dimensional truncated Z-score grammar feature vector of the cluster center of the grammar cluster node is used as the species representative vector of the plant species in the cross-species grammar cluster community. If a group contains multiple grammar cluster nodes, the average of the 90-dimensional truncated Z-score grammar feature vectors of the cluster centers of all grammar cluster nodes in the group is used as the species representative vector of the plant species in the cross-species grammar cluster community. Finally, the species representative vectors of all plant species within the cross-species grammar cluster community are averaged to obtain the 90-dimensional community average grammar vector of the cross-species grammar cluster community. Construct a candidate community grammar matrix for cross-species grammar clusters. Each row of the candidate community grammar matrix corresponds to a cross-species grammar cluster community, and each column corresponds to a grammatical feature dimension in the 90-dimensional community average grammar vector. Step 8.5.2: Using the candidate community grammar matrix as input, perform hierarchical clustering using the pvclust function, and output the hierarchical clustering tree between cross-species grammar cluster communities and the AU value of each clustering branch; If multiple cross-species grammatical cluster communities are grouped into the same cluster branch in a hierarchical clustering tree, and the following conditions are met simultaneously, the cluster branch is considered a candidate for merging: Condition 1: The clustering branch is not the root branch that includes all cross-species grammatical cluster communities; Condition 2: The clustering branch includes at least two cross-species grammatical cluster communities; Condition 3: The AU value of the clustering branch is not lower than the set AU value threshold. Then, all candidate merge branches that do not contain each other are determined as a final merge branch; for multiple candidate merge branches that have parent-child inclusion relationships, only the candidate merge branch with the most cross-species grammar cluster communities is retained as the final merge branch. Finally, all cross-species grammatical cluster communities within each final merge branch are merged into a candidate cross-species functional unit.

5. The method for constructing and predicting functional taxa in plant intrinsic disordered regions based on sequence grammar according to claim 4, characterized in that, The process of upgrading candidate cross-species functional units to formal cross-species plant IDR functional groups pIFC specifically includes the following steps: Step 8.6.1: First, construct a summary table of functional evidence for candidate cross-species functional units. The summary table includes the plant species, taxonomic classes, grammatical cluster numbers, and significantly enriched GO annotation entries included in the candidate cross-species functional units. Then, count the number of grammatical clusters, plant species, and taxonomic classes corresponding to the same significantly enriched GO annotation entries in the summary table. Step 8.6.2: When there are significantly enriched GO annotation entries in the functional evidence summary table that meet the requirements for cross-species functional consistency, the candidate cross-species functional unit is upgraded to a formal cross-species plant IDR functional group pIFC. The requirement for cross-species functional consistency is as follows: if the proportion of plant species supporting a significantly enriched GO annotation entry and the proportion of plant taxonomic classes supporting a significantly enriched GO annotation entry in the candidate cross-species functional unit are both not lower than a preset threshold, then the significantly enriched GO annotation entry meets the requirement for cross-species functional consistency. The plant species support ratio for a significantly enriched GO annotation entry is the proportion of the number of plant species supporting a significantly enriched GO annotation entry to the total number of plant species in the candidate cross-species functional unit; the plant taxonomic class support ratio is the proportion of the number of plant taxonomic classes supporting a significantly enriched GO annotation entry to the total number of plant taxonomic classes in the candidate cross-species functional unit. Step 8.6.3: Sort and number all formal cross-species plant IDR functional groups pIFC: Condition 1: There are more significantly enriched GO annotation entries that meet the requirement of cross-species functional consistency; Condition 2: The formal cross-species plant IDR functional group pIFC covers more plant taxonomic classes; Condition 3: The formal cross-species plant IDR functional group pIFC covers more plant species; Condition 4: In the N1 community partitioning results, the total number of times that all two grammatical cluster nodes constituting internal node pairs in the formal cross-species plant IDR functional taxonomy pIFC are assigned to the same community is greater. The comparison of the next condition only proceeds when the previous condition is the same; after sorting, all formal cross-species plant IDR functional groups pIFC are numbered sequentially according to the sorting order.

6. The method for constructing and predicting functional taxa in plant intrinsic disordered regions based on sequence grammar according to claim 5, characterized in that, The sequence-level plant IDR function predictor pIFCmap obtains the pIFC function category prediction results of the IDR sequence to be tested through the following steps: Step 9.1: Construct the pIFCmap reference package for the sequence-level plant IDR functional predictor pIFCmap. The pIFCmap reference package includes the numbered official cross-species plant IDR functional groups pIFC, as well as the official GIN model, official GIN-Bank, and functional annotation result table of each grammatical cluster in the official GIN-Bank for each plant species; the pIFCmap reference package also includes 54-dimensional compositional background statistical parameters for each plant species. Step 9.2: The sequence-level plant IDR function predictor pIFCmap cleans the IDR sequence to be tested; Step 9.3: The sequence-level plant IDR function predictor pIFCmap then calculates the 90-dimensional truncated Z-score grammatical feature vector of the cleaned IDR sequence to be tested relative to the 54-dimensional compositional background statistical parameters of different plant species. Step 9.4: The sequence-level plant IDR function predictor pIFCmap uses the formal GIN model of each plant species in the pIFCmap reference package to calculate the mapping distance from the IDR sequence to be tested to the cluster center of each grammatical cluster in the formal GIN-Bank. Then, the mapping distance from the IDR sequence to be tested to the cluster center of all grammatical clusters in the formal GIN-Bank of all plant species is sorted from smallest to largest to obtain the sorting table of the grammatical clusters of the IDR sequence to be tested. Step 9.5: The sequence-level plant IDR function predictor pIFCmap obtains the top 3 syntax clusters in the sorting table of the IDR sequence to be tested, and the corresponding function annotation result table are the function prediction reference syntax clusters and function prediction results of the IDR sequence to be tested. Step 9.6: The sequence-level plant IDR function predictor pIFCmap determines whether the function prediction reference syntax cluster of the IDR sequence to be tested belongs to the formal cross-species plant IDR functional group pIFC. If one or more syntax clusters in the function prediction reference syntax cluster of the IDR sequence to be tested belong to the formal cross-species plant IDR functional group pIFC, then the pIFC function category prediction result of the IDR sequence to be tested is obtained by deduplicating all the formal cross-species plant IDR functional groups pIFC to which the function prediction reference syntax cluster of the IDR sequence to be tested belongs; if the top 3 syntax clusters in the ranking table of the syntax clusters of the IDR sequence to be tested do not belong to any formal cross-species plant IDR functional group pIFC, then there is no pIFC function category prediction result for the IDR sequence to be tested. Step 9.7: The sequence-level plant IDR function predictor pIFCmap outputs the pIFCmap prediction result table of the IDR sequence to be tested. The pIFCmap prediction result table of the IDR sequence to be tested includes the sorting table of the grammar clusters of the IDR sequence to be tested, the function prediction reference grammar clusters and function prediction results of the IDR sequence to be tested, and the pIFC function category prediction results of the IDR sequence to be tested.

7. The method for constructing and predicting functional taxa in plant intrinsic disordered regions based on sequence grammar according to claim 6, characterized in that, The 90-dimensional truncated Z-score grammatical feature vector of the IDR sequence includes the truncated Z-score value of the 90-dimensional Z-score grammatical feature vector of the IDR sequence, and the 90-dimensional Z-score grammatical feature vector of the IDR sequence includes the 54-dimensional composition class Z-score of the IDR sequence and the 36-dimensional permutation class Z-score of the IDR sequence. The 54-dimensional component class Z-score of the IDR sequence is obtained in the following way: First, the 54-dimensional compositional background statistical parameters of each plant species are pre-calculated: for each IDR sequence of a plant species, the original feature values ​​of each of the 54 compositional features are calculated; for each compositional feature, the set of original feature values ​​of all IDR sequences of the plant species on each compositional feature is taken as the background distribution of the plant species on the corresponding compositional feature; then the mean and standard deviation of the background distribution on each compositional feature are calculated, and a total of 54-dimensional compositional background statistical parameters of the plant species are obtained. Secondly, calculate the 54-dimensional component class Z-score for each IDR sequence. The 54-dimensional component class Z-score for each IDR sequence includes the Z-score of each IDR sequence on each of the 54 component class features. The Z-score of the IDR sequence on a single constituent class feature is obtained as follows: the original feature value of the IDR sequence on a single constituent class feature is subtracted from the mean of the background distribution of the corresponding constituent class feature of the plant species to which the IDR sequence belongs, and then divided by the standard deviation of the background distribution of the constituent class feature of the plant species to which the IDR sequence belongs. The 36-dimensional permutation class Z-score of the IDR sequence is obtained through the following steps: Step 3.3, Residue Type Identification: The amino acid composition of the IDR sequence is divided into 8 residue types according to physicochemical properties. The 8 residue types are polar residues, hydrophobic residues, positively charged residues, negatively charged residues, aromatic residues, alanine, proline and glycine. Calculation of original residue arrangement features: The 8 residue types are paired to obtain 36 residue combinations; the arrangement parameters in the linear IDR sequence of each residue combination are calculated as the original residue arrangement features of the corresponding residue combination in the IDR sequence, resulting in a total of 36 original residue arrangement features of the IDR sequence. Step 3.4: The residue type ratio is calculated as follows: Count the number of residues of each of the 8 residue types in the IDR sequence, and divide the number of residues of each residue type in the IDR sequence by the total length of the IDR sequence to obtain the residue ratio of each residue type in the IDR sequence. For any residue arrangement class original feature in the 36-dimensional residue arrangement class original features, if the residue type ratio of any one of the two residue types corresponding to the residue arrangement class original feature in the IDR sequence is lower than a set threshold, then the corresponding residue arrangement class original feature is marked as a low-frequency arrangement feature, and the residue combination corresponding to the low-frequency arrangement feature is a low-frequency residue combination. Homogeneous random shuffling: While keeping the amino acid composition of the IDR sequence unchanged, the IDR sequence is randomly shuffled N2 times. Each random shuffling yields a random sequence, resulting in a total of N2 random sequences of the IDR sequence. These N2 random sequences together constitute the set of homogeneous random background sequences of the IDR sequence itself. Calculation of 36-dimensional permutation-class random background statistical parameters: For any single-class combination of residues in the IDR sequence, calculate the permutation parameters of the single-class combination of residues in all random sequences in the random background sequence set of the same composition of the IDR sequence. Then calculate the mean and standard deviation of all permutation parameters of the single-class combination of residues to obtain the permutation-class random background statistical parameters of the single-class combination of residues. Calculate the permutation-class random background statistical parameters for each of the 36 classes of residues to obtain the 36-dimensional permutation-class random background statistical parameters of the IDR sequence. Step 3.5, 36-dimensional permutation class Z-score calculation: Calculate the 36-dimensional permutation class Z-score for each IDR sequence: The 36-dimensional permutation class Z-score for each IDR sequence includes the Z-score for each IDR sequence on the 36 classes of residue combinations; The Z-scores of the IDR sequence on single-class residue combinations are obtained as follows: First, the Z-score of the IDR sequence on low-frequency residue combinations is set to 0; then, the original characteristics of the residue arrangement class of the IDR sequence on the other single-class residue combinations (excluding low-frequency residue combinations) are subtracted from the mean of the random background statistical parameters of the IDR sequence on the corresponding residue combination, and then divided by the standard deviation of the IDR sequence on the random background statistical parameters of the corresponding residue combination. When calculating the Z-score of the 36-dimensional permutation class, each amino acid in the IDR sequence is converted into an integer number sequence according to a preset encoding rule. Then, based on the integer number sequence corresponding to the IDR sequence, residue type identification, original feature calculation of residue permutation class, residue type ratio calculation, random shuffling of the same composition, calculation of random background statistical parameters of the 36-dimensional permutation class, and calculation of the Z-score of the 36-dimensional permutation class are performed. The calculation is accelerated by compilation, parallel computing, or vectorized computing. The entire IDR sequence of all plant species is then divided into multiple computational blocks according to a preset number. Each computational block includes multiple IDR sequences. Within the same computational block, each IDR sequence independently constructs its own set of homologous random background sequences based on its own amino acid composition. Then, the 36-dimensional permutation random background statistical parameters and the 36-dimensional permutation Z-score of each IDR sequence are calculated. Finally, after the calculation of each computational block is completed, the corresponding block result file is output. After all computational blocks are completed, the block result files are merged to obtain the 36-dimensional permutation Z-score of the entire IDR sequence of all plant species.

8. The method for constructing and predicting functional taxa in plant intrinsic disordered regions based on sequence grammar according to claim 7, characterized in that, The candidate GIN models were selected in the following manner: Step 4.1: Select all IDR sequences within the plant species whose length is within the set length range to construct a training subset of the GIN model within the species; extract the 90-dimensional truncated Z-score grammatical feature vector of each IDR sequence in the training subset of the GIN model within the species to form the training feature matrix of the GIN model of the plant species. Each row of the training feature matrix corresponds to an IDR sequence, and each column of the training feature matrix corresponds to a grammatical feature dimension in the 90-dimensional truncated Z-score grammatical feature vector. Step 4.2: Use GIN models with different K values ​​to perform KMeans clustering on the training feature matrix, and take the clustering result with the smallest inertia value as the clustering result corresponding to the K value; Step 4.3: Plot the inertia curve as the inertia value changes with the K value. The K values ​​corresponding to the inflection points of the inertia curve are K1 and K2, respectively, identified by the Kneedle algorithm and the tail-fitelbow algorithm. When K1 and K2 are different, calculate the arithmetic mean of K1 and K2 and round it to the nearest integer to get K3. If K1 and K2 are not consecutive integers, take K1, K2 and K3 as three candidate K values. If K1 and K2 are consecutive integers, K3 is the larger of K1 and K2 plus 1. If K3 exceeds the preset range of K values, K3 is the smaller of K1 and K2 minus 1. When K1 and K2 are the same, K0-2, K0 and K0+2 are taken as three candidate K values, and K0=K1=K2; Step 4.4: The three candidate K values ​​of the plant species are used as the three candidate GIN models for the plant species.

9. The method for constructing and predicting functional taxa in plant intrinsic disordered regions based on sequence grammar according to claim 8, characterized in that, The significantly enriched GO annotation entries, significantly enriched KEGG pathways, and their respective corrected p-values ​​and enrichment folds are obtained through the following steps: Step 6.1: Extract the genes to which all IDR sequences belong in the syntax cluster to form the gene set of the syntax cluster; IDR sequences whose start position is less than 10 amino acid residues from the N-terminus of a protein or whose termination position is less than 10 amino acid residues from the C-terminus of a protein are marked as non-linked IDR sequences; otherwise, they are marked as linked IDR sequences. exist IDR sequences with a length of ≥1.5 were selected, and non-linked IDR sequences with a length of 35 amino acid residues or linked IDR sequences with a length of 45 amino acid residues and their corresponding genes were constructed as functional enrichment analysis subsets. For candidate GIN-Bank, all non-redundant genes that entered the functional enrichment analysis subset were defined as the background gene set of the candidate GIN-Bank. For any grammar cluster in the candidate GIN-Bank, the non-redundant genes that entered the functional enrichment analysis subset within the gene set of the grammar cluster were defined as the gene set to be tested for the grammar cluster. Step 6.2 First, obtain the original GO annotation table: using the gene set to be tested of the syntax cluster and the gene numbers involved in the background gene set of the candidate GIN-Bank to which the syntax cluster belongs as the search objects, obtain the correspondence between genes and GO annotation entries through Ensembl Plants BioMart to form the original GO annotation table; Secondly, obtain the GO ontology file and the plant GO slim definition file; Next, the GO annotation entries in the original GO annotation table are mapped to the GO annotation entries in the plant GO slim definition file to obtain the plant GO slim annotation table; Finally, the plant GO slim annotation tables were organized into two types of files: GO-TERM2GENE and GO-TERM2NAME. Step 6.3: Perform GO enrichment analysis on the set of genes to be tested for the grammatical clusters: First, the non-redundant genes in the plant GO slim annotation table that are the genes to be tested of the grammar cluster are defined as the GO gene set of the grammar cluster, and the non-redundant genes in the plant GO slim annotation table that are the background genes of the candidate GIN-Bank to which the grammar cluster belongs are defined as the GO background gene set. Secondly, using the GO background gene set as the statistical background, the enricher function was employed, with GO-TERM2GENE and GO-TERM2NAME as reference files, to perform enrichment analysis on the GO gene set to be tested for the grammar cluster. The original p-values ​​and enrichment folds of each GO annotation entry corresponding to the GO gene set to be tested were obtained. The original p-values ​​of each GO annotation entry were corrected using the Benjamini-Hochberg method with multiple tests to obtain the corrected p-values ​​of each GO annotation entry. GO annotation entries with corrected p-values ​​not greater than the first set p-value threshold were marked as significantly enriched GO annotation entries of the grammar cluster. Step 6.4: For plant species with KEGG species IDs, obtain the KEGG pathway list of plant species, and obtain the correspondence between the genes involved in the gene set to be tested of the syntax cluster and the background gene set of the candidate GIN-Bank to which the syntax cluster belongs and the KEGG pathway, and organize them into local snapshots KEGG-TERM2GENE and KEGG-TERM2NAME. Step 6.5: Perform KEGG enrichment analysis on the set of genes to be tested for the grammatical cluster: First, non-redundant genes with KEGG pathway annotations in the gene set to be tested of the grammar cluster are defined as the KEGG gene set to be tested of the grammar cluster; non-redundant genes with KEGG pathway annotations in the background gene set of the candidate GIN-Bank to which the grammar cluster belongs are defined as the KEGG background gene set. Secondly, using the KEGG background gene set as the statistical background, the enricher function was employed, with local snapshots KEGG-TERM2GENE and KEGG-TERM2NAME as reference files, to perform KEGG pathway enrichment analysis on the KEGG gene set to be tested in the syntax cluster. The original p-values ​​and enrichment folds of each KEGG pathway corresponding to the KEGG gene set to be tested in the syntax cluster were obtained. The original p-values ​​of each KEGG pathway were corrected using the Benjamini-Hochberg method with multiple tests to obtain the corrected p-values ​​of each KEGG pathway. KEGG pathways with corrected p-values ​​not greater than a second set p-value threshold were marked as significantly enriched KEGG pathways of the syntax cluster.

10. The method for constructing and predicting functional taxa in plant intrinsic disordered regions based on sequence grammar according to claim 9, characterized in that, The species-specific final cluster number K value of the plant species, the corresponding formal GIN model, and the formal GIN-Bank are obtained through the following steps: First, a candidate K comprehensive evaluation table is constructed, which includes candidate K value, proportion of low-discrimination mapping entries, number of syntactic clusters with significant functional annotations, and functional annotation coverage. Secondly, based on the candidate K comprehensive evaluation table, the species-specific final cluster number K value is selected from the three candidate K values ​​of plant species according to the following conditions: Condition 1: The number of syntactic clusters with more significant functional annotations; Condition 2: It has broader functional annotation coverage; Condition 3: It has a lower proportion of low-discrimination mapping entries; The comparison will only proceed to the next condition if the comparison results are the same as the previous condition. Among them, the proportion of low-discrimination mapping entries in the candidate GIN-Bank is The proportion of IDR sequences with a value less than 1.5 to the total number of IDR sequences in the plant species; a grammatical cluster with significant functional annotation is a grammatical cluster that is significantly enriched with at least one GO annotation entry or significantly enriched with the KEGG pathway; functional annotation coverage is the proportion of grammatical clusters with significant functional annotation to the total number of grammatical clusters in the candidate GIN-Bank. The candidate GIN models and candidate GIN-Banks corresponding to the species-specific final cluster number K are respectively the formal GIN model, the formal GIN-Bank, and the functional annotation results of each grammatical cluster in the formal GIN-Bank.

Images