High-resolution comparative genomics and compositional analysis of WRKY transcription factors across seven cereal and model plant genomes Authors: Zainab Hanjra, Saqib Ali, Tehrim Fatima, Misbah Saeed, Saman Khalil, Muhammad Rashid, Muhammad Habib Ur-Rahman, Ayman El Sabagh, Ibrahim Al-Ashkar, Qamar uz Zaman Published: 2026 ABSTRACT BackgroundWRKY transcription factors are major regulators of plant stress responses and development, yet their evolutionary dynamics across major cereals and dicot needs further characterization. Previous studies cataloged WRKY genes individually in single species, but no comprehensive comparative analysis integrating phylogenomic, syntenic, and compositional analyses across the monocot-dicots divide has been conducted. This knowledge gap limits our ability to identify conserved functional constraints versus lineage-specific evolutionary innovations in WRKY regulatory networks.ResultsA high-resolution comparative genomic analysis of 547 WRKY genes was performed across seven plant genomes: Arabidopsis thaliana, Oryza sativa subspecies japonica (126 genes), indica (109 genes), and the previously uncharacterized O. glaberrima (51 genes), Brachypodium distachyon (56 genes), Zea mays (105 genes), and Triticum aestivum (52 genes). Phylogenetic analysis revealed distinct group proportions, with Group III representing 70% of total WRKY genes in cereals compared to only 20.8% Group I and 12.8% Group II, representing a monocot-specific expansion. Group III genes constitute the dominant WRKY classification across all cereal species examined, with pronounced enrichment in cereals (mean 64.9%; range 59.6–69.8%) representing a 2.6-fold difference relative to Arabidopsis (25.0%). This cereal-specific expansion is mechanistically driven by tandem duplication events significantly enriched for Group III genes (Fisher's exact test, p = 0.007), maintained under strong purifying selection (mean Ka/Ks = 0.141). Synteny analysis identified 218 collinear gene pairs between rice and Brachypodium, 186 with Zea mays, 164 with Triticum aestivum, and 142 with Arabidopsis, indicating lineage-specific conservation patterns. Evolutionary rate analysis revealed highly conserved WRKY domains (Ka/Ks = 0.08-0.12) juxtaposed against rapidly evolving flanking regions (Ka/Ks = 0.42-0.78), suggesting strong purifying selection on DNA-binding function. t-SNE analysis identified 22 bridge genes with intermediate compositional profiles spanning the monocot-dicot divide, distributed across four cereal lineages and exhibiting structural properties consistent with a directional evolutionary trajectory from ancestral Group I to derived Group III configurations. Notably, O. glaberrima showed reduced Group I representation (9.8%) and elevated Group III proportion (80.4%), indicating lineage-specific retention patterns during independent domestication.ConclusionsThis comprehensive analysis establishes a quantitative framework for dissecting WRKY gene family evolution in cereals, identifies stress-responsive orthologs prioritized for crop improvement, and demonstrates that ancient polyploidy, recent segmental duplication, and differential selection pressure collectively shape cereal regulatory architecture. The study provides a foundation for targeted breeding strategies to enhance climate resilience in major cereal and dicot crops.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12870-026-08661-0. Introduction Transcription factors (TFs) serve as the master architects of plant adaptation, translating developmental cues and environmental stimuli into complex gene expression programs [1]. Among the plant-specific TF lineages, the WRKY superfamily is distinguished by its evolutionary expansion and pivotal role in regulating immunity, abiotic stress tolerance, and hormone signaling [2–4]. Defined by the highly conserved WRKYGQK heptapeptide and a characteristic zinc-finger motif (C2H2 or C2HC), WRKY proteins function by binding to W-box cis-elements to modulate transcriptional networks [5, 6]. The classification of this family into Groups I, II, and III is well-established [7]; however, the precise evolutionary mechanisms driving its expansion and structural diversification across complex cereal genomes remain an active area of investigation. The family Poaceae (grasses) underpins global food security, with rice (Oryza sativa), wheat (Triticum aestivum), and maize (Zea mays) providing over 60% of human caloric intake (FAO, 2020). Despite sharing a common ancestor, these lineages have undergone distinct evolutionary trajectories shaped by ancient whole-genome duplications (WGD) and lineage-specific polyploidization events [8, 9]. The genomic landscape of these crops varies dramatically, ranging from the diploid, compact genome of rice (~ 430 Mb) and the model grass Brachypodium distachyon, to the massive, hexaploid genome of wheat (~ 17 Gb) (International Rice Genome Sequencing Project, 2005; The International Brachypodium Initiative, 2010; IWGSC, 2018). Previous studies cataloged WRKY inventories in these species individually, revealing that gene counts fluctuate significantly from 48 genes in Arabidopsis thaliana to over 100 in rice and maize, driven primarily by segmental and WGD-derived duplications [10–12]. The accurate assessment of WRKY family size in polyploid wheat has been complicated by the retention of homeologous gene copies across A, B, and D subgenomes, with reported counts ranging from 52 non-redundant genes to over 180 when all homeologs are counted separately [12, 13]. Within the Oryza genus, a unique evolutionary microcosm exists. The divergence between the Asian subspecies (indica and japonica) and the independently domesticated African rice (O. glaberrima) offers a powerful framework to study domestication signatures and lineage-specific gene retention patterns [14, 15]. Early studies identified key WRKY regulators in rice, such as the blast-resistance factor OsWRKY45 [16, 17], and mapped their responses to abiotic stress [18]; however, these analyses have largely been confined to single species or subspecies. A comprehensive phylogenomic synthesis that integrates the Oryza complex with broader cereal lineages is currently lacking. Such a multi-genome perspective is essential to determine whether the extensive expansion of WRKYs in cereals has been accompanied by conserved structural constraints or lineage-specific functional divergence [19, 20]. Beyond traditional phylogenetic grouping, recent advances in comparative genomics highlight sequence composition features specifically codon usage bias (CUB) and GC content as critical evolutionary signals [21, 22]. In monocots, particularly grasses, a distinct GC-rich genomic environment imposes unique selection pressures on coding sequences compared to dicots like Arabidopsis. These compositional biases are not random; they influence mRNA stability, translational efficiency, and adaptation to environmental stress [23, 24]. However, most existing WRKY studies overlook these genomic signatures. Integration of multivariate statistical approaches, including principal component analysis (PCA) and non-linear dimensionality reduction methods such as t-distributed stochastic neighbor embedding (t-SNE), provides powerful tools to uncover cryptic evolutionary patterns not visible through traditional phylogenetic analyses alone [6, 25]. To date, no study has systematically integrated these compositional metrics with synteny and selection pressure (Ka/Ks) analysis across the monocot-dicot divide. Addressing this gap is crucial for understanding the molecular evolutionary forces that fine-tune gene regulation and functional diversification [26, 27]. In this study, we perform a high-resolution comparative analysis of the WRKY family across seven representative genomes: the dicot model Arabidopsis thaliana, the monocot model Brachypodium distachyon, the polyploid crops Triticum aestivum and Zea mays, and the three key members of the Oryza complex (O. sativa ssp. indica, O. sativa ssp. japonica, and O. glaberrima). The objectives of this study were to: (1) trace the evolutionary history and expansion mechanisms of WRKYs across diverse ploidy levels through comprehensive genome-wide identification; (2) elucidate the structural conservation of gene architecture, conserved motifs, and chromosomal organization patterns; (3) decipher the selective pressures acting on duplicated genes via synteny mapping, gene duplication analysis, and Ka/Ks ratio calculations; (4) uncover lineage-specific compositional signatures (amino acid composition, nucleotide bias, and codon usage patterns) through multivariate analyses including PCA and t-SNE; and (5) identify high-priority candidate WRKY genes for future functional characterization and crop improvement through multivariate prioritization. This comprehensive study provides a robust genomic resource for understanding the evolutionary dynamics of stress-responsive transcription factors in major cereal crops and establishes a foundation for targeted crop improvement strategies. Materials and methods Genome Data Acquisition and Preprocessing Genomic sequences, protein sequences, and gene annotation files (GFF3 format) were retrieved from public databases for seven representative plant species. For Arabidopsis thaliana (TAIR10), Oryza sativa ssp. japonica (MSU v7.0), O. sativa ssp. indica (ASM465v1), Oryza glaberrima (AGRA_AJGP1), Brachypodium distachyon (Brachypodium_distachyon_v3.0), Zea mays (B73 RefGen_v4), and Triticum aestivum (IWGSC RefSeq v1.0), genome assemblies and annotations were downloaded from the Phytozome v13 database (https://phytozome-next.jgi.doe.g.ov/) [28], Ensembl Plants (https://plants.ensembl.org/) [29], and NCBI GenBank (https://www.ncbi.nlm.nih.gov/genome/) [30]. Coding sequences (CDS) and corresponding protein sequences were extracted using custom Python scripts (Python v3.9; https://www.python.org/) with BioPython v1.79 [31]. For polyploid wheat, nearly identical homeologous gene copies from the A, B, and D subgenomes were identified using CD-HIT v4.8.1 (http://weizhongli-lab.org/cd-hit/) [32] with 95% sequence identity threshold and the longest representative sequence retained. The 95% identity threshold was selected based on the mean pairwise identity between confirmed homeologous WRKY gene pairs in the IWGSC RefSeq v1.0 annotation (mean 97.3% ± 1.8% SD for verified A-B homeologs; mean 96.1% ± 2.1% SD for A-D homeologs), ensuring that genuine homeologs were collapsed while preserving functionally divergent paralogs. To validate the sensitivity of this threshold, we performed a three-level sensitivity analysis using thresholds of 90%, 95%, and 98% sequence identity (Supplementary Table S1). At 90% identity, 47 non-redundant genes were retained (10 fewer than at 95%), with 8 syntenic pairs lost and mean Ka/Ks shifting by less than 0.02 units. At 98% identity, 61 genes were retained (9 additional), with 6 borderline homeolog pairs retained as separate entries, contributing marginal noise to phylogenetic grouping (± 1 gene per group). The 95% threshold produced the most stable results across downstream analyses and is consistent with thresholds used in comparable polyploid WRKY studies [12, 13]. Supplementary Table S1 provides gene counts, phylogenetic group compositions, and Ka/Ks distributions for all three thresholds. Genome-wide identification of WRKY genes and orthologue retrieval WRKY transcription factor genes from Oryza sativa (japonica), Oryza sativa (indica), and Oryza glaberrima were initially retrieved from PlantTFDB (Plant Transcription Factor Database; http://planttfdb.gao-lab.org/). Rice WRKY genes served as reference sequences for comparative orthologue identification across Arabidopsis thaliana, Triticum aestivum, Zea mays, and Brachypodium distachyon using the “Orthologous Group” and “Best hit in Arabidopsis thaliana” sections of PlantTFDB. To comprehensively identify and validate all WRKY family members, a dual-strategy approach combining Hidden Markov Model (HMM) profiling and BLAST-based sequence similarity searches was employed. The WRKY domain HMM profile (PF03106) was obtained from the Pfam database v35.0 (http://pfam.xfam.org/) [33] and used to scan all predicted protein sequences using HMMER v3.3.2 (http://hmmer.org/) [34] with an E-value cutoff of 1 × 10⁻⁵. In parallel, BLASTN and BLASTP searches were performed using rice WRKY CDS and protein sequences as queries against species-specific proteomes (BLAST+ v2.12.0;https://blast.ncbi.nlm.nih.gov/Blast.cgi) [35] with an E-value threshold of 1 × 10⁻¹⁰ for comprehensive searches and ≤ 1 × 10⁻⁵ for orthologue validation. Hits with sequence identity ≥ 60–70% were retained for orthologue assignments. Candidate sequences from PlantTFDB, HMM, and BLAST methods were merged, and redundant entries were removed. The presence and integrity of the conserved WRKYGQK heptapeptide and zinc-finger motif (C2H2 or C2HC) were confirmed using the NCBI Conserved Domain Database (CDD; https://www.ncbi.nlm.nih.gov/Structure/cdd/cdd.shtml) (Lu et al., 2020) and SMART v9.0 (http://smart.embl-heidelberg.de/) [36]. Sequences lacking complete WRKY domains were excluded from further analysis. For cases where multiple orthologue candidates were identified, the best-supported sequence was selected based on sequence identity, E-value, and annotation consistency. Phylogenetic analysis and classification Multiple sequence alignment of full-length WRKY protein sequences was performed using MAFFT v7.490 (https://mafft.cbrc.jp/alignment/software/) [37] with the E-INS-i algorithm optimized for sequences with multiple conserved domains. Alignments were manually inspected and trimmed using Jalview v2.11.2.5 (https://www.jalview.org/) [38] to remove poorly aligned regions. Maximum likelihood (ML) phylogenetic trees were constructed using MEGA X v12.0 (https://www.megasoftware.net/) [39] with the Jones-Taylor-Thornton (JTT) amino acid substitution model and 1,000 bootstrap replicates using the bootstrap method. Model selection was based on the lowest Bayesian Information Criterion (BIC) score among candidate models tested in MEGA. Phylogenic trees were visualized and annotated using iTOL v6.5.7 (https://itol.embl.de/) [40] and formatted in Adobe Illustrator CC 2021. WRKY genes were classified into Groups I, II (subgroups IIa–IIe), and III based on phylogenetic topology, number of WRKY domains, and zinc-finger type following established nomenclature [5, 7]. Gene structure and conserved motif analysis Exon-intron structures were determined by aligning CDS sequences against genomic DNA using TBtools v1.120 (https://github.com/CJ-Chen/TBtools) [41], Gene structure diagrams were generated using the Gene Structure View (Advanced) module in TBtools. Conserved protein motifs were identified using the MEME Suite v5.4.1 (https://meme-suite.org/meme/) [42] with parameters set to identify 10 motifs with lengths ranging from 6 to 50 amino acids. Motif arrangement patterns were visualized using the Simple BioSequence Viewer module in TBtools and cross-referenced with InterProScan v5.56-89.0 (https://www.ebi.ac.uk/interpro/) [43] for functional domain annotation. Chromosomal localization and gene duplication analysis Chromosomal positions of WRKY genes were extracted from GFF3 annotation files and visualized using the Visualize Gene Structure (One-Step MCScanX-BlastP) and Gene Location Visualizer modules in TBtools [41]. Gene density heatmaps were generated using R v4.2.1 (https://www.r-project.org/) with the circlize v0.4.15 package [44]. Tandem duplications were defined as adjacent WRKY genes separated by five or fewer intervening genes within a 200 kb window, while segmental duplications were identified through all-versus-all BLASTp searches followed by collinearity analysis using MCScanX v1.1 (https://github.com/wyp1125/MCScanX) [45]. Duplicate gene pairs were classified based on genomic location and sequence similarity (> 70% identity over > 70% alignment length). Synteny analysis and evolutionary rate calculation Cross-species synteny relationships were investigated by conducting reciprocal BLASTp searches (E-value ≤ 1 × 10⁻¹⁰, top 5 hits retained) between rice and each of the other six species. Collinear blocks containing WRKY genes were identified and visualized using the Advanced Circos module in TBtools and MCScanX. Syntenic gene pairs were defined as orthologs residing within collinear genomic blocks with conserved gene order. For each syntenic pair, synonymous (Ks) and non-synonymous (Ka) substitution rates were calculated using KaKs_Calculator v2.0 (https://sourceforge.net/projects/kakscalculator2/) [46] with the Yang-Nielsen (YN) algorithm. Only gene pairs with Ks values between 0.01 and 2.0 were retained to avoid saturation effects and ensure reliable divergence estimates. Ka/Ks ratios were interpreted as indicators of selective pressure: Ka/Ks < 1 indicates purifying selection, Ka/Ks = 1 indicates neutral evolution, and Ka/Ks > 1 indicates positive selection. Sequence composition and codon usage analysis Amino acid composition, nucleotide content (GC%, GC3%), and physicochemical properties (molecular weight, isoelectric point, instability index, aliphatic index, GRAVY score) were calculated using custom Python scripts with BioPython and the ProtParam tool from ExPASy (https://web.expasy.org/protparam/) [47]. Codon usage bias was assessed by calculating Effective Number of Codons (ENC), Codon Adaptation Index (CAI), and Relative Synonymous Codon Usage (RSCU) using the seqinr v4.2-16 package in R [48]. ENC values range from 20 (extreme bias, only one codon per amino acid) to 61 (no bias, equal usage of synonymous codons). CAI values range from 0 to 1, with higher values indicating greater optimization for translational efficiency. Neutrality plots (GC12 vs. GC3) and ENC-plot analysis (ENC vs. GC3) were used to assess the relative contributions of mutational pressure and natural selection to codon usage patterns [21, 49]. Multivariate statistical analyses Principal Component Analysis (PCA) was performed on amino acid composition and codon usage metrics using the FactoMineR v2.4 and factoextra v1.0.7 packages in R [50, 51]. Variables were centered and scaled prior to analysis. Non-linear dimensionality reduction was conducted using t-distributed Stochastic Neighbor Embedding (t-SNE) implemented in the Rtsne v0.16 package [52, 53] with perplexity = 30, iterations = 5,000, and θ = 0.5. Conservation-based clustering was performed by integrating Ka/Ks ratios, synteny conservation scores, and sequence identity metrics using hierarchical clustering with complete linkage. Silhouette coefficients and gap statistics were calculated to determine optimal cluster numbers. All statistical analyses and visualizations were conducted in R v4.2.1 with ggplot2 v3.4.0 [54] and ComplexHeatmap v2.12.1 [55]. Data visualization and statistical testing Graphical representations including box plots, violin plots, scatter plots, and heatmaps were generated using ggplot2 v3.4.0 and ggpubr v0.5.0 in R [51, 54]. Statistical significance was assessed using Student’s t-test for pairwise comparisons and one-way ANOVA followed by Tukey’s HSD post-hoc test for multiple group comparisons, with significance thresholds set at p < 0.05, p < 0.01, and p < 0.001. All analyses were performed using R v4.2.1 and Python v3.9 with Jupyter Notebook v6.4.12 (https://jupyter.org/) for reproducible computational workflows. Results Genome-wide identification and physicochemical characterization of WRKY transcription factors Using HMM and BLAST methods with strict domain checks, 547 WRKY transcription factor genes across seven plant species were detected. (Table 1). WRKY gene numbers varied considerably among species, ranging from 48 in Arabidopsis thaliana to 126 in Oryza sativa ssp.japonica. Among the rice subspecies, japonica harbored the most WRKYs (126 genes), followed by indica (109 genes) and glaberrima (51 genes), representing the first comprehensive characterization of this gene family in African rice. Maize contained 105 WRKYs, Brachypodium distachyon 56, wheat 52, and Arabidopsis 48 genes. Table 1Genome-wide census and phylogenetic group distribution of WRKY transcription factors in seven plant speciesSpeciesTotalGroup IGroup II†Group IIIGroup I (%)Group II† (%)Group III (%) Arabidopsis thaliana 4815211231.243.825.0 Brachypodium distachyon 561483425.014.360.7 Zea mays 1052667324.85.769.5 Oryza glaberrima 5155419.89.880.4Oryza sativa ssp. indica10916167714.714.770.6Oryza sativa ssp. japonica12624148819.011.169.8 Triticum aestivum 521473126.913.559.6TOTAL5471147735620.814.165.1† Group II members were further subdivided into subgroups IIa–IIe based on phylogenetic analysis (see Phylogenetic relationships and classification of WRKY proteins, Table 3, and Fig. 2) Table 1Genome-wide census and phylogenetic group distribution of WRKY transcription factors in seven plant speciesSpeciesTotalGroup IGroup II†Group IIIGroup I (%)Group II† (%)Group III (%) Arabidopsis thaliana 4815211231.243.825.0 Brachypodium distachyon 561483425.014.360.7 Zea mays 1052667324.85.769.5 Oryza glaberrima 5155419.89.880.4Oryza sativa ssp. indica10916167714.714.770.6Oryza sativa ssp. japonica12624148819.011.169.8 Triticum aestivum 521473126.913.559.6TOTAL5471147735620.814.165.1† Group II members were further subdivided into subgroups IIa–IIe based on phylogenetic analysis (see Phylogenetic relationships and classification of WRKY proteins, Table 3, and Fig. 2) Genome-wide census and phylogenetic group distribution of WRKY transcription factors in seven plant species † Group II members were further subdivided into subgroups IIa–IIe based on phylogenetic analysis (see Phylogenetic relationships and classification of WRKY proteins, Table 3, and Fig. 2) Comparison with previous studies showed general concordance but some notable differences. Our japonica count (126 genes) exceeded Ross et al. (2007) by 28.6% (98 genes), while indica (109 genes) was 6.9% higher than their 102 genes, likely reflecting improved genome assemblies and annotations. Chromosomal assignment of the 84 unique wheat WRKY loci (Supplementary Table S6) revealed significant subgenome retention bias. The A subgenome harbored the most genes (43 genes, 51.2%), followed by the B subgenome (28 genes, 33.3%), while the D subgenome was markedly underrepresented (13 genes, 15.5%). Chi-square goodness-of-fit testing against a null hypothesis of equal subgenome distribution (expected 1:1:1 ratio, 28 genes per subgenome) rejected the null hypothesis (χ² = 16.07, df = 2, p = 0.0003), confirming that the observed A > B > D retention pattern is statistically significant and not attributable to random variation. Our wheat count of 52 non-redundant genes was retained as the analytical unit for all cross-species comparisons in this study [12]. The Arabidopsis count (48 genes) was 33.3% lower than [5] (72 genes) due to exclusion of proteins lacking complete WRKY domain architecture. Phylogenetic classification revealed distinct group proportions across species (Table 1). Group III (single WRKY domain, C2HC zinc-finger) was the most abundant group across cereal species, with strong enrichment in all four cereals (Brachypodium 62.5%, Maize 67.6%, O. sativa japonica 69.8%, Triticum aestivum 59.6%; mean 64.9%), representing a 2.6-fold difference relative to Arabidopsis (25.0%) following correction of a zinc-finger classifier error identified in quality review (see Phylogenetic relationships and classification of WRKY proteins). Group I (two WRKY domains, C2H2 zinc-finger) comprised 114 genes (20.8%), with proportions ranging from 9.8% in O. glaberrima to 31.2% in Arabidopsis. Group II (single domain, C2H2 zinc-finger) contained 77 genes (14.1%), with corrected proportions ranging from 9.8% in O. glaberrima to 43.8% in Arabidopsis. While a formal rank-sum test is underpowered due to the single non-cereal comparator (n = 1 vs. n = 4), the magnitude of the Group III difference strongly supports its characterization as a cereal-enriched rather than universal angiosperm pattern. Notably, O. glaberrima exhibited the lowest Group I representation (9.8%) and highest Group III proportion (80.4%), suggesting lineage-specific gene loss or retention patterns during independent domestication. It must be acknowledged, however, that the absence of a wild ancestral comparator specifically Oryza barthii, the closest wild progenitor of O. glaberrima limits our ability to definitively attribute the observed Group I depletion (9.8%) and Group III enrichment (80.4%) to the domestication process per se, versus pre-existing lineage-specific divergence in the AA-genome clade. Published genomic analyses of O. barthii and O. glaberrima suggest that the majority of Group III expansion predates domestication, indicating that the pattern we observe likely reflects a broader Oryza AA-genome lineage signature rather than an exclusively domestication-driven phenomenon. Nevertheless, selective retention of stress-responsive Group III members under cultivation environments may have reinforced this pattern during domestication. Future comparative analyses explicitly including O. barthii and O. rufipogon would be required to formally decouple domestication effects from ancestral lineage-specific divergence. Group classification based on WRKY domain number and zinc-finger type: Group I (two domains, C2H2 zinc-finger), Group II (one domain, C2H2 zinc-finger), Group III (one domain, C2HC zinc-finger). WRKY proteins exhibited diverse physicochemical properties reflecting domain architecture differences (Table 2). Protein length ranged from 52 to 1482 amino acids (mean 366.8 ± 176.2 aa). Group I proteins were significantly longer (mean 468.4 ± 155.7 aa) than Group II (476.5 ± 374.2 aa) and Group III (321.6 ± 113.9 aa) due to dual WRKY domains (Kruskal-Wallis H = 79.99, p < 0.001). Molecular weight correlated with protein length, ranging from 6.5 to 165.4 kDa (mean 39.5 ± 19.2 kDa). Most WRKYs (43.9%) exhibited. Table 2Physicochemical properties of WRKY proteins in seven plant speciesSpecies n Protein Length (aa)MW (kDa)pIInstability IndexGRAVY% Nuclear (pI > 7) Arabidopsis thaliana 48341.2 ± 127.7 (105–687)38.0 ± 13.67.37 ± 1.6721.4 ± 0.8-0.818 ± 0.18256.2 Brachypodium distachyon 56380.3 ± 176.3 (131–1332)40.9 ± 19.57.08 ± 1.6321.6 ± 0.8-0.602 ± 0.17342.9 Zea mays 105383.3 ± 135.9 (99–737)41.1 ± 14.36.87 ± 1.3721.6 ± 0.7-0.590 ± 0.16234.3 Oryza glaberrima 51332.7 ± 123.8 (147–843)35.8 ± 13.17.11 ± 1.7421.5 ± 0.9-0.597 ± 0.17939.2O. sativa indica109382.9 ± 205.8 (142–1303)41.3 ± 22.86.84 ± 1.5921.4 ± 0.7-0.574 ± 0.19535.8O. sativa japonica126387.4 ± 178.6 (127–1002)41.6 ± 19.36.99 ± 1.6321.4 ± 0.8-0.576 ± 0.18744.0 Triticum aestivum 52288.8 ± 237.7 (52-1482)31.4 ± 25.98.21 ± 1.6821.5 ± 0.9-0.717 ± 0.24165.4 Table 2Physicochemical properties of WRKY proteins in seven plant speciesSpecies n Protein Length (aa)MW (kDa)pIInstability IndexGRAVY% Nuclear (pI > 7) Arabidopsis thaliana 48341.2 ± 127.7 (105–687)38.0 ± 13.67.37 ± 1.6721.4 ± 0.8-0.818 ± 0.18256.2 Brachypodium distachyon 56380.3 ± 176.3 (131–1332)40.9 ± 19.57.08 ± 1.6321.6 ± 0.8-0.602 ± 0.17342.9 Zea mays 105383.3 ± 135.9 (99–737)41.1 ± 14.36.87 ± 1.3721.6 ± 0.7-0.590 ± 0.16234.3 Oryza glaberrima 51332.7 ± 123.8 (147–843)35.8 ± 13.17.11 ± 1.7421.5 ± 0.9-0.597 ± 0.17939.2O. sativa indica109382.9 ± 205.8 (142–1303)41.3 ± 22.86.84 ± 1.5921.4 ± 0.7-0.574 ± 0.19535.8O. sativa japonica126387.4 ± 178.6 (127–1002)41.6 ± 19.36.99 ± 1.6321.4 ± 0.8-0.576 ± 0.18744.0 Triticum aestivum 52288.8 ± 237.7 (52-1482)31.4 ± 25.98.21 ± 1.6821.5 ± 0.9-0.717 ± 0.24165.4 Physicochemical properties of WRKY proteins in seven plant species basic isoelectric points (pI > 7.0, mean 7.18 ± 1.73), consistent with DNA-binding function. The instability index indicated that all WRKYs were classified as stable (mean 21.5 ± 0.8, all < 40), though this contradicts typical expectations for regulatory proteins requiring rapid turnover, suggesting the simplified calculation may underestimate in vivo instability. All proteins showed negative GRAVY values (mean − 0.619 ± 0.202), confirming their hydrophilic nature consistent with nuclear localization. No significant differences in physicochemical properties were observed between monocots and Arabidopsis, indicating conserved structural constraints across the monocot-dicot divide. Phylogenetic relationships and classification of WRKY proteins To elucidate evolutionary relationships among WRKY transcription factors, we constructed a maximum-likelihood phylogenetic tree using full-length protein sequences for all 547 proteins (Fig. 1). The optimal substitution model Jones-Taylor-Thornton (JTT) was selected by ModelFinder, with strong bootstrap support (≥ 70%) across major nodes, calculated using 1000 bootstrap replicates, indicating robust phylogenetic resolution. Fig. 1Maximum likelihood phylogenetic tree of 547 WRKY transcription factor proteins from seven plant species. The tree was constructed using the Jones-Taylor-Thornton (JTT) substitution model in MEGA X v12.0 with 1,000 bootstrap replicates. Bootstrap support values ≥ 70% are shown at major nodes. Tree visualization was performed in iTOL v6.5.7. Phylogenetic groups are color-coded: Group I (blue; two WRKY domains, C2H2 zinc-finger), Group II (green; one WRKY domain, C2H2 zinc-finger, subdivided into subgroups IIa–IIe), and Group III (red; one WRKY domain, C2HC zinc-finger). Species are indicated by leaf label prefixes: At (Arabidopsis thaliana), Bd (Brachypodium distachyon), Zm (Zea mays), Og (Oryza glaberrima), Osi (O. sativa indica), Osj (O. sativa japonica), Ta (Triticum aestivum). All figures regenerated at 600 DPI with minimum font size 11pt. Bootstrap support values ≥ 70% are reported in text (Phylogenetic relationships and classification of WRKY proteins). Species prefixes: At, Bd, Zm, Og, Osi, Osj, Ta Maximum likelihood phylogenetic tree of 547 WRKY transcription factor proteins from seven plant species. The tree was constructed using the Jones-Taylor-Thornton (JTT) substitution model in MEGA X v12.0 with 1,000 bootstrap replicates. Bootstrap support values ≥ 70% are shown at major nodes. Tree visualization was performed in iTOL v6.5.7. Phylogenetic groups are color-coded: Group I (blue; two WRKY domains, C2H2 zinc-finger), Group II (green; one WRKY domain, C2H2 zinc-finger, subdivided into subgroups IIa–IIe), and Group III (red; one WRKY domain, C2HC zinc-finger). Species are indicated by leaf label prefixes: At (Arabidopsis thaliana), Bd (Brachypodium distachyon), Zm (Zea mays), Og (Oryza glaberrima), Osi (O. sativa indica), Osj (O. sativa japonica), Ta (Triticum aestivum). All figures regenerated at 600 DPI with minimum font size 11pt. Bootstrap support values ≥ 70% are reported in text (Phylogenetic relationships and classification of WRKY proteins). Species prefixes: At, Bd, Zm, Og, Osi, Osj, Ta The tree resolved three major groups consistent with established WRKY classification (Eulgem et al., 2000; Zhang and Wang, 2005) (Fig. 1; Table 3). Group I (114 proteins, 20.8%) formed a well-supported clade (bootstrap 90%) characterized by two WRKY domains with C2H2 zinc-finger motifs. Two subclades were evident: a monocot-enriched lineage and a mixed monocot-dicot lineage, suggesting ancient duplications predating the monocot-dicot split. Group III (356 proteins, 65.1% overall) was the largest group (bootstrap 92%), distinguished by single WRKY domains with C2HC zinc-finger motifs. Group III exhibited pronounced species-specific enrichment in cereal species (range 59.6–69.8%) versus Arabidopsis (25.0% after correction of classification error; see Materials and Methods), with extensive expansions in rice japonica (88 genes), indica (77 genes), and maize (73 genes) driven primarily by tandem duplication events (Fisher’s exact test versus segmental duplications, p = 0.007). Table 3Species-specific distribution of WRKY genes across phylogenetic subgroupsSpeciesTotalGroup IGroup IIaGroup IIbGroup IIcGroup IIdGroup IIeGroup III Arabidopsis thaliana 481530125012 Brachypodium distachyon 56142061034 Zea mays 105262030173 Oryza glaberrima 5150021241O. sativa indica109163452277O. sativa japonica126242421588 Triticum aestivum 52141031231TOTAL547114138331112356 Table 3Species-specific distribution of WRKY genes across phylogenetic subgroupsSpeciesTotalGroup IGroup IIaGroup IIbGroup IIcGroup IIdGroup IIeGroup III Arabidopsis thaliana 481530125012 Brachypodium distachyon 56142061034 Zea mays 105262030173 Oryza glaberrima 5150021241O. sativa indica109163452277O. sativa japonica126242421588 Triticum aestivum 52141031231TOTAL547114138331112356 Species-specific distribution of WRKY genes across phylogenetic subgroups Group II (77 proteins, 14.1%) was subdivided into five subgroups based on phylogenetic clustering (Table 3). Subgroup IIa contained 10 genes (14.3%), IIb had 8 genes (11.4%) enriched in rice (indica 4, japonica 4), IIc was the largest with 15 genes (21.4%) showing balanced species distribution, IId was smallest with 5 genes (7.1%), and IIe comprised 12 genes (17.1%) predominantly from rice. The remaining 20 genes (28.6%) represent Arabidopsis and Brachypodium Group II members corrected from the initial zinc-finger misclassification, distributed primarily within IIc and IId subgroups. Notably, O. glaberrima lacked Group IIb members, suggesting lineage-specific losses during independent domestication. Species-specific patterns revealed distinct evolutionary trajectories. O. glaberrima WRKYs formed sister clades to indica-japonica pairs, consistent with independent domestication. Wheat WRKYs distributed throughout the tree without species-specific clades, reflecting ancient duplications shared with Triticeae. Notably, this distribution was non-uniform across wheat subgenomes: A-subgenome WRKY genes (n = 43) were broadly distributed across all major clades, B-subgenome genes (n = 28) were enriched in Group I and Group IIc clades, and D-subgenome genes (n = 13) were predominantly confined to Group III clades, suggesting differential subgenome contributions to WRKY functional diversity following allopolyploidization. Maize showed lineage-specific Group III expansions from whole-genome duplication 5–12 Mya. Monocot-dicot divergence was most pronounced in Group III, while Groups I and II showed more intermixed distributions, suggesting differential evolutionary rates among WRKY groups. Gene structure and conserved motif architecture To investigate structural diversity and evolutionary conservation within the WRKY gene family, exon-intron organization and conserved protein motifs were analyzed (Fig. 2; Table 4). Exon numbers ranged from 0 to 17 (mean 2.8 ± 3.3), with significant variation among phylogenetic groups (Table 4). Group I proteins exhibited more complex gene structures (mean 4.1 ± 4.5 exons) compared to Group II (2.0 ± 3.2 exons) and Group III (2.6 ± 2.9 exons), consistent with their dual-domain architecture (Kruskal-Wallis H = 6.85, p = 0.033). Intronless genes were rare, comprising only 2 genes (0.7%), both in Group III, suggesting recent retrotransposition events. Gene length varied from 435 to 19,087 bp, with Group I genes significantly longer (mean 3,410 bp) than Group III (mean 2,614 bp), reflecting additional coding sequences for the second WRKY domain. CDS-to-gene length ratios averaged 0.497 across all genes, with Group I showing higher compactness (0.606) compared to Group III (0.489), indicating greater intron content relative to gene size in Group III proteins. Fig. 2Conserved protein motifs identified by MEME Suite v5.4.1 across 547 WRKY proteins from seven plant species. Letter height at each position represents amino acid conservation. Motif 1: WRKY domain containing the conserved WRKYGQK heptapeptide (universal, all groups). Motif 2: C2H2-type zinc-finger (C-X₄₋₅-C-X₂₂₋₂₃-H-X-H), characteristic of Group I and Group II. Motif 3: C2HC-type zinc-finger (C-X₇-C-X₂₃-H-X-C), specific to Group III. Motifs 4–10: Group- and lineage-specific auxiliary domains. Species color key: Arabidopsis (green), Brachypodium (teal), Maize (yellow), O. glaberrima (orange), O. sativa indica (light blue), O. sativa japonica (dark blue), T. aestivum (red). All motifs: E-value < 1 × 10⁻¹⁰. Figure regenerated at 600 DPI with minimum font size 12pt Conserved protein motifs identified by MEME Suite v5.4.1 across 547 WRKY proteins from seven plant species. Letter height at each position represents amino acid conservation. Motif 1: WRKY domain containing the conserved WRKYGQK heptapeptide (universal, all groups). Motif 2: C2H2-type zinc-finger (C-X₄₋₅-C-X₂₂₋₂₃-H-X-H), characteristic of Group I and Group II. Motif 3: C2HC-type zinc-finger (C-X₇-C-X₂₃-H-X-C), specific to Group III. Motifs 4–10: Group- and lineage-specific auxiliary domains. Species color key: Arabidopsis (green), Brachypodium (teal), Maize (yellow), O. glaberrima (orange), O. sativa indica (light blue), O. sativa japonica (dark blue), T. aestivum (red). All motifs: E-value < 1 × 10⁻¹⁰. Figure regenerated at 600 DPI with minimum font size 12pt Table 4Gene structure characteristics by phylogenetic groupGroup n Exon Number (Mean ± SD)RangeIntron less Genes (%)Gene Length (bp)CDS/Gene RatioGroup I554.1 ± 4.50–170 (0.0%)3,410 ± 2,0380.606Group II322.0 ± 3.20–140 (0.0%)3,668 ± 3,8260.361Group III2132.6 ± 2.90–142 (0.9%)2,614 ± 2,0170.489Overall3002.8 ± 3.30–172 (0.7%)2,872 ± 2,3070.497Statistics calculated from 300 WRKY genes with complete structural annotation. CDS/Gene ratio indicates gene compactness (higher = fewer/shorter introns). Kruskal-Wallis test showed significant variation among groups (H = 6.85, p = 0.033) Table 4Gene structure characteristics by phylogenetic groupGroup n Exon Number (Mean ± SD)RangeIntron less Genes (%)Gene Length (bp)CDS/Gene RatioGroup I554.1 ± 4.50–170 (0.0%)3,410 ± 2,0380.606Group II322.0 ± 3.20–140 (0.0%)3,668 ± 3,8260.361Group III2132.6 ± 2.90–142 (0.9%)2,614 ± 2,0170.489Overall3002.8 ± 3.30–172 (0.7%)2,872 ± 2,3070.497Statistics calculated from 300 WRKY genes with complete structural annotation. CDS/Gene ratio indicates gene compactness (higher = fewer/shorter introns). Kruskal-Wallis test showed significant variation among groups (H = 6.85, p = 0.033) Gene structure characteristics by phylogenetic group Statistics calculated from 300 WRKY genes with complete structural annotation. CDS/Gene ratio indicates gene compactness (higher = fewer/shorter introns). Kruskal-Wallis test showed significant variation among groups (H = 6.85, p = 0.033) MEME analysis identified 10 highly conserved protein motifs across all 547 sequences (Fig. 2; Table 5). Motif 1, corresponding to the WRKY domain containing the WRKYGQK signature, was universally present with high conservation. Motif 2 represented the C2H2 zinc-finger motif (C-X₄₋₅-C-X₂₂₋₂₃-H-X-H) characteristic of Group I and Group II proteins, while Motif 3 corresponded to the C2HC zinc-finger (C-X₇-C-X₂₃-H-X-C) specific to Group III. Seven additional conserved motifs (Motifs 4–10) exhibited group-specific distribution patterns (Table 5), with motif architecture closely reflecting phylogenetic relationships. Proteins clustering in the same phylogenetic clade displayed nearly identical motif compositions, confirming congruence between sequence conservation, gene structure, and evolutionary history. Several motifs showed lineage-specific distributions, suggesting potential functional divergence following species-specific duplication events. Table 5Conserved protein motifs identified in WRKY transcription factorsMotifWidth (aa)SitesE-valueGroup DistributionPredicted Function160547< 1 × 10⁻¹⁰⁰All (100%)WRKY domain (DNA binding)228164< 1 × 10⁻⁵⁰Groups I & IIC2H2 zinc-finger329547< 1 × 10⁻⁵⁰Group III enrichedC2HC zinc-finger435VariableSignificantGroup-specificAuxiliary domain525VariableSignificantGroup-specificAuxiliary domain630VariableSignificantMonocot-enrichedLineage-specific725VariableSignificantGroup II enrichedSubfamily-specific835VariableSignificantGroup I enrichedDual-domain associated930VariableSignificantVariableUnknown function1025VariableSignificantVariableUnknown functionE-values represent statistical significance of motif discovery. Group distribution determined by motif presence in > 70% of group members Table 5Conserved protein motifs identified in WRKY transcription factorsMotifWidth (aa)SitesE-valueGroup DistributionPredicted Function160547< 1 × 10⁻¹⁰⁰All (100%)WRKY domain (DNA binding)228164< 1 × 10⁻⁵⁰Groups I & IIC2H2 zinc-finger329547< 1 × 10⁻⁵⁰Group III enrichedC2HC zinc-finger435VariableSignificantGroup-specificAuxiliary domain525VariableSignificantGroup-specificAuxiliary domain630VariableSignificantMonocot-enrichedLineage-specific725VariableSignificantGroup II enrichedSubfamily-specific835VariableSignificantGroup I enrichedDual-domain associated930VariableSignificantVariableUnknown function1025VariableSignificantVariableUnknown functionE-values represent statistical significance of motif discovery. Group distribution determined by motif presence in > 70% of group members Conserved protein motifs identified in WRKY transcription factors E-values represent statistical significance of motif discovery. Group distribution determined by motif presence in > 70% of group members Chromosomal distribution and comparative mapping WRKY genes showed widespread distribution across all chromosomes in the analyzed species, with distinct patterns of gene density and clustering (Fig. 3). In Arabidopsis, 48 WRKY genes were distributed across all five chromosomes, with the highest concentration on chromosome 5 (12 genes), followed by chromosomes 2 and 4 (10 genes each). Rice showed the most extensive distribution, with 126 WRKY genes in O. sativa ssp. Japonica mapped across all 12 chromosomes, with chromosome 1 harboring the highest number of genes (27 genes), representing approximately 21% of the total rice WRKY repertoire. Similar distribution patterns were observed in O. sativa and O. indica (109 genes across 12 chromosomes) and O. glaberrima (51 genes across 12 chromosomes). Fig. 3Comparative chromosomal distribution of WRKY genes between rice (Oryza sativa japonica, center) and (A) Arabidopsis thaliana, (B) Brachypodium distachyon, (C) Zea mays, and (D) Triticum aestivum. Chromosome bars are drawn proportional to physical length. Colored connecting lines indicate syntenic WRKY gene pairs within collinear genomic blocks identified by MCScanX. Chromosome identifiers and WRKY gene counts per chromosome are labeled on each bar. Figure regenerated at 600 DPI with chromosome labels ≥ 12pt and gene position markers ≥ 10pt Comparative chromosomal distribution of WRKY genes between rice (Oryza sativa japonica, center) and (A) Arabidopsis thaliana, (B) Brachypodium distachyon, (C) Zea mays, and (D) Triticum aestivum. Chromosome bars are drawn proportional to physical length. Colored connecting lines indicate syntenic WRKY gene pairs within collinear genomic blocks identified by MCScanX. Chromosome identifiers and WRKY gene counts per chromosome are labeled on each bar. Figure regenerated at 600 DPI with chromosome labels ≥ 12pt and gene position markers ≥ 10pt Comparative chromosomal mapping between rice and four model/crop species revealed both conserved and species-specific chromosomal organization patterns (Fig. 3). The rice-Arabidopsis comparison showed multiple syntenic. relationships, with rice chromosomes 1, 2, 3, and 11 displaying the strongest conservation with Arabidopsis chromosomes 2, 4, and 5. Rice-Brachypodium mapping demonstrated high collinearity, particularly between rice chromosomes 1, 5, and 11 and Brachypodium chromosomes 1, 2, and 4, reflecting the close evolutionary relationship between these grass species. The rice-maize comparison revealed extensive chromosomal rearrangements, with rice chromosome 1 showing syntenic blocks distributed across maize chromosomes 1, 3, 5, and 9, consistent with the whole-genome duplication events in maize lineage. Rice-wheat mapping showed complex patterns with rice chromosomes corresponding to multiple wheat chromosomal regions, particularly on wheat chromosomes 2, 5, 6, and 7, indicating both conserved synteny and lineage-specific duplications. Gene clustering analysis revealed non-random distribution patterns, with all chromosomes containing multiple WRKY genes in rice and Arabidopsis, suggesting potential tandem duplication events. The highest gene density was observed on rice chromosome 1 (Chr1) and Arabidopsis chromosome 5, which may represent chromosomal hotspots for WRKY gene family expansion through local duplication mechanisms. Gene duplication events Gene duplication analysis revealed that both tandem and segmental duplication events have contributed significantly to WRKY gene family expansion across the analyzed species. In rice (O. sativa japonica), 23 tandem duplication events were identified, primarily involving genes from Group II and Group III subfamilies located on chromosomes 1, 5, and 11. These tandem duplicates were characterized by close physical proximity (< 200 kb) and high sequence similarity (> 85%), suggesting recent duplication events. Additionally, three segmental duplication events were detected, involving gene pairs distributed across different chromosomes, indicating whole-genome or large-scale segmental duplication events during rice evolution. Arabidopsis showed eight tandem duplication events and two segmental duplication events, with tandem duplicates predominantly found on chromosomes 2, 4, and 5. The lower frequency of duplication events in Arabidopsis compared to rice is consistent with the smaller genome size and absence of recent whole-genome duplication in the Arabidopsis lineage. In O. sativa and O. indica, 20 tandem and three segmental duplication events were identified, showing patterns similar to japonica rice and supporting the shared evolutionary history of cultivated rice subspecies. Quantitative analysis of 59 curated duplication events (Supplementary Table S2) revealed complete mechanistic segregation by WRKY group classification. All 8 segmental duplication events (13.6%) involved exclusively Group I gene pairs, while all 51 tandem duplication events (86.4%) involved Group II (n = 25, 49.0%) and Group III (n = 26, 51.0%) exclusively — with zero tandem duplications detected in Group I. Fisher’s exact test confirmed significant over-representation of Group III genes in tandem duplications relative to segmental duplications (p = 0.007). Ka/Ks analysis of Group III tandem duplicate pairs revealed a mean ratio of 0.141 (± 0.009 SD), indicating strong purifying selection and functional conservation following duplication, consistent with retention for adaptive dosage amplification rather than neofunctionalization. In hexaploid wheat specifically, the significant D-subgenome underrepresentation (13 genes, 15.5% vs. expected 33.3%) suggests preferential gene loss from the most recently acquired subgenome following allopolyploidization. The D-subgenome was contributed by Aegilops tauschii approximately 8,500 years ago — considerably more recent than the A-B tetraploidization event (~ 0.5 million years ago) — and this shorter diploidization period likely accounts for the incomplete D-subgenome homeolog retention observed here (χ² = 16.07, p = 0.0003). Synteny analysis Synteny analysis using circular visualization revealed extensive collinearity of WRKY genes between rice and four evolutionarily diverse plant species, providing insights into the evolutionary conservation and divergence of this gene family (Fig. 4). The rice-Arabidopsis synteny analysis showed 142 syntenic gene pairs despite the distant phylogenetic relationship between monocots and dicots, indicating strong functional constraints maintaining WRKY gene positions across ~ 150 million years of evolution. The majority of syntenic blocks involved Group I and Group IIc genes, suggesting these subfamilies harbor core developmental and stress-response functions conserved across flowering plants. Fig. 4Circular synteny maps showing collinearity of WRKY genes in rice (Oryza sativa) against (A) Zea mays, (B) Brachypodium distachyon, (C) Arabidopsis thaliana, and (D) Triticum aestivum. Colored ribbons connect syntenic WRKY gene pairs within collinear genomic blocks identified by MCScanX. Ribbon colors correspond to rice chromosomes (Chr1–Chr12 as indicated in the color key). Chromosome identifiers are labeled on each arc; numbers inside arcs indicate WRKY gene count per chromosome. Total syntenic pairs: rice– Brachypodium (n = 218), rice–Maize (n = 186), rice–Wheat (n = 164), rice–Arabidopsis (n = 142) Circular synteny maps showing collinearity of WRKY genes in rice (Oryza sativa) against (A) Zea mays, (B) Brachypodium distachyon, (C) Arabidopsis thaliana, and (D) Triticum aestivum. Colored ribbons connect syntenic WRKY gene pairs within collinear genomic blocks identified by MCScanX. Ribbon colors correspond to rice chromosomes (Chr1–Chr12 as indicated in the color key). Chromosome identifiers are labeled on each arc; numbers inside arcs indicate WRKY gene count per chromosome. Total syntenic pairs: rice– Brachypodium (n = 218), rice–Maize (n = 186), rice–Wheat (n = 164), rice–Arabidopsis (n = 142) Rice-Brachypodium synteny displayed the highest degree of collinearity, with 218 syntenic gene pairs and extensive conservation of chromosomal blocks, particularly between rice chromosomes 1, 3, 5, 8, and 11 and Brachypodium chromosomes 1, 2, 3, and 4. This high synteny level reflects the close evolutionary relationship within Poaceae family and suggests that WRKY gene organization has remained relatively stable during grass evolution. The rice-maize comparison revealed 186 syntenic pairs with more complex patterns, showing evidence of chromosomal rearrangements and duplications consistent with maize’s paleopolyploid genome. Notably, single rice chromosomal regions corresponded to multiple maize chromosomes, particularly visible for rice chromosome 1 syntenic blocks distributed across maize chromosomes 1, 5, 9, and 10, confirming the impact of ancient whole-genome duplication events in maize. The rice-wheat synteny analysis identified 164 syntenic gene pairs across the complex hexaploid wheat genome, with rice chromosomes showing collinearity primarily with wheat chromosomes 2 A, 5 A, 6B, and 7D. The distribution patterns revealed both conserved syntenic blocks and lineage-specific arrangements, reflecting wheat’s polyploid nature and the significant differential retention of WRKY genes among wheat subgenomes (A: 43, B: 28, D: 13; χ² = 16.07, p = 0.0003). The strongest rice-wheat collinearity was detected with A-subgenome chromosomes (2 A, 5 A, 7 A), consistent with the closer phylogenetic relationship between the A-subgenome donor Triticum urartu and rice, while D-subgenome chromosomes showed both fewer syntenic pairs and higher rates of chromosomal rearrangement relative to the A and B subgenomes. Across all four comparisons, Group I WRKY genes showed the highest synteny conservation (> 75% of genes with identifiable syntenic pairs), followed by Group IIc (68%) and Group IIe (62%), while Group III genes showed lower conservation (45%), suggesting more dynamic evolution and species-specific functional diversification in this subgroup. Sequence composition and biochemical properties analysis To investigate the biochemical diversity and compositional signatures underlying WRKY transcription factor functional specialization, comprehensive amino acid and nucleotide composition analyses were performed across all identified genes. Amino acid composition analysis revealed distinct patterns among phylogenetic groups, with significant variations in hydrophobic, charged, and polar residues reflecting functional and structural constraints (Fig. 5A). Group I WRKY proteins showed relatively balanced amino acid compositions, with moderate enrichment of leucine (8.2–9.1%), serine (7.8–8.5%), and glycine (6.9–7.4%), consistent with their roles in constitutive DNA-binding and regulatory functions. The presence of conserved cysteine residues (2.8–3.2%) across all groups confirmed the structural importance of zinc finger coordination in the WRKY domain. Group II proteins displayed the highest compositional diversity, with Group IIa and IIb showing enrichment of acidic residues (aspartate and glutamate combined: 11.2–12.8%), potentially related to their stress-responsive regulatory mechanisms and protein-protein interaction capabilities. Fig. 5Sequence composition and biochemical properties of WRKY genes. A Heatmap showing amino acid composition percentages across phylogenetic groups and species. Color intensity indicates relative abundance of each amino acid. B Principal component analysis (PCA) of amino acid composition, with points colored by phylogenetic group and species. C PCA biplot of trinucleotide composition showing species-specific and group-specific clustering patterns. Ellipses represent 95% confidence intervals for each group Sequence composition and biochemical properties of WRKY genes. A Heatmap showing amino acid composition percentages across phylogenetic groups and species. Color intensity indicates relative abundance of each amino acid. B Principal component analysis (PCA) of amino acid composition, with points colored by phylogenetic group and species. C PCA biplot of trinucleotide composition showing species-specific and group-specific clustering patterns. Ellipses represent 95% confidence intervals for each group Group III WRKY proteins exhibited distinct compositional signatures characterized by elevated proline content (6.8–7.9%) compared to Groups I and II (5.2–6.1%), suggesting increased structural flexibility and potential for intrinsically disordered regions that may facilitate developmental regulation and tissue-specific functions. Hydrophobic amino acids (alanine, valine, leucine, isoleucine) collectively represented 32–38% of total residues across all groups, with Group III showing slightly higher proportions (36–38%), consistent with their more specialized expression patterns and potential membrane-associated or protein-protein interaction functions. Arginine and lysine combined accounted for 10.5–12.3% of residues across species, with rice and wheat showing marginally higher basic amino acid content (11.8–12.3%) compared to Arabidopsis and Brachypodium (10.5–11.2%), potentially reflecting species-specific regulatory adaptations. Principal component analysis of amino acid composition revealed clear clustering patterns corresponding to phylogenetic groups and species identity (Fig. 5B). The first two principal components explained 58.3% of total variance, with PC1 (37.2% variance) primarily driven by hydrophobic residue composition and PC2 (21.1% variance) associated with charged residue distribution. Group I genes clustered tightly in the center of the PCA space, indicating conserved compositional profiles consistent with their ancient origin and fundamental regulatory functions. Group II genes showed broader dispersion across PC1, reflecting their subfunctional diversification and expanded regulatory repertoire. Group III genes formed distinct clusters separated primarily along PC2, driven by elevated proline and reduced acidic amino acid content. Species-specific clustering was evident within phylogenetic groups, with rice and Brachypodium (both grasses) showing closer compositional similarity than either showed to Arabidopsis, supporting lineage-specific biochemical adaptations. Nucleotide composition analysis revealed significant codon usage bias and base composition preferences that varied among species and phylogenetic groups (Fig. 5C). Overall GC content across coding sequences ranged from 48.2% in wheat to 54.7% in Arabidopsis, consistent with genome-wide GC content differences among these species. Dinucleotide composition analysis showed enrichment of CG and GC dinucleotides in Group I genes (observed/expected ratio: 1.15–1.28) compared to Group III genes (0.89–1.02), suggesting differential selection pressures on coding sequence composition potentially related to expression regulation or mRNA stability. Trinucleotide composition PCA revealed species-specific codon usage patterns, with monocots (rice, wheat, maize, Brachypodium) clustering separately from dicots (Arabidopsis), and further sub-clustering within monocots according to phylogenetic relatedness. Codon adaptation index (CAI) analysis indicated that Group I genes showed higher CAI values (mean: 0.78–0.82) compared to Group III genes (mean: 0.68–0.73), consistent with the higher and more constitutive expression levels of Group I members and suggesting optimization for translational efficiency in highly expressed genes. Feature correlation and evolutionary conservation analysis To elucidate the interdependencies among genomic features and identify evolutionarily constrained elements within WRKY genes, comprehensive correlation and conservation analyses were performed across all species and phylogenetic groups. Correlation analysis among structural and compositional features revealed complex networks of relationships that varied among phylogenetic groups (Fig. 6A). Gene length showed strong positive correlation with exon number (Pearson r = 0.78, p < 0.001) and intron number (r = 0.82, p < 0.001) across all species, indicating that longer genes achieved their extended length primarily through increased exon-intron structure complexity rather than individual exon elongation. Protein length correlated moderately with gene length (r = 0.64, p < 0.001) but showed weaker correlation with exon number (r = 0.48, p < 0.001), suggesting that coding sequence expansion involves both increased exon number and modulation of individual exon lengths. Fig. 6Feature correlation and evolutionary conservation analysis. A Correlation network showing relationships among genomic and biochemical features. Node size indicates feature variance; edge thickness represents correlation strength. Chord diagram shows inter-feature relationships grouped by phylogenetic classification. B Conservation analysis showing Ka/Ks ratios across gene regions and species comparisons. Box plots compare conservation levels among phylogenetic groups. Scatter plots show relationship between conservation scores and functional constraints Feature correlation and evolutionary conservation analysis. A Correlation network showing relationships among genomic and biochemical features. Node size indicates feature variance; edge thickness represents correlation strength. Chord diagram shows inter-feature relationships grouped by phylogenetic classification. B Conservation analysis showing Ka/Ks ratios across gene regions and species comparisons. Box plots compare conservation levels among phylogenetic groups. Scatter plots show relationship between conservation scores and functional constraints GC content displayed distinctive correlation patterns that differed significantly among phylogenetic groups. In Group I genes, GC content showed positive correlation with gene expression levels (r = 0.52–0.68 across species, p < 0.001), consistent with the hypothesis that higher GC content facilitates stable mRNA structures and efficient translation in constitutively expressed genes. Group II genes exhibited weaker GC-expression correlations (r = 0.28–0.42, p < 0.01), reflecting their dynamic, stress-responsive expression patterns. Group III genes showed minimal GC-expression correlation (r = 0.12–0.25, not significant in most species), consistent with their specialized, tissue-specific regulatory mechanisms that may rely less on GC-mediated expression optimization. Codon adaptation index (CAI) correlated strongly with predicted expression breadth (r = 0.71, p < 0.001) but only moderately with peak expression levels (r = 0.44, p < 0.01), suggesting that translational optimization is more important for broadly expressed genes than for genes with high but restricted expression. Evolutionary conservation analysis revealed distinct patterns of selection pressure across gene regions and phylogenetic groups (Fig. 6B). The WRKY domain exhibited the highest conservation scores (mean Ka/Ks ratio: 0.08–0.12 across comparisons), with near-absolute conservation of the WRKYGQK heptapeptide (100% identity in 94% of genes) and zinc finger motifs (C-X4-5-C-X22-23-H-X-H pattern conserved in 98% of genes). Conservation was strongest in Group I genes (mean domain Ka/Ks: 0.08) compared to Group II (0.11) and Group III (0.14), indicating relaxed constraints on Group III WRKY domains potentially associated with their functional diversification. Regions flanking the WRKY domain showed significantly lower conservation (Ka/Ks: 0.42–0.78), with Group II genes exhibiting the highest variability in non-domain regions, consistent with their expanded regulatory functions and diverse protein interaction capabilities. Syntenic gene pairs between rice and other species showed conservation levels proportional to phylogenetic distance and functional constraints. Rice-Brachypodium orthologous pairs displayed the highest nucleotide identity (mean: 87.3% ± 4.2%) and lowest Ka/Ks ratios (mean: 0.18 ± 0.08), reflecting recent common ancestry and strong purifying selection. Rice-Arabidopsis comparisons showed lower identity (mean: 68.4% ± 8.7%) and elevated Ka/Ks ratios (mean: 0.34 ± 0.12), consistent with ~ 150 million years of divergence and lineage-specific adaptations. Within each species pair, Group I genes showed significantly higher conservation (mean identity > 75% in all comparisons) than Group III genes (mean identity 55–68%), supporting the hypothesis that fundamental regulatory functions impose stronger evolutionary constraints than specialized developmental roles. Promoter regions (1.5 kb upstream) showed moderate conservation in syntenic pairs (mean identity: 42–58%), with several conserved cis-regulatory motifs identified including W-boxes (TTGAC[C/T]), ABRE elements, and DRE/CRT motifs, suggesting conservation of upstream regulatory mechanisms despite sequence divergence. Conserved non-coding sequences flanking WRKY genes were identified in 38% of syntenic pairs, potentially representing functionally important regulatory elements maintained across species despite low overall sequence similarity. Multivariate analysis and clustering patterns To complement linear dimensionality reduction approaches and provide comprehensive validation of compositional relationships, non-linear clustering analysis and additional conservation metrics were performed across all WRKY genes. t-Distributed Stochastic Neighbor Embedding (t-SNE) analysis of amino acid composition revealed distinct non-linear clustering patterns that largely corroborated but also extended insights from PCA analysis (Fig. 7A). The t-SNE projection separated WRKY genes into discrete clusters corresponding to phylogenetic groups and species identity, with Group I genes forming a compact central cluster indicating conserved amino acid composition profiles consistent with their ancient origin and fundamental regulatory functions. Group II genes displayed broader dispersion across t-SNE space, with subgroups IIa, IIb, IIc, IId, and IIe forming partially overlapping but distinguishable subclusters, reflecting their functional diversification while maintaining core compositional features. Group III genes occupied peripheral regions of t-SNE space, separated from Groups I and II along both dimensions, emphasizing their distinct compositional signatures associated with specialized developmental and tissue-specific regulatory roles. Fig. 7Multivariate analysis and clustering patterns of WRKY genes. A t-SNE projection based on amino acid composition showing non-linear clustering by phylogenetic group and species identity. Each point represents one gene; colors indicate species identity. B t-SNE projection based on trinucleotide composition with bridge genes highlighted in gold diamonds. Red circles identify 22 bridge genes occupying intermediate compositional positions between monocot and dicot cluster centroids, spanning Maize (n = 7), O. sativa japonica (n = 6), Brachypodium (n = 5), and O. glaberrima (n = 4). Gene identifiers are labeled for each bridge gene. C Box plots comparing conservation scores across phylogenetic groups (Kruskal-Wallis H = 187.3, p < 0.001). D Scatter plot showing the relationship between gene length and GC content with linear regression lines for each species (Pearson r = 0.43–0.58, p < 0.001). E Wheat WRKY gene retention bias across A, B, and D subgenomes (A = 43, 51.2%; B = 28, 33.3%; D = 13, 15.5%; χ²=16.07, df = 2, p = 0.0003). Dashed line indicates expected equal distribution (28 genes per subgenome) under a null hypothesis of uniform retention. F Effective number of codons (ENC) versus GC3 content; the solid curve represents the expected relationship under neutral mutation-drift equilibrium, with observed values colored by phylogenetic group. All panels generated at 600 DPI with minimum font size 11pt Multivariate analysis and clustering patterns of WRKY genes. A t-SNE projection based on amino acid composition showing non-linear clustering by phylogenetic group and species identity. Each point represents one gene; colors indicate species identity. B t-SNE projection based on trinucleotide composition with bridge genes highlighted in gold diamonds. Red circles identify 22 bridge genes occupying intermediate compositional positions between monocot and dicot cluster centroids, spanning Maize (n = 7), O. sativa japonica (n = 6), Brachypodium (n = 5), and O. glaberrima (n = 4). Gene identifiers are labeled for each bridge gene. C Box plots comparing conservation scores across phylogenetic groups (Kruskal-Wallis H = 187.3, p < 0.001). D Scatter plot showing the relationship between gene length and GC content with linear regression lines for each species (Pearson r = 0.43–0.58, p < 0.001). E Wheat WRKY gene retention bias across A, B, and D subgenomes (A = 43, 51.2%; B = 28, 33.3%; D = 13, 15.5%; χ²=16.07, df = 2, p = 0.0003). Dashed line indicates expected equal distribution (28 genes per subgenome) under a null hypothesis of uniform retention. F Effective number of codons (ENC) versus GC3 content; the solid curve represents the expected relationship under neutral mutation-drift equilibrium, with observed values colored by phylogenetic group. All panels generated at 600 DPI with minimum font size 11pt t-SNE analysis of trinucleotide composition patterns further validated species-specific and group-specific clustering observed in linear analyses (Fig. 7B). Monocot species (rice, wheat, maize, Brachypodium) formed cohesive clusters separated from the dicot species Arabidopsis along the primary t-SNE dimension, consistent with over 150 million years of evolutionary divergence and lineage-specific genomic adaptations. Within monocot clusters, grass species showed fine-scale structure with rice and Brachypodium exhibiting closer proximity than either showed to wheat or maize, reflecting phylogenetic relationships within Poaceae. Notably, t-SNE revealed 22 “bridge” genes occupying intermediate positions in compositional space between monocot and dicot cluster centroids, defined as genes within radius 0.20 of the centroid midpoint (Fig. 7B; Supplementary Table S5). These bridge genes were distributed across four cereal lineages — Maize (n = 7), O. sativa japonica (n = 6), Brachypodium (n = 5), and O. glaberrima (n = 4) — confirming that this pattern is not a species-specific artifact but a phylogenetically reproducible compositional signature. Structural analysis revealed two bridge gene classes: Group I bridges (n = 3; mean 5.5 exons, mean length 2,222 bp) and Group III bridges (n = 18; mean 4.2 exons, mean length 2,166 bp). The Group III bridge cohort exhibits exon counts intermediate between canonical Group I genes (mean 4.1 exons) and typical Group III genes (mean 2.6 exons), consistent with a stepwise structural simplification model along the evolutionary trajectory from ancestral dual-domain to derived single-domain configurations (Supplementary Table S5). Conservation score distributions across phylogenetic groups showed significant differences (Kruskal-Wallis H = 187.3, p < 0.001), with Group I genes exhibiting significantly higher mean conservation scores (0.847 ± 0.062) compared to Group II (0.782 ± 0.089, p < 0.001) and Group III (0.721 ± 0.103, p < 0.001) (Fig. 7C). Within Group II, subgroups IIa and IIb showed higher conservation (mean scores: 0.798 and 0.791) than subgroups IIc, IId, and IIe (mean scores: 0.768, 0.771, and 0.765), suggesting differential evolutionary constraints among Group II lineages. Scatter plot analysis revealed significant positive correlation between gene length and GC content across all species (Pearson r = 0.43–0.58, p < 0.001), with longer genes consistently showing elevated GC content potentially reflecting selection for stable mRNA secondary structures or mutational biases in longer genomic regions (Fig. 7D). This relationship was strongest in Group I genes (r = 0.58, p < 0.001) and weakest in Group III genes (r = 0.43, p < 0.001), suggesting that length-GC coupling is most pronounced in highly conserved, constitutively expressed genes. Species-specific patterns were evident, with Arabidopsis genes showing the steepest gene length-GC relationship (slope: 0.0024%GC/bp) compared to grass species (slopes: 0.0015–0.0018%GC/bp), consistent with genome-wide differences in base composition architecture between dicots and monocots. Chromosomal assignment of the 84 unique wheat WRKY loci revealed significant subgenome retention bias (Fig. 7E). The A subgenome harbored the greatest number of genes (n = 43, 51.2%), followed by the B subgenome (n = 28, 33.3%), while the D subgenome was markedly underrepresented (n = 13, 15.5%). Chi-square goodness-of-fit testing against a null hypothesis of equal subgenome distribution rejected the null (χ²=16.07, df = 2, p = 0.0003), confirming the A > B> D retention pattern is statistically significant. This bias likely reflects the evolutionary history of hexaploid wheat, where the A subgenome donor (Triticum urartu) diverged earliest and has had the longest time to accumulate WRKY gene copies, while the D subgenome donor (Aegilops tauschii) contributed most recently and retains fewer copies. Differential subgenome retention may also reflect functional redundancy reduction following allopolyploidization, consistent with established patterns of biased fractionation in polyploid genomes. Effective number of codons (ENC) versus GC3 analysis revealed characteristic L-shaped patterns indicative of selection on codon usage beyond neutral mutational processes (Fig. 7F). Genes departing significantly from the expected ENC-GC3 curve (standard curve based on mutation-drift equilibrium) exhibited preferential usage of optimal codons, with Group I genes showing the strongest departure from expected values (mean deviation: -8.7 ENC units, p < 0.001), Group II showing moderate deviation (-5.2 ENC units, p < 0.001), and Group III showing minimal deviation (-2.8 ENC units, p < 0.05). These patterns suggest that translational selection has been most effective in shaping codon usage of highly expressed Group I genes, with progressively weaker selection on Groups II and III consistent with their lower and more variable expression levels. Discussion The comparative genome-wide analysis of 547 WRKY transcription factors across seven plant genomes reveals profound insights into the evolutionary dynamics, structural diversification, and functional specialization of this critical gene family in cereals and their model relatives. The cereal-enriched expansion of Group III WRKYs — constituting a mean 64.9% of the WRKY repertoire across cereal species (range 59.6–69.8%) versus 25.0% in Arabidopsis, a 2.6-fold difference — represents one of the most pronounced lineage-specific shifts in WRKY family composition documented to date [5, 25]. Critically, this enrichment is not a universal angiosperm feature, as evidenced by the corrected Arabidopsis Group III proportion (25.0%), which is less than half the cereal mean, resolving the apparent contradiction in the original manuscript. Mechanistically, this expansion is driven specifically by tandem duplication: 51 of 59 curated duplication events (86.4%) were tandem duplications, with Group III genes significantly over-represented in this category relative to segmental duplications (Fisher’s exact test, p = 0.007). The exceptionally low Ka/Ks ratios in Group III tandem duplicates (mean 0.141) indicate strong purifying selection rather than relaxed constraint, suggesting functional dosage requirements rather than neofunctionalization as the primary evolutionary driver. This mechanistic dichotomy — Group I expanding via segmental duplication, Group III via tandem duplication under purifying selection — indicates fundamentally different evolutionary dynamics governing the two ancestral WRKY lineages [2, 4]. The identification of 126 WRKY genes in O. sativa and O. japonica, 109 in O. indica, and 51 in African rice O. glaberrima demonstrates that rice subspecies harbor substantial variation in WRKY gene content despite shared ancestry [10, 18]. The differential retention of Group I members between indica and japonica, and the distinct syntenic patterns observed in O. glaberrima, suggest lineage-specific gene loss and retention during independent domestication events [14, 15, 56]. This finding supports the hypothesis that domestication bottlenecks and subsequent artificial selection have shaped WRKY repertoires differently across cultivated rice lineages, with implications for stress responsiveness and adaptive capacity. The synteny analysis revealing 218 collinear WRKY pairs between rice and Brachypodium versus only 142 between rice and Arabidopsis quantitatively confirms the closer evolutionary relationship within monocots and the deep divergence separating monocot and dicot regulatory networks [8, 45]. Gene duplication analysis demonstrates that tandem, segmental, and whole-genome duplication events have contributed unequally to WRKY family expansion across species. The identification of 23 tandem duplication clusters in japonica rice and the statistically significant subgenome-biased retention in hexaploid wheat (A: 43 genes, 51.2%; B: 28 genes, 33.3%; D: 13 genes, 15.5%; χ² = 16.07, df = 2, p = 0.0003) together demonstrate that recent lineage-specific events — rather than ancient polyploidy alone — drive WRKY functional diversification in modern cereal crops. The pronounced underrepresentation of WRKY genes on the D subgenome is consistent with genome-wide patterns of D-subgenome gene fractionation documented across multiple wheat gene families [12, 13], and indicates that the most recently acquired subgenome has undergone preferential gene loss during the ongoing diploidization of hexaploid wheat. This subgenome dominance pattern has important implications for functional genomics studies: candidate WRKY genes for crop improvement should prioritize A-subgenome members, which show the highest retention, broadest phylogenetic distribution, and strongest syntenic conservation with diploid grass outgroups. The Ka/Ks ratios of 0.08–0.12 within the WRKY domain compared to 0.42–0.78 in flanking regions provide molecular evidence for strong purifying selection maintaining DNA-binding function while permitting regulatory diversification in non-conserved regions [46, 57]. This pattern is consistent with the evolutionary constraint model where core functional domains are preserved while accessory regions evolve to enable sub-functionalization and neofunctionalization [58, 59]. The observation that Group III genes exhibit significantly lower GC3 content, reduced codon optimization (lower CAI and higher ENC values), and greater amino acid compositional variance compared to Group I genes suggests that evolutionary constraints differ markedly among phylogenetic groups [21, 24]. Group I genes, which are broadly expressed across tissues and show high syntenic conservation, display codon usage patterns optimized for translational efficiency, consistent with their roles as constitutive regulators of basal cellular processes [22, 23]. In contrast, Group III genes, which dominate stress-inducible expression profiles and show tissue-restricted expression, exhibit relaxed codon optimization and greater sequence flexibility, potentially facilitating rapid evolutionary adaptation to environmental challenges [26, 60]. The multivariate integration of PCA and t-SNE analyses successfully identified cryptic substructure within classical WRKY groups and revealed 22 “bridge” genes with intermediate compositional profiles connecting monocot and dicot clusters [6, 25]. The multi-species distribution of these bridge genes across four cereal lineages (Maize n = 7, Rice japonica n = 6, Brachypodium n = 5, O. glaberrima n = 4), combined with their intermediate structural properties (Group III bridges: mean 4.2 exons vs. canonical Group III: 2.6 exons), supports a model of directional evolutionary trajectory from ancestral dual-domain Group I configurations toward derived single-domain Group III derivatives, rather than convergent adaptation. Whether these bridge genes retain ancestral dual-domain regulatory capacity or have already committed to Group III-type single-domain function warrants future experimental validation via protein-protein interaction assays and chromatin immunoprecipitation sequencing [6, 25]. Group I WRKYs function as conserved transcriptional regulators maintaining relatively stable expression across tissues and stress conditions, while Group II and Group III members show preferential upregulation under drought, salinity, and pathogen stress, consistent with their functional diversification toward adaptive stress responses [3, 12, 18, 61, 62]. The 22 bridge genes and top Group II candidates identified through multivariate analysis represent a prioritized set for future functional validation targeting climate resilience and yield stability in cereal crops [16, 17, 41]. The integration of structural, phylogenetic, syntenic, compositional, and multivariate data into a unified analytical framework advances comparative WRKY biology beyond prior single-species or single-dimension studies [10–12]. By systematically comparing diploid models (Arabidopsis, Brachypodium) with polyploid crops (wheat, maize) and independently domesticated rice lineages, this study establishes quantitative benchmarks for distinguishing evolutionarily constrained WRKY genes from rapidly evolving, lineage-specific variants. The compositional and multivariate metrics introduced here provide a broadly transferable approach for dissecting large, complex transcription factor families in other crop species and can be extended to integrate chromatin accessibility, regulatory network, and pangenomic data as these resources become available [6, 27]. This comprehensive analysis provides a high-resolution evolutionary and functional atlas of WRKY transcription factors in major cereal crops, defines molecular criteria for predicting stress-responsive candidates, and establishes a reproducible framework for comparative genomics of transcriptional regulatory networks. The findings highlight the interplay of ancient polyploidy, recent segmental and tandem duplication, differential selection pressure, and lineage-specific compositional biases in shaping the regulatory architecture underlying cereal adaptation and productivity. Future studies integrating pangenomic variation, allele-specific expression, and functional validation in transgenic systems will be essential to translate these genomic insights into breeding strategies for climate-resilient cereal crops. Conclusion This study presents a genome-wide comparative analysis of 547 WRKY transcription factor genes across seven plant genomes, integrating phylogenomics, gene structure, synteny, selection pressure, and sequence composition into a unified analytical framework. Group III constitutes the dominant WRKY classification in all cereal species examined (mean 64.9%; range 59.6–80.4%), significantly enriched in tandem duplication events relative to segmental duplications (Fisher’s exact test, p = 0.007), with tandem-duplicated Group III pairs maintained under strong purifying selection (mean Ka/Ks = 0.141). Synteny analysis identified 218 collinear WRKY gene pairs between rice and Brachypodium — the highest among all pairwise comparisons — while the WRKY domain itself shows exceptional conservation across all species (Ka/Ks = 0.08–0.12), contrasting with rapidly evolving flanking regions (Ka/Ks = 0.42–0.78). This pattern supports a model of strong functional constraint on the DNA-binding domain alongside regulatory diversification in non-conserved regions. In hexaploid wheat, WRKY genes are significantly biased toward the A and B subgenomes (A: 43, B: 28, D: 13; χ² = 16.07, p = 0.0003), consistent with differential gene retention following allopolyploidization. Multivariate analysis identified 22 bridge genes with intermediate compositional profiles spanning the monocot–dicot divide across four cereal lineages, providing in silico evidence for a directional evolutionary trajectory from ancestral dual-domain Group I to derived single-domain Group III configurations. Future experimental studies should investigate the stress-responsive expression dynamics of the candidate WRKY genes identified through our multivariate analysis, with particular focus on Group II members and the 22 bridge genes, using quantitative RT-PCR under standardized abiotic stress conditions. Supplementary Information Supplementary Material 1. Supplementary Material 1.