The rice resistance genes Xa3/Xa26, xa5, Xa7, Xa21, Xa23, Xa27 confer resistance to numerous strains of rice bacterial blight. The C-terminal kinase domain of Xa3/Xa26 is essential for its resistance to BLB. Notably, pathogen infection does not induce the expression of Xa3/Xa26 (Cao et al., 2007). OsTPI1.1 encodes a triosephosphate isomerase (TPI) that catalyzes the reversible conversion of dihydroxyacetone phosphate to glyceraldehyde-3-phosphate. Downregulation of OsTPI1.1 significantly impairs XA3/XA26-mediated resistance. The defense response involves OsTPI1.1 through its TPI activity, which is notably enhanced upon interaction with XA3/XA26 (Liu et al., 2018).

Xa5 encodes the γ subunit of the general transcription factor TFIIA (Høiby et al., 2007). A mutation at position 39, where valine is substituted by glutamic acid (resulting in a susceptible protein), impairs the function of Xa5 (Jiang et al., 2006). This mutation may confer resistance to Xoo by suppressing bacterial transmission without affecting its proliferation (Lyer-Pascuzzi et al., 2008). The recessive xa5 allele encodes a TFIIAγ variant that prevents TALE binding, thereby blocking SWEET11 transcriptional activation. TALE-induced expression of susceptibility genes (e.g., Xa13 and Xa41) is prevented in xa5 plants or TFIIAγ5-RNAi lines, conferring enhanced resistance to BLB. In the absence of TFIIAγ5, another TFIIAγ gene, OsTFIIAγ1, can be activated by PthXo7, explaining why PthXo7-containing Xoo strains overcome xa5-mediated resistance (Ma et al., 2018). Additionally, TFIIAγ5 is required for TALE-mediated transcriptional activation of key defense-related genes such as Xa27 and Xa23, which are crucial for plant immunity against this disease (Yuan et al., 2016).

The Xa7 gene encodes a protein of unknown function, composed of 113 amino acid residues. Its promoter contains an effector-binding element (EBE) recognized by the AvrXa7 and PthXo3 effectors. Constitutive expression of Xa7 leads to continuous activation of immune responses but also inhibits normal rice growth (Chen et al., 2021).

The molecular mechanisms of Xa21-mediated immunity are extensively characterized. The pathogen-derived RaxX protein, a tyrosine-sulfated secretory molecule exported through the type I secretion system of Xoo, has been identified as the cognate ligand triggering XA21-mediated immune activation. This ligand-receptor interaction establishes the molecular basis for pathogen recognition in the XA21 signaling pathway (Pruitt et al., 2015). The tyrosine-sulfated RaxX (AvrXa21) demonstrates high-affinity binding to XA21 (Luu et al., 2019). OsSERK2 functions as a regulatory component in multiple receptor kinase-mediated immune signaling pathways, including those governed by XA21, XA3, and OsFLS2. XA21-OsSERK2 complex mutually phosphorylates to regulate rice immunity. OsSERK2-knockdown rice lines exhibit compromised XA21 and XA3/XA26-mediated immunity against Xoo (Chen et al., 2014).

Several Xa21-interacting proteins (XB) have been identified. XB3, which encodes an E3 ubiquitin ligase, is crucial for maintaining Xa21 protein stability. Upon pathogen infection, Xa21 phosphorylates XB3, leading to the ubiquitination and degradation of negative regulators involved in the immune response, thereby activating the disease resistance signaling pathway (Wang et al., 2006). XA21 interacts with a WRKY transcription factor, OsWRKY62 (XB10), which encodes two splice variants (OsWRKY62.1 and OsWRKY62.2). These variants are partially localized in the nucleus and act as negative regulators of rice resistance. They are involved in the modulation of both basal immunity and race-specific immune responses (Peng et al., 2008). XB15 encodes the PP2C protein phosphatase, which dephosphorylates the auto-phosphorylated Xa21, thus negatively regulating the immune response mediated by Xa21 (Park et al., 2008). The auxin-like protein XB21 also interacts with XA21. Overexpression of XB21 enhances resistance to Xoo, potentially through a mechanism involving endocytosis mediated by clathrin (Park et al., 2017). XB24 encodes an ATPase that interacts with Xa21 through its ATPase domain. This interaction promotes the phosphorylation of Xa21 while simultaneously inhibiting the immune response mediated by Xa21 (Chen et al., 2010b). In the absence of infection, the ATPase XB24 constitutively interacts with the regulatory juxtamembrane domain of XA21, facilitating phosphorylation of conserved serine/threonine motifs that stabilize the inactive formation. Pathogen recognition triggers dissociation of the XA21-XB24 complex, activating XA21 kinase activity and initiating phosphorylation-dependent defense signaling cascades (Chen et al., 2010b).

Xa23 shares the same open reading frame (ORF) as the susceptible allele xa23, with a polymorphism of 7-bp in the promoter region. This difference allows Xa23 to be induced by the AvrXa23 effector, while the xa23 allele remains unresponsive. Upon expression, Xa23 triggers a hypersensitive response (HR), leading to cell death, thereby limiting pathogen proliferation and inhibiting further Xoo invasion (Wang et al., 2015). AvrXa23 is widely present in 38 prevalent Xoo strains, which likely contributes to the resistance conferred by Xa23 (Wang et al., 2013).

The Xa27 gene, derived from tetraploid Oryza rufipogon, confers resistance to Xoo strains from various countries (Gu et al., 2005). The BLB AvrXa27 effector binds to the promoter of the Xa27 gene and induces its expression. The Xa27 protein is then secreted into the apoplast, leading to inhibition of bacterial growth. The correct localization of Xa27 is required for resistance to BLB. Xa27 and the susceptible allele xa27 share an identical coding sequence, with only two differences in the promoter region (Wu et al., 2008). Additionally, the disease resistance mediated by Xa27 relies on the function of the basic transcription factor Xa5 (Gu et al., 2009).

The interactions between rice resistance proteins and BLB avirulence proteins adheres to the “gene-for-gene” theory.In rice, various genes have been found to correspond to BLB avirulence proteins: Xa1, Xa10, xa13, xa25, xa41.

The Xa1 encodes a protein of 1,802-amino acid, containing a nucleotide-binding domain (NBS) and a leucine-rich repeat (LRR) region (Belkhadir et al., 2004). Both mechanical injury and pathogen infection induce the expression of Xa1 (Yoshimura et al., 1996). Xa1 recognizes multiple TALEs secreted by Xoo, including PthXo1, Tal4, and Tal9d, leading to the activation of downstream gene expression and conferring resistance to the pathogen in rice. A truncated interfering TALE (iTALE) that lacks the transcriptional activation domain is capable of disrupting the resistance conferred by Xa1 (Ji et al., 2016).

The Xa10 gene encodes a small protein, consisting of 126 amino acid residues, conferring resistance to certain Philippine races of Xoo, was first identified from rice cultivar Cas 209 (Lee et al., 2003; Gu et al., 2007). The Xa10 promoter region includes an effector-binding element (EBE) recognized by the AvrXa10 effector, which initiates Xa10 expression. Xa10 specifically recognizes the AvrXa10 protein from the pathogen, disrupts intracellular Ca²⁺ homeostasis, triggers programmed cell death and restricts further pathogen invasion (Tian et al., 2014).

The xa13, xa25, and xa41 genes are recessive resistance genes against BLB, encoding the Sugar Will Eventually be Exported Transporter (SWEET) proteins. These SWEET proteins are involved in the process of sugar efflux from cells, a mechanism that plays a crucial role in various biological processes, including nectar secretion and the transport of sugars to progeny tissues for their development (Chen, 2013). In rice, 22 genes belong to the SWEET gene family (Chen et al., 2010a). A key difference of 18 base pairs in the promoter region prevents the hijacking of xa13 by the TAL effectors PthXo1 secreted by Xoo, resulting in the expression of a resistant phenotype (Chu et al., 2006). Xa13 facilitates pathogen-induced copper redistribution in rice by mediating interactions between OsCOPT1 and OsCOPT5. This interaction reduces xylem Cu²⁺ concentration, alleviating copper toxicity to Xoo and promoting its proliferation (Yuan et al., 2010). The promoter region of Xa25 contains sequences that can be recognized and activated by PthXo2-like TALE proteins secreted by Xoo. When susceptible rice plants are infected with Xoo strains expressing PthXo2, the PthXo2 protein activates the expression of Xa25, facilitating the transport of sucrose into the apoplast, thereby supporting bacterial infection of rice tissues. Conversely, the promoter of the resistant xa25 allele differs from that of the susceptible Xa25 allele, lacking the PthXo2 EBE and thereby conferring resistance (Liu et al., 2011). The xa41 gene (Os11N3, OsSWEET14) encodes a sucrose transporter protein of 303 amino acids. Similar to xa13 and xa25, it can be induced by TALE proteins secreted by various Xoo strains. The promoter region of xa41 exhibits an 18-base pair deletion, which prevents activation of Xa41 protein expression by TALE effectors (Hutin et al., 2015).

The Xa4 gene encodes a cell wall-associated kinase that promotes cellulose synthesis and inhibits cell wall loosening, thereby increasing cell wall strength. The Xa4 gene confers resistance to BLB by reinforcing the cell wall, which also enhances stem mechanical strength. Xa4 confers pleiotropic effects on important agronomic traits, increasing grains per panicle while reducing panicle number per plant and 1000-grains weight. Consequently, plant height is reduced, which enhances lodging resistance without compromising yield (Hu et al., 2017). The widespread use of Xa4 can be attributed to its association with multiple beneficial agronomic traits.

Ullah et al. introduce four BLB resistance genes Xa4, xa5, xa13, and Xa21 into the indica rice variety Basmati-385, an indica rice cultivar with much sought-after qualitative and quantitative grain traits. The pyramided lines with quadruplet or triplet R genes showed the highest level of resistance compared to doublet or singlet R genes. Further, The pyramided lines with quadruplet and triplet R genes (Xa4, xa5, xa13) exhibited high level resistance to BLB, without any yield penalty (Ullah et al., 2022). BLB resistance genes are often pyramided with blast resistance genes to achieve durable resistance. MAS successfully pyramided the blast resistance gene Pi9 and the BLB resistance gene Xa23 in the thermosensitive male-sterile line GZ63S and its hybrid progeny Liangyou6326, conferring high resistance to all eight tested Xoo strains and broad-spectrum resistance to both leaf and neck blast (Ni et al., 2015). Pyramiding three bacterial leaf blight resistance genes (xa5, xa13, Xa21) and two blast resistance genes (Pi9, Pb1) into the rice variety BRRI dhan63 resulted in lines with diverse gene combinations. All pyramided lines showed significantly enhanced resistance to BLB or blast compared to the original cultivar (Nihad et al., 2023). MAS has been successfully applied in the development of multiple rice varieties with enhanced disease resistance and provides a reliable method for obtaining highly effective BLB-resistant rice lines.

Emerging genome-editing technologies, including TALENs, CRISPR/Cas9, and prime editing (PE), allow precise modification of crop genomes. These technologies have been successfully applied in model species and hold great potential for enhancing plant disease resistance. Gene editing confers disease resistance in rice primarily through two strategies: modifying susceptibility genes or engineering resistance proteins based on structure. TALENs represent one of the earliest engineered nuclease-based tools. TALEN-based technique was employed to knockout the Xa5 gene, resulting in increased BLB resistance in the edited rice lines. Different knockout variants exhibited varying levels of resistance, suggesting that amino acid substitutions at different positions within the Xa5 gene can modulate the rice’s resistance to BLB (Han et al., 2020). But the design and construction of TALENs are labor-intensive and technically demanding, requiring the synthesis of custom-designed DNA-binding domains for each target sequence. These high costs and time demands make widespread adoption difficult.

Beyond directly editing susceptibility genes, enhancing disease resistance in rice can be achieved through knock-in or knock-out EBE element in resistance genes and susceptibility genes respectively. CRISPR/Cas9 has emerged as the most widely adopted genome-editing technology because of its simplicity, high efficiency, and ease of use. A 149-bp sequence containing a 31-bp pathogen-responsive element was edited in the promoter of the sugar transporter gene Xa13 using the CRISPR/Cas9 system. The edited rice lines exhibited resistance to Xoo infection due to the failure to activate susceptibility genes, without compromising key agronomic traits or fertility (Li et al., 2019a). Off-target effects, where Cas9 cleaves unintended sites, remain a primary concern, especially for targets with high sequence similarity. Despite the development of high-fidelity Cas9 variants to mitigate this, the risk is still non-negligible. Prime editing (PE) is a precision genome-editing technology that represents a significant advance over CRISPR/Cas9. It employs a nCas9-reverse transcriptase fusion and a pegRNA to execute targeted nucleotide substitutions, insertions, and deletions with high accuracy. PE was employed to knock-in the EBE-binding element of the Xa41 into the promoter of dysfunctional Xa23, enabling Xoo-induced, TALE-dependent BLB resistance (Gupta et al., 2023). Prime editing (PE) remains an emerging technology and still requires comprehensive validation across a wide range of experimental models and applications. In conclusion, gene pyramiding is a widely adopted strategy for durable resistance, while gene editing offers a precise alternative.

ML and DL facilitate scalable, objective screening with reproducible and quantitative outputs (Merrick et al., 2022). The Support Vector Machine (SVM) regression model achieved prediction accuracy comparable to or exceeding that of traditional parametric models (Merrick et al., 2022). A recent study utilized kinship matrices integrated into three machine learning frameworks—Random Forest (RF), Support Vector Machine (SVM), and Light Gradient Boosting Machine (LightGBM)—to predict resistance to multiple diseases in rice and wheat, including blast, stripe rust, black-streaked dwarf virus, and sheath blight. This approach effectively mitigated confounding effects associated with population structure in predictive modeling. Furthermore, feature selection was optimized by incorporating SNPs identified through genome-wide association studies (GWAS), achieving an optimal balance between marker density and computational efficiency (Liu et al., 2024). These advances highlight the potential of integrating genomic data with machine learning algorithms to accelerate the identification and utilization of disease-resistant germplasm in rice breeding.

Accurate PS are critical steps in the development of disease-resistant varieties. These processes support selection decisions by quantifying resistance levels, guiding which lines are retained as parental materials or advanced in the breeding pipeline. Traditionally, phenotypic selection relied on breeders to visually assess and empirically evaluate plants based on simple metrics, such as lesion area or disease incidence (González-Camacho et al., 2018). Although these methods are cost-effective and require minimal equipment, they are inherently low-throughput, labor-intensive, and subjective. The phenotypic characteristics of plant diseases can be represented by diverse data types, including counts, multivariate counts, and images. Deep Learning (DL) has demonstrated outstanding performance in characterizing and predicting these phenotypic data for disease resistance in crops. To address the diverse nature of plant disease damage, researchers have developed various DL algorithms. These include Bayesian regularized neural networks—which are capable of genome-based prediction using Bayesian regression for ordinal data—and Poisson-based deep neural networks (Montesinos-Lopez et al., 2021). While portable spectrometers, UAVs, and satellite remote sensing enable non-destructive measurement of disease-resistance traits, the predictive accuracy of these data varies markedly. Linear regression models with marker matrices achieve high accuracy (Zhu et al., 2021), in contrast to the lower performance of NIRS-based phenomic methods. This progression signifies a broader transformation in plant phenomics, driven by the integration of IoT and AI for high-throughput trait quantification in precision breeding (DeSalvio et al., 2022; Jackson et al., 2023).

A deep learning-based video detection system was proposed for the identification and diagnosis of plant diseases and pests. This system enables rapid recognition by decomposing video sequences into frames analyzed via Faster R-CNN and YOLO v3 for real-time processing (Li et al., 2020). The methodology involves decomposing video sequences into individual frames, which are then analyzed using the Faster R-CNN architecture for object detection. Complementing broad-spectrum monitoring, specialized AI systems have been developed for precise, in-field evaluation of specific diseases. For instance, a multi-task framework introduced for accurate classification and severity analysis of BLB. This approach first generates multiclass segmentation masks of lesions, then employs a dual-path attention mechanism to refine features from both the original RGB image and the segmentation mask. These enhanced features are fused and processed by a lightweight MobileNetV2 classifier, achieving a test accuracy of 96.23% for severity prediction and demonstrating strong robustness in diverse environmental conditions, confirming its real-world deployment potential. Its design supports deployment of edge devices, facilitating real-time, in-field disease management (K. M et al., 2025). In parallel, a hyperspectral imaging-based approach was developed for in-field detection of plant diseases, integrating a line-scan hyperspectral camera with a convolutional neural network (CNN) trained directly on field-acquired spectral data (Almoujahed et al., 2025). These advancements highlight the potential of combining deep learning and hyperspectral imaging for enhancing the accuracy and efficiency of plant disease in agricultural practices.