SNP markers associated with tobacco leaf number traits and uses thereof
By mining and screening high-contribution SNP markers related to the number of tobacco leaves, and constructing a prediction model using the XGBoost regression model, the problem of rapid screening of tobacco leaf number traits in the early stages of breeding was solved, achieving efficient material identification and accelerating the breeding process.
Patent Information
- Application Number
- CN202610797723.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-04
- Publication Date
- 2026-07-14
AI Technical Summary
Existing technologies make it difficult to quickly and accurately screen and genetically improve the number of tobacco leaves in the early stages of breeding. Traditional methods are greatly affected by the environment, the effect of molecular marker-assisted selection is limited, and there is a lack of stable genetic marker sets.
By mining high-contribution SNP markers closely related to the number of tobacco leaves, a prediction model was constructed using the XGBoost regression model. Significant and stable SNP sites were screened out, and a kit and prediction model were developed to achieve rapid prediction of the number of tobacco leaves and material screening.
It enables rapid and accurate prediction of tobacco leaf number and early material identification, reduces breeding costs, accelerates the development of new lines, and improves breeding efficiency.
Smart Images
Figure CN122382248A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of biotechnology and relates to SNP markers related to tobacco leaf number traits and their applications. Background Technology
[0002] Tobacco leaf number refers to the number of effective leaves formed by a single tobacco plant, and is one of the important agronomic traits reflecting the level of vegetative growth and plant structure. Leaf number directly affects tobacco yield formation, field management methods, and curing and grading results, making it a key indicator in tobacco variety breeding and production applications. An appropriate leaf number helps achieve a rational plant structure, uniform light distribution, and improved leaf yield and utilization. Therefore, effectively regulating the tobacco leaf number trait has always been an important goal in tobacco genetic improvement and breeding practices.
[0003] Currently, the leaf number trait in tobacco is mainly evaluated through field surveys and manual counting. This method needs to be carried out in the later stages of plant growth or around the time of budding, which has a long evaluation cycle and is greatly affected by factors such as cultivation conditions, climate, and management measures. It is difficult to quickly and accurately screen a large number of materials in the early stages of breeding, which restricts the breeding efficiency of improving the leaf number trait.
[0004] To improve breeding efficiency, existing technologies have attempted to use molecular markers and candidate gene methods to conduct genetic analysis of some agronomic traits in tobacco. However, the number of tobacco leaves is not controlled by a single gene; its formation involves the coordination of vegetative and reproductive growth, apical meristem development, hormone signal regulation, and the combined effects of multiple physiological pathways. The genetic basis of this trait is complex, influenced by the cumulative effects of small multi-gene interactions and gene interactions, limiting the effectiveness of traditional genetic analysis and marker-assisted selection methods in improving the leaf number trait.
[0005] Current research on tobacco traits mainly focuses on quality traits, alkaloid content, or disease resistance, while a dedicated genetic marker system for the leaf number trait remains lacking. At present, genome-wide association studies (GWAS) primarily rely on statistical significance screening for SNP loci associated with leaf number. However, because leaf number is significantly influenced by environmental conditions and its genetic effects are dispersed, the number of significantly associated loci obtained is limited, the explanatory power of a single locus is low, and the stability and reproducibility across different genetic backgrounds or cultivation environments are insufficient, making it difficult to form a stable set of molecular markers that can be directly used in breeding practices.
[0006] Meanwhile, existing technologies have not yet been able to identify key genetic loci that make major contributions to the leaf number trait in tobacco from whole-genome high-dimensional SNP data, and have not yet established an effective set of genetic markers that can be used for leaf number trait prediction, material screening or molecular marker-assisted selection.
[0007] In conclusion, developing SNP molecular markers that are closely related to the leaf number trait in tobacco, have stable genetic effects, and have high application value is of great significance for achieving precise improvement of the leaf number trait in tobacco and efficient breeding. Summary of the Invention
[0008] To address the shortcomings of existing technologies and practical needs, this invention provides SNP markers related to the number of tobacco leaves and their applications, mines SNP markers related to the number of tobacco leaves, and further develops a prediction scheme based on deep learning to achieve rapid prediction of the number of leaves in tobacco materials.
[0009] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides SNP markers related to the number of tobacco leaves, the SNP markers including the SNP sites shown in Table 1.
[0010] This invention identifies high-contribution SNP markers closely related to the number of tobacco leaves, which can be used for rapid prediction of the number of leaves in tobacco materials, material screening, and molecular marker-assisted selection, thereby enabling early identification of materials with the target number of leaves, reducing breeding costs, and accelerating the development of new varieties.
[0011] Table 1 Secondly, the present invention provides a method for screening SNP markers related to the tobacco leaf number trait described in the first aspect, the screening method comprising: Different varieties of tobacco were collected, and their whole genome SNP data and leaf number were detected. The XGBoost regression model was trained using the SNP data and the number of blades to construct a prediction model; Based on the prediction model, the feature contribution of each SNP is analyzed, and the SNPs are sorted from largest to smallest contribution. A specified number of top-ranked SNP sites are selected as SNP markers related to the number of tobacco leaves.
[0012] Optionally, the screening method further includes a step of quality control of SNP data, including filtering out sites with a deletion rate greater than 10%, deleting low polymorphic sites with a MAF (minor allele frequency) of less than 0.05, and appropriately imputing missing genotypes.
[0013] Thirdly, the present invention provides the application of the SNP markers and / or their detection reagents related to the tobacco leaf number trait described in the first aspect in predicting the tobacco leaf number.
[0014] Fourthly, the present invention provides a kit for predicting the number of tobacco leaves trait, the kit comprising reagents for detecting SNP markers associated with the number of tobacco leaves trait as described in the first aspect.
[0015] Fifthly, the present invention provides a tobacco leaf number trait prediction model, which is obtained by training a machine learning model using SNP markers related to the tobacco leaf number trait as described in the first aspect and leaf number trait data in tobacco.
[0016] Optionally, the machine learning model includes the XGBoost regression model.
[0017] Optionally, the training includes using a 5x cross-validation hyperparameter.
[0018] Optionally, the hyperparameters include n_estimators, learning_rate, and max_depth.
[0019] Optionally, n_estimators is 200, learning_rate is 0.01, and max_depth is 8.
[0020] Sixthly, the present invention provides a method for predicting the number trait of tobacco leaves, the prediction method comprising: Information on SNP markers related to the tobacco leaf number trait as described in the first aspect of the method is obtained, and the SNP information is input into the tobacco leaf number trait prediction model described in the fifth aspect to predict the tobacco leaf number.
[0021] In a seventh aspect, the present invention provides an electronic device comprising one or more processors and a memory for storing executable instructions, the one or more processors being configured to invoke the executable instructions stored in the memory to implement the steps of the tobacco leaf number trait prediction method described in the sixth aspect.
[0022] Eighthly, the present invention provides a computer-readable storage medium having stored thereon computer program instructions, which, when executed by a processor, implement the steps of the tobacco leaf number trait prediction method described in the sixth aspect.
[0023] Compared with the prior art, the present invention has at least the following beneficial effects: This invention identifies single nucleotide polymorphism (SNP) sites located in different chromosomes and gene regions. The nucleotide variations of these sites are conventional SNP types such as A / T, A / G, C / T, C / G, or G / T. These SNPs can be genotyped using conventional molecular detection methods. In the XGBoost model of this invention, these sites all show significant and stable feature contribution values, which can effectively reflect the genetic influence of leaf number. This model can be used for rapid prediction of leaf number in tobacco materials, material screening, and molecular marker-assisted selection, thereby enabling early identification of materials with target leaf number, reducing breeding costs, and accelerating the development of new lines. Attached Figure Description
[0024] Figure 1 This is a schematic diagram of the technical route of the present invention.
[0025] Figure 2 A graph showing the mean absolute SHAP value results for each SNP.
[0026] Figure 3 The graph shows the prediction performance of the prediction model on the test sets of all features and core SNP features, respectively.
[0027] Figure 4 This is a field planting experiment diagram of 200 tobacco materials in Example 3.
[0028] Figure 5 The values are for the tobacco leaf material in Example 3. Detailed Implementation
[0029] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments. However, the following examples are merely simplified examples of the present invention and do not represent or limit the scope of protection of the present invention. The scope of protection of the present invention is determined by the claims.
[0030] Where specific techniques or conditions are not specified in the examples, they shall be performed in accordance with the techniques or conditions described in the literature in this field, or in accordance with the product instructions. Reagents or instruments whose manufacturers are not specified are all conventional products that can be purchased from legitimate channels.
[0031] Unless otherwise defined, scientific and technical terms and their abbreviations used in conjunction with this invention shall have the meanings commonly understood by one of ordinary skill in the art to which this invention pertains. Some of the terms and abbreviations used in this invention are listed below.
[0032] SNP: Single nucleotide polymorphism.
[0033] XGBoost: The ultimate gradient boosting tree model.
[0034] n_estimators: The number of weak learners.
[0035] learning_rate: learning rate.
[0036] max_depth: Maximum depth.
[0037] SHAP value: Shapley additive interpretation value.
[0038] This invention identifies high-contribution SNP genetic markers closely related to tobacco leaf number (hereinafter referred to as the "leaf number-related SNP marker set"). These SNP loci are derived from a systematic analysis of tobacco whole-genome data and can significantly explain the genetic variation in tobacco leaf number. The overall technical approach is as follows: Figure 1 As shown.
[0039] Based on 5500 tobacco germplasm resources, genomic DNA was extracted, and whole-genome SNP data were obtained using high-throughput sequencing, with an initial total of 1,326,450 SNPs. Subsequently, quality control was performed on the SNP data, including filtering out sites with a deletion rate greater than 10%, deleting low-polymorphism sites with a MAF (minor allele frequency) below 0.05, and appropriately imputing missing genotypes, ultimately retaining 95,308 high-quality SNP sites. Furthermore, using leaf number phenotypic data corresponding to the aforementioned materials, an XGBoost regression model was constructed to predict leaf number. Model training employed 5-fold cross-validation, with key hyperparameters such as learning rate (0.01–0.3), maximum depth (3–12), and subsample ratio (0.5–1.0) determined through grid search. After model training, the tree structure within XGBoost automatically captures nonlinear interaction effects and additive effects between SNPs. The feature contribution of each SNP is extracted based on the trained XGBoost model, including the feature importance provided by the model itself and the SHAP value calculated by the SHAP framework for each SNP, so as to obtain the average contribution of each site in leaf number prediction.
[0040] Using the top 0.1% of SNPs by contribution as candidate loci (Table 1), further independent population validation, effect direction consistency checks, and contribution stability analysis were conducted to finally obtain the leaf number-related SNP marker set described in this invention. The top 0.1% of high-contribution SNPs were then used for secondary model training, and the XGBoost regression model was reconstructed using these SNPs as input. Experiments show that using this simplified SNP set as input improves the model's prediction accuracy compared to the whole-genome model and significantly reduces computational resource consumption during model training and inference, demonstrating the advantages of this SNP marker set in practical applications.
[0041] In one specific embodiment of the present invention, a kit for predicting the number of tobacco leaves is provided, the kit comprising reagents for detecting SNP markers associated with the number of tobacco leaves. Specifically, these may be high-throughput sequencing reagents.
[0042] In one specific embodiment of the present invention, a method for predicting the number of tobacco leaves is provided. The prediction method includes: obtaining information on SNP markers related to the number of tobacco leaves in the tobacco to be predicted, inputting the SNP information into a tobacco leaf number prediction model, and performing tobacco leaf number prediction.
[0043] In another specific embodiment of the present invention, an electronic device is provided, the electronic device including one or more processors and a memory for storing executable instructions, the one or more processors being configured to invoke the executable instructions stored in the memory to implement the steps of the tobacco leaf number prediction method.
[0044] Those skilled in the art will understand that the device of this application can be obtained using various forms of hardware, software, firmware, dedicated processors, or combinations thereof.
[0045] In another specific embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored, which, when executed by a processor, implement the steps of the tobacco leaf number prediction method.
[0046] Example 1 This implementation involves obtaining the SNP tag set related to the number of blades.
[0047] (1) Experimental materials and phenotypic data Using 5500 tobacco germplasm resources, covering various types including flue-cured, burley, and aromatic tobacco, representing a broad genetic background, leaf number phenotypic data were obtained using the whole-plant counting method. After the tobacco reached maturity, five representative plants were randomly selected from each germplasm resource, and the number of effective leaves was counted one by one from bottom to top. The criteria for an effective leaf were: leaf length greater than 20 cm and leaf width greater than 10 cm. Finally, the average of the five plants was calculated as the actual leaf number phenotypic value for that material.
[0048] (2) Acquisition and quality control of SNP data DNA extraction and sequencing: Genomic DNA was extracted from the material, and simplified genome sequencing was used to obtain whole-genome SNP data. The initial total number of SNPs was 1,326,450.
[0049] Genotype coding: SNP genotypes are digitized using a 0 / 1 / 2 additive model.
[0050] Quality control: Sites with a deletion rate >10% were filtered out, low polymorphism sites with a minor allele frequency (MAF) of less than 0.05% were removed, and missing genotypes were imputed. Finally, 95,308 high-quality SNP sites were retained for subsequent analysis.
[0051] Example 2 This embodiment describes the construction and performance evaluation of the initial model.
[0052] Data partitioning: The 5,500 materials in Example 1 were randomly divided into a training set (80%, 4,400 materials) and a test set (20%, 1,100 materials).
[0053] XGBoost Model Construction: An XGBoost regression model was constructed, and grid search optimization was performed on hyperparameters such as learning rate and maximum depth using 5x cross-validation. The final optimal hyperparameters were determined as: n_estimators = (200), learning_rate = (0.01), max_depth = (8).
[0054] Performance evaluation: The model made predictions on 1100 test sets, and the performance metric was a Pearson correlation coefficient of 0.52.
[0055] SHAP value calculation: Based on the trained initial XGBoost model, the SHAP framework is used to calculate the SHAP values of 95308 SNPs. The mean absolute SHAP value of each feature is calculated as its final contribution to leaf number prediction.
[0056] Screening: Select the top 0.1% of SNPs by contribution as the candidate label set. Figure 2 The final set of high-contribution SNP markers and their details are shown in Table 1.
[0057] Secondary training: Using only the high-contribution SNPs selected above as input features, the XGBoost regression model is reconstructed and trained.
[0058] Performance Comparison: The model was tested on the same 1100 test samples, and its performance was compared with the initial model. Results showed that using core SNPs significantly improved prediction accuracy from 0.52 to 0.65 with a substantial reduction in the number of features, and also increased model training and inference speed by 3 times, confirming the high efficiency and accuracy of this marker set in practical breeding applications. Figure 3 ).
[0059] Example 3 This embodiment verifies the prediction of leaf number-related SNP marker sets in independent breeding materials.
[0060] An additional 200 tobacco breeding materials with different genetic backgrounds from the model training set (5500 materials) were selected as an independent validation set. None of these materials were used in the model construction and optimization process.
[0061] High-throughput sequencing or specific SNP genotyping techniques were used to obtain the genotypic information of leaf number-related SNP marker sets (SNPs screened in Example 2) in these 200 materials.
[0062] The SNP genotype information obtained above was input into the XGBoost regression model trained twice based on the core SNP set in Example 2 to predict the number of leaves in these 200 materials. Based on the model's predictions, the number of leaves in these 200 materials was sorted from lowest to highest.
[0063] These 200 samples were used in a field planting trial. Figure 4 After the tobacco plants reached maturity, five representative plants were randomly selected from each sample, and the number of effective leaves was counted from bottom to top. The criteria for an effective leaf were: leaf length greater than 20 cm and leaf width greater than 10 cm. The average number of leaves from the five plants was then calculated as the actual number of leaves for each sample.
[0064] Based on the ranking of model predictions, the 50 materials with the highest predicted number of leaves were identified. Upon comparison, 44 of the 50 predicted multi-leaved materials (i.e., an 88% accuracy rate) were actually confirmed as multi-leaved materials by actual measurement. Figure 5 (Unit: leaf / plant). This fully demonstrates that this marker set can serve as a core tool for marker-assisted selection, greatly accelerating the breeding process of new high-yielding, multi-leaved tobacco varieties.
[0065] In summary, this invention identifies leaf number-related SNP markers covering single nucleotide polymorphism (SNP) sites located in different chromosomes and gene regions. These SNPs exhibit common SNP types such as A / T, A / G, C / T, C / G, or G / T. All SNPs can be genotyped using conventional molecular detection methods. The leaf number-related SNP markers consist of several SNPs located in gene regions (including coding regions, promoters, enhancers, and introns) and key regulatory regions of metabolic pathways. These sites show significant and stable feature contribution values in the XGBoost model of this invention, with their SHAP values ranking in the top 0.1% of all SNP features. This effectively reflects the genetic influence of leaf number and can be used for rapid prediction of leaf number in tobacco materials, material screening, and marker-assisted selection. This leads to early identification of materials with target leaf numbers, reduces breeding costs, and accelerates the development of new lines.
[0066] The applicant declares that the above description is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Those skilled in the art should understand that any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention fall within the protection and disclosure scope of the present invention.
Claims
1. A SNP marker associated with the number of tobacco leaves, characterized in that, The SNP markers include the SNP sites shown in Table 1.
2. The method for screening SNP markers related to the number trait of tobacco leaves as described in claim 1, characterized in that, The screening method includes: Different varieties of tobacco were collected, and their whole genome SNP data and leaf number were detected. The XGBoost regression model was trained using the SNP data and the number of blades to construct a prediction model; Based on the prediction model, the feature contribution of each SNP is analyzed, and the SNPs are sorted from largest to smallest contribution. A specified number of top-ranked SNP sites are selected as SNP markers related to the tobacco leaf number trait.
3. The application of the SNP markers and / or their detection reagents related to the number of tobacco leaves as described in claim 1 in predicting the number of tobacco leaves.
4. A kit for predicting the number of tobacco leaves, characterized in that, The kit includes a reagent for detecting the SNP markers associated with the number of tobacco leaves as described in claim 1.
5. A tobacco leaf number trait prediction model, characterized in that, The tobacco leaf number trait prediction model is obtained by training a machine learning model using SNP markers related to the tobacco leaf number trait as described in claim 1 and leaf number data.
6. The tobacco leaf number trait prediction model according to claim 5, characterized in that, The machine learning model includes the XGBoost regression model.
7. The tobacco leaf number trait prediction model according to claim 5 or 6, characterized in that, The training includes: using a 5x cross-validation hyperparameter; Optionally, the hyperparameters include n_estimators, learning_rate, and max_depth.
8. A method for predicting the number trait of tobacco leaves, characterized in that, The prediction method includes: Information on SNP markers related to the number of tobacco leaves as described in claim 1 is obtained in the tobacco to be predicted. The SNP information is then input into the tobacco leaf number trait prediction model as described in claim 5 to predict the number of tobacco leaves.
9. An electronic device comprising one or more processors and a memory for storing executable instructions, characterized in that, The one or more processors are configured to invoke executable instructions stored in the memory to implement the steps of the tobacco leaf number trait prediction method of claim 8.
10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the steps of the tobacco leaf number trait prediction method according to claim 8.