SNP molecular marker combination related to apple flesh crispness, primer and application

Abstract

The present application belongs to the technical field of molecular markers, and discloses a SNP molecular marker combination related to apple flesh crispness, primers and application. The SNP molecular marker combination related to apple flesh crispness of the present application is composed of 100 SNP sites, these sites show a high SHAP contribution value in the XGBoost model, and can effectively reflect the genetic influence of apple flesh crispness. Using the Top100 SNP marker set, an apple flesh crispness prediction model with high prediction performance can be constructed, and the 5-fold cross-validation Pearson correlation coefficient of the Top100 SNP model is 0.7221, the prediction correlation is higher than 0.4291 of the full feature model. It is shown that the SNP molecular marker combination of the present application contains most of the information of the genetic variation of apple flesh crispness, and can be applied to the breeding practice of the crispness trait as an efficient and low-cost molecular marker combination.

Description

Technical Field

[0001] This invention belongs to the field of molecular marker technology and relates to SNP molecular marker combinations, primers and applications related to the crispness of apple flesh. Background Technology

[0002] Apple flesh crispness is a crucial textural indicator affecting the commercial quality, market acceptance, and storage and processing performance of fresh apples, and is also an important target trait for apple quality breeding. Apple flesh crispness is jointly regulated by multiple factors, including the integrity of the plant's cell wall structure, intercellular adhesion, cell turgor pressure, the fruit's ripening and senescence process, and the post-harvest storage environment, making it a complex quantitative trait. Currently, the industry's conventional methods for identifying apple flesh crispness primarily rely on phenotypic determination, mainly including texture analyzer puncture, compression testing, and manual sensory evaluation. However, these traditional detection techniques have the following limitations: firstly, traditional phenotypic identification can only be conducted after the fruit is fully ripe, making it impossible to predict traits in the early stages of breeding; secondly, the test results are easily affected by external factors such as planting year, cultivation location, fruit harvest maturity, post-harvest storage duration, and instrument measurement parameters, resulting in poor stability and repeatability. This hinders efficient and rapid screening of large-scale progeny populations, limiting the efficiency of breeding high-quality apple crispness germplasm.

[0003] To overcome the technical bottlenecks of traditional phenotypic identification, current research is gradually focusing on the genetic analysis of apple flesh texture traits. Techniques such as molecular marker screening, candidate gene analysis, QTL mapping, and genome-wide association studies are employed to explore the genetic patterns of texture traits like flesh firmness and crispness. However, existing research systems still have the following shortcomings: Because apple flesh crispness involves the synergistic effects of multiple physiological pathways, including cell wall metabolism and turgor pressure regulation, and is significantly influenced by multi-gene interactions and the environment, current QTL or GWAS-based genetic locus analysis results often suffer from excessively large confidence intervals, insufficient validation of candidate gene functions, and a lack of high-contribution effect sites specific to crispness. This makes it difficult to directly translate these results into high-precision functional molecular markers usable in breeding practice, failing to meet the practical breeding needs for precise genotyping of large-scale offspring populations.

[0004] Therefore, developing a combination of SNP molecular markers that is closely related to the crispness trait of apple flesh and has a high model contribution and good predictive efficacy is of great significance for achieving accurate prediction of apple flesh crispness trait, efficient screening of high-crispness germplasm, scientific evaluation of breeding parents, and molecular marker-assisted precision breeding. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a combination of SNP molecular markers related to apple flesh crispness, primers, and their applications. The SNP molecular marker combination for apple flesh crispness in this invention consists of the top 100 SNP loci in terms of contribution. These loci exhibit high SHAP contribution values ​​in the XGBoost model, effectively reflecting the genetic influence of apple flesh crispness. Using this Top 100 SNP marker set, a high-performance apple flesh crispness prediction model can be constructed. The 5-fold cross-validation Pearson correlation coefficient of the Top 100 SNP model is 0.7221, higher than the 0.4291 of the full-feature model. This indicates that the SNP molecular marker combination of this invention contains most of the information on the genetic variation of apple flesh crispness and can be applied as an efficient and low-cost molecular marker combination in breeding practices for the crispness trait.

[0006] On the one hand, the present invention provides a combination of SNP molecular markers related to the crispness of apple flesh, the combination of SNP molecular markers consisting of 100 SNP sites; The location information of the SNP molecular marker combination on the apple reference genome Cameo_hap1 (http: / / 10.5281 / zenodo.22789943) is as follows:

[0007] On the other hand, the present invention seeks protection for primer combinations for amplifying the above-mentioned SNP molecular marker combinations, characterized in that the primer combinations comprise a pair of amplification primers designed for each SNP site; The primer combination consists of the first primer pair composed of SEQ ID NO.1 and SEQ ID NO.2 to the hundredth primer pair composed of SEQ ID NO.199 and SEQ ID NO.200, wherein the nth primer pair is used to amplify the DNA fragment containing the nth SNP site.

[0008]

[0009] On the other hand, the present invention seeks protection for the use of the above primer combination in the preparation of reagents for detecting the genotype of SNP loci related to apple flesh crispness.

[0010] On the other hand, the present invention claims protection for a kit for detecting the genotype of SNP loci related to apple flesh crispness, comprising the above-mentioned primer combination.

[0011] Furthermore, the kit also contains at least one of PCR reaction buffer, dNTPs, DNA polymerase, positive control, and negative control.

[0012] On the other hand, the present invention seeks protection for the application of the above-mentioned SNP molecular marker combination or primer combination in molecular marker-assisted breeding of apple flesh crispness trait.

[0013] On the other hand, this invention claims protection for a method for constructing an apple flesh crispness prediction model, comprising: Obtain a training sample set, which includes genotype data of the above-mentioned multiple apple material SNP molecular marker combinations and their corresponding flesh crispness phenotype data; Using the genotype data as input features and the flesh crispness phenotype data matching the genotype as the prediction target, a prediction model is constructed using the XGBoost regression algorithm.

[0014] On the other hand, this invention claims protection for an apple flesh crispness prediction model, which is obtained by the above-described construction method.

[0015] Furthermore, this invention claims protection for a method for screening high-crisp apple germplasm, comprising: Using the genomic DNA of the apple material to be tested as a template, PCR amplification was performed using the above primer combination. The genotype of the amplification products was detected to determine the genotype of each SNP site in the above SNP molecular marker combination. The genotype is input into the apple flesh crispness prediction model to obtain a crispness prediction value. The higher the crispness prediction value, the crisper the apple flesh. High-crispness germplasm is selected based on the crispness prediction value.

[0016] Compared with the prior art, the technical solution provided by the present invention has at least the following beneficial effects or advantages: This invention presents a set of 100 SNP molecular markers related to apple flesh crispness. These sites exhibit high SHAP contributions in the XGBoost model, effectively reflecting the genetic influence of apple flesh crispness. A high-performance apple flesh crispness prediction model can be constructed using only this Top100 SNP marker set. The 5-fold cross-validation Pearson correlation coefficient of the Top100 SNP model is 0.7221. Compared to the full-feature model, the Top100 SNP model reduces the number of features from 1,169,417 to 100, while increasing the predictive correlation from 0.4291 to 0.7221. This indicates that the Top100 SNP marker set enriches the main predictive information related to apple flesh crispness and has high breeding application value. This SNP molecular marker combination can be used for rapid prediction of apple flesh crispness, screening of high-crispness materials, evaluation of breeding parents, and molecular marker-assisted selection, thereby enabling early identification of high-crispness apple materials, reducing phenotypic identification costs, and accelerating the breeding of new high-crispness apple varieties or lines. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 The result is a graph showing the average absolute SHAP values ​​of 1,169,417 SNPs.

[0019] Figure 2 The graph shows the prediction performance of the full feature model combined with SNP molecular markers on apple flesh crispness. Detailed Implementation

[0020] The technical solution of the present invention will be described below with reference to embodiments. However, the present invention is not limited to the following embodiments. Unless otherwise specified, the experimental methods and detection methods described in each embodiment are conventional methods; unless otherwise specified, the reagents and materials can be purchased commercially. Unless otherwise specified, the percentages in the following embodiments refer to mass percentages.

[0021] XGBoost: The ultimate gradient boosting tree model.

[0022] num_boost_round (referred to as n_estimators in some interfaces): The number of iterations for the boosting tree, i.e., the number of regression trees involved in the prediction.

[0023] eta (also known as learning_rate): The learning rate, used to reduce the magnitude of the model's predictions updated by each new regression tree.

[0024] max_depth: The maximum depth that a single regression tree is allowed to reach, used to control the model's complexity and its ability to learn non-linear interaction relationships.

[0025] SHAP value: Shapley additive interpretation value, used to represent the positive or negative contribution of a single SNP feature to the prediction result of a single material.

[0026] reg:squarederror: The objective function for the squared error of the XGBoost regression model, with the squared error between the measured and predicted values ​​of fruit crispness as the main loss.

[0027] tree_method: The algorithm for constructing the regression tree; hist indicates that a histogram approximation algorithm is used to find candidate split points.

[0028] max_bin: The maximum number of bins used by the histogram algorithm when discretizing continuous feature values.

[0029] subsample: The proportion of samples randomly drawn when training each regression tree.

[0030] colsample_bytree: The proportion of SNP features randomly selected during the training of each regression tree.

[0031] lambda(reg_lambda): The L2 regularization coefficient for the leaf node weights, used to suppress model overfitting.

[0032] alpha (reg_alpha): The L1 regularization coefficient for the leaf node weights, used to control the sparsity of the model.

[0033] gamma: The minimum loss descent threshold required to generate new leaf nodes, used to control the complexity of the tree structure.

[0034] seed: Random seed, used in fixed sample subdivision and random processes during model training to ensure repeatable results.

[0035] Example 1 This embodiment obtains SNP molecular marker combinations related to fruit crispness through a prediction model, specifically including the following steps: 1. Obtaining SNP molecular marker combinations through prediction models S1: Initial File Acquisition. The 705 apple materials shown in Table 1 were used as the research subjects (collected from the Luochuan Apple Experimental Station and Xunyi Apple Experimental Station of Northwest A&F University). These apple materials included apple hybrids, breeding populations, or germplasm resources. Whole-genome SNP data of the apple materials were analyzed to obtain the original full-volume VCF file, containing 9,673,954 SNP records. Fruit flesh crispness phenotypic data of the apple materials were analyzed, using multi-year, multi-environmental fruit flesh crispness as the target trait to characterize the differences in fruit flesh crispness among different apple materials. A linear mixture model (LMM) was constructed using the original phenotypic data as input. In this model, environmental factors (year, location, and their interactions) were set as fixed effects, and variety (germplasm materials) were set as random effects. The variance components of each random effect were estimated using the restricted maximum likelihood method (REML), and then the mixed model equation was solved to obtain the best linear unbiased prediction value (BLUP) of the flesh crispness of each variety, namely the global clonal value (GCV). The results are shown in Table 2. These results were used for subsequent genotype-phenotype association analysis and prediction model construction.

[0036] S2: Quality control of SNP data: PASS sites and biallelic SNP sites were retained. Sites with a deletion rate greater than 20% were filtered out, and low polymorphism sites with a MAF below 0.05 were removed. Missing genotypes were imputed according to the mean genotype of the corresponding SNP site. SNP genotypes were digitally encoded using an additive model to construct a sample × SNP genotype matrix. After quality control, 1,169,417 high-quality SNP features were finally obtained.

[0037] S3: Construct an apple flesh crispness prediction model. The sample × SNP genotype matrix and flesh crispness phenotypic data were matched to obtain 705 apple samples with both genotype and flesh crispness phenotypic data. An XGBoost regression model (full feature model) was constructed using the sample × SNP genotype matrix and flesh crispness phenotypic data to predict apple flesh crispness. The XGBoost regression model is formed by stacking multiple regression trees in successive rounds. Model parameters include: objective function reg:squarederror, tree_method hist, max_bin 32, eta 0.03, max_depth 4, subsample 0.85, colsample_bytree 0.85, lambda 1.0, alpha 0.0, gamma 0, num_boost_round 200, and random seed 20260610. The model training employed 5-fold cross-validation for performance evaluation. In 5-fold cross-validation, 705 apple samples were randomly shuffled and divided into 5 non-overlapping subsets using a random seed 20260610. In each round, 4 folds were used as the training set and the remaining 1 fold as the validation set, and this process was repeated for 5 rounds, ensuring that each sample served as a validation sample for exactly one round. All out-of-fold predicted values ​​were then used to calculate the Pearson correlation coefficient. The results showed that the original VCF full-feature XGBoost model had a 5-fold cross-validation Pearson correlation coefficient of 0.4291, indicating that the whole-genome SNP data contained genetic information that could predict the crispness of apple flesh.

[0038] S4: Calculate the SHAP contribution value. After the model training is completed, calculate the SHAP contribution value of each SNP site based on the XGBoost model, and use the average absolute SHAP value of each SNP site ( Figure 1 The value of ) represents the contribution of this locus to the prediction of flesh crispness. For the i-th apple sample, the predicted value of its flesh crispness is expressed as: (1) In formula (1), This represents the SNP genotype vector obtained by additive encoding the i-th apple material using 0, 1, and 2. This represents the predicted value of the crispness of the flesh in the i-th apple sample. This represents the initial baseline prediction of the model before adding the regression tree (in the absence of sample-specific initial predictions, this value is a global constant shared by all materials, which can be determined by XGBoost based on the target value of fruit pulp crispness in the training set). This represents the t-th regression tree; T This represents the total number of regression trees, in this embodiment. TIt equals num_boost_round, which is 200; This represents the learning rate, which is 0.03 in this embodiment.

[0039] In the t-th training round, the XGBoost model determines the new regression tree by minimizing the sum of the prediction loss and the tree structure complexity. Its objective function is expressed as: (2) In formula (2), n Indicates the number of training samples; y i This represents the measured value of the crispness of the pulp in the i-th sample. This represents the predicted values ​​obtained from the first t-1 regression trees; l is the loss function. In this embodiment, the objective function is reg:squarederror, therefore... It is the square of the difference between the measured value and the predicted value; Let be the regularization term for the t-th regression tree.

[0040] (3) In formula (3), L t This represents the number of leaf nodes in the t-th regression tree; w tk This represents the weight of the k-th leaf node; γ This represents the complexity penalty coefficient for leaf nodes; λ and α These represent the L2 regularization coefficient and the L1 regularization coefficient, respectively. In this embodiment... γ =0, λ It is 1.0. α It is 0.0.

[0041] For the j-th SNP site, its average absolute SHAP value across all n samples is used as the characteristic contribution, calculated as follows: (4) In formula (4), This represents the SHAP contribution of the j-th SNP site to the predicted value of the i-th apple material. I j This represents the mean absolute SHAP value of the j-th SNP site. The prediction result of the XGBoost model for the i-th material can be decomposed into baseline values. The sum of the SHAP contributions of each SNP feature. I j The larger the value, the greater the overall contribution of that SNP locus to the prediction of fruit crispness. This invention follows... I jSNP sites were sorted from largest to smallest and the top 100 were selected (Table 3).

[0042] Table 1. Sources of 705 apple samples

[0043] Note: In Table 1, "-" indicates that the parental information is missing for hybrid combinations.

[0044] Table 2 Global Clonal Values ​​(GCVs) of Flesh Crisp Phenotypic Data from 705 Apple Accessions

[0045] Table 3 Locations and mutation types of the top 100 SNP molecular markers

[0046] Example 2 This embodiment provides a predictive performance evaluation of Top100 SNP tags.

[0047] The XGBoost regression model (quadratic model) was reconstructed using the 100 SNP sites selected in Example 1 as input features. The quadratic model and the full-feature model used the exact same sample subdivision method, objective function, model parameters, model hyperparameters (including eta, max_depth, subsample, colsample_bytree, and num_boost_round), and random seed. The model prediction performance results are as follows: Figure 2 As shown.

[0048] Depend on Figure 2The Pearson correlation coefficient of the 5-fold cross-validation of the Top100 SNP model is 0.7221. Compared with the full-feature model, the Top100 SNP model reduces the number of features from 1,169,417 to 100, while increasing the predictive correlation from 0.4291 to 0.7221. This indicates that the Top100 SNP marker set enriches the main predictive information related to apple flesh crispness and has high breeding application value.

[0049] As described above, the basic principles, main features, and advantages of the present invention have been well described. The above embodiments and specifications are merely descriptions of preferred embodiments of the present invention, and the present invention is not limited to the above embodiments. Various changes and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit and scope of the present invention should fall within the protection scope defined by the present invention.

Claims

1. A combination of SNP molecular markers associated with apple flesh crispness, characterized in that, The SNP molecular marker combination consists of 100 SNP sites; The location information of the SNP molecular marker combinations on the apple reference genome Cameo_hap1 is as follows: 。 2. The primer combination for amplifying the SNP molecular marker combination of claim 1, characterized in that, The primer combination includes a pair of amplification primers designed for each SNP site; The primer combination consists of the first primer pair composed of SEQ ID NO.1 and SEQ ID NO.2 to the hundredth primer pair composed of SEQ ID NO.199 and SEQ ID NO.200, wherein the nth primer pair is used to amplify the DNA fragment containing the nth SNP site in claim 1.

3. The application of the primer combination according to claim 2 in the preparation of reagents for detecting the genotype of SNP loci related to apple flesh crispness.

4. A kit for detecting the genotype of SNP loci related to apple flesh crispness, characterized in that, It includes the primer combination as described in claim 2.

5. The reagent kit according to claim 4, characterized in that, The kit also contains at least one of the following: PCR reaction buffer, dNTPs, DNA polymerase, positive control, and negative control.

6. The application of the SNP molecular marker combination of claim 1 or the primer combination of claim 2 in marker-assisted breeding of apple flesh crispness trait.

7. A method for constructing a prediction model for apple flesh crispness, characterized in that, include: Obtain a training sample set, which includes genotypic data of multiple apple materials for the SNP molecular marker combination described in claim 1 and their corresponding flesh crispness phenotype data; Using the genotype data as input features and the flesh crispness phenotype data matching the genotype as the prediction target, a prediction model is constructed using the XGBoost regression algorithm.

8. A model for predicting the crispness of apple flesh, characterized in that, It is obtained by the construction method described in claim 7.

9. A method for screening apple germplasm with high crispness, characterized in that, include: Using the genomic DNA of the apple material to be tested as a template, PCR amplification was performed using the primer combination described in claim 2. The genotype of the amplification product was detected to determine the genotype of each SNP site in the SNP molecular marker combination described in claim 1. The genotype is input into the apple flesh crispness prediction model of claim 8 to obtain a crispness prediction value. The higher the crispness prediction value, the crisper the apple flesh. High-crispness germplasm is selected based on the crispness prediction value.

Images