Fast-Ridge-based rapid multi-population genome prediction method
By constructing a multi-population adaptive weighted Ridge model using the Fast-Ridge method, and combining the dual-mode conjugate gradient algorithm and parallel Bayesian optimization, the computational complexity and prediction bias issues in large-scale data processing are resolved, achieving efficient and accurate genome prediction.
Patent Information
- Application Number
- CN202511918425.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-18
- Publication Date
- 2026-03-03
AI Technical Summary
Existing genome prediction methods have high computational complexity when processing large-scale data, and ignore the differences in genetic structure between populations when predicting across populations, which leads to increased prediction bias. There is an urgent need to develop multi-population genome prediction methods that balance computational efficiency and prediction accuracy.
We employ the Fast-Ridge method for rapid multi-population genome prediction. By constructing a multi-population adaptive weighted Ridge model, we utilize a dual-mode conjugate gradient algorithm and a parallel Bayesian optimization method to optimize regularization parameters and population weights, thereby achieving efficient sparse matrix solving and improved prediction accuracy.
It significantly enhances the ability to process large-scale data, improves the efficiency and accuracy of genome prediction, and is applicable to the prediction of complex quantitative traits in animal and plant breeding, thereby improving the accuracy and computational efficiency of cross-population prediction.
Smart Images
Figure CN121601036A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the interdisciplinary field of computational genomics and bioinformatics, specifically involving a rapid multi-population genome prediction method based on Fast-Ridge. Background Technology
[0002] Genomic prediction (GP) is a key technology that uses genome-wide marker data (such as single nucleotide polymorphisms, SNPs) to build statistical learning models to predict unobserved phenotypes (e.g., yield, disease resistance) in individuals. This technology can effectively shorten the breeding cycle and reduce the cost of traditional phenotypic selection.
[0003] Existing GP methods are mainly divided into two categories: genetic statistics methods and machine learning methods. The former includes optimal linear unbiased estimation (GBLUP), Bayesian methods (such as BayesR), Ridge regression, and LASSO regression, while the latter includes support vector machines (SVM), random forests (RF), regenerating kernel Hilbert spaces (RKHS), and deep learning (DL).
[0004] Traditional GBLUP models are solved using a mixture model equation based on the genome relation matrix (GRM). The computational complexity of the sample size Larger ( When ), it faces a serious computational bottleneck. Although Ridge regression is theoretically equivalent to GBLUP, it has two key technical defects: (1) regularization parameter Grid search optimization is usually used, which has low computational efficiency; (2) Existing software packages generally use coordinate descent to solve the problem, which is not suitable for high-dimensional feature data. This leads to a sharp decrease in convergence speed. Furthermore, existing cross-population prediction methods often simply merge heterogeneous population data, ignoring differences in genetic structure between populations, resulting in increased prediction bias. Therefore, there is an urgent need to develop a multi-population genome prediction method that balances computational efficiency and prediction accuracy. Summary of the Invention
[0005] The technical problem to be solved by this invention is to provide a fast multi-population genome prediction method based on Fast-Ridge, which can improve the efficiency and accuracy of genome prediction.
[0006] The technical solution adopted by this invention to solve the above-mentioned technical problems is: a rapid multi-population genome prediction method based on Fast-Ridge, comprising the following steps: S1: Construct a multi-population adaptive weighted Ridge model; S2: The dual-mode conjugate gradient algorithm is used to solve the multi-population adaptive weighted Ridge model, and the dual-mode intelligent selection solution path is constructed based on the relationship between sample and feature dimensions. S3: Train the multi-group adaptive weighted Ridge model sequentially from single-group to multi-group, with prediction accuracy, i.e. Pearson correlation coefficient, as the optimization target. Use parallel Bayesian optimization method, based on Gaussian process surrogate model and expected improvement criterion to search for optimal parameters in parallel, and fine-tune the regularization parameters. S4: Validate the predictive performance of the parameter-tuned multi-population adaptive weighted Ridge model sequentially from single-population to multi-population.
[0007] According to the above scheme, the specific steps in step S1 are as follows: set up For the first Adaptive weights for each group For regularization parameters, and The first Genotype matrix and phenotypic vector of each population, For the first The number of individuals contained in each population; then the multi-population adaptive weighted Ridge model is: .
[0008] According to the above scheme, the specific steps in step S2 are as follows: S21: Based on optimization theory, the equivalent linear equations of the multi-population adaptive weighted Ridge model are derived: S22: Solve linear equations using an improved CG algorithm combined with sparse matrix optimization techniques, and accelerate matrix operations using sparse matrix multiplication.
[0009] Furthermore, in step S21, let n be the number of samples and d be the number of features; when At that time, the multi-population adaptive weighted Ridge model is equivalent to solving a system of linear equations: ; when In this case, using the dual form, the multi-population adaptive weighted Ridge model is equivalent to solving a system of linear equations:
[0010] ; in: For the kernel matrix, , This is a block diagonal weight matrix. .
[0011] According to the above scheme, the specific steps in step S3 are as follows: In a single-population training model, the objective is to maximize the average Pearson correlation coefficient obtained from five-fold cross-validation. In the multi-population training model, a rotating validation population strategy is adopted, with the goal of maximizing the average Pearson correlation coefficient of each validation population.
[0012] Furthermore, in step S3, in the single-population training model, the parallel Bayesian optimization method is used for parameter tuning, and compared with the parallel grid search with hot start to verify the parameter tuning accuracy and efficiency of the parallel Bayesian optimization method; the hot start parallel grid search divides the logarithmic space into equally spaced grids, and uses the solution of the previous round for initialization to accelerate the convergence of the conjugate gradient method, and performs the calculation of each fold cross-validation in parallel.
[0013] According to the above scheme, the specific steps in step S4 are as follows: In single-group scenarios, within a group or across groups, the highest Pearson correlation coefficient is used as the evaluation index, and the prediction performance is tested using a repeated 8 / 2 splitting pattern. In multi-population scenarios, the highest Pearson correlation coefficient was used as the evaluation index. The leave-one-out population validation method was used to compare the predictive performance of the Bayesian optimized weighted model with the unit weighted model, the genetic correlation weighted model, and the GBLUP weighted model.
[0014] Furthermore, the specific steps of the leave-one-out group validation method are as follows: The weights in the unit weighted model are set to 1; The weights of each training group are determined by calculating the genetic relationship matrix and its correlation among the groups; In the Bayesian optimization weighted model, the weights and regularization parameters are combined for Bayesian optimization parameter tuning. To compare the performance of the Bayesian optimized weighted model with the unit weighted model, the genetic correlation weighted model, and the GBLUP model, one population was selected as the target test population and the remaining populations were used as the training population. The Pearson correlation coefficient of each model was calculated, and the weighted model with the highest Pearson correlation coefficient was considered to have the best predictive performance.
[0015] A rapid multi-population genome prediction system based on Fast-Ridge. The modeling submodule is used to build a multi-population adaptive weighted Ridge model; The solution submodule is used to solve the multi-population adaptive weighted Ridge model using the dual-mode conjugate gradient algorithm, and constructs a dual-mode intelligent selection solution path based on the relationship between the sample and feature dimensions. The optimization submodule is used to train the multi-group adaptive weighted Ridge model sequentially from a single group to multiple groups. The prediction accuracy, i.e., the Pearson correlation coefficient, is used as the optimization target. The parallel Bayesian optimization method is adopted, and the optimal parameters are searched in parallel based on the Gaussian process surrogate model and the expected improvement criterion. The regularization parameters are then tuned. The validation submodule is used to validate the predictive performance of the parameter-tuned multi-population adaptive weighted Ridge model sequentially from a single population to multiple populations.
[0016] A computer memory storing a computer program executable by a computer processor, the computer program performing a Fast-Ridge-based method for rapid multi-population genome prediction.
[0017] The beneficial effects of this invention are as follows: 1. The Fast-Ridge-based rapid multi-population genome prediction method of this invention is an efficient Ridge regression solution method (FAST-Ridge-MP) for multi-population genome prediction. Through innovative collaborative design algorithm architecture and adaptive hyperparameter tuning, it reduces the complexity of traditional algorithms by using an efficient sparse matrix solver based on the conjugate gradient method (CG). ) down to This significantly improves the ability to process large-scale data; it innovatively proposes a parallel Bayesian optimization (PBO) framework, using the prediction accuracy (PCC) of five-fold cross-validation as an evaluation index to achieve intelligent collaborative optimization of regularization parameters and population weights; and it realizes the function of improving the efficiency and accuracy of predicted genomes.
[0018] 2. This invention establishes a hierarchical performance verification system and constructs a new multi-population weighted prediction paradigm. Through systematic comparison of the equal-weighted model (Uni_w), the genetic correlation weighted model (GRM_w), and the PBO optimized weighted model, it is confirmed that the PBO weighted scheme has the best prediction performance and is suitable for genome prediction of complex quantitative traits in plant and animal breeding, improving the accuracy and computational efficiency of cross-population prediction.
[0019] 3. This invention solves the problem of balancing computational efficiency and model generalization in multi-population genome prediction, and provides a solution for large-scale genome selection that combines high computational efficiency and prediction robustness, which has important application value in fields such as agricultural breeding.
[0020] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description
[0021] 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0022] Figure 1 This is a flowchart of an embodiment of the present invention.
[0023] Figure 2 This is an overall architecture diagram of the FAST-Ridge-MP algorithm, including single-population and multi-population methods, according to an embodiment of the present invention.
[0024] Figure 3 This is a flowchart of a dual-mode CG algorithm with adaptive dimension switching logic according to an embodiment of the present invention.
[0025] Figure 4 This is a flowchart of the parallel Bayesian optimization PBO parameter tuning algorithm according to an embodiment of the present invention.
[0026] Figure 5 This is a comparison chart of intra-population prediction PCC and computation time for a single population according to an embodiment of the present invention.
[0027] Figure 6 This is a comparison chart of the PCC and computation time for predicting other populations based on population 1 according to an embodiment of the present invention.
[0028] Figure 7 This is a comparison chart of the PCC and computation time for predicting other populations based on population 2 according to an embodiment of the present invention.
[0029] Figure 8 This is a comparison chart of the PCC and computation time for predicting other populations based on population 3 according to an embodiment of the present invention.
[0030] Figure 9 This is a comparison chart of the PCC and computation time for predicting other populations based on population 4 according to an embodiment of the present invention.
[0031] Figure 10 This is a comparison chart of PCC and computation time for cross-population prediction based on multiple populations according to an embodiment of the present invention. Detailed Implementation
[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0033] Example 1 See Figure 1 and Figure 2 The specific steps of the Fast-Ridge-based rapid multi-population genome prediction method are as follows: S1: Constructing a multi-population adaptive weighted Ridge model: (1) set up For the first Adaptive weights for each group For regularization parameters, and The first Genotype matrix and phenotypic vector of each population, For the first The number of individuals contained in a group; this model is achieved through... It degenerates into the classic Ridge model to maintain theoretical compatibility; S2: The dual-mode conjugate gradient (CG) algorithm is used to solve the multi-population adaptive weighted Ridge model, constructing a dual-mode intelligent solution path selection based on the sample-feature dimension relationship; see also Figure 3 The specific steps are as follows: S21: Let n be the number of samples and d be the number of features; based on optimization theory, the equivalent linear equations of the multi-population adaptive weighted Ridge model are derived: when At that time, the multi-population adaptive weighted Ridge model is equivalent to solving a system of linear equations: (2) when In this case, using the dual form, the multi-population adaptive weighted Ridge model is equivalent to solving a system of linear equations:
[0034] (3) in: For the kernel matrix, , This is a block diagonal weight matrix. ; S22: Using the improved CG algorithm and combined with sparse matrix optimization techniques, solve the linear equation system (2) or (3); Also includes: Set the maximum number of iterations. Precision parameter for algorithm termination ; Accelerate matrix operations using sparse matrix multiplication (such as sklearn.utils.extmath.safe_sparse_dot).
[0035] S3: Optimize regularization parameters through parallel parameter tuning strategies The optimization scope is The optimization objective is to achieve the prediction accuracy (PCC) of five-fold cross-validation; specifically, it includes two strategies: See Figure 4 Parallel Bayesian Optimization (PBO): Based on a Gaussian process surrogate model and the expected improvement (EI) criterion, it searches for optimal parameters in parallel; PBO iterations can be set with 5 initial points and a maximum of 10 iterations (single population); or with 10 initial points and a maximum of 80 iterations (multi-population). Parallel Grid Search with Warm Startup (WPGS): On a log-space equally divided grid, the convergence of the conjugate ladder is accelerated by initialization using the previous solution, and parallel computation is performed on each fold; WPGS uses np.logspace(-2,2,100) to generate 100 log-divided grid points; S4: Establish a hierarchical, progressive validation framework from single-population to multi-population scenarios, sequentially validating the prediction performance in single-population (within-population / cross-population) and multi-population scenarios; the specific steps are as follows: See Figure 5 In single-group scenarios, the prediction patterns within and across groups are distinguished; a 10-repeated 80 / 20 split test is used, with Pearson correlation coefficient (PCC) as the evaluation metric. See Figures 6 to 10 In multi-population scenarios, the system compares three weighting strategies: unit weight (Uni_W), genetic correlation (GRM_W), and PBO optimization; and adopts a leave-one-population-out strategy. S5: Construct a multi-population weighted prediction paradigm, and determine the optimal weight allocation scheme by comparing and analyzing equal-weighted, genetically related weighted, and PBO optimized weighted models; the specific steps are as follows: S51: The genetic correlation weighted model calculates population weights through the following steps. : Computational Groups with the group GRM matrix: , ; Calculate the correlation between groups: ; According to the target group The following strategy is used to calculate the weights:
[0036] S52: In the PBO optimization weighted scheme, the weights With regularization parameters Joint optimization, parameter tuning range is .
[0037] Experiments have demonstrated that the PBO optimized weighted scheme has the best prediction performance. In the single-population prediction paradigm, it improves the computational efficiency by 4-10 times while maintaining comparable computational accuracy. In the multi-population prediction paradigm, it improves the computational accuracy compared to the traditional GBLUP method (PCC is improved by 1%-2%), and the computation time is less than half that of GBLUP.
[0038] This embodiment, through innovative collaborative design algorithm architecture and adaptive hyperparameter tuning, utilizes an efficient sparse matrix solver based on the conjugate gradient method (CG) to reduce the complexity of traditional algorithms from... ) down to This significantly improves the ability to process large-scale data; it innovatively proposes a parallel Bayesian optimization (PBO) framework, using the prediction accuracy (PCC) of five-fold cross-validation as an evaluation index to achieve intelligent collaborative optimization of regularization parameters and population weights; and it realizes the function of improving the efficiency and accuracy of predicted genomes.
[0039] Example 2 The steps in this embodiment are the same as in Embodiment 1, except that each step is applied to a specific instance. Specifically, it includes the following steps: 1. Experimental Setup Development environment: Python 3.7 + NumPy / SciPy / scikit-learn; Test data: 44,349 SNP markers from 4 maize populations (n1=5808, n2=1337, n3=1346, n4=858); Data preprocessing: genotype matrix column standardization (μ=0, σ²=1).
[0040] 2. Parameter Settings Conjugate gradient: Maximum number of iterations Precision parameter for algorithm termination ; Regular expression parameters:
[0041] WPGS: np.logspace(-2,2,100).
[0042] PBO: Iteration settings include 5 initial points and a maximum of 10 iterations (single population); or 10 initial points and a maximum of 80 iterations (multi-population). 3. Verification Plan Single population (within the population): 10 replicates of 80 / 20 split validation Cross-population: Leave-One-Population-Out validation 4. CG Algorithm Optimization Considering the sparsity of the matrix, the matrix multiplication and matrix-vector multiplication in the CG iteration are performed using the safe_sparse_dot function from Python's sklearn library to improve the computational efficiency of CG. 5. Weighted Strategy 5.1 The genetic correlation weighted model is calculated using the following formula: Step 1: Calculate the population with the group GRM matrix: , ; Step 2: Calculate the correlation between groups: ; Step 3: Calculate the group weights Based on the target group Calculate the weights: ; 5.2 In the PBO weighted scheme, the weights The parameter tuning range is , and regularization parameters Joint optimization.
[0043] See Figures 5 to 10 This is a comparison chart of the experimental results of maize population prediction in an embodiment of the present invention. In the single-population prediction scenario, when the training sample size N < 1500, the new algorithm maintains comparable prediction accuracy to the GBLUP model (PCC difference is less than 2%); while when N > 5000, the computational efficiency is improved by 4-10 times (PBO mode takes only 1 / 10 of GBLUP time); in cross-population prediction, the new algorithm consistently maintains a speed advantage of more than 10 times.
[0044] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.
[0045] Example 3 This embodiment is used to implement the principle of the above method embodiment to construct a fast multi-population genome prediction system based on Fast-Ridge, including a modeling submodule, a solution submodule, and an optimization submodule.
[0046] The modeling submodule is used to build a multi-population adaptive weighted Ridge model; The solution submodule is used to solve the multi-population adaptive weighted Ridge model using the dual-mode conjugate gradient algorithm, and constructs a dual-mode intelligent selection solution path based on the relationship between the sample and feature dimensions. The optimization submodule is used to optimize the regularization parameters by using the Pearson correlation coefficient, which is the prediction accuracy of five-fold cross-validation, as the optimization target.
[0047] Each submodule is mainly used to implement the various steps of the method implementation, which will not be elaborated here.
[0048] It should be noted that, depending on the implementation needs, the various steps / components described in this application can be broken down into more steps / components, or two or more steps / components or parts of the operation of steps / components can be combined into new steps / components to achieve the purpose of this invention.
[0049] This embodiment also includes a processor, a communication interface, a memory, and a communication bus; wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory stores a computer program, which, when executed by the processor, causes the processor to perform the steps of the Fast-Ridge-based rapid multi-population genome prediction method.
[0050] This embodiment also provides a computer-readable storage medium storing executable instructions that, when executed by a processor, enable the processor to implement a fast multi-population genome prediction method based on Fast-Ridge.
[0051] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects.
[0052] Furthermore, this application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0053] This application is described with reference to the flowchart of the method and computer program product according to Embodiment 1 and the block diagram of the device (system) according to Embodiment 3. It should be understood that each step or block in the flowchart or block diagram, as well as combinations of steps or blocks in the flowchart or block diagram, can be implemented by computer program instructions.
[0054] These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions, which are executable by the processor of the computer or other programmable data processing device, produce instructions for implementing the process. Figure 1 One or more processes or boxes Figure 1 A Fast-Ridge-based system for rapid multi-population genome prediction, specifying the functions in one or more boxes.
[0055] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes or boxes Figure 1 The function specified in one or more boxes.
[0056] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes or boxes Figure 1 The steps of the Fast-Ridge-based rapid multi-population genome prediction method are specified in one or more boxes.
[0057] The above embodiments are only used to illustrate the design concept and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes or modifications made based on the principles and design ideas disclosed in the present invention are within the protection scope of the present invention.
Claims
1. A rapid multi-population genome prediction method based on Fast-Ridge, characterized by: Includes the following steps: S1: Construct a multi-population adaptive weighted Ridge model; S2: The dual-mode conjugate gradient algorithm is used to solve the multi-population adaptive weighted Ridge model, and the dual-mode intelligent selection solution path is constructed based on the relationship between sample and feature dimensions. S3: Train the multi-group adaptive weighted Ridge model sequentially from single-group to multi-group, with prediction accuracy, i.e. Pearson correlation coefficient, as the optimization target. Use parallel Bayesian optimization method, based on Gaussian process surrogate model and expected improvement criterion to search for optimal parameters in parallel, and fine-tune the regularization parameters. S4: Validate the predictive performance of the parameter-tuned multi-population adaptive weighted Ridge model sequentially from single-population to multi-population.
2. The method for rapid multi-population genome prediction based on Fast-Ridge according to claim 1, characterized in that: The specific steps in step S1 are as follows: set up For the first Adaptive weights for each group For regularization parameters, and The first Genotype matrix and phenotypic vector of each population, For the first The number of individuals contained in each population; then the multi-population adaptive weighted Ridge model is: 。 3. The method for rapid multi-population genome prediction based on Fast-Ridge according to claim 1, characterized in that: The specific steps in step S2 are as follows: S21: Based on optimization theory, the equivalent linear equations of the multi-population adaptive weighted Ridge model are derived: S22: Solve linear equations using an improved CG algorithm combined with sparse matrix optimization techniques, and accelerate matrix operations using sparse matrix multiplication.
4. The method for rapid multi-population genome prediction based on Fast-Ridge according to claim 3, characterized in that: In step S21, let n be the number of samples and d be the number of features; when At that time, the multi-population adaptive weighted Ridge model is equivalent to solving a system of linear equations: ； when In this case, using the dual form, the multi-population adaptive weighted Ridge model is equivalent to solving a system of linear equations: ； in: For the kernel matrix, , This is a block diagonal weight matrix. .
5. The method for rapid multi-population genome prediction based on Fast-Ridge according to claim 1, characterized in that: The specific steps in step S3 are as follows: In a single-population training model, the objective is to maximize the average Pearson correlation coefficient obtained from five-fold cross-validation. In the multi-population training model, a rotating validation population strategy is adopted, with the goal of maximizing the average Pearson correlation coefficient of each validation population.
6. The method for rapid multi-population genome prediction based on Fast-Ridge according to claim 5, characterized in that: In step S3, in the single-population training model, the parallel Bayesian optimization method is used for parameter tuning, and compared with the parallel grid search with hot start to verify the parameter tuning accuracy and efficiency of the parallel Bayesian optimization method; the hot start parallel grid search divides the logarithmic space into equally spaced grids and uses the solution of the previous round for initialization to accelerate the convergence of the conjugate gradient method, and performs the calculation of each fold cross-validation in parallel.
7. The method for rapid multi-population genome prediction based on Fast-Ridge according to claim 1, characterized in that: The specific steps in step S4 are as follows: In single-group scenarios, within a group or across groups, the highest Pearson correlation coefficient is used as the evaluation index, and the prediction performance is tested using a repeated 8 / 2 splitting pattern. In multi-population scenarios, the highest Pearson correlation coefficient was used as the evaluation index. The leave-one-out population validation method was used to compare the predictive performance of the Bayesian optimized weighted model with the unit weighted model, the genetic correlation weighted model, and the GBLUP weighted model.
8. The method for rapid multi-population genome prediction based on Fast-Ridge according to claim 7, characterized in that: The specific steps of the leave-one-out group validation method are as follows: The weights in the unit weighted model are set to 1; The weights of each training group are determined by calculating the genetic relationship matrix and its correlation among the groups; In the Bayesian optimization weighted model, the weights and regularization parameters are combined for Bayesian optimization parameter tuning. To compare the performance of the Bayesian optimized weighted model with the unit weighted model, the genetic correlation weighted model, and the GBLUP model, one population was selected as the target test population and the remaining populations were used as the training population. The Pearson correlation coefficient of each model was calculated, and the weighted model with the highest Pearson correlation coefficient was considered to have the best predictive performance.
9. A rapid multi-population genome prediction system based on Fast-Ridge, characterized in that: The modeling submodule is used to build a multi-population adaptive weighted Ridge model; The solution submodule is used to solve the multi-population adaptive weighted Ridge model using the dual-mode conjugate gradient algorithm, and constructs a dual-mode intelligent selection solution path based on the relationship between the sample and feature dimensions. The optimization submodule is used to train the multi-group adaptive weighted Ridge model sequentially from a single group to multiple groups. The prediction accuracy, i.e., the Pearson correlation coefficient, is used as the optimization target. The parallel Bayesian optimization method is adopted, and the optimal parameters are searched in parallel based on the Gaussian process surrogate model and the expected improvement criterion. The regularization parameters are then tuned. The validation submodule is used to validate the predictive performance of the parameter-tuned multi-population adaptive weighted Ridge model sequentially from a single population to multiple populations.
10. A computer memory, characterized in that: It contains a computer program that can be executed by a computer processor, which performs the Fast-Ridge-based rapid multi-population genome prediction method as described in any one of claims 1 to 8.