Model for predicting muscle selenium content based on wool selenium content and construction method
By establishing a prediction model based on the selenium content of wool, the problem of difficulty in evaluating the selenium content in mutton in the prior art is solved, and accurate prediction of the selenium content of muscle is achieved, which improves food safety and nutritional value.
Patent Information
- Application Number
- CN202411771812.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-04
- Publication Date
- 2025-03-14
AI Technical Summary
The prior art is difficult to effectively evaluate the selenium deposit content in mutton, and it is impossible to accurately predict the intake of animal selenium, which affects food safety and nutritional value.
Based on the model of predicting muscle selenium content of wool selenium, a prediction model was established by obtaining sample data of wool and muscle selenium content, and Spearman correlation analysis and support vector machine were used.
It realizes convenient and rapid prediction of muscle selenium content, helps farmers manage animal selenium intake and improves mutton quality, and has high prediction accuracy and consistency of the model.
Smart Images
Figure CN119626342A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of methods for predicting selenium content in mutton, and particularly relates to a model for predicting muscle selenium content based on wool selenium content and a construction method thereof. Background Art
[0002] Selenium is a trace element that is essential to human and animal health. It plays an important role in anti-oxidation, immune regulation, reproductive health and muscle function. Due to the uneven distribution of selenium in the soil, the selenium content of food in some areas is low, which may affect the selenium intake of residents. In order to ensure that livestock products contain a stable selenium content, supplementing selenium in livestock diets is becoming an important means for the human body to supplement selenium.
[0003] An appropriate level of selenium is necessary to maintain the normal physiological functions of the animal body. However, excessive or low levels of selenium in the diet will have adverse consequences. Excessive intake of selenium can cause selenium poisoning in ruminants. Selenium content in feed higher than 20-30 mg / kg can cause acute selenium poisoning, and higher than 3-5 mg / kg can cause chronic and subacute selenium poisoning. When the selenium content in the ruminant diet is lower than 0.05 mg / kg DM, selenium deficiency will occur, leading to oxidative stress, loss of appetite, white muscle disease, abortion in pregnant females, fetal mulberry heart and other symptoms that affect the growth and development of ruminants.
[0004] Since the boundary between selenium toxicity and deficiency doses is small, it is necessary to formulate a recommended selenium addition dose to guide the production of ruminants. Countries around the world have introduced different recommended selenium addition doses in diets based on the nutritional status of their respective ruminants. So far, the European Union has not yet established a recommended daily intake of selenium, and the maximum allowable intake of total selenium is 0.5 mg / g. Selenium is added as a trace element, and the appropriate amount of selenium deposition in meat products is difficult to assess conveniently and quickly. Therefore, developing a method that can accurately predict and assess muscle selenium content is of great practical significance for ensuring food safety and nutritional value.
[0005] In the current production process of selenium-rich mutton, there is no effective and scientific method to evaluate the selenium deposition content in mutton before slaughter, and the selenium intake of animals cannot be well grasped. The selenium content in mutton can only be tested after slaughter.
[0006] As a component of animals, the selenium content of wool may be related to the metabolism and distribution of selenium in animals. Studies have shown that there is a significant correlation between the selenium content in wool and the serum selenium content, which provides a possible basis for predicting muscle selenium content by wool selenium content.
[0007] Based on the above reasons, we proposed a model and construction method for predicting muscle selenium content based on wool selenium content. Summary of the invention
[0008] The object of the present invention is to provide a model for predicting muscle selenium content based on wool selenium content, and the model prediction formula is as follows: Muscle selenium content (mg / kg) = 0.2121 + 0.0367 × wool selenium content (mg / kg) Another object of the present invention is to provide a method for constructing a model for predicting muscle selenium content based on wool selenium content, which is used to construct the above-mentioned model for predicting muscle selenium content based on wool selenium content, comprising the following steps: S1 obtains sample data of wool selenium content and muscle selenium content; S2 integrated the data of wool selenium content and muscle selenium content, and used Spearman correlation analysis to study the correlation between wool selenium content and muscle selenium content; S3 used support vector machine to establish a model for predicting muscle selenium content by wool selenium content.
[0009] Preferably, the step S1 is specifically as follows: S1.1 Shave the wool on the chest and back of the test sheep, and keep the wool shaved at the same position for each individual; S1.2 Supplementation of selenium supplements in the diet of experimental sheep; S1.3 After the feeding period, collect the newly grown wool from the shaved area and test the selenium content in the collected wool samples; S1.4 Slaughter the test sheep corresponding to the collected wool, keep the muscle collection sites of all test sheep consistent, collect the longissimus dorsi muscle and leg muscle, and test the selenium content in the muscle samples.
[0010] Preferably, in step S2, the correlation analysis result between wool selenium content and muscle selenium content is: The Spearman rank correlation coefficient (rho) > 0.7, P < 0.01, indicating that the selenium content in wool was significantly correlated with the selenium content in muscle.
[0011] Preferably, in S3, the specific steps are as follows: S3.1 Data preprocessing: Use the preProcess function to standardize the wool data column, including mean centering and scaling; S3.2 Grid search parameter setting: Create a grid search table containing different combinations of sigma and C parameters for parameter tuning when training the support vector machine model. sigma controls the width of the radial basis kernel (RBF) function, and C is the penalty term coefficient, which determines the tolerance to errors. S3.3 Create training control: Use 5-fold cross validation for model evaluation, that is, divide the data into 5 subsets, use one subset as the test set each time, and the remaining subsets as the training set; S3.4 Model training and parameter tuning: Use the train function to train the model through support vector machine to predict the selenium content in muscle, and perform parameter tuning based on grid search and cross-validation during the training process; S3.5 Output the best parameters: Print out the best sigma and C parameters found by the model after grid search and cross validation; S3.6 performs prediction: Use the trained support vector machine model to predict the data set and obtain svm_predictions as the prediction result of the model.
[0012] As a preference, the method further includes S3.7 evaluating the model performance: including calculating the mean square error (MSE) and the coefficient of determination (R 2 ), to measure the overall prediction error of the model. By calculation, we get MSE=0.0008, R 2 >0.7, indicating that the model fits the data well.
[0013] Preferably, S3.8 is also included to evaluate the model performance: including calculating the residuals, and the residual mean = 0.0025 is obtained by calculation, indicating that the predicted values of the model have no systematic deviation in general, and the residual standard deviation = 0.0281, indicating that the prediction effect of the model is stable.
[0014] The present invention has the following beneficial effects: The present invention provides a model and construction method for predicting muscle selenium content based on wool selenium content. By constructing a model to quantify the use of wool selenium content to predict muscle selenium content, compared with detecting muscle selenium content after slaughter, wool is easy to collect and does not damage animals. It can quickly and conveniently predict selenium deposition in muscles, help farmers better manage the selenium intake of animals, and help better control and improve the quality of mutton. In the present invention, the mean square error (MSE) between the predicted value and the actual observed value of the wool selenium content prediction model for muscle selenium content is 0.0008, and the determination coefficient (R²) value is greater than 0.7, indicating that the model has a good fit to the data, and the residual mean is 0.0025, indicating that the predicted value of the model has no systematic deviation overall, and the residual standard deviation is 0.0281, indicating that the model's prediction is stable. In summary, the present invention uses wool selenium content as an indicator for predicting muscle selenium content, which is not only based on scientific correlation research, but also has the potential for practical application. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] Figure 1 This is a correlation analysis between the selenium content in sheep wool and the selenium content in sheep muscle in the present invention.
[0016] Figure 2 It is a scatter plot of predicted values and actual values of the model for predicting muscle selenium content in wool in the present invention.
[0017] Figure 3 It is the residual diagram of the model of predicting muscle selenium content by wool selenium content in the present invention. DETAILED DESCRIPTION
[0018] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0019] A method for constructing a model for predicting muscle selenium content based on wool selenium content, comprising the following steps: S1. Obtain wool selenium content and muscle selenium content data.
[0020] S1.1. Shave the wool on the chest (between the 4th to 8th ribs) and back (between the 3rd to 7th thoracic vertebrae) of the test sheep (shaved area: 10cm×10cm), keeping the shaved wool position consistent for each individual.
[0021] S1.2. Supplement the diet of experimental sheep with appropriate selenium additives.
[0022] S1.3. After the feeding cycle is over, collect the newly grown wool from the shaved area and test the selenium content in the collected wool samples.
[0023] S1.4. Slaughter the test sheep whose wool is to be collected, keep the muscle collection sites of all test sheep consistent, collect the longissimus dorsi muscle and leg muscle, and test the selenium content in the muscle samples.
[0024] S2. Integrate the data of wool selenium content and muscle selenium content, and use Spearman correlation analysis to study the correlation between wool selenium content and muscle selenium content; S3. Considering that support vector machine (SVM) has good performance in small sample conditions and is suitable for problems with small sample sizes, SVM was used to establish a model for predicting muscle selenium content from wool selenium content.
[0025] S3.1. Data preprocessing: Use the preProcess function to standardize the wool data column, including mean centering and scaling, to make the data more balanced when training the model.
[0026] S3.2, Grid search parameter setting: Create a grid search table containing different combinations of sigma and C parameters for parameter tuning when training the support vector machine model. sigma controls the width of the radial basis kernel (RBF) function, while C is the penalty term coefficient, which determines the tolerance to error.
[0027] S3.3. Create training control: Use 5-fold cross validation for model evaluation, that is, divide the data into 5 subsets, use one subset as the test set each time, and the remaining subsets as the training set, so as to evaluate the performance of the model more robustly.
[0028] S3.4. Model training and parameter tuning: Use the train function to train the model through the support vector machine (svmRadial is a support vector machine based on the radial basis kernel) to predict the selenium content in the muscle, and perform parameter tuning (based on grid search and cross-validation) during the training process.
[0029] S3.5. Output the best parameters: Print out the best sigma and C parameters found by the model after grid search and cross validation.
[0030] S3.6. Make predictions: Use the trained support vector machine model to predict the data set and obtain svm_predictions as the prediction result of the model.
[0031] The prediction model formula is: Muscle selenium content (mg / kg) = 0.2121 + 0.0367 × wool selenium content (mg / kg) S3.7. Evaluate model performance: Mean square error (MSE): Calculate the average of the squared errors between the predicted value and the actual value to measure the overall prediction error of the model. R² value: Coefficient of determination, used to measure the ability of the model to explain data changes. The closer the value is to 1, the better the model.
[0032] S3.8. Evaluate model performance: Calculate residuals: Calculate residuals, which are the differences between actual values and predicted values, to analyze the error distribution of the model.
[0033] A total of 99 4-month-old male lambs (32.86±2.86 kg) were selected and the basal diet was supplemented with appropriate selenium additives for the experiment. The experiment included a 14-day pre-feeding period and a 63-day test period.
[0034] The basic diet was prepared with reference to the "Standards for Sheep Feeding" (NY / T 816-2004). The experimental animals were fed the basic diet during the pre-feeding period, and no additional selenium was added to the basic diet. Selenium and other mineral elements were mixed into a premix and added to the feed.
[0035] Before the experiment officially started, all experimental sheep were dewormed and vaccinated, and the sheep houses and experimental sites were thoroughly cleaned and disinfected.
[0036] Shave the wool on the chest and back of the test sheep (shaved area: 10cm×10cm), keeping the shaved wool position consistent for each individual sheep.
[0037] The experimental sheep were fed in a single pen, and were fed twice a day at 08:00 and 18:00, with free access to food and water. After the 63-day fattening experiment, the sheep were deprived of water and feed 1 day before the end of the experiment.
[0038] The wool is shaved off the newly grown wool on the chest and back, and then the longissimus dorsi and leg muscles are taken after slaughter.
[0039] A total of 99 wool samples and 99 muscle samples were collected in this experiment (data set see Table 1). The selenium content in muscle and wool (fresh sample basis) was determined using an inductively coupled plasma mass spectrometer (ICP-MS7800, Agilent Technologies, Japan), and the method was based on the National Standard of the People's Republic of China "Determination of Selenium in Feed (GB / T 13883-2023)" and the National Food Safety Standard Determination of Multi-Elements in Food (GB 5009.268-2016). The standard solution used the selenium single element solution standard material (GSB 04-1751-2004, National Nonferrous Metals and Electronic Materials Analysis and Testing Center, Beijing).
[0040] Table 1 Dataset of wool and muscle selenium content used for model construction (unit: mg / kg) wool muscle wool muscle wool muscle 0.170 0.144 0.187 0.203 0.239 0.231 0.218 0.176 0.531 0.310 0.381 0.205 0.122 0.162 0.399 0.272 0.376 0.275 0.185 0.142 0.320 0.211 0.411 0.247 0.155 0.141 0.214 0.204 0.458 0.277 0.204 0.203 0.250 0.256 0.220 0.227 0.242 0.170 0.291 0.237 0.498 0.278 0.287 0.178 0.340 0.281 0.384 0.296 0.249 0.170 0.396 0.290 0.235 0.172 0.293 0.210 0.518 0.260 0.197 0.161 0.365 0.196 0.354 0.268 0.175 0.144 0.223 0.228 0.463 0.306 0.161 0.168 0.330 0.223 0.204 0.158 0.140 0.151 0.430 0.227 0.265 0.164 0.180 0.167 0.351 0.231 0.218 0.144 0.273 0.188 0.205 0.217 0.282 0.186 0.267 0.184 0.416 0.279 0.216 0.176 0.233 0.194 0.341 0.196 0.159 0.162 0.320 0.206 0.397 0.287 0.158 0.142 0.301 0.193 0.288 0.261 0.173 0.159 0.407 0.212 0.370 0.284 0.183 0.183 0.244 0.196 0.224 0.163 0.322 0.180 0.389 0.253 0.153 0.138 0.285 0.196 0.272 0.254 0.158 0.153 0.188 0.186 0.213 0.212 0.115 0.182 0.344 0.183 0.341 0.234 0.139 0.138 0.274 0.213 0.382 0.231 0.175 0.167 0.382 0.201 0.386 0.215 0.279 0.191 0.266 0.199 0.423 0.281 0.297 0.178 0.242 0.230 0.362 0.241 0.172 0.184 0.408 0.218 0.327 0.282 0.249 0.196 0.319 0.221 0.513 0.277 0.222 0.207 0.470 0.244 0.496 0.290 0.374 0.222 0.268 0.257 0.462 0.310 Excel 2021 was used for experimental data statistics, and SPSS 26.0 was used for data analysis. Spearman correlation analysis was used to study the correlation between wool selenium content and muscle selenium content (Spearman rank correlation coefficient rho (Spearman rank correlation coefficient) > 0.7; P < 0.01 indicated significant correlation). R version 4.4.1 was used to establish a wool selenium prediction muscle selenium model using support vector machine (R 2 > 0.6 indicates that the prediction model performs well and the model fits the data well).
[0041] 1. Analysis of the correlation between wool selenium content and muscle selenium content Supplement appropriate selenium additives in sheep diets, collect wool and muscle for correlation studies, and evaluate the feasibility of establishing a model for predicting muscle selenium content using wool selenium content. Figure 1The selenium content in sheep wool was significantly correlated with that in muscle (rho = 0.792; P < 0.001).
[0042] 2. Constructing a model to predict muscle selenium content based on wool selenium content Correlation analysis showed that the changes in selenium content in wool and muscle were correlated. Therefore, we used support vector machine to perform prediction training on the collected data on selenium content in wool and muscle (prediction formula: muscle selenium content (mg / kg) = 0.2121 + 0.0367 × wool selenium content (mg / kg)).
[0043] Firstly, the wool selenium content data were mean-centered and scaled and standardized. Then, a grid search table was created for parameter tuning when training the support vector machine model, and 5-fold cross-validation was used for model evaluation. Finally, the selenium content in muscle was predicted and trained based on the parameters tuned by grid search and cross-validation.
[0044] The results show that the mean square error between the predicted value and the actual observed value of the model using wool selenium content to predict muscle selenium content is small (MSE = 0.0008) (refer to Figure 2 ), the coefficient of determination (R²) value is 0.7207 ( Figure 2 ), which means that the model can explain 72.07% of the variation in the data, which has good explanatory power, but there are still about 27.93% of the variation not captured by the model. The mean residual is 0.0025 (refer to Figure 3 ), indicating that the model's predictions have no systematic deviation overall. The residual standard deviation is 0.0281 (refer to Figure 3 ), indicating that the model’s predictions are stable.
[0045] In summary, the results of the model for predicting muscle Se content based on wool Se content showed that the model had high prediction accuracy and consistency.
[0046] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A model for predicting muscle selenium content based on wool selenium content, characterized in that: The model prediction formula is as follows: Muscle selenium content (mg / kg) = 0.2121 + 0.0367 × wool selenium content (mg / kg).
2. A method for constructing a model for predicting muscle selenium content based on wool selenium content, which is used to construct the model for predicting muscle selenium content based on wool selenium content as claimed in claim 1, characterized in that: The following steps are involved: S1 obtains sample data of wool selenium content and muscle selenium content; S2 integrated the data of wool selenium content and muscle selenium content, and used Spearman correlation analysis to study the correlation between wool selenium content and muscle selenium content; S3 used support vector machine to establish a model for predicting muscle selenium content by wool selenium content.
3. The model construction method for predicting muscle selenium content based on wool selenium content as claimed in claim 2, characterized in that: The step S1 is specifically as follows: S1.1 Shave the wool on the chest and back of the test sheep, and keep the wool shaved at the same position for each individual; S1.2 Supplementation of selenium supplements in the diet of experimental sheep; S1.3 After the feeding period, collect the newly grown wool from the shaved area and test the selenium content in the collected wool samples; S1.4 Slaughter the test sheep corresponding to the collected wool, keep the muscle collection sites of all test sheep consistent, collect the longissimus dorsi muscle and leg muscle, and test the selenium content in the muscle samples.
4. The method for constructing a model for predicting muscle selenium content based on wool selenium content as claimed in claim 2, characterized in that: In step S2, the correlation analysis results of wool selenium content and muscle selenium content are as follows: The Spearman rank correlation coefficient (rho) > 0.7, P < 0.01, indicating that the selenium content in wool was significantly correlated with the selenium content in muscle.
5. The method for constructing a model for predicting muscle selenium content based on wool selenium content as claimed in claim 2, characterized in that: In S3, the specific steps are as follows: S3.1 Data preprocessing: Use the preProcess function to standardize the wool data column, including mean centering and scaling; S3.2 Grid search parameter setting: Create a grid search table containing different combinations of sigma and C parameters for parameter tuning when training the support vector machine model. sigma controls the width of the radial basis kernel (RBF) function, and C is the penalty term coefficient, which determines the tolerance to errors. S3.3 Create training control: Use 5-fold cross validation for model evaluation, that is, divide the data into 5 subsets, use one subset as the test set each time, and the remaining subsets as the training set; S3.4 Model training and parameter tuning: Use the train function to train the model through support vector machine to predict the selenium content in muscle, and perform parameter tuning based on grid search and cross-validation during the training process; S3.5 Output the best parameters: Print out the best sigma and C parameters found by the model after grid search and cross validation; S3.6 performs prediction: Use the trained support vector machine model to predict the data set and obtain svm_predictions as the prediction result of the model.
6. The method for constructing a model for predicting muscle selenium content based on wool selenium content as claimed in claim 5, characterized in that: Also included is S3.7 Evaluation of model performance: including calculation of mean square error (MSE) and coefficient of determination (R 2 ) to measure the overall prediction error of the model.
7. The method for constructing a model for predicting muscle selenium content based on wool selenium content according to claim 6, characterized in that: Also included is S3.8 Evaluating model performance: including calculating residuals to analyze the error distribution of the model.