Method for predicting content of free fatty acid in dairy cow blood based on infrared spectrum data in DHI daily measurement

By combining milk infrared spectral data, parity and lactation days in DHI measurement, a cable regression model was established, which solved the accuracy of predicting free fatty acid concentration in dairy cows, and achieved efficient and accurate energy balance management of dairy cows.

CN118914126APending Publication Date: 2024-11-08INST OF ANIMAL SCI & VETERINARY MEDICINE SHANDONG ACADEMY OF AGRI SCI +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411071932.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-06
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The prior art is difficult to accurately predict the concentration of free fatty acids in the blood of dairy cows in daily DHI measurements, and traditional blood collection methods are cumbersome and time-consuming, affecting the energy balance management of dairy cows.

Method used

Based on the mid-infrared spectral data of milk, a cable regression model was established, and the prediction accuracy was improved through first-order derivative processing and multiple spectral pretreatment methods, and the DHI device and R software system were used for analysis.

Benefits of technology

It realizes accurate prediction of free fatty acid concentration in cows without increasing workload, reduces costs, and improves management efficiency and prediction accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118914126A_ABST
    Figure CN118914126A_ABST
Patent Text Reader

Abstract

The invention belongs to the field of dairy cow breeding, and provides a method for predicting the concentration of free fatty acid in dairy cow blood based on DHI mid-infrared spectroscopy, and the method comprises the following steps: collecting a milk sample and a blood sample of a dairy cow, and collecting ear number, fetal number and postpartum day number data of the dairy cow; detecting the content of free fatty acid (NEFA) in the serum; performing DHI detection on the milk sample to obtain mid-infrared spectrum data; keeping wave bands related to milk fat in all the spectral data to obtain original spectral data; processing the original spectral data by using a first-order derivative to obtain a processing value of the spectral data of each wave band; according to the fetal number and the number of lactation days, obtaining grouping data of the fetal number and the number of postpartum weeks; and carrying out regression analysis by adopting an inhaul cable regression model program, and outputting the free fatty acid concentration. According to the method, the model for predicting the concentration of the NEFA in the blood of the Chinese Holstein cattle is established on the basis of the mid-infrared spectrum of the milk in combination with fetal times and lactation days during individual sampling, the content of the NEFA in the blood of the Chinese Holstein cattle is accurately predicted through daily DHI milk samples, and the method has convenience, accuracy and applicability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of dairy cow breeding, and particularly relates to a method for predicting the free fatty acid content in dairy cow blood based on infrared spectrum data measured in daily DHI. Background Art

[0002] The information disclosed in this background technology section is only intended to enhance the understanding of the overall background of the invention, and should not necessarily be regarded as an admission or any form of suggestion that the information constitutes the prior art already known to a person skilled in the art.

[0003] Negative energy balance in dairy cows is caused by high milk production and insufficient feed intake, that is, less energy intake and more energy consumption, which puts dairy cows in a state of energy deficiency, which is easy to cause a variety of energy metabolism disorders in dairy cows, resulting in huge economic losses. In order to maintain energy balance after giving birth, dairy cows will mobilize fat reserves, and free fatty acids (NEFA) decomposed from body fat can be released into the blood and quickly used by tissues. A certain concentration of NEFA in the blood is a manifestation of the normal adaptation and regulation of negative energy balance metabolism in dairy cows in the early lactation period. Excessive NEFA concentration indicates that the negative energy balance of dairy cows has intensified. Therefore, NEFA can be used as a physiological indicator to measure the negative energy balance of dairy cows. Detecting the NEFA content in the blood can determine the state of negative energy balance in dairy cows and provide scientific guidance for production. However, collecting blood samples in daily testing and conducting laboratory chemical methods for testing are cumbersome, time-consuming and laborious, and the blood collection process causes stress or damage to the cows themselves. With the popularization of dairy cow production performance measurement (DHI), more ranches are participating in DHI measurement. While measuring milk components, they can obtain mid-infrared spectral data of milk. For example, the NEFA in the blood can be predicted by using the mid-infrared spectroscopy method of daily DHI measurement. The NEFA content in the blood of individual cows can be obtained without increasing the workload, which can conveniently and effectively manage dairy cow production and prevent and treat postpartum energy balance, and is of great significance to actual production.

[0004] Patent CN115508305A discloses a method for directly predicting the concentration of monounsaturated fatty acids in milk. This method uses components in milk for prediction and the accuracy of the prediction needs to be improved. It does not involve the use of infrared data in milk to predict indicators in the blood and the problem of improving the accuracy of prediction. Summary of the invention

[0005] In order to solve the above problems, the present invention provides a method for predicting the concentration of free fatty acids (NEFA) in the blood of dairy cows based on the daily DHI measurement of milk mid-infrared spectrum. The present invention is based on the mid-infrared spectrum of milk combined with the data of parity and lactation days at the time of individual sampling, and establishes a prediction model for the concentration of NEFA in the blood of Chinese Holstein cows. It can more accurately predict the NEFA content in the blood through daily milk samples, and has convenience, accuracy and applicability.

[0006] The innovation of the present invention lies in that, unlike the method of directly predicting the concentration of monounsaturated fatty acids in milk in patent CN115508305A, the present invention is based on the data of my country's dairy cow population and takes into account that parity and lactation days are the main factors affecting blood NEFA. The prediction effects of different spectral processing methods and multiple models such as partial least squares are compared in the same experiment, and finally the optimal model is selected, which improves the accuracy and use value of NEFA prediction in theory and production.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] The first aspect of the present invention provides a method for predicting the concentration of free fatty acids in cow blood based on DHI mid-infrared spectroscopy, comprising:

[0009] Collect milk and blood samples from dairy cows, collect data on the cows’ ear numbers, parity, and days postpartum;

[0010] Detection of free fatty acid levels in serum;

[0011] Perform DHI test on milk samples to obtain mid-infrared spectrum data;

[0012] Retain the bands related to milk fat in all spectral data to obtain original spectral data;

[0013] The original spectral data is processed using the first-order derivative to obtain the processed value of the spectral data of each band;

[0014] According to parity and days of lactation, grouped data of parity and postpartum weeks were obtained;

[0015] The Lasso regression model program was used for regression analysis, and the free fatty acid concentration was output.

[0016] In some embodiments, the cow is a periparturient cow.

[0017] In some embodiments, the wavelength band associated with milk fat is: 2976.06-2767.89 cm -1 、1800.28-1684.63cm -1 and 1607.53-925.2cm -1 .

[0018] In some embodiments, parity is divided into three levels, namely, 1st parity, 2nd parity, and "3rd parity and above", each of which is 1 group.

[0019] In some embodiments, the postpartum weeks are divided into groups of 7 days each, i.e., 1-7 days, 8-14 days, 15-21 days, 22-28 days, and 29-35 days are each a group, with a total of 7 levels.

[0020] In some embodiments, the regression model formula is Y ij =PA i +DIM j +FD+LassoR, where Y ij is the predicted value of NEFA of the jth lactation day group of the i-th parity group of the individual, PA i For the i-th parity group, DIM j is the j-th lactation day group, FD is the first-order derivative processed value of the original spectrum, and LassoR is the model algorithm.

[0021] In some embodiments, the raw data are processed using standard normal transformation, second-order derivative, and convolution smoothing instead of first-order derivative.

[0022] In some embodiments, data are analyzed using 10-fold cross validation.

[0023] In some embodiments, the data are randomly divided into 7:3, where 7 folds are used for model analysis and 3 folds are used for model validation.

[0024] More specifically, they include:

[0025] 1. Collect 40mL of milk sample from each cow. Add preservatives to the sampling bottle. The amount of milk sample from each cow is 40mL. For specific sampling methods, refer to the industry standard "Technical Specifications for Determination of Dairy Cow Production Performance in China" (2007-0415).

[0026] 2. Collect cattle-related data, including ear number, parity and days postpartum of the sampled cattle.

[0027] 3. Use DHI routine equipment to test milk samples and obtain mid-infrared spectral data. Specifically, the samples were heated in a 42°C constant temperature water bath for 15-20 minutes and analyzed using a Milkoscan TM FT+milk analyzer (Foss Electric A / S, Denmark) for measurement, and the operation method refers to the industry standard "Technical Specifications for Measurement of Dairy Cow Production Performance in China" (NY / T 2007-0415).

[0028] 4. Use DHI routine equipment to test milk samples and obtain mid-infrared spectral data. Specifically, the samples were heated in a 42°C constant temperature water bath for 15-20 minutes and analyzed using a Milkoscan TM FT+milk analyzer (Foss Electric A / S, Denmark) for measurement, and the operation method refers to the industry standard "Technical Specifications for Measurement of Dairy Cow Production Performance in China" (NY / T 2007-0415).

[0029] 5. Keep 2976.06-2767.89cm in all spectral data -1 、1800.28-1684.63cm -1 and 1607.53-925.2cm -1 Band associated with milk fat.

[0030] 6. Use the first-order derivative (FD) to process the original spectral data to obtain the processed value of the spectral data in each band.

[0031] 7. According to individual parity and postpartum lactation days, the fixed effect data of parity and postpartum weeks were obtained. The parity was divided into 3 levels, namely: 1, 2, "3 parities and above" each with 1 group, a total of 3 levels; the postpartum weeks were divided into 7-day groups, namely: 1-7 days, 8-14 days, 15-21 days, 22-28 days, 29-35 days each with 1 group, a total of 7 levels.

[0032] 8. Using the R software system, regression analysis was performed using the LassoR model program. The regression model formula is: Y ij =PA i +DIM j +FD+LassoR.

[0033] Among them, Y ij is the predicted value of NEFA of the jth lactation day group of the i-th parity group of the individual, PA i For the i-th parity group, DIM j is the j-th lactation day group, FD is the first-order derivative processed value of the original spectrum, and LassoR is the model algorithm.

[0034] The second aspect of the present invention provides a system for predicting the concentration of free fatty acids in cow blood based on DHI mid-infrared spectroscopy, comprising:

[0035] A module for collecting milk and blood samples from cows, and collecting data on the cows’ ear numbers, parity, and days after calving;

[0036] A module for detecting the content of free fatty acids in serum;

[0037] A module for performing DHI detection on milk samples and obtaining mid-infrared spectral data;

[0038] A module for retaining the bands related to milk fat in all spectral data to obtain the original spectral data;

[0039] A module for processing the original spectral data using the first-order derivative to obtain the processed value of the spectral data of each band;

[0040] Module for obtaining grouped data on parity and weeks postpartum according to parity and days of lactation;

[0041] Module for outputting free fatty acid concentrations from regression analysis using the Lasso regression model procedure.

[0042] Beneficial Effects of the Invention

[0043] (1) The present invention has the advantages of low cost, simple operation, high throughput and rapidity, which are also the common advantages of MIRS analysis method;

[0044] (2) The model of the present invention has high accuracy. The milk composition and yield of dairy cows are affected by parity and days of lactation. The yield and composition of cows with different parities or days of lactation are significantly different. Adding parity and days of lactation to the model of the present invention significantly improves the accuracy of prediction.

[0045] (3) The present invention compares the effects of different spectral preprocessing methods and different model algorithms on model accuracy to make the results more convincing. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention, and the exemplary embodiments of the present invention and their description are used to explain the present invention and do not constitute improper limitations on the present invention.

[0047] Figure 1 :A flow chart, B model schematic diagram;

[0048] Figure 2 : Spectral diagram (original spectrum and 4 preprocessed spectra). DETAILED DESCRIPTION

[0049] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0050] Sources of the idea and prediction model formula of the present invention:

[0051] Studies have found that the NEFA content in the blood of dairy cows is closely related to the individual physiological state, among which parity and lactation days are important factors affecting its content. The mid-infrared spectrum in milk can be used to predict relevant indicators, but different methods of processing spectral data and different models significantly affect the accuracy of the prediction results. Based on a full analysis of the physiological factors affecting the NEFA content, the present invention combines different data processing methods with models, uses the spectral data of milk samples collected daily to establish an optimal regression equation, and obtains a more accurate value of the NEFA content in the blood of dairy cows without increasing the workload of the production line, which is used to guide production management. By collecting milk samples and participating in DHI determination in the early stage, sorting data in the mid-term, and processing and modeling analysis of the data using the relevant programs of the R software system in the later stage, the regression equation formula for the optimal prediction of blood NEFA content is screened. At the same time, blood samples of some cows are collected, and the NEFA determination is determined by the classical method. The SAS8.2 FREQ program is used for suitability test analysis to compare the significance of the difference between the two results.

[0052] 1. Sample collection and testing

[0053] 1. Blood and milk samples were collected from 405 dairy cows at a farm in Shandong Province 5-35 days after giving birth. Considering the physiological process of free fatty acids from blood to milk, blood samples were collected between 6:00 and 10:00 in the morning, and milk samples were collected between 12:00 and 4:00 in the afternoon of the same day.

[0054] 2. Collect 5 mL of procoagulant blood from each cow, place it at room temperature for 1-2 hours to promote coagulation, and then centrifuge (parameters: 2500g, 15min, 4°C) to separate the serum. The serum samples are frozen and stored at -20°C for testing.

[0055] 3. Collect 40mL of milk sample from each cow. Add preservatives to the sampling bottle. The milk sample volume for each cow is 40mL. For specific sampling methods, refer to the industry standard "Technical Specifications for Determination of Dairy Cow Production Performance in China" (2007-0415).

[0056] 4. Cattle data collection: collect data such as ear number, parity and days after delivery of the above-mentioned sampled cattle.

[0057] 5. The acyl-CoA synthetase-acyl-CoA oxidase (ACS-ACOD) method (Lab Assay TMNEFA) was used to detect the content of NEFA in serum. Each sample was measured twice. After eliminating the values ​​with a coefficient of variation greater than 15% in the two measurements, the final value was taken as the average of the two measurements to determine the data volume;

[0058] 6. Use DHI routine equipment to test milk samples and obtain mid-infrared spectral data. Specifically, the samples were heated in a 42°C constant temperature water bath for 15-20 minutes and analyzed using a Milkoscan TM FT+milk analyzer (Foss Electric A / S, Denmark) for measurement, and the operation method refers to the industry standard "Technical Specifications for Measurement of Dairy Cow Production Performance in China" (NY / T 2007-0415).

[0059] 7. Keep 2976.06-2767.89cm in all spectral data -1 、1800.28-1684.63cm -1 and 1607.53-925.2cm -1 The bands related to milk fat. The original spectral data were processed using the first-order derivative (FD) to obtain the processed values ​​of the spectral data of each band.

[0060] 8. According to individual parity and postpartum lactation days, grouped data of parity and postpartum weeks were obtained. Parity was divided into 3 levels, namely: 1, 2, and "3 parities and above" were each in 1 group, for a total of 3 levels; postpartum weeks were divided into 7-day groups, namely: 1-7 days, 8-14 days, 15-21 days, 22-28 days, and 29-35 days were each in 1 group, for a total of 7 levels.

[0061] 9. Using the R software system, the LassoR model program was used to perform regression analysis. The regression equation was Y ijmmn =PA i +DIM j +FD+LassoR.

[0062] Among them, Y ij is the predicted value of NEFA of the jth lactation day group of the i-th parity group of the individual, PA i For the i-th parity group, DIM j is the j-th lactation day group, FD is the first-order derivative processed value of the original spectrum, and LassoR is the model algorithm.

[0063] 2. Data Processing and Model Analysis

[0064] 1. Calculate the average of the two NEFA measurements, remove the individuals whose difference between the two measurements is greater than 15% of the average, and use the average of the two measurements as the final NEFA test value of the individual, retaining a total of 366 sample values.

[0065] 2. Keep 2976.06-2767.89cm in all spectral data -1 、1800.28-1684.63cm -1and 1607.53-925.2cm -1 Bands related to milk fat. The original spectral data were processed using four spectral data processing methods, namely standard normal transformation (SNV), first-order derivative (FD), second-order derivative (SD), and convolution smoothing (SG), with a window length of 5 and a polynomial order of 3 to obtain the processed values ​​of the spectral data of each band.

[0066] 3. According to individual parity and postpartum lactation days, the fixed effect data of parity and postpartum weeks were obtained. The parity was divided into 3 levels, namely: 1, 2, "3 parities and above" were each 1 group, a total of 3 levels; the postpartum weeks were divided into a group every 7 days, namely: 1-7 days, 8-14 days, 15-21 days, 22-28 days, 29-35 days were 1 group, 2 groups, etc., a total of 7 levels.

[0067] 4. Using the R software system, regression analysis was performed using four model programs: partial least squares (PLSR), Lasso regression (LassoR), ridge regression (RidgeR) and random forest (RFR). The general form of the regression equation is Y ijmmn =PA i +DIM j +MIR m +M n +e.

[0068] Among them, Y ijmmn is the NEFA detection value corresponding to the nth model algorithm of the mth spectral processing method in the jth lactation day group of the i-th parity group, PA i For the i-th parity group, DIM j is the j-th lactation day group, MIR m The spectral value of the mth spectral processing method, M n is the nth model algorithm, and e is the random effect.

[0069] 5. To improve the accuracy of the analysis, the data were analyzed using 10-fold cross validation, that is, the data were randomly allocated 7:3, of which 7 folds were used for model analysis and 3 folds were used for model validation.

[0070] 6. Through the coefficient of determination (R 2 ), root mean square error (RMSE) and relative predictive deviation (RPD) were used to evaluate and screen the models. 2It is the degree of fit between the model prediction value and the actual detection value. The larger the value, the better the model fits, otherwise it means the model fits poorly; RMSE is the deviation between the prediction value and the actual detection value. The larger the value, the worse the model fits and the lower the accuracy. Otherwise, it means the model fits well and has high accuracy; RPD is the ratio of the sum of squares of the actual detection value to the model prediction difference. The larger the value, the better the model fits, otherwise it means the model fits poorly. It is generally believed that when RPD ≥ 2, it means that the prediction model has excellent prediction effect and can be used for quantitative prediction. When the RPD value is 1.4 to 2.0, it means that the model has good prediction effect and can be used for rough quantitative prediction or qualitative analysis. When RPD < 1.4, the model prediction effect is poor and the accuracy is low, and it cannot be used for quantitative prediction. The calculation formulas are as follows:

[0071]

[0072] In the formula, y i represents the observed value; is the model prediction value; Yes i The average value of; n represents the number of samples.

[0073]

[0074] In the formula, y i represents the actual observed value of the sample; is the model predicted value of the sample; n represents the number of samples.

[0075]

[0076] Where n is the total number of samples; x i is the actual observed value of the sample; μ represents the overall mean of the sample; SD represents the standard deviation of the observed value; RMSE represents the root mean square error.

[0077] Results showed that the coefficient of determination (R 2 ), root mean square error (RMSE), and relative analytical error (RPD), with R 2 The optimal model is selected by taking the maximum, the minimum RMSE and the maximum RPD as the criteria. It can be seen that the regression fitting equations of the combination of PA-DIM-FD-LassoR and FD-LassoR achieve good results, among which the combination of PA-DIM-FD-LassoR is the best. Therefore, Y can be selected. ij =PA i +DIM j The +FD+LassoR regression equation was used as the prediction equation for blood NEFA.

[0078] The present invention provides a method for predicting the free fatty acid concentration in the blood of dairy cows based on DHI mid-infrared spectrum, comprising:

[0079] 1. Sampling: Consider the impact of factors such as parity and lactation days on the prediction model; leave some time interval between blood sampling and milk sampling;

[0080] 2. Data collection: milk samples were used to obtain MIR spectral data; blood samples were centrifuged to obtain serum and the concentration of NEFA was measured and recorded as the reference value;

[0081] 3. Spectral pre-processing: select some bands for subsequent analysis, standardize the spectrum first, and then use SNV, FD, SD, and SG methods to reduce noise and smooth the spectrum;

[0082] 4. Model fitting: Match the MIR spectral data with the reference value data, and use four algorithms: PLSR, LassoR, RidgeR, and RFR to fit the model;

[0083] 5. Model validation.

[0084] The present invention will be further described in detail below in conjunction with specific embodiments. It should be noted that the specific embodiments are intended to explain the present invention rather than to limit it.

[0085] Example 1

[0086] The present invention collects relevant data in a large-scale dairy farm in Shandong Province that participates in DHI determination, collects milk samples and blood samples, and determines mid-infrared spectral data and NFEA values.

[0087] Specific steps:

[0088] 1. Collect milk and blood samples from the experimental herd (38 cows), and record the individual parity and the number of days of lactation when sampling. The milk sampling method refers to the "Technical Specifications for Production Performance Measurement of Chinese Holstein Cattle" (NY / T 1450-2007). When collecting milk samples, about 40 mL per cow, 0.03 g of potassium dichromate is added to the bottle as a preservative before sampling, and the milk samples are stored at room temperature; 5 mL of procoagulant blood is collected from each cow, placed at room temperature for 1-2 hours to promote coagulation, and then centrifuged (parameters: 2500g, 15min, 4℃), serum is separated, and serum samples are frozen and stored at -20℃ for testing.

[0089] 2. The acyl-CoA synthetase-acyl-CoA oxidase (ACS-ACOD) method (Lab Assay TMNEFA) was used to detect the content of NEFA in serum. Each sample was measured twice. After eliminating the values ​​with a coefficient of variation greater than 15% between the two measurements, the final value was taken as the average of the two measurements. The final data volume was determined to be 32;

[0090] 3. The DHI laboratory of Shandong Academy of Agricultural Sciences used the Milkoscan TM FT+milk analyzer (Foss Electric A / S, Denmark) measured milk samples and obtained mid-infrared spectral data.

[0091] 4. Keep 2976.06-2767.89cm in all spectral data -1 、1800.28-1684.63cm -1 and 1607.53-925.2cm -1 For the bands related to milk fat, the original spectral data were processed using the first-order derivative (FD) to obtain the processed values ​​of the spectral data of each band.

[0092] 5. According to individual parity and postpartum lactation days, the fixed effect data of parity and postpartum weeks were obtained. The parity was divided into 3 levels, namely: 1, 2, 3 and above parities each as 1 group, a total of 3 levels; the postpartum weeks were divided into 7 groups every 7 days, namely: 1-7 days, 8-14 days, 15-21 days, 22-28 days, 29-35 days each as 1 group, a total of 7 levels.

[0093] 6. Using the R software system, the LassoR model program was used to perform regression analysis. The regression equation was Y ijmmn =PA i +DIM j +FD+LassoR, the predicted values ​​of NEFA of 32 individuals were obtained.

[0094] Among them, Y ij is the predicted value of NEFA of the jth lactation day group of the i-th parity group of the individual, PA i For the i-th parity group, DIM j is the j-th lactation day group, FD is the first-order derivative processed value of the original spectrum, and LassoR is the model algorithm.

[0095] Table 1 Comparison of the results of the two methods

[0096]

[0097] (8) SAS8.2 TEST program was used to perform a significance test analysis to compare the significance of the two test results. There was no significant difference between the model prediction method and the ACS-ACOD classical chemical method (P = 0.5689).

[0098] Conclusion: The significance test was performed using the PROC TEST program of SAS8.2. The results showed that there was no significant difference between the model prediction method and the ACS-ACOD classical chemical method (P>0.05). The model prediction method can achieve the accuracy of the ACS-ACOD classical chemical method and can be used in production. In addition, the method of the present invention greatly reduces the workload of the actual production process of cattle farms by applying daily DHI data for large-group detection.

[0099] Although the present invention is based on the research results obtained from cattle farms in Shandong Province, it can be widely used on a large scale. The reason is that Holstein cows around the world have a relatively consistent large-scale breeding method and population basis, and the present invention can achieve the effects described in the present invention by being widely used on a large scale.

[0100] Example 2

[0101] 1. Sampling: A total of 405 Chinese Holstein dairy cows (117 in group A, 28 in group B, 145 in group C, and 115 in group D) were randomly sampled, ranging from parity 1 to 3, and lactation days between 1 and 7 weeks. Blood samples were collected between 6 and 10 a.m., and milk samples were collected between 12 and 4 p.m. on the same day, with a delay between the two samplings.

[0102] 2. Data acquisition: Blood samples were collected from all cows after returning to the cowshed after milking in the morning. Blood was collected from the heel vein of the cow using a 5 mL coagulant tube. The blood was placed at room temperature for a period of time to promote coagulation and then placed on ice. After returning to the laboratory, the blood samples were transferred to a 4-degree refrigerator and centrifuged overnight (2500 g, 15 min, 4°C) to separate serum. The serum samples were frozen and stored at -20°C until analysis. The ACS-ACOD method (LabAssay) was used according to the manufacturer's instructions. TM The NEFA content in serum was measured by ELISA and recorded as the reference value. Each sample was measured twice. After eliminating the values ​​with a coefficient of variation greater than 15%, the final value was taken as the average of the two measurements. The data volume was also reduced from 405 to 366.

[0103] Mid-infrared analysis was performed in the DHI laboratory of the Animal Husbandry and Veterinary Research Institute of Shandong Academy of Agricultural Sciences. The milk samples were heated in a constant temperature water bath at 42°C for 15 to 20 minutes before being put into the machine. The Milkoscan TM FT+milk analyzer (FossElectric A / S, Denmark) to collect spectra, select two consecutive detection modes, and the final value is the average of the two detections. The mid-infrared spectrum of each sample represents the wave number range from 5007.6 to 925.2 cm -1The instrument also provides some other information, including the percentage of milk fat, total protein, lactose, total solids, non-fat milk solids, urea, acetone and beta-hydroxybutyric acid in milk.

[0104] 3. Spectral preprocessing: Import the infrared spectral information of milk, serum NEFA concentration, instrument prediction data, and the corresponding cow information including cow number, parity, and lactation days into R (version 4.2.1; R Core Team, 2022). The role of the cow number is to associate all the information. Select 2976.06-2767.89cm -1 、1800.28-1684.63cm -1 and 1607.53-925.2cm -1 The spectral region is used for subsequent analysis, and the remaining spectral regions are considered to be greatly affected by water molecules or do not contain spectral information related to milk components. The reference value is tested for data normal distribution by Shapiro test, and the data is normalized by logarithmic transformation in the case of non-normality. The prediction model for serum NEFA concentration based on mid-infrared spectroscopy is further referred to as a "pure spectral model", while the prediction model based on mid-infrared spectroscopy and other information (parity, lactation days and percentage of milk fat, total protein, lactose, total solids, non-fat milk solids, urea, acetone and β-hydroxybutyric acid in milk) is further referred to as a "non-pure spectral model". According to the 7:3 principle, the samples are randomly divided into training and test sets using the createDataPartition function. It is worth noting that the two sets of data need to be adjusted to have similar descriptive statistics. In order to eliminate the influence of various factors in the environment and the factors existing in the instrument itself, the collected spectral information needs to be preprocessed before modeling. In the present invention, the present invention uses five algorithms to preprocess the spectrum, including standardization, SNV, FD, SD, SG (spectral window length is 5, polynomial degree is 3).

[0105] 4. Model fitting: In the present invention, four algorithms, PLSR, LassoR, RidgeR and RFR, are used to establish prediction models.

[0106] 5. Model validation: All four algorithms use 10-fold cross validation to select the model with the best performance as the final prediction model. The indicators for evaluating the prediction model include the determination coefficient (R 2 cv) and the coefficient of determination (R 2v), root mean square error in cross validation (RMSEcv), root mean square error in external validation (RMSEv), relative analytical error in cross validation (RPDcv), and relative analytical error in external validation (RPDv). The model with the best performance was the SG-LassoR model based on MIR spectra.

[0107] Table 2: Evaluation results of model combinations for predicting free fatty acids in bovine serum based on MIR spectra using 16 algorithms combined with other information.

[0108] Table 2 Model combination evaluation results

[0109]

[0110]

[0111] Note: PA in the table is parity grouping, DIM is lactation days grouping; SNV is standard normal transformation, FD is the first-order derivative, SD is the second-order derivative, SG is convolution smoothing; R 2 cv is the coefficient of determination and RMSE of the training set cv is the root mean square error of the training set, RPD cv is the relative analysis error of the training set, R 2 v The coefficient of determination and RMSE of the validation set v Validation set root mean square error, RPD v Relative analysis error of the validation set.

[0112] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting the concentration of free fatty acids in cow blood based on DHI mid-infrared spectroscopy, characterized in that: include: Collect milk and blood samples from dairy cows, collect data on the cows’ ear numbers, parity, and days postpartum; Detection of free fatty acid levels in serum; Perform DHI test on milk samples to obtain mid-infrared spectrum data; Retain the bands related to milk fat in all spectral data to obtain original spectral data; The original spectral data is processed using the first-order derivative to obtain the processed value of the spectral data of each band; According to parity and days of lactation, grouped data of parity and postpartum weeks were obtained; The Lasso regression model program was used for regression analysis, and the free fatty acid concentration was output.

2. The method for predicting the free fatty acid concentration in cow blood based on DHI mid-infrared spectroscopy according to claim 1, characterized in that: The dairy cows are periparturient dairy cows.

3. The method for predicting the free fatty acid concentration in cow blood based on DHI mid-infrared spectroscopy according to claim 1, characterized in that: The wavelength band related to milk fat is: 2976.06-2767.89cm -1 、1800.28-1684.63cm -1 and 1607.53-925.2cm -1 .

4. The method for predicting the free fatty acid concentration in cow blood based on DHI mid-infrared spectroscopy according to claim 1, characterized in that: Parity was divided into three levels, namely, 1st parity, 2nd parity, and 3rd parity and above, each with 1 group.

5. The method for predicting the free fatty acid concentration in cow blood based on DHI mid-infrared spectrum according to claim 1, characterized in that: The postpartum weeks were divided into groups of 7 days each, i.e., 1-7 days, 8-14 days, 15-21 days, 22-28 days, and 29-35 days were each a group, for a total of 7 levels.

6. The method for predicting the free fatty acid concentration in cow blood based on DHI mid-infrared spectrum according to claim 1, characterized in that: The regression model formula is ij =PA i +DIM j +FD+LassoR, where Y ij is the predicted value of NEFA of the jth lactation day group of the i-th parity group of the individual, PA i is the i-th parity group effect, DIM j is the group effect of the j-th lactation day, FD is the first-order derivative processed value of the original spectrum, and LassoR is the model algorithm.

7. The method for predicting the free fatty acid concentration in cow blood based on DHI mid-infrared spectrum according to claim 1, characterized in that: The original data were processed by using the first-order derivative to replace the standard normal transformation, the second-order derivative, and the convolution smoothing method.

8. The method for predicting the free fatty acid concentration in cow blood based on DHI mid-infrared spectrum according to claim 1, characterized in that: The data were analyzed using 10-fold cross validation.

9. The method for predicting the free fatty acid concentration in cow blood based on DHI mid-infrared spectrum according to claim 1, characterized in that: The data were randomly divided into 7:3, 7 folds were used for model analysis and 3 folds were used for model validation.

10. A system for predicting the concentration of free fatty acids in cow blood based on DHI mid-infrared spectroscopy, characterized in that: include: A module for collecting milk and blood samples from cows, and collecting data on the cows’ ear numbers, parity, and days after calving; A module for detecting the content of free fatty acids in serum; A module for performing DHI detection on milk samples and obtaining mid-infrared spectral data; A module for retaining the bands related to milk fat in all spectral data to obtain the original spectral data; A module for processing the original spectral data using the first-order derivative to obtain the processed value of the spectral data of each band; Module for obtaining grouped data on parity and weeks postpartum according to parity and days of lactation; Module for outputting free fatty acid concentrations from regression analysis using the Lasso regression model procedure.