A method for predicting offspring growth traits based on the rumen flora structure of maternal ruminants

By sequencing microbial DNA of rumen juice of maternal ruminants and constructing random forest models, the non-invasive problem of predicting growth traits in young ruminants was solved, and efficient and accurate prediction of growth traits in offenders was achieved, avoiding the development of young ruminants. Animal damage.

CN113643750BActive Publication Date: 2024-11-29ZHEJIANG UNIV
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202110911708.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-08-09
Publication Date
2024-11-29
Estimated Expiration
2041-08-09

AI Technical Summary

Technical Problem

The prior art is difficult to effectively predict the growth traits of young ruminants through non-invasive methods, especially due to the fragile esophageal tract of young animals, and it is difficult to perform microbial determination by conventional oral collection of rumen fluid.

Method used

The total microbial DNA from the rumen fluid of maternal ruminants was extracted for 16S rDNA high-throughput sequencing, the rumen flora composition was analyzed, and a random forest biclassification model was constructed to predict the growth traits of the offspring.

Benefits of technology

It realizes efficient and accurate prediction of the growth traits of progeny ruminants, avoids damage to young animals by direct sampling, and shows high accuracy and sensitivity.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113643750B_ABST
    Figure CN113643750B_ABST
Patent Text Reader

Abstract

The present invention provides a method for predicting offspring growth traits based on the rumen flora structure of maternal ruminants, comprising the following steps: step 1, collecting offspring ruminant growth trait data and collecting maternal ruminant rumen fluid samples; step 2, extracting maternal ruminant rumen microbial DNA from the maternal ruminant rumen fluid samples, and performing 16S rDNA high-throughput sequencing; step 3, analyzing the results of 16S rDNA high-throughput sequencing; step 4, constructing a machine learning model based on the offspring ruminant growth trait data and the analysis results of the maternal ruminant rumen fluid samples, and predicting offspring growth traits using the machine learning model. The present invention analyzes the flora structure of the rumen of the maternal ruminant to be tested by 16S rDNA, and based on the construction of a machine learning model, achieves efficient and accurate prediction of offspring ruminant growth traits, avoiding damage to young ruminants caused by direct sampling.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a method for predicting growth traits of animal offspring, and in particular to a method for predicting growth traits of offspring based on the rumen flora structure of maternal ruminants. Background Art

[0002] Ruminants are important livestock (cattle, sheep, etc.) that provide humans with a large amount of important agricultural products such as meat, milk, leather, and wool. Ruminants are able to utilize plant fibers in the diet that are difficult for other mammals to digest, which is mainly due to their unique digestive organ, the rumen. The presence of a complex microbial community in the rumen helps the rumen perform its digestive physiological functions and is the main digestive organ of ruminants. Rumen microorganisms secrete a variety of digestive enzymes that act on polysaccharides such as cellulose, hemicellulose, and pectin in food, and the volatile fatty acids produced provide 60-70% of the energy source for the ruminant host. A large number of studies have shown that rumen microorganisms are highly correlated with the feed digestibility, metabolism, and health status of ruminants, and the structure of the rumen microbial community can be used to effectively predict animal feed efficiency.

[0003] Therefore, rumen flora can be used to predict the growth traits of young ruminants at an early stage. However, the esophagus of young animals is fragile, and it is difficult to measure rumen microorganisms by conventional oral collection of rumen fluid. Studies have shown that microorganisms from young animal mothers are closely related to the initial growth and development of offspring, and ruminant rumen microorganisms have certain genetic characteristics. Therefore, the use of maternal microbial structure can effectively predict the growth traits of young ruminants. Summary of the invention

[0004] The purpose of the present invention is to provide a method for predicting offspring growth traits based on the rumen flora structure of maternal ruminants. The method extracts the total microbial DNA in the rumen fluid of maternal ruminants, sequences and analyzes the rumen flora composition, and establishes a random forest binary classification model with ASV abundance as a variable to achieve the prediction of offspring growth traits.

[0005] The technical solution adopted by the present invention to solve its technical problem is:

[0006] A method for predicting offspring growth traits based on the rumen flora structure of maternal ruminants, the method comprising the following steps:

[0007] Step 1, collecting growth trait data of offspring ruminants and collecting rumen fluid samples of maternal ruminants;

[0008] Step 2, extracting maternal ruminant rumen microbial DNA from the maternal ruminant rumen fluid sample, and performing 16S rDNA high-throughput sequencing;

[0009] Step 3, analyzing the results of the 16S rDNA high-throughput sequencing;

[0010] Step 4: construct a machine learning model based on the offspring ruminant growth trait data and the analysis results of the maternal ruminant rumen fluid sample, and use the machine learning model to predict the offspring growth traits.

[0011] Preferably, in step 1, the growth trait data of the offspring ruminants include the birth weight and weaning weight of the offspring ruminants.

[0012] Preferably, in step 2, the method for extracting the rumen microbial DNA of the maternal ruminant is the CTAB method.

[0013] Preferably, in step 2, the fragment region for 16S rDNA high-throughput sequencing is the V3-V4 region, which is amplified using primers 341F and 806R.

[0014] Preferably, in step 3, the analysis method is amplicon sequence variation (ASV) analysis.

[0015] Preferably, in step 3, the analysis method comprises the following steps:

[0016] Step 3.1, remove barcodes and primer sequences to obtain the true sequence of the sample;

[0017] Step 3.2, generate the ASV table after filtering, trimming and merging the real sequences of the samples;

[0018] Step 3.3, remove chimeras, archaea, and chloroplast information, and only keep bacterial sequences;

[0019] In step 3.4, each sample sequence is flattened to a depth of 10,000 sequences.

[0020] Preferably, in step 4, the machine learning model is a random forest model.

[0021] Preferably, in step 4, constructing the machine learning model includes the following steps:

[0022] Step 4.1, define the characteristic variables of growth trait groups and models;

[0023] Step 4.2, use the RandomForestClassifier function to build a random forest model;

[0024] Step 4.3, use variable selection to select valid feature variables from the candidate feature variables, and set the threshold of valid feature variables to 1e-3;

[0025] Step 4.4, adjust the number of decision trees, the maximum number of features, and the maximum depth of the decision tree in the random forest model in turn, and select the best parameters to improve the model accuracy and prevent overfitting;

[0026] In step 4.5, the accuracy of the model is verified by 10-fold cross validation, and the random forest model is evaluated by ROC curve. After the evaluation is qualified, an effective machine learning model is obtained.

[0027] Preferably, in step 4.1, the method for defining growth trait groups is: taking the weaning weight or weaning weight gain of offspring ruminants as the growth trait, at each litter size level, defining the top 25% of offspring ruminants in growth shape as a high trait group, and the bottom 25% of offspring ruminants in growth shape as a low trait group.

[0028] Preferably, in step 4.1, the characteristic variable of the model is the relative abundance of the parent bacterial ASV after logarithmic transformation.

[0029] The beneficial effects of the present invention are:

[0030] The present invention discloses a method for predicting offspring growth traits based on the rumen flora structure of maternal ruminants. The method analyzes the flora structure of the rumen of the maternal ruminants to be tested by 16S rDNA and builds a machine learning model to achieve efficient and accurate prediction of the growth traits of offspring ruminants, thereby avoiding damage to young ruminants caused by direct sampling. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0032] Figure 1 is a ROC curve diagram of the weaning weight gain prediction model for Hu sheep lambs in Example 2 of the present invention;

[0033] Figure 2 It is a ROC curve diagram of the Hu sheep lamb weaning weight prediction model in Example 2 of the present invention. DETAILED DESCRIPTION

[0034] The technical solution of the present invention is further described in detail below through specific embodiments and in conjunction with the accompanying drawings. It should be understood that the implementation of the present invention is not limited to the following embodiments, and any form of modification and / or change made to the present invention will fall within the protection scope of the present invention.

[0035] In the present invention, unless otherwise specified, all parts and percentages are weight units, and the equipment and raw materials used can be purchased from the market or are commonly used in the art. The methods in the following embodiments, unless otherwise specified, are conventional methods in the art. The components or equipment in the following embodiments, unless otherwise specified, are universal standard parts or components known to those skilled in the art, and their structures and principles are known to those skilled in the art through technical manuals or conventional experimental methods.

[0036] Embodiment 1:

[0037] A method for predicting offspring growth traits based on the rumen flora structure of maternal ruminants, the method comprising the following steps:

[0038] Step 1, collecting growth trait data of offspring ruminants and collecting rumen fluid samples of maternal ruminants;

[0039] Step 2, extracting rumen microbial DNA of the maternal ruminant from the rumen fluid sample of the maternal ruminant, and performing 16S rDNA high-throughput sequencing;

[0040] Step 3, analyzing the results of 16S rDNA high-throughput sequencing;

[0041] Step 4: construct a machine learning model based on the growth trait data of the offspring ruminants and the analysis results of the rumen fluid sample of the maternal ruminants, and use the machine learning model to predict the growth traits of the offspring.

[0042] Through the above technical scheme, the flora structure of the rumen of the maternal ruminant to be tested is analyzed by 16S rDNA, and based on the construction of a machine learning model, efficient and accurate prediction of the growth traits of the offspring ruminants is achieved, avoiding the damage to young ruminants caused by direct sampling.

[0043] Embodiment 2:

[0044] A method for predicting offspring growth traits based on the rumen flora structure of maternal ruminants, the method comprising the following steps:

[0045] Step 1, collecting growth trait data of offspring ruminants and collecting rumen fluid samples of maternal ruminants;

[0046] Step 2, extracting rumen microbial DNA of the maternal ruminant from the rumen fluid sample of the maternal ruminant, and performing 16S rDNA high-throughput sequencing;

[0047] Step 3, analyzing the results of 16S rDNA high-throughput sequencing;

[0048] Step 4: construct a machine learning model based on the growth trait data of the offspring ruminants and the analysis results of the rumen fluid sample of the maternal ruminants, and use the machine learning model to predict the growth traits of the offspring.

[0049] In step 1, the offspring ruminant growth trait data include the birth weight and weaning weight of the offspring ruminants.

[0050] In step 2, the method for extracting the rumen microbial DNA of the maternal ruminant is the CTAB method; the fragment region for 16S rDNA high-throughput sequencing is the V3-V4 region, which is amplified using primers 341F and 806R.

[0051] In step 3, the analysis method is amplicon sequence variation (ASV) analysis, comprising the following steps:

[0052] Step 3.1, remove barcodes and primer sequences to obtain the true sequence of the sample;

[0053] Step 3.2, generate the ASV table after filtering, trimming and merging the real sequences of the samples;

[0054] Step 3.3, remove chimeras, archaea, and chloroplast information, and only keep bacterial sequences;

[0055] In step 3.4, each sample sequence is flattened to a depth of 10,000 sequences.

[0056] In step 4, the machine learning model is a random forest model. Building a random forest model includes the following steps:

[0057] Step 4.1, define the characteristic variables of growth trait groups and models;

[0058] Step 4.2, use the RandomForestClassifier function to build a random forest model;

[0059] Step 4.3, use variable selection to select valid feature variables from the candidate feature variables, and set the threshold of valid feature variables to 1e-3;

[0060] Step 4.4, adjust the number of decision trees, the maximum number of features, and the maximum depth of the decision tree in the random forest model in turn, and select the best parameters to improve the model accuracy and prevent overfitting;

[0061] In step 4.5, the accuracy of the model is verified by 10-fold cross validation, and the random forest model is evaluated by ROC curve. After the evaluation is qualified, an effective machine learning model is obtained.

[0062] In step 4.1, the method for defining growth trait groups is as follows: taking the weaning weight or weaning weight gain of offspring ruminants as the growth trait, at each litter size level, defining the offspring ruminants in the top 25% of growth shape as the high trait group, and the offspring ruminants in the bottom 25% of growth shape as the low trait group; the characteristic variable of the model is the relative abundance of maternal bacterial ASV after logarithmic transformation.

[0063] According to the above method, taking Hu sheep as an example, the growth traits of Hu sheep lambs are predicted based on the maternal Hu sheep. The specific examples are as follows.

[0064] 1. Data collection on growth traits of young ruminants

[0065] The growth trait data of young ruminants were collected from 179 Huyang female lambs in China's national mutton breeding core farm (Zhejiang Sino Ecological Agriculture Co., Ltd.). The lambs were raised in the same pen with their mothers until weaning on the 45th day. The birth weight of the lambs was measured at birth, and the weaning weight was measured at the 45th day of weaning. During this period, they were free to eat drinking water and pellets. The results showed that the birth weight of the lamb group was 3.65±0.71kg, and the weaning weight was 14.10±1.72kg.

[0066] 2. Collection of Rumen Fluid Samples from Maternal Ruminants

[0067] (1) Sample source: Rumen fluid samples were collected from 128 female Hu sheep aged 1032.7±424.62 days from a national core farm of mutton sheep breeding in China (Zhejiang Sino Ecological Agriculture Co., Ltd.). Hu sheep were raised in pens under the same feeding conditions and had free access to silage and drinking water.

[0068] (2) Rumen fluid collection: Rumen fluid was collected orally using a rumen fluid collection tube (A1320K, Kelibo Animal Husbandry Technology Co., Ltd.) before morning feeding 59.7±1.55 days after delivery of the female Hu sheep. The first 20 mL of rumen fluid was discarded to avoid saliva contamination, and the collected rumen fluid samples were quickly frozen in liquid nitrogen and then stored at -80°C.

[0069] 3. Rumen microbial DNA extraction and 16S high-throughput sequencing

[0070] (1) DNA extraction: The total DNA of maternal rumen organisms was extracted using the CTAB method. The frozen rumen fluid samples were thawed on ice, and 0.2-0.3 g of the sample was added to a 2 mL grinding tube containing 1 mL of 2% CTAB solution, 0.3 g of sterile alumina zirconium beads (0.1 mm) and 2 steel beads (2 mm). The mixture was physically broken by a grinder (JXFSTPRP-24, Shanghai Jingxin Industrial Development Co., Ltd.), running at 65 Hz for 30 s, pausing for 10 s, and repeated three times. Then centrifuged at 4°C, 14000 rpm for 15 min. The supernatant was transferred to a new 2 mL centrifuge tube, and an equal volume of PCI solution (phenol chloroform 25:24:1) was added and the mixture was shaken vigorously on an oscillator for 30 s. Centrifuged at room temperature at 14000 rpm for 10 min. The supernatant was transferred to a new centrifuge tube (1.5 mL), 0.6-0.8 times the volume of isopropanol was added, and the mixture was gently mixed by inverting. -20℃ overnight or -80℃ for 2 hours. Centrifuge at 14000rpm for 30min at room temperature. Discard the supernatant and a white DNA precipitate will be visible. Wash the precipitate with 1ml 70% ethanol and blow up the precipitate from the bottom of the tube. Centrifuge at 14000rpm for 30min at 4℃ and discard the supernatant. Dry the precipitate. Dissolve the DNA in 50μL sterile water. Add RNase to a final concentration of 0.1mg / ml. Incubate in a 37℃ water bath for 15min. Detect the quality of the DNA using 1% agarose gel electrophoresis. The extracted DNA solution is stored at -20℃.

[0071] (2) 16S rDNA sequencing: Primer 341F (5'-CCTAYGGGRBGCASCAG-3', SEQ ID No. 1) and primer 806R (5'-GGACTACNNGGGTATCTAAT-3', SEQ ID No. 2) were used to amplify the V3-V4 hypervariable region of the 16S rDNA sequence, which has 6 bp sample label barcodes. The second-generation high-throughput sequencing of the samples was performed by Beijing Novogene Technology Co., Ltd. using the Illumina NovaSeq 6000 platform 2×250 paired-end sequencing.

[0072] IV. 16S rDNA Sequencing Data Analysis

[0073] The original sequences were grouped according to the tag barcodes, and the barcodes and primer sequences were removed. The sequencing results were then analyzed for amplicon sequence variants (ASV) using the DADA2 package (version 1.16) in R software (version 4.0.2) based on a patterned process (http: / / benjjneb.github.io / dada2 / tutorial.html). That is, the sequences of each sample were filtered, trimmed, and merged to generate an ASV table and imported into the phyloseq package (version 1.34.0) to remove chimeras, archaea, and chloroplast information, retaining only bacterial sequences, and flattening the sequences of each sample to a depth of 10,000 sequences.

[0074] 5. Construction of a machine learning model to predict offspring growth traits based on maternal rumen bacterial flora structure

[0075] A random forest binary classification model was established using samples of known offspring weaning weight and weaning weight gain in the target population, and this model was used to determine whether the offspring of other individuals belonged to high growth shape individuals or low growth trait individuals (growth traits can be weaning weight, weaning weight gain, etc.). The process of building a model using known data is: at each litter size level, the top 25% of growth traits are defined as the high trait group, and the bottom 25% are defined as the low trait group. The variables of the model are the relative abundance of maternal bacterial ASVs after logarithmic transformation. The RandomForestClassifier function in the Sklearn package of python (version 0.23.2, https: / / scikit-learn.org / stable / ) was used to construct a random forest model. Through bootstrap sample resampling, n samples were randomly selected repeatedly with replacement from the training sample set N to generate a new training sample set to train the decision tree, and then m decision trees were generated according to the above steps to form a random forest, and the category with the most votes from the m classification trees was designated as the final output.

[0076] After calculating the Gini index (Gini value) of the selected features, they are sorted from large to small and used as the classification features of the root node, intermediate node and leaf node of the decision tree respectively.

[0077]

[0078] Among them, Gini (D, A m ) is divided into D according to the candidate feature A (logarithmically transformed ASV relative abundance) 1 (high production trait) and D 2 (Low production traits) Gini index of two categories, |D 1 |、|D 2| and |D| are respectively the set D 1 , D 2 , the number of samples in D.

[0079] Then use the variable selection function feature_selection to select the effective feature A variable (threshold = 1e-3). Use the param_grid function to adjust the n_estimators, max_features and max_depth parameters in turn. Among them, n_estimators is the classifier in the random forest, that is, the number of decision trees; max_features is the maximum number of features selected when building a decision tree; max_depth is the maximum depth of the decision tree. Select the best parameters to improve the accuracy of the model and prevent overfitting. Evaluate the model with the accuracy of the model under ten-fold cross validation and the ROC curve graph.

[0080] After obtaining an effective model, input the maternal rumen ASV information to be predicted to obtain the prediction results for its offspring. The average accuracy of judging high and low weaning weight gain under ten-fold cross validation is 0.7986, and the average accuracy of judging high and low weaning weight is 0.7889. The ROC curve shows that Figure 1 As shown in Figure 2, the area under the curve (AUC) of the random forest model for weaning weight gain is 0.89. Figure 2 As shown, the AUC of the random forest model for weaning weight is 0.88. The above results show that the method for predicting offspring growth traits based on the rumen flora structure of maternal ruminants in this application can effectively predict offspring weaning weight gain and weaning weight growth traits with high accuracy, sensitivity and specificity.

[0081] The above-described embodiment is only a preferred solution of the present invention and does not limit the present invention in any form. There are other variations and modifications without exceeding the technical solution described in the claims. Sequence Listing <110> Zhejiang University <120> A method for predicting offspring growth traits based on the rumen flora structure of maternal ruminants <130> ZJWL-WJK202102 <160> 2 <170> SIPOSequenceListing 1.0 <210> 1 <211> 17 <212> DNA <213> Synthetic (341F) <400> 1 cctaygggrb gcascag 17 <210> 2 <211> 20 <212> DNA <213> Synthetic (806R) <220> <221> misc_feature <222> (8)..(9) <223> n is a, c, g, t or u <400> 2 ggactacnng ggtatctaat 20

Claims

1. A method for predicting offspring growth traits based on the rumen flora structure of maternal Hu sheep, characterized in that: The method comprises the following steps: Step 1, collecting growth trait data of offspring Hu sheep and collecting rumen fluid samples of maternal Hu sheep, the growth trait data of offspring Hu sheep including birth weight and weaning weight of offspring Hu sheep; Step 2, extracting maternal Hu sheep rumen microbial DNA from the maternal Hu sheep rumen fluid sample, and performing 16S rDNA high-throughput sequencing, the method for extracting maternal Hu sheep rumen microbial DNA is the CTAB method, the fragment region of 16S rDNA high-throughput sequencing is the V3-V4 region, and primers 341F and 806R are used for amplification; Step 3, analyzing the results of the 16S rDNA high-throughput sequencing by an amplicon sequence variation (ASV) analysis; Step 4, constructing a machine learning model based on the offspring Hu sheep growth trait data and the analysis results of the maternal Hu sheep rumen fluid sample, and using the machine learning model to predict the offspring growth traits; In step 3, the analysis method comprises the following steps: Step 3.1, remove barcodes and primer sequences to obtain the true sequence of the sample; Step 3.2, generate the ASV table after filtering, trimming and merging the real sequences of the samples; Step 3.3, remove chimeras, archaea, and chloroplast information, and only keep bacterial sequences; Step 3.4: flatten each sample sequence to a depth of 10,000 sequences; In step 4, the machine learning model is a random forest model, and constructing the machine learning model includes the following steps: Step 4.1, define the growth trait groups and the characteristic variables of the model; the method for defining the growth trait groups is: taking the weaning weight or weaning weight gain of the offspring Hu sheep as the growth trait, at each litter size level, define the offspring Hu sheep in the top 25% of the growth shape as the high trait group, and the offspring Hu sheep in the bottom 25% of the growth shape as the low trait group; the characteristic variable of the model is the relative abundance of maternal bacterial ASV after logarithmic transformation; Step 4.2, use the RandomForestClassifier function to build a random forest model; Step 4.3, use variable selection to select valid feature variables from the candidate feature variables, and set the threshold of valid feature variables to 1e-3; Step 4.4, adjust the number of decision trees, the maximum number of features, and the maximum depth of the decision tree in the random forest model in turn, and select the best parameters to improve the model accuracy and prevent overfitting; In step 4.5, the accuracy of the model is verified by 10-fold cross validation, and the random forest model is evaluated by ROC curve. After the evaluation is qualified, an effective machine learning model is obtained.

Citation Information

Patent Citations

  • Bacterial populations for desirable traits in ruminating animals

    WO2021001834A1