CN114373505B

CN114373505B - System for early prediction of postpartum subclinical ketosis of dairy cow based on intestinal microorganisms

Info

Publication number: CN114373505B
Application number: CN202111639671.9A
Authority: CN
Inventors: 孙会增; 朱森林; 顾凤飞; 刘建新
Original assignee: Zhejiang University ZJU
Current assignee: Zhejiang University ZJU
Priority date: 2021-12-29
Filing date: 2021-12-29
Publication date: 2022-11-01
Anticipated expiration: 2041-12-29
Also published as: CN114373505A

Abstract

The invention discloses a system for early predicting dairy cow postpartum subclinical ketosis based on intestinal microorganisms, which predicts the dairy cow postpartum subclinical ketosis by using the abundance of intestinal prognostic marker microorganisms after dairy cows in the perinatal period, wherein the prognostic marker microorganisms comprise genus level Parabacter, shigella, cellulosilyticum, roseburia, sporobacter and Acetanerobacterium. The invention screens the 6 types of microorganisms of the prognosis markers of the subclinical ketosis, and establishes a prognosis discrimination model based on discrimination analysis. The kit has accurate and stable prediction capability on postpartum subclinical ketosis of the dairy cow, can be used for early warning and screening the dairy cow in the perinatal period by a noninvasive rectal sampling method, and provides an effective means for preventing and treating postpartum ketosis and subclinical ketosis of the dairy cow as early as possible in a dairy farm.

Description

System for early prediction of postpartum subclinical ketosis of dairy cow based on intestinal microorganisms

Technical Field

The invention relates to the field of perinatal dairy cow health, in particular to a system for predicting blood beta-hydroxybutyric acid level of a dairy cow in one week after delivery based on prenatal intestinal microorganisms so as to predict postnatal subclinical ketosis.

Background

The dairy cattle breeding industry in China develops rapidly in the twenty-first century, the annual single-yield level of the dairy cattle is increased from less than 4,000 kilograms to more than 8,000kg, but the health problem is increasingly prominent along with the continuous increase of the yield of the dairy cattle. After delivery, due to the fact that the energy supplied by the feed cannot meet the requirement of milk production, cows usually have different degrees of energy negative balance, and therefore need to be supplied with energy by moving the cows from body tissues. In this process, a large amount of long-chain fatty acids in body fat are decomposed into non-esterified fatty acids which enter the liver and are not completely oxidized, free fatty acids further form ketone bodies such as BHBA, acetoacetic acid, acetone and the like, and partial long-chain fatty acids are re-esterified to form triglyceride. When a large amount of BHBA and triglyceride are accumulated in the liver, the BHBA and the triglyceride cannot be utilized and removed in time, so that the lipid metabolism disorder of the dairy cow can be caused, and meanwhile, the BHBA diffuses into blood to cause metabolic diseases such as ketosis, fatty liver and the like. The incidence of ketosis of dairy cows in late perinatal period of China is reported to account for 10% -30% of lactating dairy cows, and the health of the dairy cows and the economic benefit of dairy farms are seriously harmed.

The most common determination index for ketosis is blood BHBA concentration, which can be divided into subclinical and clinical ketosis. Subclinical ketosis is defined as the postpartum disease of the dairy cow with no clinical symptoms but excessive ketone body content detected by organisms and the plasma BHBA concentration of more than 1.2 mmol/L. Research statistics total economic losses due to reduced milk production account for approximately 24%, mainly because ketosis causes hypoglycemia, which lowers lactose synthesis, resulting in reduced milk production. The frequency of occurrence of the subclinical ketosis is higher than that of the clinical ketosis, the subclinical ketosis is more hidden, economic losses in various aspects such as reduction of milk yield of cows, increase of milk cow elimination rate, increase of treatment cost, prolongation of calving period and the like can be caused, and subclinical ketosis can occur in the perinatal later stage and the lactation initial stage of about 30-50% of the cows according to the report.

At present, the prevention and treatment agents and products related to subclinical ketosis and ketosis of dairy cows have been researched more, and include methods such as glucose bolus injection, glucocorticoid gluconeogenesis promotion, insulin lipolysis reduction, vitamin B12/phosphorus gluconeogenesis promotion, propylene glycol as gluconeogenesis precursor and the like, but due to the higher latency of subclinical ketosis symptoms, the prevention and treatment which is not targeted can generate high cost and side effects on healthy individuals. Therefore, a simple and accurate prediction method is urgently needed to judge the dairy cattle which are prone to subclinical ketosis after delivery, so that early intervention is accurately implemented, and economic loss of the dairy cattle in the perinatal period is reduced.

The posterior intestinal microorganisms are inseparable from the health of the organism in the research, nutrient substances are absorbed in the process of chyme of the dairy cows from front to back in the digestive tract, the abundance of metabolic waste is gradually increased through an accumulation effect, and the metabolic waste is accumulated in the posterior intestinal tract. In human and monogastric animal studies, posterior intestinal microorganisms are induced by equiaxial induction of intestine-brain, intestine-liver and intestine-heart or participate in health regulation of the body, and a large number of novel microbial markers are reported. With the rapid development of the microbial 16S rRNA gene sequencing technology, it is possible to obtain the marker microorganism by microbial sequencing. Among the existing ruminant studies, the research on the microorganisms of the forestomach is very abundant, and the microorganisms in the forestomach are proved to be associated with the production performance of cows, such as milk yield, milk protein content and nitrogen metabolism efficiency, feed conversion efficiency, and the like. The posterior intestinal microorganisms are considered to be closely related to the diseases and physiological states of the cow body, but related research reports are very poor. The potential value of intestinal microorganisms on the regulation and prediction of the health of the dairy cows needs to be excavated.

The invention successfully establishes a machine learning model for predicting the blood BHBA concentration of postpartum one week based on the intestinal marker microorganisms of the prenatal cows for the first time, and has higher model efficiency (AUROC: 0.876-0.917 (95% CI 0.778-0.993)) and Accuracy (Accuracy: 0.839-0.857). The invention provides a novel solution for accurate prevention and treatment of postpartum ketosis, and has important guiding significance and application value for accurately relieving metabolic burden of high-yield dairy cows in perinatal period, improving stress state and improving health and production level.

Disclosure of Invention

The invention aims to provide a system for early predicting postpartum subclinical ketosis of a dairy cow based on intestinal microorganisms aiming at the defects of the prior art, and provides a machine learning model for predicting the concentration of blood BHBA in one week postpartum based on the marker microorganisms of the intestinal tract of the antenatal dairy cow, which has higher model efficiency and accuracy.

The purpose of the invention is realized by the following technical scheme: a system for early predicting postpartum subclinical ketosis of a cow based on intestinal microorganisms comprises a microorganism collecting module and a postpartum blood beta-hydroxybutyric acid predicting module;

the microorganism collection module is used for collecting microorganisms of 3 weeks before birth of the dairy cow and inputting the microorganisms into the postpartum blood beta-hydroxybutyric acid prediction module, wherein the microorganisms are prognosis marker intestinal microorganisms and comprise six microorganisms, namely Parabacteriaceae, shigella, cellulosilyticum, roseburia, sporobacter and Acetanerobacterium;

the postpartum blood beta-hydroxybutyric acid prediction module comprises a discrimination model for predicting the postpartum blood beta-hydroxybutyric acid level of a cow in one week, wherein the discrimination model comprises the following specific steps:

wherein g is_iFor the discrimination coefficient, the numerical values are shown in the following table; abundannce_iInputting data for a discrimination model, and obtaining the genus level abundance obtained by sequencing the 16S rRNA gene of the ith microorganism in the prenatal rectal contents of the dairy cows; score represents the Score obtained by the discriminant model after inputting genus-level abundances of 6 microorganisms. When Score is measured>At-0.004, cows are considered to be at risk of developing subclinical ketosis one week post partum.

Marker microorganism	Coefficient g_i
		Parabacteroides	0.502
Shigella	-11.255
		Cellulosilyticum	2.669
Roseburia	2.202
		Sporobacter	-19.538
Acetanaerobacterium	-8.803

Furthermore, for all microorganism samples in the intestinal tracts of the cows in three weeks before birth, a sample with the abundance higher than 0.01% in 50% of the samples is selected, all the microorganisms in three weeks before birth are subjected to characteristic screening of postpartum subclinical ketosis prediction effectiveness by a random forest Boruta method, and the first 6 microorganisms are obtained according to the minimum variable number (6) with the lowest error rate obtained by cross validation and the ranking of the mean Gini index (MeanecreaseGini).

Further, the genus-level abundance of the microorganism is the amplified abundance of the V3-V4 region of the 16S rRNA gene of the microorganism.

Further, the coefficient of discrimination g_iThe abundance of the 16S rRNA gene variable region V3-V4 region microbial marker fragment of 6 microorganisms was obtained by classical discriminant analysis (Canonical diagnostic analysis).

Further, the determination method of the abundance of the microorganism characteristic fragments in the V3-V4 region of the variable region of the 16S rRNA gene of the marker microorganism comprises but is not limited to PCR, amplicon sequencing, second-generation high-throughput sequencing, panomics or Nanostring metagenome sequencing.

The method has the advantages that the method can predict the disease condition of the subclinical ketosis of the dairy cattle after delivery and carry out early warning diagnosis by utilizing the discrimination model through detecting the abundance of the microorganism characteristic fragments of the 16S rDNA variable region V3-V4 region of the dairy cattle hindgut marker microorganisms (Parabactoides, shigella, cellulosidium, roseburia, sporobacter and Acetaceae) 3 weeks before delivery, thereby being conductive to reducing the disease condition by adopting effective prevention and control measures before diseases and laying a foundation for accurate prevention and control of the dairy cattle in the perinatal period. The invention provides a novel solution for accurate prevention and treatment of postpartum ketosis, and has important guiding significance and application value for accurately relieving metabolic burden of high-yield dairy cows in perinatal period, improving stress state and improving health and production level.

Drawings

Fig. 1 is a case where cows were grouped into values according to BHBA one week postpartum, healthy group one week postpartum and subclinical ketosis group.

Fig. 2 is a graph of error rates plotted after 5 iterations of cross validation using a random forest method, with a cut line showing that when there are 6 variables, the minimum error rate and the minimum number of variables can be guaranteed.

FIG. 3 is a graph of the importance of microbial variables ranked by the MeanGini index of the random forest model and the top 6 species at the genus level were selected.

FIG. 4 is a graph showing that there were significant differences in the abundance of 5 microorganisms between individuals who developed sub-clinical ketosis postpartum and those who did not developed sub-clinical ketosis three weeks prenatal under the differential t test.

FIG. 5 is an ROC curve for discriminant analysis prediction model prediction based on performance testing of the discriminant analysis prediction model by cross validation.

Fig. 6 is a system block diagram of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below, and the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

As shown in figure 6, the invention provides a system for predicting postnatal subclinical ketosis based on prenatal intestinal microorganisms of dairy cows, which comprises a microorganism collection module and a postnatal blood beta-hydroxybutyric acid prediction module; as shown in fig. 1, prior to predicting the total BHBA level in postpartum blood, blood sample collection and indicator determination were first performed as follows:

1. posterior intestinal content Collection

Selecting 54 cows with age, body condition and gestational age similar to Hangzhou intensive cow farm, tracking the cows to take feces from rectum before feeding in three weeks before parturition and one week after parturition, taking the feces from rectum, and immediately quenching in liquid nitrogen. All stored in an ultra low temperature freezer at-80 ℃ prior to 16SrDNA assay.

2. Plasma collection and routine blood analysis

When the fecal microorganisms are collected, blood samples are collected from the tail vein blood vessels of the cows, and the low-temperature centrifuge is started in advance in an EDTA vacuum blood collection tube at the set temperature of 4 ℃. Immediately after blood collection, the plasma was obtained by centrifugation at 3000 Xg for 15 minutes at 4 ℃ and the uppermost layer of plasma was aspirated by a 1mL pipette, and the plasma was kept in 1.5mL centrifuge tubes and placed in a-80 ℃ refrigerator. A multi-sample automated assay was initiated using Hitachi 7020 full-automatic biochemical analyzers (High-technologies corporation, tokyo, japan) and companion biochemical kits (Ningbo Medical System Biotechnology Co., ltd.) to determine total BHBA levels in plasma of each sample one week post partum. Taking the subclinical ketosis standard of 1.2mmol/L in the dairy cow industry as a two-classification grouping demarcation point, the dairy cow with the total BHBA level in blood lower than 1.2umol/L is defined as a normal dairy cow, the dairy cow with the total BHBA level higher than 1.2umol/L is defined as a dairy cow with subclinical ketosis, and the dairy cow with the clinical ketosis, namely the total BHBA level higher than 2.0mmol/L, does not appear in the group.

The method comprises the following steps of (1) screening marker microorganisms based on the microbial genus level, and collecting the screened marker microorganisms through a microorganism collecting module;

subsequent analysis was performed using the microbial genus level annotated abundance information obtained after 16S rRNA gene sequencing, and 73 genus level microbes remained after removing low abundance species with abundance below 0.01% in 50% of the samples. Since the number of subclinical ketosis and normal cows in one week after delivery is 6 respectively, and the number of samples between groups is very different, the two groups of unbalanced samples are subjected to balanced development by using the SMOTE method in the DMwR package. The parameters of the SMOTE method were chosen as perc. Over =500, perc. Under =300.

After sample equilibration, a Boruta variable screening method for random forests was performed based on the prenatal three weeks of microbes, and the variable identified as "verified" was selected. And performing random forest prediction modeling by using an R packet of RandomForest. Inputting variables into random forest function by using a random forest R packet to construct a random forest model, performing characteristic screening of postnatal subclinical ketosis prediction effectiveness on all prenatal microorganisms in three weeks by using a random forest Boruta method, repeating for 5 times by using an rfcv function of the R packet to perform cross verification, drawing an error rate curve, observing the change relationship between the error rate and the number of used Markers as shown in figure 2, and selecting 6 variables with the minimum variables and the lowest prediction error rate. After removing the genus-level microorganisms not classified under 16S rRNA gene sequencing, the top 6 microbial genera were selected as in figure 3 based on random forests in the variable importance ranking by means of meanderiesegini ranking: parabacteroides, shigella, cellulosilyticum, roseburia, sporobacter, acetanerobacterium as marker microorganisms, the functions of which are shown in Table 1 below.

TABLE 1 six markers of intestinal microbial function

The differential t-test in figure 4 shows that there are significant differences in the abundance of 5 microorganisms between postpartum healthy and subclinical ketosis individuals three weeks prior to delivery. The postpartum blood beta-hydroxybutyric acid prediction module comprises a discrimination model, and the specific construction process is as follows:

because the random forest model is an integrated model based on bagging, the model is composed of tens of hundreds of decision trees, and the model needs to be stored in a form of programming language objects, is difficult to popularize in livestock production and diagnosis, and cannot obtain a prediction result through rapid manual calculation. Therefore, the abundance matrix of 6 microorganisms collected by the microorganism collecting module is used for classical discriminant analysis (Canonical diagnostic analysis) to obtain a post-natal subclinical ketosis prediction discriminant calculation formula which can be operated based on the abundance of the marker microorganisms. The discriminatory model is defined as a linear combination of microbial abundances of marker microorganisms. The calculation formula is as follows:

in the formula, a coefficient g is discriminated_iAs shown in the above table, abundannce_iThe abundance of the microorganism characteristic fragments of the variable region V3-V4 of the 16S rRNA gene of the microorganism of the i-th prenatal three-week prognosis marker microorganism. The Score value obtained by calculation is used for calculating posterior probability through the following Bayesian formula,

P(k|X)∝P(X|k)*P(k)

where k represents the incidence probability of subclinical ketosis, P (k) is the prior probability, and the likelihood P (X | k) is the probability of the occurrence of the target predictor variable X in each class (incidence versus non-incidence). Likelihood is calculated by projecting data onto a discriminant function, and based on the distribution of discriminant function values, calculating normal distributions N (mu) dependent on different kinds of discriminant values using the fittistr function in MASS packets in R₁,σ₁ ²) And N (mu)₂,σ₂ ²) The Score cut was estimated based on the probability density function of the two types of normal distributions of onset and non-onset. Calculated at Score less than-0.004, only 0.001 probability is encountered one week after delivery. When Score is greater than-0.004, the cow is considered to be at risk of developing subclinical ketosis postpartum.

Cross validation and performance testing of the models:

and (3) carrying out performance test on the discriminant analysis prediction model by using cross validation, wherein the ROC curve of model prediction is shown in FIG. 5, and calculating the model Accuracy and AUROC value of each fold of cross validation based on the confusion matrix, and the AUROC range of the discriminant model is AUROC:0.876-0.917 (95% CI 0.778-0.993), and the Accuracy is 0.839-0.857 through the cross validation. The Accuracy of the typical discriminant prediction system in each fold cross validation is greater than 0.839 and the auc is greater than 0.870.

The implementation case is as follows:

randomly collecting the contents of the intestinal tracts of the cows in three weeks before birth in an intensive pasture in Hangzhou through a microorganism collecting module, sequencing the 16S rRNA to obtain the abundances of six microorganisms, evaluating the abundances by using a discrimination model, and predicting the accuracy to be 0.818, as shown in table 2, wherein the implementation effect of the table 2 based on 6 marker microorganisms is shown in table 2

Dairy cow numbering	1	2	3	4	5	6	7	8	9	10	11
												g__Cellulosilyticum	0.67	0.21	0.19	0.07	0.64	0.17	0.16	0.21	0.1	0.16	0.25
g__Parabacteroides	0.18	0.44	0.28	0.64	0.45	0.6	0.64	0.43	0.58	0.51	0.44
												g__Roseburia	0.09	0.26	0.27	0.11	0.24	0.23	0.24	0.16	0.31	0.3	0.2
g__Shigella	0.01	0.07	0.15	0.02	0.01	0.02	0.02	0	0.04	0.03	0.01
												g__Acetanaerobacterium	0.01	0.07	0.04	0.05	0.04	0.08	0.09	0.02	0.01	0.01	0.04
g__Sporobacter	0.06	0.07	0.04	0.06	0.06	0.03	0.04	0.03	0.02	0.03	0.04
												Score	1.51	-0.8	-1	-0.3	1.6	0.51	0.08	1.04	0.98	0.96	0.77
Actual onset of disease	0	0	0	1	1	1	1	1	1	1	1
												Conjecture of postpartum onset of disease	1	0	0	0	1	1	1	1	1	1	1

In conclusion, the method calculates and predicts the risk probability of postpartum occurrence of the test sample according to the discriminant model through the intestinal microbial abundances of the six dairy cows, namely Parabacteroides, shigella, cellulositicum, roseburia, sporobacter and Acetanaerobacter, and provides a basis for health maintenance and early warning of the dairy cows.

The above-described embodiments are intended to illustrate rather than to limit the invention, and any modifications and variations of the present invention are within the spirit of the invention and the scope of the appended claims.

Claims

1. A system for early predicting postpartum subclinical ketosis of a dairy cow based on intestinal microorganisms is characterized by comprising a microorganism acquisition module and a postpartum blood beta-hydroxybutyric acid prediction module;

the microorganism acquisition module is used for acquiring intestinal microorganisms of the dairy cows in 3 weeks before delivery and inputting the intestinal microorganisms into the postpartum blood beta-hydroxybutyric acid prediction module, and the microorganisms are prognosis mark intestinal microorganisms and comprise six microorganismsParabacteroides、Shigella、Cellulosilyticum、Roseburia、Sporobacter、Acetanaerobacterium；

wherein

In order to discriminate the coefficient(s),Parabacteroidesthe coefficient of (a) is 0.502,Shigellathe coefficient of (a) is-11.255,Cellulosilyticumthe coefficient of (a) is 2.669,Roseburiathe coefficient of (a) is 2.202,Sporobacterthe coefficient of (a) is-19.538,Acetanaerobacteriumthe coefficient of (a) is-8.803;

inputting data for discriminating model, and determining the content of prenatal rectum of milk cowiGenus-level abundance obtained by sequencing of the species microorganism 16S rRNA gene; score represents a Score calculated after the genus level abundance of 6 microorganisms is input into the discrimination model; when Score is measured>At-0.004, the cows are considered to be at risk of developing subclinical ketosis one week post-partum.

2. The system of claim 1, wherein the sample with an abundance of 50% of samples higher than 0.01% is selected for all microorganism samples in the intestinal tract of the dairy cow in three weeks before birth, and after feature screening of the efficacy of predicting the postpartum subclinical ketosis is performed on all the microorganisms in three weeks before birth by the random forest Boruta method, the first 6 microorganisms are obtained according to the minimum variable quantity of 6 with the lowest error rate obtained by cross validation and the high-low ranking of the average Gini index MeanecreaseGini.

3. The system for early predicting postpartum subclinical ketosis of a cow based on intestinal microorganisms of claim 1, wherein the genus-level abundance of the microorganisms is the amplified abundance of the V3-V4 region of the 16S rRNA gene of the microorganisms.

4. The system of claim 1, wherein the discrimination coefficients are based on intestinal microbiome to early predict postpartum subclinical ketosis of a cow

The gene is obtained by performing classical discriminant analysis on the abundance of microbial marker fragments in the variable regions V3-V4 of the 16S rRNA genes of 6 microorganisms.

5. The system for early predicting postpartum subclinical ketosis of a cow according to claim 4, wherein the determination method of the abundance of the microbial characteristic fragments in the variable region V3-V4 of the 16S rRNA gene of the microorganism comprises PCR, amplicon sequencing, next-generation high-throughput sequencing, panomics or Nanostring metagenomic sequencing.