Establishment method of animal blood pressure prediction model based on smell information

Abstract

The invention discloses an animal blood pressure prediction model establishing method based on smell information. The animal blood pressure prediction model establishing method comprises the following steps that response signals are obtained through response of an electronic nose gas sensor array to feces smell of a hypertensive rat; detecting the blood pressure of the hypertensive rat of which the excrement smell is detected; extracting characteristic response signals from the excrement smell information, wherein the characteristic response signals comprise a response mean value, a steady-state value, an integral and a differential; principal component analysis and discriminant analysis are carried out on the extracted characteristic values, and qualitative discrimination is carried out in combination with the blood pressure states (hypertension and normal blood pressure) of the rats; the extracted characteristic values serve as independent variables, multivariate linear analysis, a multi-layer perceptron neural network and partial least squares regression analysis are carried out, and quantitative prediction is carried out on the blood pressure of the female rats and the male rats. By means of the method, rat blood pressure state judgment and blood pressure detection based on excrement smell are achieved, and the problems that invasive operation difficulty is large, conscious animal individual compliance is poor, external interference factors are large, and a large amount of manpower, material resources and financial resources are consumed are solved. And a basis is provided for rapid and non-invasive evaluation of experimental animals.

Description

Technical Field

[0001] The present invention belongs to the field of animal disease diagnosis, and particularly relates to a method for establishing an animal blood pressure prediction model based on odor information. Background Art

[0002] It is estimated that by 2025, one-third of the global population will be hypertensive. Hypertension increases the risk of other diseases such as heart disease and stroke, and is one of the major diseases endangering human health. As one of the important physical sign parameters, timely monitoring of its change trend can evaluate cardiovascular function, predict the occurrence of diseases such as hypertension, and reduce the risk of death from cardiovascular diseases. Spontaneous Hypertension Rat (SHR), as the animal model closest to human essential hypertension, is used to study the pathogenesis, treatment, and prevention of hypertension. Monitoring the blood pressure of SHR rats is an important means in the experimental process. However, due to the easy influence of animals by external environmental factors, general animal blood pressure monitoring is time-consuming and laborious, and still faces significant challenges.

[0003] Currently, there are two methods for measuring animal blood pressure: invasive and non-invasive. Among them, invasive detection methods such as implantable telemetry based on hemodynamics and femoral artery intubation can directly obtain the changes in intravascular pressure, and the results are accurate. However, it is necessary to anesthetize the animal and implant the blood pressure probe, which has an infection risk and requires high technical requirements for operators, and is not very suitable for blood pressure monitoring of a large number of animals; non-invasive detection methods such as Doppler blood pressure monitors and oscillometric methods can indirectly reflect blood pressure changes through blood flow, and the operation is relatively simple. However, they face problems such as poor compliance of conscious animal individuals and large external interference factors. Animal blood pressure detection methods based on infrared sensors to detect blood flow wave changes and photoplethysmography, although they solve the problems of portability and comfort, their measurement systems are complex, and some characteristic parameters are not applicable to all animals. Therefore, it is particularly important to establish a non-invasive blood pressure detection method applicable to animals (especially a large number of animals).

[0004] Electronic nose technology is a rapid detection method for odor information developed by simulating the sense of smell of mammals. Due to its advantages such as rapidity and non-invasiveness, it has been used in the diagnosis of animal diseases such as tuberculosis, avian influenza, and Mycobacterium suis disease. Volatile organic compounds (VOCs) are released from microbial metabolites into body metabolites such as urine, feces, and sweat, and can be used to diagnose diseases. Studies have found that hypertension can affect the richness and diversity of the gut microbiota, with an increase in the number of Proteobacteria that produce lactic acid; the Firmicutes / Bacteroidetes (F / B) ratio increases, and the numbers of Actinobacteria, Bifidobacterium, Bacteroidetes, and the metabolic amount of short-chain fatty acids (SCFAs) decrease. The gut microbiota of hypertensive patients drives changes in body metabolites and changes in the contents of fecal SCFAs, trimethylamine N-oxide, corticosterone, and indoleacetic acid, resulting in changes in fecal odor information, which makes it possible to non-invasively monitor changes in animal blood pressure based on odor information. Summary of the Invention

[0005] In view of the above problems, the purpose of the present invention is to provide a method for establishing an animal blood pressure prediction model based on odor information, which can quickly discriminate and detect the blood pressure status (hypertension (grade 1 hypertension, grade 2 hypertension), normal blood pressure) of SHR rats based on metabolite odor information.

[0006] A method for establishing an animal blood pressure prediction model based on odor information includes the following steps: 1) Pump the headspace odor gas of rat feces into the electronic nose sensor chamber and contact it with the sensor, and obtain the sensor response signal by using the reaction of the sensor to the odor; 2) Detect the blood pressure of rats in different blood pressure states of normal blood pressure, grade 1 hypertension, and grade 2 hypertension; 3) Extract the response mean value, steady-state value, integral, and differential from the response signal of the fecal odor as characteristic values, and use principal component analysis or discriminant analysis multivariate statistical analysis methods to conduct qualitative analysis in combination with the rat blood pressure status; use multiple linear analysis, multi-layer perceptron neural network, or partial least squares regression analysis to establish a quantitative prediction model of rat blood pressure and conduct quantitative analysis of its blood pressure.

[0007] The headspace volume is 100 - 500 mL; the flow rate of gas pumping is 100 - 500 mL / min, the signal acquisition time is 50 - 80 s, and the sensor cleaning time is 70 - 90 s; the characteristic data are the mean value, mean value, steady-state value, integral, and differential information of the sensor response signal; the qualitative discrimination methods are principal component analysis and canonical discriminant analysis; the quantitative prediction model is multiple linear regression analysis, multi-layer perceptron neural network, and partial least squares regression analysis.

[0008] A blood pressure prediction model for SHR rats was obtained by multiple linear regression analysis: Systolic blood pressure of SHR rats = -88.103S1 + 2.688S2 + 1105.814S3 - 7.129S4 - 399.819S5 - 1.176S6 - 6.047S7 + 49.741S8 + 34.311S9 - 2.594S10 - 554.645; Diastolic blood pressure of SHR rats = -84.509S1 - 2.568S2 + 1013.674S3 + 18.410S4 - 663.024S5 - 11.447S6 + 1.398S7 + 25.898S8 - 0.148S9 + 54.997S10 - 263.436; In the formula, S1 to S10 are the odor information of aromatic components, alkanes and organic sulfides in the odor information Beneficial effects (1) Non-invasive and highly efficient: No anesthesia or invasive operation is required, the single detection time ≤ 80 seconds, suitable for large-scale animal experiments; There is no need to sacrifice experimental animals, pre-treat the detection, the operation is simple, the detection speed is fast, and the sensitivity is high. It can realize the rapid determination and prediction of different blood pressure states of SHR rats, and is suitable as a real-time and fast method for diagnosing animal hypertension diseases; (2) Precise classification: The total contribution rate of PCA / CDA to the blood pressure state of male SHR is > 87%, and the classification accuracy of females reaches 100%; (3) Provide a new paradigm for metabolite-disease association research, fill the research gap in the diagnosis of animal hypertension based on metabolite odor information, and broaden the methods for diagnosing animal diseases. Description of the drawings

[0009] Figure 1 Radar charts of the fecal odors of SHR rats at different blood pressure levels (a: male SHR rats, b: female SHR rats) Figure 2 PCA and CDA results of the electronic nose of fecal samples of SHR rats in different blood pressure states (a: male PCA, b: male CDA, c: female PCA, d: female CDA) Detailed implementation manners

[0010] The present invention will be further described below with reference to the drawings and embodiments.

[0011] Embodiment: The present invention mainly lies in non-invasively discriminating the blood pressure state and blood pressure value of rats. An electronic nose based on a metal sensor odor sensor array is adopted, and its sensor array is composed of 10 sensors. The names and performances of each sensor are shown in Table 1.

[0012] The function of these sensors is to convert the action of different flavor substances on the surface of feces of SHR rats in different blood pressure states into measurable electrical signals.

[0013] In this example, feces of SHR rats in different blood pressure states (normal blood pressure, stage 1 hypertension, stage 2 hypertension) were collected. SHR rat feces samples were placed in a 150 mL beaker, sealed for 10 min to generate headspace gas, and the headspace gas was pumped into the sensor chamber at a flow rate of 200 mL / min to contact the sensor and generate a response signal. The detection time was 60 s, and the sensor cleaning time was 80 s. 40 parallel samples were set in each group, and the response value at the 59th s of the sensor steady state was selected for analysis.

[0014] The radar chart of the odor information differences of feces of SHR rats in different blood pressure states in the example is as Figure 1 shown. It can be seen from Figure 1 that: the differences in the odor fingerprint information of feces of SHR rats in different blood pressure states are relatively small for sensors S1, S3, S4, and S5; there are relatively large differences in the odor fingerprint information for sensors S2, S6, S7, S8, S9, and S10.

[0015] Figure 2 is the two-dimensional score chart of the principal component analysis and canonical discriminant analysis of the odor of feces of SHR rats in different blood pressure states. In the PCA result, the contribution rates of the first two principal components of male SHR rats are 54.58% and 32.81% respectively, and their total contribution rate reaches 87.93%; the contribution rates of the first two principal components of female SHR rats are 64.84% and 23.93% respectively, and their total contribution rate reaches 88.77%. In the CDA result, both male and female SHR rats explain 100% of the original information. It can be seen from Figure 2 that the feces samples of SHR rats in different blood pressure states are regularly distributed, that is, normal blood pressure is in the area where Can1 < -5, and hypertension is in the area where Can1 > -5. The different blood pressure states of rats can be well distinguished by principal component analysis and canonical discriminant analysis.

[0016] On the basis of canonical discriminant analysis, multiple linear regression analysis was further used to establish the correlation between odor information and the blood pressure state of rats. The electronic nose response data of feces of male and female SHR rats were randomly divided into a modeling set of 150 (30×5) and a validation set of 35 (7×5). With systolic blood pressure and diastolic blood pressure as the outputs, the odor information of the electronic nose was used as the parameter of multiple linear regression analysis for regression to establish a prediction model for the blood pressure of SHR rats.

[0017] The prediction model for the blood pressure of SHR rats was obtained by multiple linear regression analysis: Systolic blood pressure of SHR rats = -88.103S1 + 2.688S2 + 1105.814S3 - 7.129S4 - 399.819S5 - 1.176S6 - 6.047S7 + 49.741S8 + 34.311S9 - 2.594S10 - 554.645 (1) Diastolic blood pressure of SHR rats = -84.509S1 - 2.568S2 + 1013.674S3 + 18.410S4 - 663.024S5 - 11.447S6 + 1.398S7 + 25.898S8 - 0.148S9 + 54.997S10 - 263.436 (2) In formula (1), S1 to S10 are for the odor components, alkanes, organic sulfides and other odors in the odor fingerprint information.

[0018] Coefficient of determination R of the prediction model 2 Are 0.9161 and 0.9213 respectively, indicating that the prediction models established by multiple linear regression analysis are effective.

[0019] Table 2 shows the effects of the quantitative prediction model established by multiple linear regression analysis. It can be seen from the prediction results of the model that the relationship between the odor fingerprint information and the blood pressure of rats can be established, indicating that the prediction of the blood pressure of SHR rats by the present invention is feasible.

[0020] The above has described the embodiments of the present invention in detail, but the content described is only the preferred embodiments of the present invention and is not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the scope of the application of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for establishing an animal blood pressure prediction model based on odor information, comprising the following steps: 1) Pump the headspace odor gas of rat feces into the electronic nose sensor chamber and contact it with the sensor, and use the sensor's response to the odor to obtain the sensor response signal; 2) Blood pressure was measured in rats with normal blood pressure, primary hypertension, and secondary hypertension; 3) Extract characteristic data from the response signal of fecal odor, and use principal component analysis or discriminant analysis multivariate statistical analysis methods, combined with the blood pressure status of rats for qualitative analysis; use multivariate linear analysis or multilayer perceptron neural network partial least squares regression analysis to establish a quantitative prediction model for rat blood pressure, and conduct quantitative analysis of its blood pressure.

2. The method for establishing an animal blood pressure prediction model based on odor information according to claim 1, characterized in that: The gas pump flow rate is 100-500 mL / min, the signal acquisition time is 50-80 s, and the sensor cleaning time is 70-90 s.

3. The method for establishing an animal blood pressure prediction model based on odor information according to claim 1, characterized in that: The characteristic data are the response mean, steady-state value, integral and differential of the sensor response signal.

4. The method for establishing an animal blood pressure prediction model based on odor information according to claim 1, characterized in that: The qualitative discriminant model method is principal component analysis or canonical discriminant analysis; the quantitative prediction model method is multivariate linear analysis, multilayer perceptron neural network or partial least squares regression analysis.

5. The method for establishing an animal blood pressure prediction model based on odor information according to claim 1, characterized in that: The prediction model of blood pressure in SHR rats was obtained by multivariate linear regression analysis: SHR rat systolic blood pressure = -88.103S1+2.688S2+1105.814S3-7.129S4-399.819S5-1.176S6-6.047S7+49.741S8+34.311S9-2.594S10-554.645; Diastolic blood pressure of SHR rats = -84.509S1-2.568S2+1013.674S3+18.410S4-663.024S5-11.447S6+1.398S7+25.898S8-0.148S9+54.997S10-263.436; Where S1~S10 are the aroma components, alkanes and organic sulfide odor information in the odor information.

Images