Method for identifying saline-alkaline resistance of red tilapia mossambica based on combination of serum characteristic parameters under saline-alkaline micro-stimulation and machine learning algorithm

Abstract

The invention provides a method for identifying saline-alkaline resistance of red tilapia based on combination of serum characteristic parameters under saline-alkaline micro-stimulation and a machine learning algorithm, and belongs to the technical field of bioinformatics. The invention discloses a method for establishing a red tilapia saline-alkaline resistance prediction model based on a machine learning algorithm. The method comprises the following steps: establishing a red tilapia serum characteristic parameter database under saline-alkaline stress; and extracting a serum characteristic parameter data set from the red tilapia serum characteristic parameter database, and carrying out machine learning classification model training and testing based on a LightGBM algorithm to obtain a red tilapia saline-alkaline resistance prediction model. The saline-alkaline tolerant red tilapia prediction model established by the invention can effectively identify saline-alkaline tolerant red tilapia, the identification period is short, the detection cost is low, and the detection method belongs to minimally invasive non-lethal; the saline-alkaline tolerant red tilapia parents identified by the method can be directly put into production, the saline-alkaline tolerant fry quality of red tilapia is effectively improved, and green and healthy development of the red tilapia breeding industry is promoted.

Description

Technical Field

[0001] The present invention belongs to the technical field of bioinformatics, and particularly relates to a method for identifying the saline-alkali tolerance of red tilapia based on serum characteristic parameters under saline-alkali micro-stimulation combined with a machine learning algorithm. Background Art

[0002] Tilapia generally lives in fresh water, but because it is an euryhaline fish, it has a wide salinity adaptation range and strong salt tolerance. The previous research results of the team showed that red tilapia also has excellent alkalinity tolerance, and it can be used as an important candidate material for the breeding of new varieties of saline-alkali tolerant fish. There are rich saline-alkali water areas in China. Exploring how to effectively develop and utilize saline-alkali water areas can open up a new direction for the sustainable development of tilapia farming. Most saline-alkali water areas are difficult to achieve the cultivation of freshwater tilapia into seawater tilapia through conventional methods such as transplantation and domestication. Therefore, it is crucial to identify and breed new germplasms of fish with good saline-alkali tolerance, which will lay a foundation for the efficient development and utilization of saline-alkali water areas.

[0003] At present, most studies use the 96-hour median lethal concentration (96h LC 50 ) to measure the saline-alkali tolerance of tilapia. Using this method causes great damage to the fish body, and substantial changes will occur in the physiological functions of the surviving individuals. The recovery period is long, which seriously affects the progress of the breeding work. It can be seen that there is currently a lack of a method for quickly and effectively evaluating the saline-alkali tolerance of red tilapia. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method for identifying the saline-alkali tolerance of red tilapia based on serum characteristic parameters under saline-alkali micro-stimulation combined with a machine learning algorithm. By optimizing the machine learning classification model, an efficient method for identifying saline-alkali tolerant red tilapia is obtained.

[0005] The present invention provides a method for establishing a saline-alkali tolerance prediction model of red tilapia based on a machine learning algorithm, including the following steps:

[0006] Construct a database of serum characteristic parameters of red tilapia under saline-alkali stress;

[0007] Extract a serum characteristic parameter data set from the red tilapia serum characteristic parameter database and perform machine learning classification model training and testing based on the LightGBM algorithm to obtain a saline-alkali tolerance prediction model of red tilapia.

[0008] Preferably, the construction of the serum characteristic parameter database under saline-alkali stress includes the following steps:

[0009] Perform saline-alkali micro-stimulation treatment on red tilapia, and detect the serum characteristic parameters of the serum of individual red tilapia after treatment;

[0010] Number the surviving red tilapia individuals and classify the serum characteristic parameters measured from the surviving red tilapia individuals as the data of the saline-alkali tolerance group; number the dead red tilapia individuals and classify the serum characteristic parameters measured from the dead red tilapia individuals as the data of the saline-alkali sensitive group, so as to obtain a database of serum characteristic parameters under saline-alkali stress.

[0011] Preferably, when treating red tilapia with mild saline-alkali stimulation, the salinity is 4.06‰ and the alkalinity is 11.78 mmol / L; the time for treating red tilapia with mild saline-alkali stimulation is 22 - 26 h.

[0012] Preferably, when evaluating the saline-alkali tolerance of red tilapia, the salinity is 14.20‰ and the alkalinity is 41.22 mmol / L; the duration for evaluating the saline-alkali tolerance of red tilapia is 95 - 97 h.

[0013] Preferably, the serum characteristic parameters include at least one of the following: blood ammonia, glutamine, lactate dehydrogenase, and superoxide dismutase.

[0014] Preferably, after training and testing the machine learning classification model, it further includes evaluating the performance of the obtained classification model, and selecting the classification model that can accurately distinguish saline-alkali tolerance samples and saline-alkali sensitivity samples as the prediction model for the saline-alkali tolerance of red tilapia.

[0015] Preferably, the indicators for performance evaluation include at least one of the following: accuracy rate, recall rate, precision rate, F1 score, and area under the curve.

[0016] The present invention provides a method for identifying the saline-alkali tolerance of red tilapia based on the combination of serum characteristic parameters under mild saline-alkali stimulation and machine learning algorithms, including the following steps:

[0017] Input the detection results of the serum characteristic parameters of the red tilapia to be tested into the prediction model for the saline-alkali tolerance of red tilapia established by the method, and output the saline-alkali tolerance result of the red tilapia to be tested.

[0018] Preferably, the red tilapia includes one-year-old reserve parent fish.

[0019] The present invention provides a method for establishing a prediction model of the saline-alkali tolerance of red tilapia based on machine learning algorithms, comprising the following steps: constructing a database of serum characteristic parameters of red tilapia under saline-alkali stress; extracting a dataset of serum characteristic parameters from the database of serum characteristic parameters of red tilapia and training and testing a machine learning classification model based on the LightGBM algorithm to obtain a prediction model of the saline-alkali tolerance of red tilapia. The optimal machine learning classification model based on serum characteristic parameters under mild saline-alkali stimulation of the present invention can effectively identify saline-alkali-tolerant red tilapia, with a short identification period, low detection cost, and the detection method being minimally invasive and non-lethal; the established prediction model can effectively identify the parents of saline-alkali-tolerant red tilapia, so that the identified parents of red tilapia can be directly put into production, effectively improving the quality of saline-alkali-tolerant fry of red tilapia and promoting the green and healthy development of the red tilapia aquaculture industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is the survival rate curve of red tilapia under 96-hour saline-alkali stress;

[0021] Figure 2 are the measurement results of serum characteristic parameters related to the saline-alkali tolerance of red tilapia. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] The present invention provides a method for establishing a prediction model of the saline-alkali tolerance of red tilapia based on machine learning algorithms, comprising the following steps:

[0023] Constructing a database of serum characteristic parameters of red tilapia under saline-alkali stress;

[0024] Extracting a dataset of serum characteristic parameters from the database of serum characteristic parameters of red tilapia and training and testing a machine learning classification model based on the LightGBM algorithm to obtain a prediction model of the saline-alkali tolerance of red tilapia.

[0025] The present invention constructs a database of serum characteristic parameters of red tilapia under saline-alkali stress.

[0026] In an embodiment of the present invention, taking red tilapia as the experimental object, an embodiment of the method for establishing a prediction model of the saline-alkali tolerance of red tilapia is carried out. The red tilapia preferably includes first-year reserve parents.

[0027] In the present invention, the construction of the database of serum characteristic parameters under acute semi-lethal saline-alkali stress includes the following steps: subjecting red tilapia to mild saline-alkali stimulation and respectively detecting the serum characteristic parameters of the serum of the treated red tilapia individuals;

[0028] The processed red tilapia is subjected to saline-alkali tolerance treatment, and the individual numbers of the surviving red tilapia and the serum characteristic parameters measured for the surviving red tilapia individuals are classified as data of the saline-alkali tolerance group; the individual numbers of the dead red tilapia and the serum characteristic parameters measured for the dead red tilapia individuals are classified as data of the saline-alkali sensitive group, obtaining a database of serum characteristic parameters under acute semi-lethal saline-alkali stress.

[0029] In the present invention, the serum characteristic parameters preferably include at least one of the following: blood ammonia, glutamine, lactate dehydrogenase, and superoxide dismutase. The present invention proves that the glutamine (GLN), blood ammonia content, lactate dehydrogenase (LDH), and superoxide dismutase (SOD) activities in the serum of red tilapia will change significantly with the change of saline-alkali concentration, and there are significant differences in the data of the serum characteristic parameters of red tilapia in the saline-alkali tolerance group and the saline-alkali sensitive group after acute semi-lethal saline-alkali stress. The serum GLN, blood ammonia content, and LDH activity in the saline-alkali tolerance group are significantly lower than those in the saline-alkali sensitive group, while the serum SOD activity in the saline-alkali tolerance group is significantly higher than that in the saline-alkali sensitive group. Therefore, it is considered that these four indicators can be used as characteristic indicators reflecting the saline-alkali tolerance ability of red tilapia. Therefore, this experiment first constructs a database of characteristic parameters around these four serum indicators. The present invention has no special limitation on the method for detecting the serum characteristic parameters, and commercially available detection kits for GLN, blood ammonia, LDH, and SOD well-known in the art can be used. In the examples of the present invention, the commercially available detection kits for GLN, blood ammonia, LDH, and SOD are purchased from Shanghai Langton Biotechnology Co., Ltd.

[0030] In the present invention, when the red tilapia is treated with mild saline-alkali stimulation, the salinity is preferably 4.06‰, and the alkalinity is preferably 11.78 mmol / L; the time for treating the red tilapia with mild saline-alkali stimulation is preferably 22 - 26 h. When evaluating the saline-alkali tolerance of the red tilapia, the salinity is preferably 14.20‰, and the alkalinity is preferably 41.22 mmol / L; the duration for evaluating the saline-alkali tolerance of the red tilapia is preferably 95 - 97 h. When treating the red tilapia with mild saline-alkali stimulation or evaluating the saline-alkali tolerance of the red tilapia, the breeding conditions of the red tilapia are preferably as follows: temperature 25 - 26°C, dissolved oxygen > 6.0 mg / L; during this period, commercial floating feed (crude protein 32%, crude fat 5%) is fed twice a day, morning and evening, and the feeding amount is about 3% of the fish body weight. In order to distinguish individual red tilapia, the red tilapia is marked by injecting PIT.

[0031] In the present invention, according to the survival situation of the red tilapia, it is divided into a saline-alkali tolerance group and a saline-alkali sensitive group. Preferably, the surviving individuals of the red tilapia are classified into the saline-alkali tolerance group; the dead individuals of the red tilapia are classified into the saline-alkali sensitive group.

[0032] After obtaining the red tilapia serum characteristic parameter database, the present invention extracts a serum characteristic parameter data set from the red tilapia serum characteristic parameter database and trains and tests a machine learning classification model based on the LightGBM algorithm to obtain a red tilapia salt and alkali tolerance prediction model.

[0033] In the present invention, the quantity ratio of the serum characteristic parameter data set for machine learning classification model training to the serum characteristic parameter data set for machine learning classification model testing is 6-8:4-2, and it can be 7:3. When training and testing the machine learning classification model based on the LightGBM algorithm, it is preferably completed using SPSSAU 24.0 software.

[0034] In the embodiment of the present invention, when training and testing the machine learning classification model, in order to screen out the best machine learning classification model algorithm, machine learning classification model training and testing are respectively carried out based on random forest, gradient boosting tree, support vector machine, K-nearest neighbor, BP neural network, and LightGBM algorithms. The test results show that all 6 models can obtain good classification results, but the classification model established by the LightGBM algorithm has the best effect, and the accuracy, recall rate, precision, F1 score, and area under the curve results in the performance evaluation are all 1. Therefore, the present invention selects the LightGBM algorithm to construct the machine learning classification model.

[0035] In the present invention, after training and testing the machine learning classification model, it preferably further includes performing a performance evaluation on the obtained classification model, and selecting the classification model that can accurately distinguish saline-alkali tolerance samples and saline-alkali sensitivity samples as the red tilapia salt and alkali tolerance prediction model. The performance evaluation indicators preferably include at least one of the following: accuracy, recall rate, precision, F1 score, and area under the curve.

[0036] The present invention provides a method for identifying the salt and alkali tolerance of red tilapia based on the combination of serum characteristic parameters under saline-alkali micro-stimulation and machine learning algorithms, including the following steps:

[0037] Input the detection result of the serum characteristic parameters of the red tilapia to be tested into the red tilapia salt and alkali tolerance prediction model established by the method, and output the salt and alkali tolerance result of the red tilapia to be tested.

[0038] In the present invention, the red tilapia preferably includes one-year-old reserve parent fish. The present invention does not impose special restrictions on the variety of the red tilapia. The content described in the above technical solution will not be elaborated herein. The saline-alkali tolerance results of the red tilapia to be tested are divided into two results: saline-alkali tolerance and saline-alkali sensitivity. In the embodiments of the present invention, taking the red tilapia variety as an example, the saline-alkali tolerance of the red tilapia is identified based on the established prediction model for the saline-alkali tolerance of the red tilapia. In an embodiment of the present invention, 20 red tilapia samples are verified. The results show that the accuracy rate of classifying and discriminating the saline-alkali tolerance levels of 20 red tilapia samples by selecting the best LightGBM model is 100%. This result further shows that the established machine learning classification model can quickly identify red tilapia with high saline-alkali tolerance, greatly improving the identification efficiency of saline-alkali tolerant red tilapia.

[0039] The following is a detailed description of a method for identifying the saline-alkali tolerance of red tilapia based on the combined machine learning algorithm of serum characteristic parameters under saline-alkali micro-stimulation provided by the present invention in conjunction with the embodiments, but they should not be construed as limiting the protection scope of the present invention.

[0040] Example 1

[0041] Construction of a database of serum characteristic parameters of red tilapia under saline-alkali micro-stimulation

[0042] 1. Screening of moleculars related to the saline-alkali tolerance ability of red tilapia: Through the detection of serum physiological and biochemical indexes, it is found that the levels of glutamine (GLN), ammonia, lactate dehydrogenase (LDH), and superoxide dismutase (SOD) in the serum of red tilapia change significantly with the change of saline-alkali concentration. It is considered that these four indexes can be used as characteristic indexes reflecting the saline-alkali tolerance ability of red tilapia. Therefore, this experiment first constructs a characteristic parameter database around these four serum indexes.

[0043] 2. Cultivation of the experimental population of red tilapia: The red tilapia used in this experiment [weight (225.73 ± 13.55) g] is selected from the one-year-old reserve parent fish cultivated in the Yangzhong Base of the Freshwater Fisheries Research Center, Chinese Academy of Fishery Sciences, with a total of 500 tails. The experimental fish are of uniform size, without injuries on the body surface, and are healthy and energetic. The experimental fish are placed in a freshwater environment with a temperature of 25-26 °C and a dissolved oxygen > 6.0 mg / L for 15 days of temporary cultivation. During the temporary cultivation period, commercial floating feed (crude protein 32%, crude fat 5%) is fed twice a day, morning and evening, and the feeding amount is about 3% of the fish weight.

[0044] 3. Saline-alkali micro-stimulation treatment: The experiment is carried out in a square cement pond (length × width × height 7 m × 3 m × 1.5 m, water depth 0.5 m). According to the 96-hour salinity median lethal concentration (96h SA LC 50 = 20.28‰) and 96-hour alkalinity median lethal concentration (96h CALC50 = 58.88 mmol / L) to prepare a saline-alkali mixed water body. Since the interaction generated by the mixing of salt and alkali is stronger than that of a single factor, in order to ensure the survival rate of red tilapia while having a slight saline-alkali stimulation effect, the saline-alkali concentration of the mixed water body set in this experiment is 20% of the 96-hour semi-lethal concentration of red tilapia under single-factor stress, that is, the salinity is 4.06‰ and the alkalinity is 11.78 mmol / L. The salinity of the water body was detected using a portable salinometer, and the salinity of the water body was adjusted by sea salt crystals (Chuangyi Biotechnology Co., Ltd., Shanghai). The alkalinity of the water body was detected by acid-base titration, and the alkalinity of the water body was adjusted by NaHCO3 (Sinopharm Chemical Reagent Co., Ltd., analytical pure) so that the saline-alkali degree of the water body was maintained at the set value. Other aquaculture conditions during the experiment, such as dissolved oxygen, water temperature, etc., were the same as those during the temporary culture period. The slight saline-alkali stimulation treatment lasted for 24 hours, and then the experimental fish were caught and anesthetized with 100 mg / L MS-222. Blood was collected from the caudal vein, numbered one by one according to the order of blood collection, and PIT tags were injected intramuscularly. A total of 500 fish were used. The red tilapia injected with PIT tags were put back into the fresh water environment for cultivation, and povidone iodine (0.5 mL / m 3 ) was sprinkled for disinfection to prevent wound infection of the experimental fish. The fish population was observed, and dead individuals were fished out in time. The blood samples were centrifuged at 4°C and 5000 g for 15 minutes to collect the serum, which was stored at -20°C for detecting serum characteristic parameters.

[0045] 4. Experiment for evaluating saline-alkali tolerance: The experimental fish were cultured in the fresh water environment for 1 month to fully recover. In order to evaluate the saline-alkali tolerance of the experimental fish, an acute saline-alkali semi-lethal stress experiment was carried out. The saline-alkali concentration of the water body was adjusted to 70% of the 96-hour semi-lethal concentration of red tilapia under single-factor stress, that is, the salinity was 14.20‰ and the alkalinity was 41.22 mmol / L. The experimental environment and other experimental conditions were the same as those during the slight saline-alkali stimulation treatment. The stress experiment lasted for 96 hours. The death situation of red tilapia was observed and recorded every 4 hours, and the PIT numbers of the dead individuals were scanned and recorded. After the experiment ended, the PIT numbers of the remaining surviving individuals were scanned and recorded. According to the survival situation of red tilapia, the following actual classification was carried out: The surviving individuals were classified into the saline-alkali tolerant (ST) group, and the dead individuals were classified into the saline-alkali sensitive (SS) group. The serum samples obtained under the slight saline-alkali stimulation treatment were reorganized according to the detection results of their saline-alkali tolerance and divided into two groups: ST and SS.

[0046] 5) Detect the expression levels of characteristic parameters in serum samples and construct a database of characteristic parameters: 150 serum samples were randomly selected from each of the ST and SS groups, and the expression levels of four indicators, namely GLN, blood ammonia, LDH, and SOD, were detected. The detection kits for GLN, blood ammonia, LDH, and SOD were purchased from Shanghai Langton Biotechnology Co., Ltd., and the operations were carried out according to the requirements of the kit instructions. All sample detections were completed within 24 hours. The detection data were summarized to construct a database of characteristic parameters.

[0047] The results showed that after the saline-alkali micro-stimulation treatment, blood sampling, and injection of PIT labeling, 500 red tilapia were cultured in a freshwater environment for 1 month, during which 27 fish died, that is, 473 fish were used for the saline-alkali tolerance evaluation experiment.

[0048] The survival rate curve of red tilapia in the saline-alkali stress experiment is as Figure 1 shown. In the 96-hour saline-alkali stress experiment, 268 red tilapia died and 205 survived, with a survival rate of 43.34%. Based on the saline-alkali tolerance evaluation experiment, the serum samples of red tilapia obtained under saline-alkali micro-stimulation treatment were divided into two groups, ST and SS. 150 samples were randomly selected from each group to detect the expression levels of four indicators, namely GLN, blood ammonia, LDH, and SOD. The serum GLN, blood ammonia contents, and LDH activities in the ST group were significantly lower than those in the SS group, while the serum SOD activity in the ST group was significantly higher than that in the SS group ( Figure 2 ). The detection data of the four serum characteristic parameters of red tilapia were summarized to build a serum characteristic parameter database for identifying the saline-alkali tolerance ability of red tilapia.

[0049] Example 2

[0050] Establishment of an intelligent classification technology for the saline-alkali tolerance ability of red tilapia based on machine learning

[0051] 1. Based on the constructed serum characteristic parameter database, 280 sample information (140 cases in each of the ST and SS groups) was extracted, and the category (ST or SS) of the saline-alkali tolerance ability of red tilapia was used as the target variable to establish a 280×4 data set. The entire data set was randomly divided into a training set and a test set according to a ratio of 7:3.

[0052] 2. Based on the divided training set data, machine learning classification models were trained using random forest, gradient boosting tree (GBDT), support vector machine (SVM), K-nearest neighbor (KNN), BP neural network, and LightGBM classification algorithms. The trained models were tested using the test set data, and the accuracy, recall rate, precision rate, F1 score, and area under the curve (AUC) were used to evaluate the performance of the above training models to screen the best machine learning model.

[0053] The results showed that: The 280 sample information was randomly divided into a training set and a test set according to a ratio of 7:3. There were no significant differences among the four characteristic variable groups of the 196 training sets and 84 test sets (Table 1). The performance of the classification models established by six machine learning algorithms is shown in Table 2. Based on the four characteristic variables, each model had a good classification effect. Among them, the classification model established based on the LightGBM algorithm had the best effect and could accurately distinguish the samples in the ST and SS groups.

[0054] Table 1 Comparison of characteristic variables between the training set and the test set

[0055] Number Feature variable Training set (n = 196) Test set (n = 84) P value 1 GLN (μmol / L) 432.22±104.56 427.58±100.83 0.731 2 LDH (U / mL) 21.32±3.38 21.42±3.25 0.831 3 Blood ammonia (μg / mL) 0.14±0.05 0.14±0.05 0.917 4 SOD (U / mL) 10.13±2.41 10.23±2.21 0.747

[0056] Table 2 Performance evaluation of different machine learning models

[0057] Accuracy Recall Precision F1 score AUC Random forest 0.988 0.988 0.988 0.988 1 GBDT 0.976 0.976 0.976 0.976 0.977 SVM 0.976 0.976 0.977 0.976 0.997 KNN 0.976 0.976 0.977 0.976 0.998 BP neural network 0.976 0.976 0.977 0.976 0.999 LightGBM 1 1 1 1 1

[0058] Example 3

[0059] Efficient identification, verification and application of salt-tolerant red tilapia based on the LightGBM model

[0060] Based on the performance evaluation results of different machine learning models in Example 2, the best classification model, that is, the LightGBM model, was selected to classify and discriminate 20 additional samples in the characteristic parameter database constructed in Example 1. The actual classification of the samples was based on the survival of red tilapia. Specifically, the surviving individuals after acute saline-alkali stress were divided into the saline-alkali tolerance group (ST), and the dead individuals after acute saline-alkali stress were divided into the saline-alkali sensitive group (SS).

[0061] The results showed that: The accuracy rate of classifying and discriminating the saline-alkali tolerance levels of 20 red tilapia samples by selecting the best LightGBM model was 100%. This result further indicated that the established machine learning classification model could quickly identify red tilapia with high saline-alkali tolerance, greatly improving the identification efficiency of salt-tolerant red tilapia.

[0062] Table 3 Classification verification results of the saline-alkali tolerance level of red tilapia based on the LightGBM model

[0063]

[0064] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for establishing a salt-alkalinity tolerance prediction model of red tilapia based on a machine learning algorithm, characterized in that: The following steps are involved: Build a database of serum characteristic parameters of red tilapia under saline-alkali stress; A serum characteristic parameter data set is extracted from the red tilapia serum characteristic parameter database, and a machine learning classification model is trained and tested based on the LightGBM algorithm to obtain a red tilapia salt-alkali tolerance prediction model.

2. The application according to claim 1, characterized in that: The construction of a serum characteristic parameter database under saline-alkali stress comprises the following steps: Red tilapia was treated with saline-alkali micro-stimulation, and serum characteristic parameters were tested on individual serum of the treated red tilapia. The treated red tilapia is subjected to salt-alkali tolerance treatment, and the individual numbers of the surviving red tilapia and the serum characteristic parameters measured for the surviving red tilapia are classified as salt-alkali tolerance group data; the individual numbers of the dead red tilapia and the serum characteristic parameters measured for the dead red tilapia are classified as salt-alkali sensitivity group data, so as to obtain a serum characteristic parameter database under salt-alkali stress.

3. The application according to claim 2, characterized in that: When the red tilapia is treated with the saline-alkali micro-stimulation, the salinity is 4.06‰ and the alkalinity is 11.78mmol / L.

4. The application according to claim 2, characterized in that: The time for the saline-alkali micro-stimulation treatment of red tilapia is 22 to 26 hours.

5. The application according to claim 2, characterized in that: When evaluating the salt-alkali tolerance of red tilapia, the salinity is 14.20‰ and the alkalinity is 41.22mmol / L; the duration of the salt-alkali tolerance evaluation of red tilapia is 95-97h.

6. The use according to claim 1, characterized in that: The serum characteristic parameters include at least one of the following: blood ammonia, glutamine, lactate dehydrogenase and superoxide dismutase.

7. The use according to any one of claims 1 to 6, characterized in that: After the machine learning classification model is trained and tested, the method also includes performing a performance evaluation on the obtained classification model, and selecting a classification model that can accurately distinguish between salt-alkali tolerance samples and salt-alkali sensitivity samples as a prediction model for the salt-alkali tolerance of red tilapia.

8. The use according to claim 7, characterized in that: The performance evaluation index includes at least one of the following: accuracy, recall, precision, F1 score and area under the curve.

9. A method for identifying the salt-alkali tolerance of red tilapia based on serum characteristic parameters under saline-alkali micro-stimulation combined with machine learning algorithm, characterized in that: The following steps are involved: The test results of the serum characteristic parameters of the red tilapia to be tested are input into the red tilapia salt-alkali tolerance prediction model established by the method described in any one of claims 1 to 7, and the salt-alkali tolerance results of the red tilapia to be tested are output.

10. The method according to claim 9, characterized in that: The red tilapia includes one-year-old reserve parents.

Images