Aquatic echinocystin protein recognition method and system, terminal and medium

Abstract

The invention belongs to the field of toxin protein recognition, and particularly relates to an aquatic echinocystis toxin protein recognition method, system, terminal and medium, a toxin protein recognition model is constructed, and the toxin protein recognition model comprises a feature coding module and a classifier module. The feature coding module performs toxin protein coding by using amino acid composition, a word2vec technology and normalized Moreau-Broto self-correlation description to obtain coding features, and the classifier module classifies the coding features to identify the aquatic echinocystis toxin protein; training the constructed toxin protein recognition model; and identifying the target toxin protein by using the trained toxin protein identification model so as to judge whether the target toxin protein is the aquatic echinocystis toxin protein or not. According to the method, the toxin protein of the aquatic echinocystis is identified through the model based on the random forest, the efficiency and accuracy of identifying the toxin protein of the aquatic echinocystis are effectively improved, and the cost is relatively low.

Description

Technical Field

[0001] The present invention belongs to the field of toxin protein recognition, and in particular relates to a method, system, terminal and medium for recognizing toxin proteins of aquatic cnidarians. Background Art

[0002] Cnidaria (also known as cnidarians) are a class of aquatic animals that include jellyfish, corals, sea anemones, and hydras. Their body structure is relatively simple, usually consisting of a hollow digestive cavity surrounded by tentacles or petal-like structures with toxic stinging cells used to capture prey or for defense. Cnidaria usually exhibit radial symmetry and lack a well-defined head or brain. As one of the most primitive multicellular animals, cnidarians live in a variety of marine environments, from shallow waters to the deep sea, and some species also occur in freshwater habitats. Cnidaria play an important role in the ecosystem, both as predators and prey, and are an important component of coral reef ecosystems.

[0003] Cnidaria inject nematocyst venom to elicit toxic and immune responses in the envenomed organism. These venoms contain enzymes, potent pore-forming toxins, and neurotoxins. Enzymes include lipolytic and proteolytic enzymes, which break down the tissues of prey. Therefore, the correct identification of cnidarian toxin proteins is crucial to understanding their role in the ecosystem.

[0004] Currently, some tentative technologies, such as mass spectrometry, simplified characterization bisulfite sequencing, and single-molecule real-time sequencing, have been developed to identify different types of toxin proteins. Although these technologies are helpful in the identification of aquatic cnidarian toxin proteins, they are costly in the application of large-scale sequence data. Therefore, it is urgent to develop a bioinformatics tool for the identification of aquatic cnidarian toxin proteins. Summary of the invention

[0005] To solve the above problems, the present invention provides a method, system, terminal and medium for identifying toxin proteins of aquatic cnidarians, which can effectively improve the efficiency and accuracy of identifying toxin proteins of aquatic cnidarians at a low cost by identifying toxin proteins of aquatic cnidarians based on a random forest model.

[0006] In a first aspect, the technical solution of the present invention provides a method for identifying aquatic cnidarian toxin proteins, comprising the following steps: Constructing a toxin protein recognition model, the toxin protein recognition model includes a feature encoding module and a classifier module, the feature encoding module uses amino acid composition, word2vec technology, and normalized Moreau-Broto autocorrelation description to encode toxin proteins to obtain encoding features, and the classifier module classifies the encoding features to identify aquatic cnidarian toxin proteins; Training the constructed toxin protein recognition model; The trained toxin protein recognition model is used to identify the target toxin protein to determine whether the target toxin protein is an aquatic cnidarian toxin protein.

[0007] In an optional embodiment, training the constructed toxin protein recognition model specifically includes: Step 1, obtain a cnidarian toxin protein dataset and divide it into a training dataset and a test dataset; Step 2: Describe the amino acid composition of the toxin protein in the dataset using the following formula: f(x) = N(x) / N In the formula, N(x) In the protein sequence x The number of amino acid residues, the total number of residues is N ; Step 3, use word2vec technology to describe the toxin protein sequence in the dataset in vector form; Step 4: The amino acid values ​​of the toxin protein in the dataset are described by the normalized Moreau-Broto autocorrelation method. The formula is: m = 1,2,3,…,20; In the formula, Q(m) represents the normalized Moreau-Broto autocorrelation, m represents the number of amino acid residues, P i Indicates the amino acid index i amino acid value; Step 5, converting the amino acid composition description, vector description, and autocorrelation description into a numerical feature vector; Step 6: Input the numerical feature vector into the random forest based classifier for training.

[0008] In an optional embodiment, obtaining a toxin protein dataset specifically includes: Collect several positive and negative samples from open source databases; The final toxin protein dataset was obtained by removing similar samples with a similarity threshold of 80% using CD-HIT.

[0009] In an optional embodiment, after converting the amino acid composition description, the vector description, and the autocorrelation description into a numerical feature vector, the following steps are further included: The numerical feature vector is optimized by using the variance analysis method of incremental feature selection and gradient boosting decision tree to obtain the optimized numerical feature vector.

[0010] In an optional embodiment, when the numerical feature vector is input into a random forest-based classifier for training, a cross-validation method is used to test the toxin protein recognition model.

[0011] In an optional embodiment, after training the constructed toxin protein recognition model, the following steps are further included: The performance of the trained toxin protein recognition model was evaluated using sensitivity, specificity, accuracy and Matthews correlation coefficient.

[0012] In a second aspect, the technical solution of the present invention provides an aquatic cnidarian toxin protein recognition system, comprising: Model building unit: used to build a toxin protein recognition model, the toxin protein recognition model includes a feature encoding module and a classifier module, the feature encoding module uses amino acid composition, word2vec technology, and normalized Moreau-Broto autocorrelation description to encode toxin proteins to obtain encoding features, and the classifier module classifies the encoding features to identify aquatic cnidarian toxin proteins; Model training unit: used to train the constructed toxin protein recognition model; Toxin protein recognition unit: used to identify the target toxin protein using the trained toxin protein recognition model to determine whether the target toxin protein is an aquatic cnidarian toxin protein.

[0013] In an optional embodiment, it also includes: Model evaluation unit: used to evaluate the performance of the trained toxin protein recognition model using sensitivity, specificity, accuracy and Matthews correlation coefficient.

[0014] In a third aspect, the technical solution of the present invention provides a terminal, including: A memory for storing a program for identifying aquatic cnidarian toxin proteins; A processor is used to implement the steps of the aquatic cnidarian toxin protein identification method as described in any of the above items when executing the aquatic cnidarian toxin protein identification program.

[0015] In a fourth aspect, the technical solution of the present invention provides a computer-readable storage medium, on which is stored an aquatic cnidarian toxin protein identification program, which, when executed by a processor, implements the steps of the aquatic cnidarian toxin protein identification method as described in any one of the above items.

[0016] The present invention provides a method, system, terminal and medium for identifying aquatic cnidarian toxin proteins, which have the following beneficial effects compared with the prior art: firstly, by constructing a toxin protein recognition model including a feature coding module and a classifier module, key feature information of the toxin protein can be efficiently extracted and utilized; secondly, the feature coding module integrates amino acid composition, word2vec technology and normalized Moreau-Broto autocorrelation descriptors; these multi-dimensional feature information can more comprehensively and accurately reflect the biological characteristics of the toxin protein, thereby improving the accuracy of identifying aquatic cnidarian toxin proteins; furthermore, by constructing and training a toxin protein recognition model, rapid and accurate recognition of the target toxin protein can be achieved, the experimental process can be simplified, and the experimental cost can be reduced, providing a more economical and convenient method for the research and application of toxin proteins. BRIEF DESCRIPTION OF THE DRAWINGS In order to more clearly illustrate the embodiments of the present invention or the technical solutions of 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 some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a schematic flow chart of a method for identifying aquatic cnidarian toxin protein provided by an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of the principles of the toxin protein recognition model training process.

[0019] Figure 3 It is a diagram showing the prediction accuracy of the best model built using different numbers of features.

[0020] Figure 4 is a schematic diagram of the AUC evaluation results of different classifiers, where Figure 4 (A) is a schematic diagram of the AUC evaluation results of different classifiers on the training set. Figure 4 (B) is a schematic diagram of the AUC evaluation results of different classifiers on independent datasets.

[0021] Figure 5 It is a schematic block diagram of the structure of an aquatic cnidarian toxin protein recognition system provided by an embodiment of the present invention.

[0022] Figure 6 It is a schematic diagram of the structure of a terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the scheme of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0025] The key terms appearing in the present invention are explained below.

[0026] word2vec: An effective tool for converting discrete data into numerical data, namely word embedding. It is based on a neural network language model and maps each word to a fixed-length vector space through training. These vectors can represent the relationship between words.

[0027] Moreau-Broto: The Moreau-Broto autocorrelation descriptor is a mathematical tool used to describe the characteristics of a signal or sequence.

[0028] IFS: Incremental Feature Selection, incremental feature selection.

[0029] ANOVA: Analysis of Variance, a statistical method used to analyze the contribution of variations from different sources to the total variation, thereby determining the influence of controllable factors on research results.

[0030] GBDT: Gradient Boosted Decision Tree.

[0031] RF: Random Forest.

[0032] CD-HIT: Cluster Database at High Identity with Tolerance, is a popular tool for removing redundant (highly similar) sequences from sequence collections.

[0033] Figure 1 1 is a schematic flow chart of a method for identifying aquatic cnidarian toxin protein provided by an embodiment of the present invention. Figure 1The execution subject may be an aquatic cnidarian toxin protein identification system. The aquatic cnidarian toxin protein identification method provided in the embodiment of the present invention is executed by a computer device, and accordingly, the aquatic cnidarian toxin protein identification system runs in the computer device. According to different requirements, the order of the steps in the flowchart can be changed, and some can be omitted.

[0034] like Figure 1 As shown, the method includes the following steps.

[0035] S1, Construction of toxin protein recognition model.

[0036] The toxin protein recognition model of this embodiment includes a feature encoding module and a classifier module. The feature encoding module uses amino acid composition, word2vec technology, and normalized Moreau-Broto autocorrelation description to encode toxin proteins to obtain encoding features, and the classifier module classifies the encoding features to identify aquatic cnidarian toxin proteins.

[0037] Selecting feature encodings that are informative and independent is a key step in establishing a machine learning model. Representing protein sequences as mathematical expressions is very important for identifying functional elements. In computational biology, some protein sequence encoding strategies such as physicochemical properties, binary encoding, amino acid pair spectrum encoding, and natural vectors have been applied here. In order to extract as much sequence information of aquatic cnidarian toxin proteins as possible, this embodiment uses the embedding technology 'word2vec' to describe the sequence of aquatic cnidarian toxin proteins, while fusing amino acid composition and normalized Moreau-Broto autocorrelation descriptors. These multi-dimensional feature information can more comprehensively and accurately reflect the biological characteristics of toxin proteins, thereby improving the accuracy of identifying aquatic cnidarian toxin proteins.

[0038] S2, training the constructed toxin protein recognition model.

[0039] In this embodiment, samples are extracted from a public database to form a data set, and the data set is used to train the toxin protein recognition model constructed in step S1 to obtain a toxin protein recognition model with optimal performance.

[0040] S3, using the trained toxin protein recognition model to identify the target toxin protein to determine whether the target toxin protein is an aquatic cnidarian toxin protein.

[0041] For the target toxin protein to be identified, the trained toxin protein recognition model is used for identification to determine whether the target toxin protein is an aquatic cnidarian toxin protein.

[0042] The following is a detailed description of the training of the toxin protein recognition model. Figure 2This is a schematic diagram of the training process of the toxin protein recognition model. Figure 2 As shown, the toxin protein recognition model training includes the following steps.

[0043] Step 1: Obtain a cnidarian toxin protein dataset and divide it into a training dataset and a test dataset.

[0044] First, 297 positive samples (referring to cnidarian toxin protein samples) and 389 negative samples (referring to non-cnidarian toxin protein samples) were collected from the open source database UniProt, and then CD-HIT was used to remove highly similar sequences with an 80% similarity threshold. Finally, 119 cnidarian toxin proteins and 263 non-cnidarian toxin proteins were obtained. In addition, an independent dataset including 63 cnidarian toxin protein sequences and 87 non-cnidarian toxin protein sequences was established to test the prediction model.

[0045] Step 2: Describe the amino acid composition of the toxin protein in the dataset using the following formula:

[0046] In the formula, N(x) In the protein sequence x The number of amino acid residues, the total number of residues is N .

[0047] Amino acid composition (AAC) refers to the frequency of amino acid residues in a protein sequence. The frequencies of the 20 residues f(x) It can be calculated by the above formula.

[0048] Step 3: Use word2vec technology to vectorize the toxin protein sequence in the dataset.

[0049] 'Word2Vec' is a natural language processing scheme that uses a neural network model to generate a distributed representation of vocabulary in a corpus. In this technique, word embedding is used to represent the vector of a word. Naturally, the closer the real-valued vectors encoding the meaning of words are in the vector space, the more similar their meanings are. There are two model designs for 'word2vec': continuous bag of words (CBOW) and continuous skip-gram model (CSG). The basic concept of the continuous bag of words model is to use the context vocabulary of a word (in the adjacent window) to predict the word, while the main concept of the skip-gram model is to use a word to predict its adjacent vocabulary. Algorithmically, the structure of the continuous bag of words model is conducive to organizing the information in the data set in an orderly manner. Therefore, in this embodiment, the continuous bag of words model is used to train the contextual similarity of the aquatic cnidarian toxin protein sequence. Finally, the dimension of the 'word2vec' embedding feature vector is 200.

[0050] Step 4: The amino acid values ​​of the toxin protein in the dataset are described by the normalized Moreau-Broto autocorrelation method. The formula is: m = 1,2,3,…,20; In the formula, Q(m) represents the normalized Moreau-Broto autocorrelation, m represents the number of amino acid residues, P i Indicates the amino acid index i Amino acid value.

[0051] The normalized Moreau-Broto autocorrelation descriptor is an autocorrelation function, similar to the Moran descriptor, and its mathematical expression is as shown above.

[0052] Step 5, convert the amino acid composition description, vector description, and autocorrelation description into numerical feature vectors.

[0053] Redundancy in the feature vector may lead to suboptimal performance. Therefore, selecting ideal features is an important step to eliminate irrelevant features and improve model efficiency. There are many feature selection and ranking methods to optimize features, such as ANOVA, F-score, mRMR, chi-square test, GBDT, and LGBM. ANOVA is a well-known choice that can overcome these problems because it is time-saving and produces effective results. Merging the best performing features does not guarantee the best results because these features may have high redundancy, resulting in redundant information in the features. Therefore, GBDT is an ideal choice to overcome these obstacles. In this embodiment, ANOVA and GBDT with IFS are used to obtain the best feature subset that can produce the maximum accuracy. Figure 3 It is a diagram showing the prediction accuracy of the best model built using different numbers of features.

[0054] In this example, IFS ANOVA and GBDT are used to obtain the best feature subset that produces the greatest accuracy. When using statistical analysis and machine learning techniques to optimize feature subsets to improve model accuracy, the combination of incremental feature selection (IFS) method, analysis of variance (ANOVA) and gradient boosted decision tree (GBDT) can effectively improve feature optimization.

[0055] First, ANOVA is used to perform univariate variance analysis on each candidate feature and the target variable for preliminary feature selection. Then, incremental feature selection is performed, starting from the features screened by ANOVA. Methods such as forward selection, backward elimination, or stepwise regression can be used to iteratively add or remove features, and model performance is evaluated at each iteration. Finally, GBDT is used for feature evaluation and final selection. For example, in each iteration step of IFS, a gradient boosted decision tree (GBDT) is used as a model to evaluate the performance of the current feature subset.

[0056] Step 6: Input the numerical feature vector into the random forest based classifier for training.

[0057] Classification is a form of supervised learning and plays an important role in decision making. In this example, RF was selected to build a model for identifying lipocalin and compared with three machine learning methods, namely Naive Bayes (NB), Support Vector Machine (SVM) and Ada boost (AB). RF is a well-known ensemble learning algorithm in bioinformatics, and its key principle is to combine multiple weak classifiers. The results of the model are obtained through a voting process, so its results have higher accuracy and generalization. The relevant brief description provides detailed information. In order to adjust the hyperparameters, random search CV was first used, followed by grid search CV. The task was completed using Scikit-learn software version 0.22.2, and Table 1 summarizes the ideal parameters of the constructed model. All experiments were run on a Windows operating system and an Intel Core i10 processor at 1.9 GHz. Table 1: Ideal parameters of the constructed model

[0058] Exemplarily, a 5-fold cross validation (CV) test was used to improve its equivalent machine learning limit. In 5-fold CV, the benchmark dataset was randomly divided into five groups of approximately equal size. Each group was independently tested by a model that was trained on the remaining four groups. Therefore, the 5-fold CV method was performed five times, and the average of their results was the final result. The AUC of the best model was 0.917, recorded in Figure 4 (A) and Table 2.

[0059] In this example, sensitivity (Sn), specificity (Sp), accuracy (Acc) and Matthews correlation coefficient (MCC) are used to evaluate the overall efficiency of the model, as defined in the following formula.

[0060]

[0061]

[0062]

[0063]

[0064] Where TP represents the cnidarian toxin proteins correctly identified in the benchmark data, and FP represents the cnidarian toxin proteins misclassified as non-cnidarian toxin proteins. Similarly, TN represents the non-cnidarian toxin proteins correctly identified in the data, and FN represents the non-cnidarian toxin proteins misclassified as cnidarian toxin proteins. Therefore, the receiver operating characteristic (ROC) curve is used to vividly demonstrate the efficiency of the model. The ROC curve can evaluate the outstanding ability of the proposed model on the result value. The performance of the model is evaluated by calculating the area under the curve (AUC). A good classifier has an AUC=1, while random performance has an AUC=0.5.

[0065] The numerical feature vectors were input into three machine learning classifiers, namely SVM, NB and AB, and compared with the RF-based model. Finally, each classifier achieved the ideal model based on AUC, which is shown in Figure 4 (B) and Table 2.

[0066] Table 2: Model performance of different classifiers on training sets and independent datasets

[0067] In this example, protein sequences were encoded using amino acid composition, distributed embedding techniques, and normalized Moreau-Broto autocorrelation, and then optimized using ANOVA and GBDT's IFS technique. Subsequently, these optimized encoding features were input into a RF-based classifier, and the best model was obtained using a 5-fold CV test. The estimation results of independent data showed that the constructed model exhibited excellent performance and good generalization ability. An embodiment of a method for identifying aquatic cnidarian toxin protein is described in detail above. Based on the method for identifying aquatic cnidarian toxin protein described in the above embodiment, an embodiment of the present invention also provides an aquatic cnidarian toxin protein identification system corresponding to the method.

[0068] Figure 5 This is a schematic block diagram of the structure of an aquatic cnidarian toxin protein identification system provided by an embodiment of the present invention. In this embodiment, the aquatic cnidarian toxin protein identification system 500 can be divided into multiple functional units according to the functions performed. The unit referred to in the present invention refers to a series of computer program segments that can be executed by at least one processor and can complete fixed functions, which are stored in a memory.

[0069] Model construction unit 510: used to construct a toxin protein recognition model, wherein the toxin protein recognition model includes a feature encoding module and a classifier module. The feature encoding module uses amino acid composition, word2vec technology, and normalized Moreau-Broto autocorrelation description to encode toxin proteins to obtain encoding features, and the classifier module classifies the encoding features to identify aquatic cnidarian toxin proteins.

[0070] Model training unit 520: used to train the constructed toxin protein recognition model.

[0071] The toxin protein recognition unit 530 is used to recognize the target toxin protein using the trained toxin protein recognition model to determine whether the target toxin protein is an aquatic cnidarian toxin protein.

[0072] Model evaluation unit 540: used to evaluate the performance of the trained toxin protein recognition model using sensitivity, specificity, accuracy and Matthews correlation coefficient.

[0073] The aquatic cnidarian toxin protein recognition system of the present embodiment is used to implement the aforementioned aquatic cnidarian toxin protein recognition method. Therefore, the specific implementation methods of the system can be seen in the embodiment section of the aquatic cnidarian toxin protein recognition method in the previous text. Therefore, its specific implementation methods can refer to the descriptions of the corresponding embodiments of each part and will not be elaborated here.

[0074] In addition, since the aquatic cnidarian toxin protein recognition system of the present embodiment is used to implement the aforementioned aquatic cnidarian toxin protein recognition method, its function corresponds to that of the aforementioned method and will not be described in detail here.

[0075] Figure 6 A schematic diagram of the structure of a terminal 600 provided in an embodiment of the present invention includes: a processor 610, a memory 620 and a communication unit 630. The processor 610 is used to implement the following steps when implementing the aquatic cnidarian toxin protein recognition program stored in the memory 620: Constructing a toxin protein recognition model, the toxin protein recognition model includes a feature encoding module and a classifier module, the feature encoding module uses amino acid composition, word2vec technology, and normalized Moreau-Broto autocorrelation description to encode toxin proteins to obtain encoding features, and the classifier module classifies the encoding features to identify aquatic cnidarian toxin proteins; Training the constructed toxin protein recognition model; The trained toxin protein recognition model is used to identify the target toxin protein to determine whether the target toxin protein is an aquatic cnidarian toxin protein.

[0076] The present invention also provides a computer storage medium, wherein the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0077] The computer storage medium stores an aquatic cnidarian toxin protein identification program, and when the aquatic cnidarian toxin protein identification program is executed by a processor, the following steps are implemented: Constructing a toxin protein recognition model, the toxin protein recognition model includes a feature encoding module and a classifier module, the feature encoding module uses amino acid composition, word2vec technology, and normalized Moreau-Broto autocorrelation description to encode toxin proteins to obtain encoding features, and the classifier module classifies the encoding features to identify aquatic cnidarian toxin proteins; Training the constructed toxin protein recognition model; The trained toxin protein recognition model is used to identify the target toxin protein to determine whether the target toxin protein is an aquatic cnidarian toxin protein.

[0078] The above disclosure is only a preferred embodiment of the present invention, but the present invention is not limited thereto. Any non-creative changes that can be thought of by a person skilled in the art, as well as several improvements and modifications made without departing from the principle of the present invention, should fall within the protection scope of the present invention.

Claims

1. A method for identifying aquatic cnidarian toxin proteins, characterized in that: The following steps are involved: Constructing a toxin protein recognition model, the toxin protein recognition model includes a feature encoding module and a classifier module, the feature encoding module uses amino acid composition, word2vec technology, and normalized Moreau-Broto autocorrelation description to encode toxin proteins to obtain encoding features, and the classifier module classifies the encoding features to identify aquatic cnidarian toxin proteins; Training the constructed toxin protein recognition model; The trained toxin protein recognition model is used to identify the target toxin protein to determine whether the target toxin protein is an aquatic cnidarian toxin protein.

2. The method for identifying aquatic cnidarian toxin proteins according to claim 1, characterized in that: The constructed toxin protein recognition model is trained, including: Step 1, obtain a cnidarian toxin protein dataset and divide it into a training dataset and a test dataset; Step 2: Describe the amino acid composition of the toxin protein in the dataset using the following formula: f(x) = N(x) / N In the formula, N(x) In the protein sequence x The number of amino acid residues, the total number of residues is N ; Step 3, use word2vec technology to describe the toxin protein sequence in the dataset in vector form; Step 4: The amino acid values ​​of the toxin protein in the dataset are described by the normalized Moreau-Broto autocorrelation method. The formula is: m = 1,2,3,…,20； In the formula, Q(m) represents the normalized Moreau-Broto autocorrelation, m represents the number of amino acid residues, P i Indicates the amino acid index i amino acid value; Step 5, converting the amino acid composition description, vector description, and autocorrelation description into a numerical feature vector; Step 6: Input the numerical feature vector into the random forest based classifier for training.

3. The method for identifying aquatic cnidarian toxin proteins according to claim 2, characterized in that: Get toxin protein datasets, including: Collect several positive and negative samples from open source databases; The final toxin protein dataset was obtained by removing similar samples with a similarity threshold of 80% using CD-HIT.

4. The method for identifying aquatic cnidarian toxin proteins according to claim 2, characterized in that: After converting the amino acid composition description, vector description, and autocorrelation description into numerical feature vectors, the following steps are also included: The numerical feature vector is optimized using the variance analysis method with incremental feature selection and the gradient boosting decision tree to obtain the optimized numerical feature vector.

5. The method for identifying aquatic cnidarian toxin proteins according to claim 2, characterized in that: The toxin protein recognition model was tested and trained using a cross-validation approach when the numerical feature vectors were input into a random forest-based classifier for training.

6. The method for identifying aquatic cnidarian toxin proteins according to any one of claims 1 to 5, characterized in that: After training the constructed toxin protein recognition model, the following steps are also included: The performance of the trained toxin protein recognition model was evaluated using sensitivity, specificity, accuracy and Matthews correlation coefficient.

7. A system for identifying aquatic cnidarian toxin proteins, characterized in that: include: Model building unit: used to build a toxin protein recognition model, the toxin protein recognition model includes a feature encoding module and a classifier module, the feature encoding module uses amino acid composition, word2vec technology, and normalized Moreau-Broto autocorrelation description to encode toxin proteins to obtain encoding features, and the classifier module classifies the encoding features to identify aquatic cnidarian toxin proteins; Model training unit: used to train the constructed toxin protein recognition model; Toxin protein recognition unit: used to identify the target toxin protein using the trained toxin protein recognition model to determine whether the target toxin protein is an aquatic cnidarian toxin protein.

8. The aquatic cnidarian toxin protein recognition system according to claim 7, characterized in that: Also includes, Model evaluation unit: used to evaluate the performance of the trained toxin protein recognition model using sensitivity, specificity, accuracy and Matthews correlation coefficient.

9. A terminal, characterized in that: include: A memory for storing a program for identifying aquatic cnidarian toxin proteins; A processor is used to implement the steps of the aquatic cnidarian toxin protein identification method as described in any one of claims 1 to 6 when executing the aquatic cnidarian toxin protein identification program.

10. A computer-readable storage medium, characterized in that: The readable storage medium stores an aquatic cnidarian toxin protein identification program, and when the aquatic cnidarian toxin protein identification program is executed by a processor, the steps of the aquatic cnidarian toxin protein identification method according to any one of claims 1 to 6 are implemented.

Images