Quick antifreeze peptide prediction method and device based on position coding

AFPepPredictor, an anti-freeze peptide prediction tool, was constructed through a bidirectional long and short-term memory network algorithm based on location encoding, which solved the problems of low efficiency and high cost of anti-freeze peptide prediction in the existing technology, and achieved rapid and efficient anti-freeze activity prediction, which is suitable for Food industry and biomedicine field.

CN118898031AInactive Publication Date: 2024-11-05OCEAN UNIV OF CHINA

Patent Information

Application Number
CN202411378899.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-30
Publication Date
2024-11-05
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

There is a lack of efficient antifreeze peptide prediction model in the prior art, and the traditional methods are inefficient and costly, making it difficult to meet the needs of the food industry and biomedicine field.

Method used

Using a bidirectional long and short-term memory network algorithm based on location encoding, the anti-freeze peptide prediction tool AFPepPredictor is constructed through data augmentation and deep learning models, and the peptide sequence prediction is predicted using amino acid characteristics, which improves the accuracy and efficiency of prediction.

Benefits of technology

It realizes rapid and efficient screening and prediction of anti-freeze activity short peptides, reduces R&D costs, provides more accurate prediction results for anti-freeze peptides, and is suitable for industrial applications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118898031A_ABST
    Figure CN118898031A_ABST
Patent Text Reader

Abstract

The invention discloses an antifreeze peptide rapid prediction method and device based on position coding, and relates to the technical field of bioactive peptides. The method is used for screening, predicting and controlling the antifreeze activity of natural oligopeptides in advance, and verifying the activity of the predicted polypeptide through various experiments. Compared with the prior art, the method has the advantages that the oligopeptide with anti-freezing activity can be quickly, efficiently and accurately screened, the research and development cost is greatly reduced, the method can be widely applied to exploration and prediction of the anti-freezing oligopeptide, and the technical and cost problems of complex process, high price and large error in the prior art are solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention discloses a method and a device for quickly predicting antifreeze peptides based on position coding, and relates to the technical field of bioactive peptides. Background Art

[0002] Cryogenics has extremely wide applications in the fields of food industry and biomedicine, such as food cryopreservation, frozen processing, blood and organ preservation, storage of vaccines and biological products, etc. However, the growth of ice crystals during cryopreservation and recrystallization during low-temperature transportation will inevitably cause mechanical damage to organisms. In addition, during the freezing process, the formation of extracellular ice will cause the water in the cells to overflow, and will also cause the cells to be damaged by osmotic damage due to excessive contraction.

[0003] At present, the main preventive measure is to add antifreeze before freezing. Phosphates and sugars are commonly used in food, and glycerol and dimethyl sulfoxide are commonly used in biomedicine. However, phosphates may bring alkaline and astringent taste to food, and excessive phosphate intake will endanger the health of consumers. The high calories of sugars do not conform to the current low-sugar consumption trend. Substances such as glycerol and dimethyl sulfoxide may have certain toxicity to cells. Antifreeze proteins are a class of biological macromolecules that can prevent body fluids from freezing in cold environments. They can inhibit the growth and recrystallization of ice crystals to protect organisms from ice crystals. However, due to their long molecular sequences and complex tertiary spatial structures, they are difficult to produce by artificial methods, resulting in high costs for their extraction and production processes, and difficult to be used on a large scale in the industry. Generally, a chain of less than 50 amino acids is called a polypeptide. Studies have shown that only specific polypeptide chain domains in antifreeze proteins, rather than the entire protein molecule, exhibit antifreeze activity. Therefore, compared with antifreeze proteins, antifreeze peptides show more advantages in industrial applications.

[0004] Antifreeze peptides are polypeptides that can lower the freezing point of intercellular fluids in some organisms such as fish, insects, and spiders that grow in cold regions. Studies have shown that antifreeze peptides can inhibit the growth of ice crystals in food during freezing, maintain the freshness and taste of food, and extend the shelf life of food. Antifreeze peptides also help maintain the activity of cells and tissues, which is crucial for medical procedures such as organ transplantation. Antifreeze peptides not only have excellent antifreeze activity, but also have relatively stable physical and chemical properties. Therefore, antifreeze peptides have broad application prospects in the food industry and biomedicine.

[0005] Traditionally, the active peptides need to be obtained by enzymatic hydrolysis or microbial fermentation of the protein, and then the activity is evaluated by combining purification, characterization, in vivo and in vitro experiments. This method is not only very time-consuming and economically expensive, but also inefficient. In order to meet the needs of the food industry and the biomedical field for antifreeze peptides, active peptide prediction tools constructed using artificial intelligence technology are increasingly becoming the preferred method for quickly predicting active peptides. However, the models currently reported are all prediction models for antifreeze proteins, and no research reports or development of prediction models for antifreeze peptides have been reported. Prior art CN115785224A provides an antifreeze peptide screening method based on molecular docking, but the prediction method based on molecular docking is inefficient and not accurate. Therefore, the development of an efficient antifreeze peptide prediction method is a technical problem that needs to be solved urgently. Summary of the invention

[0006] In view of the problems of the prior art, the present invention provides a method and device for rapid prediction of antifreeze peptides based on position coding, and the technical solution adopted is: In a first aspect, a method for rapid prediction of antifreeze peptides based on position coding is provided, the method comprising: S1, collect antifreeze peptide and non-antifreeze peptide data, and construct training set and test set respectively; S2, based on data enhancement and bidirectional long short-term memory network algorithm, the antifreeze peptide prediction tool AFPepPredictor was established; S3, predicting antifreeze peptides according to the antifreeze peptide prediction tool AFPepPredictor to obtain antifreeze activity; S4, verifying the antifreeze performance of the antifreeze activity by measuring.

[0007] In some implementations, the S1 specifically includes: S11, defining an empty list, storing the feature vectors of amino acids through the empty list, and encoding the amino acids to form an array; S12, according to the polypeptide sequence, obtaining a corresponding feature array through the parameter data of the amino acids, and merging them into a multidimensional array, wherein the multidimensional array includes the polypeptide sequence features; S13, according to the multidimensional array, training is performed through train_test_split and the training set and the test set are divided into 8:2; S14, sampling positive samples and negative samples through a data enhancement algorithm, randomly selecting existing samples through an oversampling function and adding random noise to increase the number of samples; merging the original features and labels of the oversampled positive samples and negative samples with the training set.

[0008] In some implementations, the S2 specifically includes: S21, building and training a deep learning model through the Keras library according to the training set; S22, compiling a model using an Adam optimizer and a binary cross entropy loss function, training the deep learning model according to the training set, and verifying it using the test set; S23, calculating the loss and accuracy through the model.evaluate function to evaluate the performance of the deep learning model; S24, according to the deep learning model, predict and evaluate the test set, and command the deep learning model as an antifreeze peptide prediction tool AFPPepPredictor.

[0009] In some implementations, the S12 specifically includes: S121, obtaining the AAC characteristics of the polypeptide according to the relative content of the amino acids in the polypeptide sequence; S122, obtain the DPC characteristics of the peptide based on the relative frequency of each pair of adjacent amino acids; S123, according to the polypeptide sequence, the CKS characteristics of the polypeptide are obtained through the conservative amino acid combination at each amino acid position; S124, based on the peptide sequence, obtains the K-mer features of the peptide through the dipole character and side chain volume of the amino acids; S125, merging the AAC features, DPC features, CKS features and K-mer features into a multidimensional array.

[0010] In a second aspect, an embodiment of the present invention provides a device for rapid prediction of antifreeze peptides based on position coding, the device comprising: The dataset module is used to collect antifreeze peptide and non-antifreeze peptide data and construct training and test sets respectively; The prediction building module is used to build the antifreeze peptide prediction tool AFPepPredictor based on data enhancement and bidirectional long short-term memory network algorithm; A performance prediction module, used to predict antifreeze peptides according to the antifreeze peptide prediction tool AFPepPredictor to obtain antifreeze activity; The performance verification module is used to verify the antifreeze performance of the antifreeze activity through measurement.

[0011] In some implementations, the data set module specifically includes: An encoding unit, used to define an empty list, store the feature vectors of amino acids through the empty list, and encode the amino acids to form an array; An array merging unit, used to obtain corresponding feature arrays according to the polypeptide sequence through the parameter data of the amino acids, and merge them into a multidimensional array, wherein the multidimensional array includes polypeptide sequence features; A data set partitioning unit, used to perform training through train_test_split according to the multidimensional array and partition the training set and the test set into 8:2; The data set processing unit is used to sample positive samples and negative samples through a data enhancement algorithm, randomly select existing samples through an oversampling function and add random noise to increase the number of samples; and merge the original features and labels of the oversampled positive and negative samples with the training set.

[0012] In some implementations, the prediction building module specifically includes: A model building unit, used to build and train a deep learning model through the Keras library according to the training set; A data set verification unit, used to compile the model through the Adam optimizer and the binary cross entropy loss function, train the deep learning model according to the training set, and verify it through the test set; A model evaluation unit, used to evaluate the performance of the deep learning model by calculating the loss and accuracy through the model.evaluate function; A model processing unit is used to predict and evaluate the test set according to the deep learning model, and command the deep learning model as an antifreeze peptide prediction tool AFPPepPredictor.

[0013] In some implementations, the array merging unit specifically includes: The first characteristic unit is used to obtain the AAC characteristics of the polypeptide according to the relative content of the amino acids in the polypeptide sequence; The second feature unit is used to obtain the DPC feature of the polypeptide according to the relative frequency of each pair of adjacent amino acids; The third feature unit is used to obtain the CKS feature of the polypeptide through the conservative amino acid combination at each amino acid position according to the polypeptide sequence; The fourth feature unit is used to obtain the K-mer features of the polypeptide according to the polypeptide sequence through the dipole character and side chain volume of the amino acid; The fifth feature unit is used to merge the AAC feature, DPC feature, CKS feature and K-mer feature into a multidimensional array.

[0014] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a memory and a processor, wherein the memory is used to store one or more computer instructions, wherein when the one or more computer instructions are executed by the processor, the method described in the first aspect above is implemented.

[0015] In a fourth aspect, an embodiment of the present invention provides a computer storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it is used to implement the method described in the first aspect.

[0016] One or more embodiments of the present invention can bring at least the following beneficial effects: The present invention provides a method and device for quickly predicting antifreeze peptides based on position coding, which are used to screen, predict and control the antifreeze activity of natural short peptides in advance, and verify the activity of the predicted polypeptides through multiple experiments. Compared with the prior art, the present invention has the advantage that it can quickly, efficiently and accurately screen short peptides with antifreeze activity, greatly reducing the research and development cost, and can be widely used in the discovery and prediction of antifreeze short peptides, solving the technical and cost problems of the prior art that the process is complicated, the price is high and the error is large. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0018] Figure 1 A confusion matrix diagram of the antifreeze peptide rapid prediction model constructed by using an oversampling function and a bidirectional long short-term memory network (Bidirectional LSTM, Bi-LSTM) algorithm in Example 2 of the present invention; Figure 2 The figure is a ROC curve diagram of the antifreeze peptide rapid prediction model constructed by the oversampling function and the bidirectional long short-term memory network (Bidirectional LSTM, Bi-LSTM) algorithm in Example 2 of the present invention; Figure 3 The freezing differential scanning calorimetry (DSC) curves of the sucrose solution containing the predicted antifreeze peptide and the control solution at -1.5°C in Experimental Verification Example 1 of the present invention are shown; Figure 4 The freezing DSC curves of the sucrose solution containing the predicted antifreeze peptide and the control solution at -1.7°C in Experimental Verification Example 1 of the present invention; Figure 5 The freezing DSC curves of the sucrose solution containing the predicted antifreeze peptide and the control solution at -2°C in Experimental Verification Example 1 of the present invention; Figure 6 The ice crystal morphology of the control solution without antifreeze peptide in Experimental Verification Example 2 of the present invention, wherein A is the ice crystal morphology before the ice recrystallization inhibition activity of the solution without antifreeze peptide is measured, and B is the ice crystal morphology after the ice recrystallization inhibition activity of the solution without antifreeze peptide is measured; Figure 7 3. The ice crystal morphology of the solution containing the predicted antifreeze peptide in Experimental Verification Example 2 of the present invention, wherein A is the ice crystal morphology of the solution containing the predicted antifreeze peptide before the ice recrystallization inhibition activity is measured, and B is the ice crystal morphology of the solution containing the predicted antifreeze peptide after the ice recrystallization inhibition activity is measured; Figure 8 This is a comparison chart of the average diameter of ice crystals of the control solution and the solution containing the predicted antifreeze peptide in Experimental Verification Example 2 of the present invention; Fig. 9 This is a diagram of the freezing process of the control pure water solution in Experimental Verification Example 3 of the present invention; Fig.10 This is a diagram of the freezing process of the solution containing the predicted antifreeze peptide in Experimental Verification Example 3 of the present invention; Fig.11 This is a comparison chart of the ice crystal growth rates of the control pure water solution and the solution containing the predicted antifreeze peptide in Experimental Verification Example 3 of the present invention; Fig.12 This is a diagram of the gyration radius of water molecules in the control and antifreeze peptide groups in Experimental Verification Example 3 of the present invention; Fig.13 The radial distribution function diagram of oxygen atoms of the control and antifreeze peptide groups in Experimental Verification Example 3 of the present invention; Fig.14 This is a radial distribution function diagram of hydrogen atoms of the control and antifreeze peptide groups in Experimental Verification Example 3 of the present invention. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0020] Embodiment 1:

[0021] The method for rapid prediction of antifreeze peptides based on position coding provided in this embodiment includes: S1, collect antifreeze peptide and non-antifreeze peptide data, and construct training set and test set respectively; S2, based on data enhancement and bidirectional long short-term memory network algorithm, the antifreeze peptide prediction tool AFPepPredictor was established; S3, predicting antifreeze peptides according to the antifreeze peptide prediction tool AFPepPredictor to obtain antifreeze activity; S4, verifying the antifreeze performance of the antifreeze activity by measuring.

[0022] First, according to S1, antifreeze peptides and non-antifreeze peptides were collected from existing Chinese and English papers, and training and test sets were constructed respectively. According to the antifreeze peptide and non-antifreeze peptide data, the peptides of amino acids were retained by removing the peptide segments of amino acids; the antifreeze peptide label was defined as 1, and the non-antifreeze peptide label was defined as 0. Specifically, it includes: S11, define an empty list, store the feature vectors of amino acids through the empty list, encode various amino acids in the polypeptide according to the following table, Table 1, and finally convert them into an array; Table 1 is a coding sequence table of amino acids in polypeptides;

[0023] Next, according to S12, according to the polypeptide sequence, the corresponding feature array is obtained through the parameter data of the amino acids, and is merged into a multidimensional array, wherein the multidimensional array includes the polypeptide sequence features; According to S121, the AAC (Amphiphilic Amino Acid Composition) feature of the polypeptide is obtained by calculating the relative content of various amino acids in the polypeptide sequence; according to S122, the DPC (DiPe Composition) feature of the polypeptide is obtained by calculating the relative frequency of each pair of adjacent amino acids; according to S123, the CKS (Consensus Knowledge Score) feature of the polypeptide is obtained by calculating the probability of conservative amino acid combinations at each amino acid position in the polypeptide sequence; according to S124, the K-mer feature of the polypeptide is obtained by the dipole and side chain volume of the amino acids in the polypeptide sequence; Next, according to S13, according to the multidimensional array, training is performed through train_test_split and the training set and the test set are divided into 8:2; Next, according to S14, the positive samples (antifreeze peptides) and negative samples (non-antifreeze peptides) are oversampled respectively through the data augmentation algorithm, the existing samples are randomly selected through the oversampling function and random noise is added to increase the number of samples, and the original features and labels of the oversampled positive and negative samples are merged with the original training set to improve the generalization ability of the model.

[0024] Next, according to S2, based on data enhancement and bidirectional long short-term memory network algorithm, the antifreeze peptide prediction tool AFPepPredictor was established; specifically, According to S21, a deep learning model is constructed and trained by using the Keras library based on the training set; The deep learning model is built and trained through the Keras library. First, the main input layer is defined to receive the main input data, with an input length of maxlen and a data type of int32. The main input layer is embedded through the Embedding Layer, with an embedding dimension of embedding_dim and an input length of maxlen. The embedded result is processed through a bidirectional long short-term memory network (Bidirectional LSTM, Bi-LSTM), and the output dimension is 64. The auxiliary input layer is defined to receive the auxiliary input data through the auxiliary input layer. The Bi-LSTM output and the auxiliary input are concatenated through the concatenation layer. The concatenation result is subjected to Dropout processing by setting the Dropout layer (DropoutLayer) to prevent overfitting, with a Dropout rate of 0.5. The fully connected layer is set to perform 4 full-connection processing on the concatenated result, which is connected to the fully connected layer of 64 neurons, and ReLU is used as the activation function. The main output layer is defined. Layer), the output dimension is 1, and Sigmoid is used as the activation function; Next, according to S22, the model is compiled by using the Adam optimizer and the binary cross entropy loss function, the deep learning model is trained according to the training set, and the deep learning model is verified by the test set; Next, S23, after the model training is completed, the model.evaluate function is used to calculate the loss and accuracy to evaluate the model performance. The accuracy is the proportion of samples that the model predicts correctly on the given data. Finally, the ROC curve is drawn using the matplotlib function. The roc_curve function is used to calculate the FPR (false positive rate) and TPR (true positive rate). The auc(fpr, tpr) function is used to calculate the AUC value. The calculation formula is as follows:

[0025] True positive TP, false positive FP, false negative FN, true negative TN, M positive samples, N negative samples; Next, according to S24, according to the deep learning model, prediction and evaluation are performed on the test set, and the deep learning model is commanded as an antifreeze peptide prediction tool AFPPepPredictor.

[0026] Use the model to predict and evaluate the test set and call the model as AFPPepPredictor.

[0027] In some implementations, the S12 specifically includes: S121, obtaining the AAC characteristics of the polypeptide according to the relative content of the amino acids in the polypeptide sequence; S122, obtain the DPC characteristics of the peptide based on the relative frequency of each pair of adjacent amino acids; S123, according to the polypeptide sequence, the CKS characteristics of the polypeptide are obtained through the conservative amino acid combination at each amino acid position; S124, based on the peptide sequence, obtains the K-mer features of the peptide through the dipole character and side chain volume of the amino acids; S125, merging the AAC features, DPC features, CKS features and K-mer features into a multidimensional array.

[0028] Embodiment 2:

[0029] On the basis of Example 1, this example collects and summarizes the antifreeze peptide database, uses the enhanced sampling method to divide the training set and the test set, and then uses the Keras library to build a deep learning model including a main input layer, an embedding layer, a bidirectional LSTM layer, an auxiliary input layer, a concatenation layer, a Dropout layer and a fully connected layer. The main output layer with a Sigmoid activation function is combined to train a binary classification model of highly active antifreeze peptides; finally, the model is predicted and evaluated in the test set, and the best model is obtained and named AFPPepPredictor.

[0030] Reference Figure 1: Confusion matrix diagram of the antifreeze peptide rapid prediction model constructed by the oversampling function and the Bidirectional LSTM (Bi-LSTM) algorithm: The prediction accuracy of antifreeze peptides is 0.75, and the prediction accuracy of non-antifreeze peptides is 0.85, indicating that the model obtained in this example has a higher prediction performance.

[0031] Reference Figure 2 : ROC curve of the antifreeze peptide rapid prediction model constructed by the oversampling function and the bidirectional long short-term memory network (Bidirectional LSTM, Bi-LSTM) algorithm. From the figure, we can see that the AUC value of the algorithm is 0.88. The closer the AUC value is to 1, the stronger the classification ability of the model is, and the more it can distinguish between positive and negative classes. Therefore, this algorithm has a high accuracy in distinguishing antifreeze peptides from non-antifreeze peptides.

[0032] Experimental verification example 1: The deep learning model of the above embodiment 1 is verified: 1. Peptide prediction and experimental verification (1) Random generation and prediction of peptides Ten peptide sequences were randomly constructed, and the antifreeze peptide prediction model AFPPepPredictor was used for activity prediction. The prediction showed that Gly-Lys-Gly-Pro-Lys-Gly-Pro-Gly-Pro-Gln-Phe-Thr-Pro-Arg of the 10 peptide sequences had antifreeze activity.

[0033] (2) Purchase of raw materials The peptides were synthesized using solid-phase peptide synthesis technology, and the pure peptides (≥98%) were provided by Sangon Biotech (Shanghai) Co., Ltd.

[0034] (3) DSC thermal hysteresis curve determination The experiment was carried out using NETZSCH DSC300. 10 μL of 20 mg / mL antifreeze peptide solution was sealed in an aluminum crucible and heated from -25°C to 25°C at a rate of 5°C / min to determine the solid-liquid coexistence temperature, i.e., the retention temperature. The antifreeze peptide solution was then heated from -25°C to the retention temperature (-1.5°C, -1.7°C, -2°C), maintained for 5 min, and then cooled to -10°C at a rate of 1°C / min.

[0035] 2. For step S3, the thermal hysteresis curve of Gly-Lys-Gly-Pro-Lys-Gly-Pro-Gly-Pro-Gln-Phe-Thr-Pro-Arg was measured. The results are as follows: Figure 3 (-1.5℃), Figure 4 (-1.7℃) and Figure 5 As shown in Figure 2, compared with the solution without antifreeze peptide, the cooling curve of the solution containing the predicted antifreeze peptide showed obvious hysteresis at three different retention temperatures. The above results indicate that the predicted antifreeze peptide has significant antifreeze activity.

[0036] Experimental verification example 2: The deep learning model of the above embodiment 1 is verified: 1. Peptide prediction and experimental verification (1) Random generation and prediction of peptides Ten peptide sequences were randomly constructed, and the antifreeze peptide prediction model AFPPepPredictor was used to predict their activity. The prediction results showed that the peptide sequence Gly-Lys-Gly-Pro-Lys-Gly-Pro-Gly-Pro-Gln-Phe-Thr-Pro-Arg had antifreeze activity.

[0037] (2) Purchase of raw materials The peptides were synthesized using solid-phase peptide synthesis technology, and the pure peptides (≥98%) were provided by Sangon Biotech (Shanghai) Co., Ltd.

[0038] (3) Determination of ice recrystallization inhibition activity The antifreeze peptide was dissolved in a 200 mg / mL sucrose solution to prepare a 20 mg / mL sample solution. 2 μL of the sample was titrated on a glass slide and covered with a glass coverslip. The slide was quickly frozen to -20°C at a rate of 20°C / min and kept for 1 min. The sample temperature was then raised to -6°C at a rate of 10°C / min and then kept for 3 min, and the ice crystal image was recorded.

[0039] 2. For step S3, the ice recrystallization inhibitory activity of Gly-Lys-Gly-Pro-Lys-Gly-Pro-Gly-Pro-Gln-Phe-Thr-Pro-Arg was determined. The results are shown in the attached Figure 6 and attached Figure 7 As shown in Figure 2, compared with the solution without antifreeze peptides, the predicted antifreeze peptides showed significant ice recrystallization inhibition activity. Under the microscope, it can be found that the ice crystals containing the antifreeze peptide group are kept at -6°C for a period of time ( Figure 7 B), which is significantly higher than the ice crystals in the group without antifreeze peptides ( Figure 6 B) is smaller and denser. Figure 8 As shown, the average diameter of ice crystals in the group without antifreeze peptides is about twice that of the group with antifreeze peptides. The above results indicate that the predicted antifreeze peptides have significant antifreeze activity.

[0040] Experimental verification example 3: The deep learning model of the above embodiment 1 is verified: 1. Construction of peptide model and simulation system (1) constructing the structural model of the above-mentioned Gly-Lys-Gly-Pro-Lys-Gly-Pro-Gly-Pro-Gln-Phe-Thr-Pro-Arg by Alphafold3, and obtaining the pdb file. The website of the Alphafold3 server is: https: / / alphafoldserver.com / ; (2) constructing an ice structure model using Genice2 software. The download address of Genice2 is: https: / / pypi.org / project / genice2 / ; (3) VMD software was used to combine the peptide structure without / with the above Gly-Lys-Gly-Pro-Lys-Gly-Pro-Gly-Pro-Gln-Phe-Thr-Pro-Arg with ice and water molecules into a system. The initial state is as follows: Fig. 9 0ns and Fig.10 The download address of the VMD software is: https: / / www.ks.uiuc.edu / Development / Download / download.cgi?PackageName=VMD; (4) The above structures were imported into Gromacs software and simulated for 200 ns respectively, and then the results were processed by Gromacs and VMD. The download address of the Gromacs software is: https: / / www.gromacs.org / ; 2. For step S4, the groups containing Gly-Lys-Gly-Pro-Lys-Gly-Pro-Gly-Pro-Gln-Phe-Thr-Pro-Arg and those not containing the peptide were subjected to ice crystal growth rate analysis, water molecule gyration radius analysis (Rg), oxygen atom radial distribution function analysis, and hydrogen atom radial distribution function analysis, respectively.

[0041] The ice crystal growth process Fig. 9 and Fig.10 As shown in Figure 2, it can be found that the growth rate of ice crystals in the group without antifreeze peptides is significantly faster than that in the group with antifreeze peptides. At the end of the 200 ns simulation, the water molecules in the group without antifreeze peptides ( Fig. 9 ) are almost completely frozen, while the water molecules containing the antifreeze peptide group ( Fig.10 ) Only a small part of it was frozen. The growth rate of ice crystals in the group without antifreeze peptides was about twice that of the group with antifreeze peptides ( Fig.11).

[0042] The results of the water molecule gyration radius analysis (Rg) are as follows Fig.12 As shown, the gyration radius of a molecule can reflect its compactness, and the freezing of water molecules will cause its volume to expand and its compactness to decrease, so the Rg value becomes larger. Since the freezing process of water is inhibited in the group containing antifreeze peptides, the Rg value is smaller than that of the control group. The radial distribution function analysis of oxygen atoms and hydrogen atoms is as follows Fig.13 and Fig.14 As shown in the figure, the radial distribution function (RDF) can reflect the distribution characteristics of other particles around a specific particle. Fig.13 It can be found that the peak intensity of the curve fluctuates significantly near r = 0.3, 0.45 and 0.6; Fig.14 It can be found that the peak intensity of the curve fluctuates significantly near r=0.3, 0.45 and 0.7; this is because the antifreeze peptide inhibits the freezing process of water molecules and changes the distribution state of water molecules in the system. Therefore, compared with the control group, the distribution characteristics between oxygen atoms and hydrogen atoms have changed significantly. The above results show that the predicted antifreeze peptide has significant antifreeze activity.

[0043] The above verification examples show that the predicted polypeptide has antifreeze activity, which indicates that this example can accurately predict whether a polypeptide has antifreeze activity, further proving the effectiveness of the present invention.

[0044] In summary, the present invention provides and develops a rapid prediction method for antifreeze peptides based on deep learning, which has the advantages of high efficiency, convenience and rapidity, and can accurately predict whether an unknown polypeptide has antifreeze activity, providing an efficient, accurate and convenient strategy for the development of antifreeze products. In addition, the present invention provides an experimental verification example to further prove the effectiveness of the developed antifreeze peptide prediction model AFPPepPredictor, which makes the present invention more persuasive.

[0045] Embodiment three: In a second aspect, an embodiment of the present invention provides a device for rapid prediction of antifreeze peptides based on position coding, the device comprising: The dataset module is used to collect antifreeze peptide and non-antifreeze peptide data and construct training and test sets respectively; The prediction building module is used to build the antifreeze peptide prediction tool AFPepPredictor based on data enhancement and bidirectional long short-term memory network algorithm; A performance prediction module, used to predict antifreeze peptides according to the antifreeze peptide prediction tool AFPepPredictor to obtain antifreeze activity; The performance verification module is used to verify the antifreeze performance of the antifreeze activity through measurement.

[0046] In some implementations, the data set module specifically includes: An encoding unit, used to define an empty list, store the feature vectors of amino acids through the empty list, and encode the amino acids to form an array; An array merging unit, used to obtain corresponding feature arrays according to the polypeptide sequence through the parameter data of the amino acids, and merge them into a multidimensional array, wherein the multidimensional array includes polypeptide sequence features; A data set partitioning unit, used to perform training through train_test_split according to the multidimensional array and partition the training set and the test set into 8:2; The data set processing unit is used to sample positive samples and negative samples through a data enhancement algorithm, randomly select existing samples through an oversampling function and add random noise to increase the number of samples; and merge the original features and labels of the oversampled positive and negative samples with the training set.

[0047] In some implementations, the prediction building module specifically includes: A model building unit, used to build and train a deep learning model through the Keras library according to the training set; A data set verification unit, used to compile the model through the Adam optimizer and the binary cross entropy loss function, train the deep learning model according to the training set, and verify it through the test set; A model evaluation unit, used to evaluate the performance of the deep learning model by calculating the loss and accuracy through the model.evaluate function; A model processing unit is used to predict and evaluate the test set according to the deep learning model, and command the deep learning model as an antifreeze peptide prediction tool AFPPepPredictor.

[0048] In some implementations, the array merging unit specifically includes: The first characteristic unit is used to obtain the AAC characteristics of the polypeptide according to the relative content of the amino acids in the polypeptide sequence; The second feature unit is used to obtain the DPC feature of the polypeptide according to the relative frequency of each pair of adjacent amino acids; The third feature unit is used to obtain the CKS feature of the polypeptide through the conservative amino acid combination at each amino acid position according to the polypeptide sequence; The fourth feature unit is used to obtain the K-mer features of the polypeptide according to the polypeptide sequence through the dipole character and side chain volume of the amino acid; The fifth feature unit is used to merge the AAC feature, DPC feature, CKS feature and K-mer feature into a multidimensional array.

[0049] Embodiment 4:

[0050] This embodiment further provides an electronic device, including a memory and a processor, wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions implement the method of embodiment 1 when executed by the processor; In practical applications, the processor can be an application specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a controller, a microcontroller unit (MCU), a microprocessor or other electronic components to execute the methods in the above embodiments.

[0051] The method implemented in this embodiment is as described in Embodiment 1.

[0052] Embodiment five:

[0053] This embodiment further provides a computer storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by one or more processors, the method of the first embodiment is implemented; Among them, the computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0054] The method implemented in this embodiment includes: The method implemented in this embodiment is as described in Embodiment 1.

[0055] In the several embodiments provided in the embodiments of the present invention, it should be understood that the disclosed system and method can also be implemented in other ways. The system and method embodiments described above are merely illustrative.

[0056] It should be noted that, in this article, the terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. The terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a..." does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0057] Although the embodiments disclosed in the present invention are as above, the above contents are only embodiments adopted for facilitating the understanding of the present invention and are not intended to limit the present invention. Any technician in the technical field to which the present invention belongs can make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in the present invention, but the patent protection scope of the present invention shall still be subject to the scope defined in the attached claims.

Claims

1. A rapid prediction method for antifreeze peptides based on position coding, It is characterized in that The method comprises: S1, collect antifreeze peptide and non-antifreeze peptide data, and construct training set and test set respectively; S2, based on data enhancement and bidirectional long short-term memory network algorithm, the antifreeze peptide prediction tool AFPepPredictor was established; S3, predicting antifreeze peptides according to the antifreeze peptide prediction tool AFPepPredictor to obtain antifreeze activity; S4, verifying the antifreeze performance of the antifreeze activity by measuring.

2. The method according to claim 1, It is characterized in that The S1 specifically includes: S11, defining an empty list, storing the feature vectors of amino acids through the empty list, and encoding the amino acids to form an array; S12, according to the polypeptide sequence, obtaining a corresponding feature array through the parameter data of the amino acids, and merging them into a multidimensional array, wherein the multidimensional array includes the polypeptide sequence features; S13, according to the multidimensional array, training is performed through train_test_split and the training set and the test set are divided into 8:2; S14, sampling positive samples and negative samples through a data enhancement algorithm, randomly selecting existing samples through an oversampling function and adding random noise to increase the number of samples; merging the original features and labels of the oversampled positive samples and negative samples with the training set.

3. The method according to claim 2, It is characterized in that The S2 specifically includes: S21, building and training a deep learning model through the Keras library according to the training set; S22, compiling a model using an Adam optimizer and a binary cross entropy loss function, training the deep learning model according to the training set, and verifying it using the test set; S23, calculating the loss and accuracy through the model.evaluate function to evaluate the performance of the deep learning model; S24, predicting and evaluating the test set according to the deep learning model, and commanding the deep learning model as an antifreeze peptide prediction tool AFPPepPredictor.

4. The method according to claim 3, It is characterized in that The S12 specifically includes: S121, obtaining the AAC characteristics of the polypeptide according to the relative content of the amino acids in the polypeptide sequence; S122, obtain the DPC characteristics of the peptide based on the relative frequency of each pair of adjacent amino acids; S123, according to the polypeptide sequence, the CKS characteristics of the polypeptide are obtained through the conservative amino acid combination at each amino acid position; S124, based on the peptide sequence, obtains the K-mer features of the peptide through the dipole character and side chain volume of the amino acids; S125, merging the AAC features, DPC features, CKS features and K-mer features into a multidimensional array.

5. A rapid prediction device for antifreeze peptides based on position coding, It is characterized in that The device comprises: The dataset module is used to collect antifreeze peptide and non-antifreeze peptide data and construct training and test sets respectively; The prediction building module is used to build the antifreeze peptide prediction tool AFPepPredictor based on data enhancement and bidirectional long short-term memory network algorithm; A performance prediction module, used to predict antifreeze peptides according to the antifreeze peptide prediction tool AFPepPredictor to obtain antifreeze activity; The performance verification module is used to verify the antifreeze performance of the antifreeze activity through measurement.

6. The device according to claim 5, It is characterized in that The data set module specifically includes: An encoding unit, used to define an empty list, store the feature vectors of amino acids through the empty list, and encode the amino acids to form an array; An array merging unit, used to obtain corresponding feature arrays according to the polypeptide sequence through the parameter data of the amino acids, and merge them into a multidimensional array, wherein the multidimensional array includes polypeptide sequence features; A data set partitioning unit, used to perform training through train_test_split according to the multidimensional array and partition the training set and the test set into 8:2; The data set processing unit is used to sample positive samples and negative samples through a data enhancement algorithm, randomly select existing samples through an oversampling function and add random noise to increase the number of samples; and merge the original features and labels of the oversampled positive and negative samples with the training set.

7. The device according to claim 6, It is characterized in that The prediction establishment module specifically includes: A model building unit, used to build and train a deep learning model through the Keras library according to the training set; A data set verification unit, used to compile the model through the Adam optimizer and the binary cross entropy loss function, train the deep learning model according to the training set, and verify it through the test set; A model evaluation unit, used to evaluate the performance of the deep learning model by calculating the loss and accuracy through the model.evaluate function; A model processing unit is used to predict and evaluate the test set according to the deep learning model, and command the deep learning model as an antifreeze peptide prediction tool AFPPepPredictor.

8. The device according to claim 7, It is characterized in that The array merging unit specifically includes: The first characteristic unit is used to obtain the AAC characteristics of the polypeptide according to the relative content of the amino acids in the polypeptide sequence; The second feature unit is used to obtain the DPC feature of the polypeptide according to the relative frequency of each pair of adjacent amino acids; The third feature unit is used to obtain the CKS feature of the polypeptide through the conservative amino acid combination at each amino acid position according to the polypeptide sequence; The fourth feature unit is used to obtain the K-mer features of the polypeptide according to the polypeptide sequence through the dipole character and side chain volume of the amino acid; The fifth feature unit is used to merge the AAC feature, DPC feature, CKS feature and K-mer feature into a multidimensional array.

9. An electronic device, It is characterized in that The method comprises a memory and a processor, wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions, when executed by the processor, implement the method as described in any one of claims 1 to 4.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it is used to implement the method as described in any one of claims 1 to 4.

Citation Information

Patent Citations

  • Fish scale antifreeze peptide as well as bioinformatics-based screening method and application thereof

    CN115785224A

  • Prokaryotic acetylation site prediction method based on information fusion and deep learning

    CN111063393A

  • Anticancer peptide prediction method based on bidirectional long-short-term memory network and feature fusion

    CN114863997A

  • Carbon fiber reinforcement construction process

    CN118334269A

  • Method, System and Computer Program Product for Determining Presentation Likelihoods of Neoantigens

    US20230298692A1

Cited By

  • Method for auxiliary screening of grease antioxidant peptides based on artificial intelligence

    CN119694449A