Multi-feature coding fusion-based bitter peptide identification method and device

Abstract

The invention relates to a bitter peptide recognition method and device based on multi-feature coding fusion. According to the application, the peptide structural characteristics and the peptide biochemical characteristics of the protein sequence are combined, so that the characteristics of the protein sequence can be more comprehensively analyzed, and the bitter peptide can be more accurately identified. Besides, by utilizing the combination of a plurality of primary classifiers and secondary classifiers, the over-fitting risk of a single model can be reduced, the generalization ability of the model can be improved, and meanwhile, by combining the prediction results of a plurality of classifiers, the advantages of the classifiers can be integrated, and the prediction precision higher than that of the single classifier under some conditions can be obtained. Finally, compared with a bitter peptide recognition method based on biological experiments, the bitter peptide recognition method is higher in recognition speed and low in cost, and has higher economic value.

Description

Technical Field

[0001] The present application relates to the field of machine learning technology, and in particular to a bitter peptide identification method and device based on multi-feature coding fusion. Background Art

[0002] Bitterness is one of the five basic human senses and plays an important role in daily life. By perceiving bitterness, humans are able to detect potential dangers, and after a long period of evolution, we have gradually formed the perception that bitter substances are difficult to swallow and may be harmful. The ability to perceive bitterness helped our ancestors survive and has become an indispensable part of modern life. Most medicines are bitter, and adding bitter agents to small objects can also effectively prevent infants and young children from accidentally ingesting them. In addition, bitter peptides have been found to have medicinal effects such as regulating blood sugar, anti-oxidation, and promoting cardiovascular health. Given the important role of bitter peptides, it is crucial to accurately determine which peptides have bitter taste. By locating bitter peptides, the drug manufacturing process can be improved to reduce the bitterness of drugs, thereby encouraging patients to be more willing to take medicines, and food manufacturers can achieve the ideal taste profile by adjusting the content of bitter peptides. However, the discovery and identification process of bitter peptides is very complicated, time-consuming and expensive. Therefore, it is crucial to find a method that can predict the bitterness of peptides.

[0003] There are two main directions of existing bitter peptide identification technology: one is identification based on biological experiments, that is, qualitative analysis through experimental methods such as mass spectrometry, liquid chromatography-mass spectrometry, and sensory evaluation; the other is identification based on machine learning or deep learning.

[0004] However, identification based on biological experiments requires wet experimental operations, which are costly and time-consuming, and sensory evaluation is not objective and accurate enough; identification based on machine learning or deep learning relies heavily on traditional feature encoding technology. In addition, most existing predictors are based on SMILES (Simplified molecular input line entry system). The few remaining models that use protein sequence information for prediction have major defects in feature extraction, generalization, and accuracy. Summary of the invention

[0005] Based on this, it is necessary to provide a bitter peptide identification method and device based on multi-feature coding fusion to address the above technical problems, which can quickly, efficiently and accurately identify bitter peptides.

[0006] In the first aspect, the present application provides a bitter peptide identification method based on multi-feature coding fusion. The method comprises:

[0007] The protein sequence is input into the first feature extraction module to obtain peptide structural features;

[0008] The protein sequence is input into the second feature extraction module to obtain the peptide biochemical features;

[0009] Based on the structural characteristics and biochemical characteristics of peptides, at least two first-level classifiers are used to make predictions to obtain the first-level prediction outputs corresponding to each first-level classifier; a combination strategy of each first-level prediction output is generated using a second-level classifier to obtain a target prediction output for locating bitter peptides in protein sequences.

[0010] In one embodiment, the first feature extraction module includes a multi-layer Transformer encoder;

[0011] The protein sequence is input into the first feature extraction module to obtain peptide structural features including:

[0012] The first feature extraction module is used to convert the protein sequence into a numerical vector representation, and a multi-layer encoder is used to capture the features in the protein sequence and the long-term dependencies between the features, and output the peptide structural features.

[0013] In one embodiment, the biochemical characteristics of peptides include: molecular weight, average hydrophobicity index, polarity ratio, isoelectric point, relative abundance of amino acids, frequency of adjacent amino acid pairs and amino acid distribution position characteristics.

[0014] In one embodiment, after obtaining the peptide biochemical characteristics, the method further comprises:

[0015] Recursive feature elimination combined with cross-validation was used to screen the peptide biochemical features, and the retained peptide biochemical features were input into the primary classifier.

[0016] In one embodiment, the first-level classifier includes at least random forest, extreme random trees, and gradient boosting.

[0017] In one embodiment, the method also includes: constructing a recognition model including a first feature extraction module, a second feature extraction module, a primary classifier and a secondary classifier, pre-training the recognition model using training data containing bitter peptide sequences and non-bitter peptide sequences, and using accuracy, sensitivity, specificity, Matthews correlation coefficient and area under the curve as evaluation indicators to evaluate the performance of the recognition model.

[0018] In a second aspect, the present application also provides a bitter peptide recognition device based on multi-feature coding fusion.

[0019] The device includes:

[0020] The first feature extraction module is used to obtain peptide structural features based on protein sequences;

[0021] The second feature extraction module is used to obtain peptide biochemical features based on protein sequences;

[0022] The prediction output module is used to make predictions based on peptide structural features and peptide biochemical features using at least two first-level classifiers to obtain the first-level prediction outputs corresponding to each first-level classifier; a combination strategy of each first-level prediction output is generated using a second-level classifier to obtain a target prediction output for locating bitter peptides in protein sequences.

[0023] In a third aspect, the present application further provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps in the bitter peptide identification method based on multi-feature coding fusion when executing the computer program.

[0024] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps in the bitter peptide identification method based on multi-feature coding fusion when executed by a processor.

[0025] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the steps in the bitter peptide identification method based on multi-feature coding fusion when executed by a processor.

[0026] The bitter peptide identification method and device based on multi-feature coding fusion inputs the protein sequence into the first feature extraction module to obtain the peptide structural features; inputs the protein sequence into the second feature extraction module to obtain the peptide biochemical features; based on the peptide structural features and the peptide biochemical features, at least two first-level classifiers are used for prediction to obtain the first-level prediction output corresponding to each first-level classifier; a combination strategy of generating each first-level prediction output using a second-level classifier is used to obtain the target prediction output for locating the bitter peptide in the protein sequence. The present application combines the peptide structural features and the peptide biochemical features of the protein sequence, which is conducive to a more comprehensive analysis of the characteristics of the protein sequence, thereby more accurately identifying the bitter peptide. In addition, the combination of multiple first-level classifiers and second-level classifiers can reduce the risk of overfitting of a single model, help improve the generalization ability of the model, and by combining the prediction results of multiple classifiers, the advantages of each classifier can be combined to obtain a higher prediction accuracy than a single classifier in some cases. Finally, compared with the bitter peptide identification method based on biological experiments, the present application has a faster recognition speed and low cost, and has a higher economic value. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Schematic diagram of a bitter peptide identification method based on multi-feature coding fusion in one embodiment;

[0028] Figure 2 is a performance comparison diagram of various primary classifiers in one embodiment;

[0029] Figure 3 A performance comparison chart of different feature extraction combinations in an embodiment;

[0030] Figure 4 This is a diagram of SHAP analysis results in one embodiment. DETAILED DESCRIPTION

[0031] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0032] The present application embodiment provides a bitter peptide identification method based on multi-feature coding fusion, comprising the following steps:

[0033] Step 102: Input the protein sequence into a first feature extraction module to obtain peptide structural features.

[0034] The protein sequence contains one or more peptide chains, which are chain structures formed by amino acid residues connected by peptide bonds. The first feature extraction module analyzes the position distribution of peptides in the protein sequence to obtain the structural features of the peptides.

[0035] Step 104: input the protein sequence into a second feature extraction module to obtain peptide biochemical features.

[0036] The second feature extraction module adopts the traditional feature extraction method, that is, extracting peptides, hydrophilicity, polarity, isoelectric point, amino acid composition frequency, amino acid transferability, amino acid distribution and other methods used in the past to qualitatively identify bitter peptide characteristics, to obtain the biochemical characteristics of peptides.

[0037] The biochemical characteristics of peptides include: molecular weight, average hydrophobicity index, polarity ratio, isoelectric point, relative abundance of amino acids, frequency of adjacent amino acid pairs and amino acid distribution position characteristics.

[0038] Among them, Molecular Weight Calculation (MWC) is a tool used to calculate the total molecular weight of an amino acid sequence. Molecular weight is the sum of the relative atomic masses of amino acids and can reflect the mass of a protein or peptide chain. The process of calculating molecular weight is to add up the molecular weight of each amino acid in the sequence to get the total. The formula is as follows:

[0039]

[0040] Where MW is the molecular weight, n is the number of amino acids in the protein sequence, and w i is the molecular weight of the ith amino acid.

[0041] The average hydrophobicity index calculation (HC) is to calculate the average hydrophobicity index of the amino acid sequence, which is used to evaluate the hydrophobicity characteristics of the sequence. Each amino acid has a corresponding hydrophobicity value, and the average hydrophobicity can be calculated by the following formula:

[0042]

[0043] In the formula, n is the number of amino acids, h i is the hydrophobicity value of the ith amino acid.

[0044] Polarity ratio calculation (PC) is to calculate the ratio of polar and non-polar amino acids in a sequence to understand the polarity characteristics of the sequence. The formula is:

[0045]

[0046] Where n polar and n nonpolar Represents the number of polar and nonpolar amino acids, n total is the total number of amino acids in the protein sequence.

[0047] Isoelectric Point Calculation (IPC) is used to calculate the isoelectric point of an amino acid sequence, which is the pH value at which the net charge of the protein in solution is zero. The isoelectric point is determined by the pKa value of the amino acid and its charge state, and is obtained by a recursive algorithm or computational fit.

[0048] Amino Acid Composition Calculation (AACC) is to calculate the relative abundance of each amino acid in the sequence and is used to analyze the ratio of each amino acid. The formula is as follows:

[0049]

[0050] Where n i Represents amino acid a i The number of total is the total number of amino acids in the protein sequence.

[0051] Tetramer Frequency Calculation (TFC) is the frequency of adjacent amino acid pairs in a sequence, which is used to evaluate the occurrence of different amino acid pairs. The formula is:

[0052]

[0053] Where n ij Indicates amino acid pair (a i , a j ) appears, n totalpairs is the total number of adjacent amino acid pairs in the protein sequence.

[0054] Amino Acid Distribution Calculation (AADC) is to calculate the distribution of each amino acid in the sequence in different fragments (first 25%, middle 50%, last 25%), which is used to analyze the position characteristics of amino acids in the sequence. The distribution calculation formula is as follows:

[0055]

[0056] Where AADC k (a i ) represents amino acid a i The relative distribution in the kth segment, n k i Amino acid a i The number of segments in k, n k segment represents the total number of amino acids in fragment k.

[0057] Step 106, based on the peptide structural features and peptide biochemical features, at least two first-level classifiers are used to perform predictions to obtain the first-level prediction outputs corresponding to each first-level classifier; a combination strategy of each first-level prediction output is generated using a second-level classifier to obtain a target prediction output for locating bitter peptides in a protein sequence.

[0058] The peptide structural features and peptide biochemical features are simultaneously input into each first-level classifier to obtain the first-level prediction output corresponding to each first-level classifier. Multiple first-level prediction outputs are then spliced ​​and fused and input into the second-level classifier. The second-level classifier performs secondary learning on these features to obtain the final target prediction output and predict bitter peptides in protein sequences.

[0059] Different types of models may perform better on different data distributions or features. This diversity helps capture features of different dimensions in the data. In addition, different models may perform poorly in some cases but perform well in other cases. By combining the prediction results of these models, we can complement each other's weaknesses and improve the robustness and prediction accuracy of the overall model. By combining the output results of multiple first-level classifiers, we can reduce the risk of overfitting of a single model and help improve the generalization ability of the model.

[0060] The mathematical expression of the secondary classifier is:

[0061] y=f(g 1 (x), g 2 (x), ..., g T (x))

[0062] In the formula, f is the secondary classifier, y is the target prediction output, and g 1 , g 2 ,……,g T is the output of each first-level classifier. The second-level classifier can learn how to select the optimal combination strategy in different situations, thereby improving the prediction performance of the overall model.

[0063] In one embodiment, the first feature extraction module includes a multi-layer Transformer encoder. Step 102 includes: using the first feature extraction module to convert the protein sequence into a numerical vector representation, capturing the features in the protein sequence and the long-term dependencies between the features through the multi-layer encoder, and outputting the peptide structure features.

[0064] Existing machine learning methods for proteins rely heavily on traditional feature encoding techniques, which require complex feature engineering steps and suffer from poor feature extraction effects.

[0065] Therefore, this embodiment adopts a model with multiple layers of Transformer encoder stacks, each Transformer encoder includes a self-attention head and a fully-feed fully-connected layer, and generates high-quality and highly discriminative sequence features through the point-product attention mechanism and the fully-feed fully-connected layer. These features can then be used for downstream tasks such as protein function prediction.

[0066] In the above process, the encoding process of protein sequence can be expressed as:

[0067]

[0068] Where Q, K, and V are query vector, key vector, and value vector, respectively. k is the dimension of the key vector.

[0069] This embodiment uses a multi-layer encoder to learn the representation of protein sequences. By cleverly combining position encoding, multi-head self-attention mechanism and feedforward neural network, the multi-layer Transformer encoder can highly effectively capture and understand the complex contextual relationships in language, thereby obtaining information at different positions in the input protein sequence and then extracting peptide structural features. At the same time, the Transformer encoder is easier to implement than traditional encoding technology.

[0070] In one embodiment, after obtaining the peptide biochemical features, the method further comprises: performing feature screening on the peptide biochemical features by using recursive feature elimination combined with cross-validation, and inputting the screened and retained peptide biochemical features into a primary classifier.

[0071] Recursive Feature Elimination with Cross-Validation (RFECV) is a machine learning method for feature selection. It recursively eliminates unimportant features and combines cross-validation to evaluate the performance of the model, thereby screening out the features that contribute most to the model performance. When processing high-dimensional data, RFECV can effectively reduce the number of features and reduce the processing complexity of the first-level classifier.

[0072] In one embodiment, the first-level classifier includes at least random forest, extreme random trees, and gradient boosting.

[0073] Among them, Random Forest (RF) is an ensemble learning model based on decision trees, which improves overall performance by combining multiple decision trees. Each decision tree in the random forest randomly extracts a subset from the original data set for training during training, thereby achieving model diversity. Each tree votes during prediction, and the final result is determined by the majority vote of all trees.

[0074] To build a random forest, we first randomly sample multiple subsets from the data set, and then train a decision tree on each subset. When predicting, the random forest integrates the prediction results of all decision trees and obtains the final classification result through majority voting. In RF, the variance of the model is reduced by introducing the Bagging algorithm, that is, the Bootstrap self-help method plus the aggregation strategy.

[0075] The final output formula of random forest is as follows:

[0076]

[0077] In the formula, represents the final prediction value, T represents the number of decision trees in the forest, and h i (x) represents the predicted value of the i-th tree.

[0078] ExtraTrees (ETree) is an ensemble learning-based algorithm similar to random forests, but different from them in that ETree increases the diversity of trees through extreme randomness. The main feature of ExtraTrees is that it randomly selects split points and features during training to maximize the differences between trees, thereby improving the generalization ability of the model.

[0079] The generation process of each tree is more randomized than that of random forest to reduce the risk of overfitting. The final prediction for the input sample is the average or vote of the predictions of all trees.

[0080] Gradient Boosting (GB) is an ensemble algorithm that integrates weak learners (usually decision trees) into strong learners. Gradient boosting trains a series of decision trees in an iterative manner, and each tree is improved on the basis of the previous tree, so that the residual error of the model gradually decreases.

[0081] Specifically, gradient boosting gradually improves the performance of the model by minimizing the loss function. Each time you train, the goal of the new tree is to reduce the residual of the previous round of tree models. The model update formula is as follows:

[0082] F m (x) = F m-1 (x)+η·h m (x)

[0083] In the formula, F m (x) represents the model of the mth iteration, η is the learning rate, which controls the sharing of each tree to the overall model, and h m (x) represents the predicted value of each tree.

[0084] In one embodiment, the method also includes: constructing a recognition model including a first feature extraction module, a second feature extraction module, a primary classifier and a secondary classifier, pre-training the recognition model using training data containing bitter peptide sequences and non-bitter peptide sequences, and using accuracy, sensitivity, specificity, Matthews correlation coefficient and area under the curve as evaluation indicators to evaluate the performance of the recognition model.

[0085] A reliable and high-quality benchmark dataset is crucial, especially since the first five models all use the same benchmark dataset, called BTP640. This dataset has been verified and widely recognized in actual use, providing a basis for fair comparison with all previous models. BTP640 is a balanced benchmark dataset with a 1:1 ratio of bitter peptides to non-bitter peptides, containing a total of 640 peptide chains. The dataset used for training contains 512 peptide chains (256 bitter peptides and 256 non-bitter peptides), and the remaining 128 peptide chains (64 bitter peptides and 64 non-bitter peptides) are used for independent testing. The ratio of training to independent testing is 8:2.

[0086] In order to evaluate the performance of the recognition model, this embodiment uses five commonly used evaluation indicators, including accuracy (ACC), sensitivity (Sn), specificity (Sp), Matthews correlation coefficient (MCC, Matthews correlation coefficient) and receiver operating characteristic (ROC, Receiver Operating Characteristic curve) area under the curve (AUC, Area Under Curve). The expression formulas of these indicators are as follows:

[0087]

[0088] Among them, TP, TN, FN, and FP represent the number of true positives, true negatives, false negatives, and false positives, respectively. Sn represents the proportion of correctly identified positive samples. Sp represents the proportion of correctly identified negative samples. ACC represents the proportion of all samples that are correctly classified. MCC measures the correlation between the true value and the predicted value, ranging from [-1,1]. In addition, in order to comprehensively compare the performance of different models, it is necessary to calculate the evaluation index based on the area under the receiver operating characteristic curve, which shows the ratio of true positives to false positives. The AUC value ranges from 0 to 1, and the higher the AUC value, the better the predictive performance of the underlying model. Generally speaking, the higher the values ​​of these five indicators, the better the model performance.

[0089] In one embodiment, Figure 1As shown, the recognition model includes a first feature extraction module, a second feature extraction module, a first-level classifier and a second-level classifier. Among them, the first feature extraction module is the upper half of B in the figure, the second feature extraction module is the lower half of B in the figure, the first-level classifier is the left half of C in the figure, and the second-level classifier is the right half of C in the figure. 320 sequences that are experimentally verified to be bitter peptides were found from existing articles, and 320 sequences that are not bitter peptides were found from the BIOPEP database. The found sequences were deduplicated, and the remaining sequence data were shuffled and divided into training sets and test sets in a ratio of 4:1. The training set is input into the first feature extraction module for feature extraction. The module includes a Transformer encoder with 6 layers, 12 layers or more layers. Each Transformer encoder contains a multi-head self-attention layer, a feedforward network layer, a residual connection layer, etc., to fully mine the peptide structure information and extract the peptide structure features. The training set is input into the second feature extraction module for feature extraction. The extraction methods include HC, IPC, TFC, PC, MWC, AACC and AADC to obtain the biochemical features of peptides. RFECV is used to screen the biochemical features of peptides, retain some important features, and input these retained features together with the peptide structural features into the first-level classifier, which includes GB, RF and ETree. The three first-level classifiers splice and fuse the corresponding first-level prediction outputs (probability values) and input them into the second-level classifier to finally obtain the target prediction output. Among them, the second-level classifier can be a logistic regression.

[0090] In order to demonstrate the superiority of the scheme of the present invention, it was compared with previously existing models, including iBitter-SCM, iBitter-Fuse, Bert4Bitter, iBitter-DRLF, Bitter-RF and CPM-BP (BTP640). As shown in Table 1, the scheme is basically superior to all previously constructed models in terms of Sn (sensitivity), ACC (accuracy), AUC (area under the receiver operating characteristic curve) and MCC (Matthews correlation coefficient), which shows that the scheme effectively mines the local and global information of peptide sequences, so that it shows excellent ability in predicting bitter peptides.

[0091] Table 1 Comparison of the identification model of this scheme with the existing model

[0092] ACC AUC Sn Sp MCC iBitter-SCM 0.844 0.904 0.844 0.844 0.688 iBitter-Fuse 0.930 0.933 0.938 0.922 0.859 Bert4Bitter 0.922 0.964 0.938 0.906 0.844 iBitter-DRLF 0.945 0.977 0.922 0.969 0.892 Bitter-RF 0.938 0.978 0.938 0.938 0.875 CPM-BP(BTP640) 0.836 0.836 0.773 0.903 0.680 This program 0.961 0.978 0.984 0.938 0.923

[0093] In order to integrate powerful and powerful machine learning classifiers so that our final model can achieve strong performance, we selected 13 classifiers, including logistic regression (LR), ridge regression classifier (RC), decision tree (DT), random forest (RF), extreme tree (ETree), gradient boosting (GB), Gaussian naive Bayes (GNB), linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), support vector machine (SVM), linear support vector classifier (LSVC), K nearest neighbor (KNN), and nearest neighbor classifier (NC) for performance measurement. Figure 2 As shown in the figure, the comprehensive values ​​of random forest, extreme tree, gradient boosting and support vector machine in discriminative performance are higher than those of other classifiers, and they are also used as the basis for selecting the base classifier of our integration.

[0094] In order to verify the effect of the fusion features we used, the present invention was used to compare the performance indicators under four feature combinations: features extracted from ESM-2 (Esm), features extracted from traditional descriptors (TD), features extracted from traditional descriptors after RFECV processing (TDRFECV), and our integrated features (ETD). Figure 3 As shown, in terms of ACC, our integrated features (ETD) improved by 4.06%, 4.06% and 4.87% compared to TD, TDRFECV and Esm, respectively. In terms of Sn, ETD improved by 6.35%, 7.99% and 5.27% compared to TD, TDRFECV and Esm, respectively. In terms of Sp, ETD improved by 1.66%, 0.00% and 5.00% compared to TD, TDRFECV and Esm, respectively. As for MCC, ETD improved by 8.57%, 8.53% and 10.17% compared to TD, TDRFECV and Esm, respectively. This shows that the fused features we use can greatly improve the performance of our model compared to the two separate feature types that are not fused.

[0095] In order to verify the performance of the features extracted by the traditional method, we used SHAP to analyze the 39 features that have the greatest impact on individual samples. Figure 4 As shown in the figure, the features extracted by the traditional method have made outstanding contributions to the prediction of individual samples, and the prediction results are biased. Combining the samples and the results of previous wet experiments, it further proves the correctness of the traditional feature extraction method we used and the correctness of the previous experiments in identifying bitter peptides.

[0096] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0097] Based on the same inventive concept, the embodiment of the present application also provides a bitter peptide identification device based on multi-feature coding fusion for implementing the bitter peptide identification method based on multi-feature coding fusion involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the bitter peptide identification device based on multi-feature coding fusion provided below can refer to the limitations of the bitter peptide identification method based on multi-feature coding fusion above, and will not be repeated here.

[0098] In one embodiment, a bitter peptide recognition device based on multi-feature coding fusion is provided, comprising:

[0099] The first feature extraction module is used to obtain peptide structural features based on protein sequences;

[0100] The second feature extraction module is used to obtain peptide biochemical features based on protein sequences;

[0101] The prediction output module is used to make predictions based on peptide structural features and peptide biochemical features using at least two first-level classifiers to obtain the first-level prediction outputs corresponding to each first-level classifier; a combination strategy of each first-level prediction output is generated using a second-level classifier to obtain a target prediction output for locating bitter peptides in protein sequences.

[0102] Each module in the bitter peptide identification device based on multi-feature coding fusion can be implemented in whole or in part by software, hardware and a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module.

[0103] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in all the above method embodiments when executing the computer program.

[0104] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in all the above method embodiments are implemented.

[0105] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in all the above method embodiments when executed by a processor.

[0106] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0107] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited thereto. The processors involved in each embodiment provided in this application may be general-purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, data processing logic devices based on quantum computing, etc., but are not limited thereto.

[0108] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0109] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A bitter peptide identification method based on multi-feature coding fusion, characterized in that: The method comprises: The protein sequence is input into the first feature extraction module to obtain peptide structural features; Inputting the protein sequence into a second feature extraction module to obtain peptide biochemical features; Based on the peptide structural features and the peptide biochemical features, at least two first-level classifiers are used to perform predictions to obtain first-level prediction outputs corresponding to each of the first-level classifiers; a second-level classifier is used to generate a combination strategy for each of the first-level prediction outputs to obtain a target prediction output for locating bitter peptides in the protein sequence.

2. The method according to claim 1, characterized in that The first feature extraction module includes a multi-layer Transformer encoder; Inputting the protein sequence into the first feature extraction module to obtain peptide structural features includes: The protein sequence is converted into a numerical vector representation by using the first feature extraction module, the features in the protein sequence and the long-term dependencies between the features are captured by the multi-layer encoder, and the peptide structural features are output.

3. The method according to claim 1, characterized in that The biochemical characteristics of the peptides include: molecular weight, average hydrophobicity index, polarity ratio, isoelectric point, relative abundance of amino acids, frequency of adjacent amino acid pairs and amino acid distribution position characteristics.

4. The method according to claim 1, characterized in that: After obtaining the peptide biochemical characteristics, the method further comprises: The peptide biochemical features are screened by recursive feature elimination combined with cross-validation, and the peptide biochemical features retained after screening are input into the primary classifier.

5. The method according to claim 1, characterized in that The primary classifier includes at least random forest, extreme random tree and gradient boosting.

6. The method according to claim 1, characterized in that The method also includes: constructing a recognition model including the first feature extraction module, the second feature extraction module, the first-level classifier and the second-level classifier, pre-training the recognition model using training data containing bitter peptide sequences and non-bitter peptide sequences, and using accuracy, sensitivity, specificity, Matthews correlation coefficient and area under the curve as evaluation indicators to evaluate the performance of the recognition model.

7. A bitter peptide recognition device based on multi-feature coding fusion, characterized in that: The device comprises: The first feature extraction module is used to obtain peptide structural features based on protein sequences; A second feature extraction module is used to obtain peptide biochemical features based on the protein sequence; A prediction output module is used to make predictions based on the peptide structural features and the peptide biochemical features using at least two first-level classifiers to obtain the first-level prediction outputs corresponding to each of the first-level classifiers; and to generate a combination strategy for each of the first-level prediction outputs using a second-level classifier to obtain a target prediction output for locating bitter peptides in the protein sequence.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images