A method and system for umami peptide prediction based on deep learning
By constructing the Umami-Transformer deep learning framework and integrating physicochemical features and molecular docking technology, the problems of low efficiency and insufficient prediction accuracy of existing umami peptide screening are solved, and efficient and accurate umami peptide prediction and verification are achieved.
Patent Information
- Application Number
- CN202510817727.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-18
- Publication Date
- 2026-01-02
- Estimated Expiration
- 2045-06-18
AI Technical Summary
Existing umami peptide screening technologies are costly, time-consuming, and have low throughput. Furthermore, existing pre-trained models lack generalization ability in umami peptide prediction, making it difficult to support practical industrial applications.
We constructed the Umami-Transformer deep learning framework, which integrates the Transformer architecture with eight key physicochemical features. We used an exhaustive screening method to select dipeptides to pentapeptides and analyzed the interaction mechanism between candidate peptides and umami receptors T1R1/T1R3 through molecular docking technology.
Significantly improves the prediction accuracy of umami peptides, enabling a full-process study from theoretical prediction to biological experimental verification, and providing an innovative technical path for the efficient discovery and functional application of umami peptides.
Smart Images
Figure CN120766752B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application relates to a deep learning-based umami peptide prediction method and system, and belongs to the technical field of food science and artificial intelligence. BACKGROUND
[0002] Umami peptides are a class of bioactive peptides that can significantly enhance the umami taste of food, usually derived from protein hydrolysates. They not only enhance the perception of food umami, but also effectively reduce bitterness and improve overall flavor palatability. For example, known umami peptides such as NKF and LGLR can interact with taste receptors to enhance the perception of saltiness in low-salt solutions, providing a potential solution for salt-reduced food development. Currently, umami peptides are widely present in natural organisms such as edible fungi and oysters, and their research not only focuses on food flavor optimization, but also involves reducing sodium salt intake and developing healthy functional foods, which has important value for food science and public health.
[0003] Traditional umami peptide screening techniques mainly rely on sensory-guided separation combined with mass spectrometry analysis. Although this method can effectively identify active peptides, it has problems such as high cost, long time consumption, and low throughput, and the active peptides are easily lost during the complex separation process. In recent years, with the development of artificial intelligence technology, machine learning models have been gradually applied to peptide segment flavor classification. For example, multilayer perception (MLP) and long short-term memory network (LSTM) models have been used for umami peptide prediction, but RNN-based models have difficulty in effectively processing long sequence data due to the gradient vanishing problem. In addition, although pre-trained language models such as BERT perform well in natural language processing, they are difficult to directly migrate to the peptide sequence prediction field due to the limited size of the umami peptide database, and the sequence correlation between different peptide segments is significantly different from natural language, further limiting the application effect of pre-trained models.
[0004] The Transformer neural network has shown significant advantages in emerging bioinformatics due to its unique multi-head attention mechanism and parallel computing architecture. Its multi-head attention mechanism can simultaneously capture the local and long-range dependencies of amino acids in peptide sequences, accurately analyzing complex sequence characteristics. The parallel computing design significantly improves training efficiency and has no risk of gradient vanishing or explosion, making it particularly suitable for processing variable-length peptide sequence prediction tasks. However, existing research has focused on single sequence information, and few have integrated physical and chemical characteristics of peptide segments such as lipophilicity and charge distribution to optimize model performance. In addition, due to the scarcity of umami peptide data and research costs, existing prediction models generally have insufficient generalization ability and limited prediction accuracy, making it difficult to support actual industrial applications. SUMMARY
[0005] In view of the deficiencies of the prior art, the present application provides a umami peptide prediction method and system based on deep learning, and proposes a deep learning framework of Umami-Transformer, which can efficiently process peptide sequence information and significantly improve the prediction accuracy of umami peptides by fusing the Transformer architecture and 8 key physical and chemical characteristics (including molecular lipophilicity, charge distribution, and polarity parameters, etc.). Further, all dipeptides to pentapeptides are screened by the exhaustive method, and the interaction mechanism of the candidate peptides and the umami receptor T1R1 / T1R3 is analyzed by molecular docking technology, so as to realize the whole process research from theoretical prediction to biological experimental verification, and provide an innovative technical path for efficient mining and functional application of umami peptides.
[0006] Specifically, the technical scheme comprises the following:
[0007] In a first aspect, the present application provides a umami peptide prediction method based on deep learning, comprising the following steps:
[0008] S1, constructing a data set: collecting umami peptides and non-umami peptide sequences verified by experiments to construct a UMP499 data set containing umami peptides and non-umami peptides;
[0009] S2, feature selection and processing: standardizing, truncating and padding the peptide sequence, and calculating 8 kinds of physical and chemical characteristics;
[0010] S3, constructing an Umami-Transformer model: constructing an Umami-Transformer model containing a feature processing and generation module, a sequence information processing module, a feature information processing module and a result output module; wherein the sequence information processing module adopts a Transformer encoder; the feature information processing module adopts a multi-layer fully connected network; the result output module adopts an MLP network with a Sigmoid activation function, and outputs the probability of umami peptide;
[0011] S4, model training: training the Umami-Transformer model by using a 5-fold cross-validation method, using a BCELoss function as the loss function, until the classification accuracy of the Umami-Transformer model on the test set is more than 95%;
[0012] S5, umami peptide prediction: inputting the peptide sequence to be tested into the trained Umami-Transformer model to obtain the probability prediction result of being a umami peptide.
[0013] In an embodiment of the present application, the step S1 of constructing a data set specifically comprises:
[0014] S11, data collection: from public literature and BIOPEP-UWM database, collect experimentally verified umami peptides and non-umami peptides sequences, construct UMP499 dataset containing 212 umami peptides and 287 non-umami peptides;
[0015] S12, data preprocessing, including:
[0016] Removing outliers: remove peptide sequences containing non-standard amino acid symbols to ensure the accuracy and consistency of the data;
[0017] Removing redundancy: remove duplicate sequences through sequence alignment tools to eliminate redundancy and avoid the influence of repeated data on model training;
[0018] Hierarchical division: divide the dataset into UMP-CV training set and UMP-TS test set according to the proportion of 80:20 to ensure the consistency of the distribution of training set and test set.
[0019] In an embodiment of the present application, the feature selection and processing in step S2 specifically include:
[0020] S21, feature calculation: using RDKit chemical informatics toolkit to calculate 8 physical and chemical characteristics of each peptide sequence, including: molecular weight related characteristics: BCUT2D_MWLOW; charge distribution indicators: PEOE_VSA14, VSA_EState5, VSA_EState6, VSA_EState7, MinEStateIndex; polarity parameters: SMR_VSA1; molecular lipophilicity: MolLogP;
[0021] S22, sequence standardization:
[0022] Standardize and fill the variable-length peptide sequences to ensure that all peptide sequences have the same length, and the specific operation is: unify all peptide sequences to 50 in length, and perform learnable coding on 20 standard amino acids to generate 16-dimensional feature vectors;
[0023] S23, position encoding: using sine-cosine position encoding strategy to generate 16-dimensional encoding vectors for each amino acid position to compensate for the lack of position information of the Transformer architecture and ensure that the model can effectively capture the position characteristics in the amino acid sequence.
[0024] In an embodiment of the present application, in step S3:
[0025] The feature processing and generating module is used for processing and generating related features to provide rich feature data for the model; the input layer of the feature processing and generating module receives 16-dimensional encoding vectors and 8 physical and chemical characteristics of the peptide sequence, and the physical and chemical characteristics are spliced with the sequence encoding after standardization processing;
[0026] The sequence information processing module adopts a 12-layer Transformer encoder, each layer containing 8 attention heads, and compensates the position information through sine-cosine position encoding; the peptide sequence information is deeply analyzed and efficiently processed through the multi-head attention mechanism to extract the correlation between amino acids; the Transformer encoder can simultaneously focus on different parts of the sequence to capture the complex relationships between amino acids, including local key structural features and long-distance associations;
[0027] The feature information processing module adopts a multi-layer fully connected network to further optimize and integrate the input 8 physical and chemical characteristics, improve the expression ability and effectiveness of the characteristics; the Dropout mechanism is introduced in the training process to prevent overfitting and enhance the generalization ability of the model;
[0028] The result output module adopts an MLP network with a Sigmoid activation function, and the Sigmoid activation function outputs a probability value in the range of [0, 1].
[0029] In an embodiment of the present application, in the result output module, the peptides are classified by setting a probability threshold of 0.5, and a probability greater than 0.5 is determined as a umami peptide, and otherwise as a non-umami peptide.
[0030] In an embodiment of the present application, the model training in step S4 specifically includes:
[0031] S41, training method: the Umami-Transformer model is trained by using a 5-fold cross-validation method, and a BCELoss function is used as a loss function to measure the difference between the model prediction value and the true value; during the training process, the convergence of the loss curve is monitored, the model performance is optimized by adjusting the hyperparameters, and the classification accuracy of the Umami-Transformer model on the test set is greater than 95% until the model is optimized;
[0032] S42, model optimization, including:
[0033] Learning rate adjustment: dynamically adjust the learning rate according to the loss change in the training process to ensure that the model can quickly converge and achieve optimal performance;
[0034] Batch size setting: balance the training speed and model stability to improve the training efficiency;
[0035] Apply regularization techniques to prevent model overfitting and enhance its generalization ability on new data.
[0036] In an embodiment of the present application, it further includes:
[0037] S51, performance evaluation index: a variety of evaluation indexes are used to comprehensively quantitatively evaluate the trained model, including accuracy, precision, recall, F1 score and Matthew correlation coefficient; the above indexes reflect the classification performance of the model from different angles, and ensure its reliability and effectiveness in actual application;
[0038] S52, ablation experiment: the influence of feature selection on the performance of the model is verified through the ablation experiment;
[0039] S53, input all dipeptides to pentapeptides into the model to judge umami peptides, and finally select DD, DDE, DDED and DDEDD for synthesis verification;
[0040] S54, analyze the taste profile of the synthesized peptide by electronic tongue and sensory evaluation, and analyze the interaction mechanism of the candidate peptide and umami receptor T1R1 / T1R3 by molecular docking technology.
[0041] In a second aspect, the present application provides a umami peptide prediction system based on deep learning, comprising:
[0042] Data construction module: collect umami peptides and non-umami peptide sequences verified by experiments, and construct UMP499 data set containing multiple umami peptides and multiple non-umami peptides;
[0043] Feature selection and processing module: used for standardizing, truncating and filling the peptide sequence, and calculating various physical and chemical characteristics;
[0044] Model construction module: used for constructing Umami-Transformer model containing feature processing and generation module, sequence information processing module, feature information processing module and result output module; wherein the sequence information processing module adopts Transformer encoder; the feature information processing module adopts multi-layer full connection network; the result output module adopts MLP network with Sigmoid activation function, and outputs umami peptide probability;
[0045] Model training module: the Umami-Transformer model is trained by using 5-fold cross-validation method, and BCELoss function is used as loss function until the classification accuracy of Umami-Transformer model on test set is more than 95%;
[0046] Umami peptide prediction module: input the peptide sequence to be tested into the trained Umami-Transformer model, and get the probability prediction result of umami peptide.
[0047] In a third aspect, the present application provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the deep learning-based umami peptide prediction method when executing the computer program.
[0048] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the deep learning-based umami peptide prediction method when executed by a processor.
[0049] The present application has the following beneficial effects:
[0050] The present application proposes a deep learning framework of Umami-Transformer, which can efficiently process peptide sequence information and significantly improve the prediction accuracy of umami peptides by fusing the Transformer architecture with eight key physicochemical characteristics (including molecular lipophilicity, charge distribution, and polarity parameters, etc.). Further combined with the exhaustive method to screen all dipeptides to pentapeptides, and through the molecular docking technology to analyze the interaction mechanism of candidate peptides and umami receptor T1R1 / T1R3, the whole process research from theoretical prediction to biological experimental verification is finally realized, which provides an innovative technical path for efficient mining and functional application of umami peptides. BRIEF DESCRIPTION OF DRAWINGS
[0051] Figure 1 Fig. 1 is a schematic diagram of the architecture of the Umami-Transformer model.
[0052] Figure 2 Fig. 4 is a schematic diagram of the loss curve and classification accuracy change during the model training process.
[0053] Figure 3 Fig. 6 is a schematic diagram of the ablation experiment results.
[0054] Figure 4 Fig. 7 is a schematic diagram of the performance comparison of different umami peptide prediction models.
[0055] Figure 5 Fig. 8 is a schematic diagram of the molecular docking interaction between umami peptides and umami receptor T1R1 / T1R3.
[0056] Figure 6 Fig. 9 is a schematic diagram of the sensory evaluation and electronic tongue analysis results of the synthesized umami peptides.
[0057] Figure 7 Fig. 10 is a schematic diagram of the network server interface of the Umami-Transformer model. DETAILED DESCRIPTION
[0058] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of the present application.
[0059] As shown in Figures 1 to 7 The present application provides a deep learning-based umami peptide prediction method and system. Among them, Figure 1 is an architecture diagram of the Umami-Transformer model, showing the four core components of the model and their functions. Figure 2 is a diagram showing the loss curve and classification accuracy during model training, indicating that the model converges quickly during training and achieves more than 95% classification accuracy on the test set. Figure 3 is an ablation experiment result diagram, showing the influence of different features on the performance of the model, verifying the importance of the selected features. Figure 4 is a performance comparison diagram of different umami peptide prediction models, highlighting the superiority of the Umami-Transformer model in various evaluation indicators. Figure 5 is a molecular docking interaction diagram of umami peptides and umami receptors T1R1 / T1R3, revealing the binding mechanism between umami peptides and receptors. Figure 6 is a sensory evaluation and electronic tongue analysis result diagram of synthetic umami peptides, verifying the accuracy of the model prediction. Figure 7 is a network server interface diagram of the Umami-Transformer model, showing the online prediction function of the model.
[0060] Embodiment 1
[0061] The present embodiment provides a deep learning-based umami peptide prediction method, comprising the following steps:
[0062] S1, constructing a data set: collecting experimentally verified umami peptides and non-umami peptides, and constructing a UMP499 data set containing multiple umami peptides and multiple non-umami peptides;
[0063] S2, feature selection and processing: standardizing, truncating and padding the peptide sequence, and calculating various physicochemical features;
[0064] S3, constructing an Umami-Transformer model: constructing an Umami-Transformer model comprising a feature processing and generation module, a sequence information processing module, a feature information processing module, and a result output module; wherein the sequence information processing module adopts a Transformer encoder; the feature information processing module adopts a multi-layer fully connected network; the result output module adopts an MLP network with a Sigmoid activation function, and outputs the umami peptide probability;
[0065] S4, model training: training the Umami-Transformer model using a 5-fold cross-validation method, using the BCELoss function as the loss function, until the classification accuracy of the Umami-Transformer model on the test set is more than 95%;
[0066] S5, umami peptide prediction: inputting the peptide sequence to be tested into the trained Umami-Transformer model to obtain the probability prediction result of being an umami peptide.
[0067] Optionally, the step S1 of constructing a data set specifically comprises:
[0068] S11, data collection: collecting experimentally verified umami peptides and non-umami peptides from public literature and BIOPEP-UWM database, and constructing a UMP499 data set comprising 212 umami peptides and 287 non-umami peptides;
[0069] S12, data preprocessing, including:
[0070] Removing outliers: removing peptide sequences containing non-standard amino acid symbols (such as "B", "U", "X", "Z") to ensure the accuracy and consistency of the data;
[0071] Removing redundancy: removing duplicate sequences by sequence alignment tools (such as CD-HIT) to eliminate redundancy and avoid the influence of repeated data on model training;
[0072] Layered division: dividing the data set into UMPCV training set and UMPTS test set according to the ratio of 80:20 to ensure the distribution consistency of the training set and the test set, so as to improve the generalization ability of the model.
[0073] Optionally, the feature selection and processing in step S2 specifically comprises:
[0074] S21, Feature calculation: 8 physical and chemical features of each peptide sequence are calculated using the RDKit chemical informatics toolkit, including: molecular weight-related features: BCUT2D_MWLOW; charge distribution indicators: PEOE_VSA14, VSA_EState5, VSA_EState6, VSA_EState7, MinEStateIndex; polarity parameters: SMR_VSA1; molecular lipophilicity: MolLogP; the above features can comprehensively describe the physical and chemical properties of the peptide from different angles, providing rich information for the model.
[0075] S22, Sequence normalization:
[0076] The variable-length peptide sequence is standardized, truncated and padded to ensure that all peptide sequences have the same length for processing by the machine learning model. Specifically, all peptide sequences are unified to a length of 50, and 20 standard amino acids are learnable coded to generate a 16-dimensional feature vector.
[0077] S23, Position encoding: A sine-cosine position encoding strategy is used to generate a 16-dimensional encoding vector for each amino acid position to compensate for the lack of position information in the Transformer architecture, ensuring that the model can effectively capture the position features in the amino acid sequence.
[0078] Optionally, in the step S3:
[0079] The feature processing and generation module is used to process and generate related features to provide rich feature data for the model. The input layer of the feature processing and generation module receives the 16-dimensional encoding vector of the peptide sequence and the 8 physical and chemical features. The physical and chemical features are normalized (Z-Score normalization) and then spliced with the sequence encoding.
[0080] The sequence information processing module uses a 12-layer Transformer encoder, each layer containing 8 attention heads, and compensates for the position information through sine-cosine position encoding. Through the multi-head attention mechanism, the peptide sequence information is analyzed and processed efficiently, and the correlation between amino acids is accurately extracted. The Transformer encoder can simultaneously focus on different parts of the sequence and capture the complex relationships between amino acids, whether it is local key structural features or long-range associations.
[0081] The feature information processing module uses a multi-layer fully connected network to further optimize and integrate the input 8 physical and chemical features, improving the expression ability and effectiveness of the features. In the training process, the Dropout mechanism is introduced to prevent overfitting and enhance the generalization ability of the model.
[0082] The result output module adopts an MLP network with a Sigmoid activation function, the Sigmoid activation function outputs a probability value in the range of [0, 1], and the peptides are classified by setting a probability threshold of 0.5, and the probability greater than 0.5 is determined as a umami peptide, and otherwise is determined as a non-umami peptide.
[0083] Optionally, the model training in the step S4 specifically includes:
[0084] S41, training method: the Umami-Transformer model is trained by using a 5-fold cross-validation method, a BCELoss function is used as a loss function to measure the difference between the predicted value and the true value of the model; in the training process, the convergence of the loss curve is monitored, the model performance is optimized by adjusting the hyperparameters, and until the classification accuracy of the Umami-Transformer model on the test set is more than 95% (such as Figure 2 );
[0085] S42, model optimization, including:
[0086] learning rate adjustment, dynamically adjusting the learning rate according to the loss change in the training process, to ensure that the model can quickly converge and reach the optimal performance;
[0087] batch size setting, selecting a suitable batch size to balance the training speed and model stability, and improving the training efficiency;
[0088] regularization techniques, applying L2 regularization and other regularization techniques to prevent model overfitting and enhance its generalization ability on new data.
[0089] Optionally, the umami peptide prediction method further includes:
[0090] S51, performance evaluation index: a variety of evaluation indexes are used to comprehensively quantitatively evaluate the trained model, including accuracy, precision, recall, F1 score and Matthew correlation coefficient; the above indexes reflect the classification performance of the model from different angles, to ensure its reliability and effectiveness in actual application;
[0091] S52, ablation experiment: the influence of feature selection on the performance of the model is verified by the ablation experiment, and the results show that the selected 8 kinds of physical and chemical characteristics have significant contribution to the performance indexes of the model, and the VSA_EState6 and BCUT2D_MWLOW characteristics have the most critical influence on the performance of the model (such as Figure 3 );
[0092] S53, all dipeptides to pentapeptides are input into the model to judge the umami peptides, and finally DD, DDE, DDED and DDEDD are selected for synthesis verification;
[0093] S54, analyze the taste profile of the synthesized peptide by electronic tongue and sensory evaluation, and analyze the interaction mechanism of the candidate peptide with umami receptor T1R1 / T1R3 by molecular docking technology;
[0094] S55, comparison with other models: compared with existing other umami peptide prediction models (such as iUmami-SCM, Umami_YYDS, Umami-BERT), Umami-Transformer performs well in accuracy, precision, recall, Matthew correlation coefficient and F1 score, reaching 0.965, 1.0, 0.825, 0.889 and 0.903 respectively, which is significantly better than other models (such as Figure 4
[0095] Example 2
[0096] This embodiment provides a prediction and screening of umami peptides based on the deep learning-based umami peptide prediction method described in Example 1, comprising:
[0097] 2.1, large-scale peptide prediction:
[0098] All dipeptides to pentapeptides are comprehensively predicted by using Umami-Transformer model, a total of 3,368,400 peptides (400 2 peptides, 8,000 tripeptides, 160,000 tetrapeptides, and 3,200,000 pentapeptides). Through prediction, peptides with potential umami activity are screened out to provide candidates for subsequent experimental verification and application development.
[0099] 2.2, molecular docking analysis:
[0100] Molecular docking analysis is performed on the top-ranked peptides, and the results show that these peptides can stably bind to umami receptor T1R1 / T1R3, and the binding energy is low. For example, the binding energy-CDOCKER ENERGY of dipeptide DD, tripeptide DDE, tetrapeptide DDED and pentapeptide DDEDD is 59.3507 kcal / mol, 80.4485 kcal / mol, 95.5644 kcal / mol and 97.9330 kcal / mol, respectively, and the-CDOCKER INTERACTION ENERGY of dipeptide DD, tripeptide DDE, tetrapeptide DDED and pentapeptide DDEDD is 38.0929 kcal / mol, 55.1156 kcal / mol, 75.3756 kcal / mol and 65.6673 kcal / mol, respectively (as shown in Figure 5
[0101] 2.3, peptide synthesis and taste verification:
[0102] 2.3.1, peptide synthesis
[0103] The top 4 peptides (DD, DDE, DDED, DDEDD) were synthesized with purity not less than 95%.
[0104] 2.3.2, Sensory evaluation
[0105] The taste characteristics of the synthesized peptides were verified through sensory evaluation experiments. The results showed that these peptides had significant umami and salty taste, among which DDE and DDED had umami intensity exceeding 3% monosodium glutamate, and their taste thresholds were lower, at 0.29 mg / mL and 0.19 mg / mL, respectively. The sensory evaluation of the peptides is shown in Table 1 below:
[0106] Table 1. Sensory evaluation of peptides
[0107]
[0108] 2.3.3, Electronic tongue analysis
[0109] The electronic tongue device was used to quantitatively analyze the taste characteristics of the synthesized peptides, and the results were consistent with the sensory evaluation, further verifying the accuracy of the model prediction. For example, the umami intensity score of DDE was 12.06, higher than that of 3 mg / mL monosodium glutamate (11.56), close to the umami value of 5 mg / mL (12.60), as shown in Table 2. Figure 6
[0110] Example 3
[0111] This example provides a deep learning-based umami peptide prediction system, in which the system implementation and deployment include network server development: developing a network server for the Umami-Transformer model, providing a user-friendly interface and comprehensive documentation to facilitate researchers to efficiently analyze umami peptides. The server supports submission of single or batch peptide sequences, users can download standardized input templates, and after submitting the sequences, the analysis results can be obtained in the "Result" section, as shown in Figure 2. Figure 7
[0112] A deep learning-based umami peptide prediction system, comprising:
[0113] A data construction module: collects experimentally verified umami peptide and non-umami peptide sequences, and constructs a UMP499 dataset containing multiple umami peptides and multiple non-umami peptides;
[0114] A feature selection and processing module: used for standardizing, truncating and padding the peptide sequences, and calculating various physicochemical characteristics;
[0115] The model construction module is configured to construct an Umami-Transformer model comprising a feature processing and generation module, a sequence information processing module, a feature information processing module, and a result output module. The sequence information processing module adopts a Transformer encoder. The feature information processing module adopts a multi-layer fully connected network. The result output module adopts an MLP network with a Sigmoid activation function to output the umami peptide probability.
[0116] The model training module is configured to train the Umami-Transformer model using a 5-fold cross-validation method, using a BCELoss function as the loss function, until the classification accuracy of the Umami-Transformer model on the test set is greater than 95%.
[0117] The umami peptide prediction module is configured to input the peptide sequence to be tested into the trained Umami-Transformer model to obtain the probability prediction result of the umami peptide.
[0118] Example 4
[0119] The present embodiment provides the results of the docking between DD, DDE, DDED, DDEDD and umami receptors.
[0120] 4.1, Docking method
[0121] The Discovery Studio 2019 software was used to perform molecular docking experiments. The homology model of the T1R1 / T1R3 umami receptor was used as the target for docking analysis of the four umami peptides DD, DDE, DDED, and DDEDD. Before docking, the receptor and ligand were optimized for energy minimization to ensure the stability of the structure. During the docking process, the CDOCKER program was used to find the optimal binding conformation between the ligand and the receptor.
[0122] 4.2, Docking results with umami receptors
[0123] The binding of glutamate to umami receptors on the membrane of taste cells initiates an intracellular signaling cascade that mediates ion channel activation, particularly through transient receptor potential (TRPM) channels. This signal transduction process subsequently induces calcium-dependent depolarization of the taste receptor cells, ultimately leading to neural transmission to the taste cortical areas through afferent fibers of the facial and glossopharyngeal nerves. The perception of taste is significantly influenced by the T1R1 / T1R3 receptor, which is mainly located in the anterior 15 of the tongue.
[0124] Homology modeling and computational molecular docking are advanced methods for determining possible binding sites between umami peptides and taste receptors. In the binding with the umami receptor T1R1 / T1R3, DD adopts four different forms of hydrogen bonds: salt bridges, van der Waals, traditional hydrogen bonds, and carbon-hydrogen bondsFigure 2 The carbon-hydrogen bond is a covalent bond with high bond energy, approximately 413 kJ / mol, while the strength of a traditional hydrogen bond is typically between 5 and 30 kJ / mol. The strength of a salt bridge is approximately 5 to 6 kJ / mol. The strength of a van der Waals force is typically in the range of 0.4 kJ / mol to 4 kJ / mol. The oxygen atom in the R group on the Asp residue forms a strong carbon-hydrogen bond with ARG 712, GLY 695, and GLN 853.
[0125] The establishment of a salt bridge between the oxygen atom and ARG 712 is of significant importance. According to the research results, ARG 712 constitutes an important binding site. Meanwhile, ARG 712 of T1R1 / T1R3 also binds with the umami peptides PQFAPEED, EEHPVLLTEA, and DQAIPNKPEE. As shown in FIG. 1, the oxygen atom of the R group on the Asp residue forms a strong carbon-hydrogen bond with ARG 712, GLY 695, and GLN 853. Figure 3 As shown in FIG. 1, DDE binds with the umami receptor T1R1 / T1R3 at several key sites, including ARG 712, TYR 694, HIS 915, SER 693, ALA 856, VAL 714, GLY 695, HIS 914, GLY 911, GLY 855, LEU 912, ARG 854, PRO 715, LEU 852, GLN 853, LEU 967, ARG 591, and PHE 851.
[0126] The main interactions formed include van der Waals forces, carbon-hydrogen bonds, attractive charge-charge interactions, unfavorable positive-positive interactions, traditional hydrogen bonds, and pi-anion interactions. Attractive charge-charge interactions are stronger than carbon-hydrogen bonds, while pi-anion interactions are stronger than salt bridges but weaker than traditional hydrogen bonds. The research shows that hydrogen bonds play an important role in the perception process caused by the binding of umami peptides with umami receptors. Notably, ARG 591 and ARG 712 form attractive charge-charge interactions, TYR 694 and HIS 915 form pi-anion interactions with DDE. In addition, GLY 695 forms a strong traditional hydrogen bond with the oxygen atom on the R group of Asp. ARG 712 forms an unfavorable positive-positive interaction with the N atom of DDE, which refers to the repulsive force between two positive charges, requiring additional energy to maintain this interaction.
[0127] The binding force of the tetrapeptide DDED to T1R1 / T1R3 includes all favorable interactions of the dipeptide DD and the tripeptide DDE, such as van der Waals forces, traditional hydrogen bonds, salt bridges, carbon-hydrogen bonds, attractive charge-charge interactions, and pi-anion interactions, as shown in FIG. 1. Figure 4ARG591, ARG854, HIS915 and ARG712 form salt bridges, attractive charge interactions and Pi anion interactions with DDED. The presence of D / E residues enhances the electrostatic interactions in the umami peptides, reduces the HOMO-LUMO energy gap, and modulates the receptor recognition pathway through orbital repartitioning effects. Compared to DDED, the pentapeptide DDEDD forms fewer salt bridge interactions with the umami receptor. Overall, VAL714, TYR694, SER693, LEU912, LEU852, HIS915, HIS914, GLY911, GLY855, GLY695, GLN853, ARG854 and ARG712 form binding interactions with all four peptides, which are likely to be the key sites for umami perception Figure 5 Figure 6 Figure 7 Figure 5 Figure 5 Figure 5 Figure 5 In addition, DDE, DDED and DDEDD also bind to PRO715, PHE851 and LEU967 on the umami receptor, which can lead to their higher umami intensity compared to DD. These findings suggest that there is no correlation between peptide chain length and umami receptor activation efficiency, and shorter peptides (2-5 residues) exhibit excellent binding affinity, which can be due to their optimal spatial compatibility with the receptor hydrophobic pocket.
[0128] Example 5
[0129] This embodiment provides an application case of the deep learning-based umami peptide prediction method described in Example 1, including:
[0130] 1. Low-sodium food development: By predicting and screening umami peptides with salty taste enhancing effect, such as DDE, DDED and DDEDD, they can be used as functional salt substitutes to partially replace sodium chloride in food, achieving the development of low-sodium food.
[0131] 2. Taste regulation research: Using the umami peptides predicted by the Umami-Transformer model, the molecular mechanism of umami perception is further studied, providing theoretical support for the research and development of food flavorings.
[0132] In summary, to address the low efficiency and high cost of traditional umami peptide screening methods, the present application combines the Transformer architecture with eight physicochemical descriptors (including molecular lipophilicity, charge distribution, and molecular weight-related features) to construct a high-precision umami peptide prediction model. The model is trained using five-fold cross-validation, with a classification accuracy of 0.965, an F1 score of 0.903, and a Matthews correlation coefficient of 0.889, significantly better than existing methods. Further based on the model, four candidate umami peptides (DD, DDE, DDED, and DDEDD) are screened from all 2-5 peptides. After solid-phase synthesis and sensory evaluation verification, their umami intensity exceeds 3% MSG, among which DDE and DDED have umami thresholds of 0.29 mg / mL and 0.19 mg / mL, respectively, and exhibit a synergistic salt-enhancing effect, which can be applied to low-sodium food development. Molecular docking shows that the Asp / Glu residues at the ends of the peptide chains bind stably to the umami receptors T1R1 / T1R3 through hydrogen bonds, salt bridges, and charge complementation, revealing the taste perception mechanism. The present application provides an innovative technical solution for efficient prediction, flavor optimization, and salt reduction strategies of umami peptides, with significant application value in the food industry.
[0133] The present application also has the following beneficial effects:
[0134] 1. Expanding peptide diversity: Further expanding the training data set to cover more types of peptide sequences, improving the generalization ability and prediction accuracy of the model.
[0135] 2. In-depth study of taste mechanism: Combining computational simulation and experimental techniques, in-depth exploration of the interaction mechanism between umami peptides and taste receptors, providing a basis for model optimization and development of new taste modulators.
[0136] 3. Interdisciplinary cooperation: Strengthening interdisciplinary cooperation in food science, bioinformatics, chemical informatics, and other disciplines, promoting innovation and development in the field of taste research, and bringing more innovative products and solutions to the food industry.
[0137] Through the above specific implementation schemes, the present application successfully establishes a machine learning-based umami peptide prediction method and system, providing an efficient and accurate tool for the research and application of umami peptides, and promoting the progress of taste regulation and low-sodium food development in the field of food science.
[0138] Further, the present application also provides a computer device, which can include a processor, a memory, a network interface and a database connected through a system bus. Wherein, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, causes the processor to perform the steps of the deep learning-based umami peptide prediction method according to any one of the above embodiments.
[0139] The working process, working details and technical effects of the computer device provided by the present embodiment can be referred to the above embodiments of the deep learning-based umami peptide prediction method, which will not be repeated here.
[0140] Further, the present application also provides a computer readable storage medium having a computer program stored thereon, and the computer program, when executed by a processor, implements the steps of the deep learning-based umami peptide prediction method according to any one of the above embodiments. Wherein, the computer readable storage medium refers to a carrier for storing data, which can include, but is not limited to, floppy disks, optical discs, hard disks, flash memories, USB flash disks and / or memory sticks, etc. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices.
[0141] The working process, working details and technical effects of the computer readable storage medium provided by the present embodiment can be referred to the above embodiments of the deep learning-based umami peptide prediction method, which will not be repeated here.
[0142] Those skilled 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. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, storage, database or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0143] Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims
1. A deep learning-based method for predicting umami peptides, characterized in that, Includes the following steps: S1. Constructing the dataset: Collect experimentally verified umami peptide and non-umami peptide sequences to construct the UMP499 dataset containing umami peptides and non-umami peptides. S2. Feature selection and processing: Standardize and truncate the peptide sequence and fill it in, and calculate 8 physicochemical features; S3. Constructing the Umami-Transformer Model: A Umami-Transformer model is constructed, comprising a feature processing and generation module, a sequence information processing module, and a result output module. The sequence information processing module uses a Transformer encoder; the feature information processing module uses a multi-layer fully connected network; the result output module uses an MLP network with a Sigmoid activation function to output the probability of umami peptides. The feature processing and generation module processes and generates relevant features, providing rich feature data for the model. The input layer of the feature processing and generation module receives a 16-dimensional encoding vector of the peptide sequence and eight physicochemical features. The physicochemical features are standardized and then concatenated with the sequence encoding. The sequence information processing module uses a 12-layer Transformer encoder, each layer containing eight attention heads, and compensates for positional information through sine-cosine positional encoding. A multi-head attention mechanism is used to deeply analyze and efficiently process peptide sequence information, extracting correlations between amino acids. The Transformer encoder can simultaneously focus on different parts of the sequence, capturing complex interrelationships between amino acids, including local key structural features and long-distance associations. S4. Model Training: The Umami-Transformer model is trained using a 5-fold cross-validation method, with the BCELoss function as the loss function, until the classification accuracy of the Umami-Transformer model on the test set exceeds 95%. S5. Umami Peptide Prediction: Input the peptide sequence to be tested into the trained Umami-Transformer model to obtain the probability prediction result that it is an umami peptide.
2. The method for predicting umami peptides based on deep learning according to claim 1, characterized in that, The specific steps of constructing the dataset in step S1 include: S11. Data collection: Collect experimentally validated umami peptide and non-umami peptide sequences from public literature and the BIOPEP-UWM database, and construct the UMP499 dataset containing 212 umami peptides and 287 non-umami peptides. S12. Data preprocessing, including: Outlier removal: Peptide sequences containing non-standard amino acid symbols are removed to ensure data accuracy and consistency; Redundancy removal: Repeated sequences are removed using sequence alignment tools to eliminate redundancy and avoid the impact of duplicate data on model training; Stratified partitioning: The dataset is divided into UMP-CV training set and UMP-TS test set in an 80:20 ratio to ensure the consistency of the distribution between the training set and the test set.
3. The method for predicting umami peptides based on deep learning according to claim 2, characterized in that, The feature selection and processing in step S2 specifically includes: S21. Feature Calculation: Using the RDKit cheminformatics toolkit, calculate eight physicochemical features for each peptide sequence, including: molecular weight association feature: BCUT2D_MWLOW; charge distribution indices: PEOE_VSA14, VSA_EState5, VSA_EState6, VSA_EState7, MinEStateIndex; polarity parameter: SMR_VSA1; molecular lipophilicity: MolLogP. S22, Sequence Standardization: Standardize and padded variable-length peptide sequences to ensure that all peptide sequences have the same length. Specifically, all peptide sequences are standardized to a length of 50, and 20 standard amino acids are encoded in a learnable manner to generate a 16-dimensional feature vector. S23. Position Encoding: A sine-cosine position encoding strategy is adopted to generate a 16-dimensional encoding vector for each amino acid position to compensate for the lack of position information in the Transformer architecture and ensure that the model can effectively capture the position features in the amino acid sequence.
4. The method for predicting umami peptides based on deep learning according to claim 3, characterized in that, In step S3: The feature information processing module employs a multi-layer fully connected network to further optimize and integrate the eight input physicochemical features, thereby improving the expressive power and effectiveness of the features. During training, a Dropout mechanism is introduced to prevent overfitting and enhance the model's generalization ability. The output module uses an MLP network with a Sigmoid activation function, and the output probability value of the Sigmoid activation function is in the range of [0,1].
5. The method for predicting umami peptides based on deep learning according to claim 4, characterized in that, In the result output module, peptides are classified by setting a probability threshold of 0.
5. Peptides with a probability greater than 0.5 are identified as umami peptides, and those with a probability less than 0.5 are identified as non-umami peptides.
6. The method for predicting umami peptides based on deep learning according to claim 5, characterized in that, The model training in step S4 specifically includes: S41. Training Method: The Umami-Transformer model was trained using a 5-fold cross-validation method, with the BCELoss function used as the loss function to measure the difference between the model's predicted values and the true values. During training, the convergence of the loss curve was monitored, and the model performance was optimized by adjusting hyperparameters until the classification accuracy of the Umami-Transformer model on the test set exceeded 95%. S42. Model optimization, including: Learning rate adjustment: The learning rate is dynamically adjusted based on the changes in loss during training to ensure that the model can converge quickly and reach optimal performance. Batch size settings can balance training speed and model stability, thereby improving training efficiency. Regularization techniques are applied to prevent model overfitting and enhance its generalization ability on new data.
7. The method for predicting umami peptides based on deep learning according to claim 6, characterized in that, Also includes: S51. Performance Evaluation Metrics: A variety of evaluation metrics are used to comprehensively and quantitatively evaluate the trained model, including accuracy, precision, recall, F1 score, and Matthews correlation coefficient. These metrics reflect the classification performance of the model from different perspectives, ensuring its reliability and effectiveness in practical applications. S52. Ablation Experiment: The impact of feature selection on model performance is verified through ablation experiments; S53. Input all dipeptides to pentapeptides into the model to determine umami peptides, and finally select DD, DDE, DDED, and DDEDD for synthesis verification. S54. The flavor profile of the synthetic peptide was analyzed by electronic tongue and sensory evaluation, and the interaction mechanism between the candidate peptide and the umami receptors T1R1 / T1R3 was analyzed by molecular docking technology.
8. A deep learning-based umami peptide prediction system, characterized in that, include: Data construction module: Collect experimentally validated umami peptide and non-umami peptide sequences to construct a UMP499 dataset containing multiple umami peptides and multiple non-umami peptides; Feature selection and processing module: used to standardize and truncate peptide sequences and calculate various physicochemical features; The model building module is used to construct an Umami-Transformer model, which includes a feature processing and generation module, a sequence information processing module, and a result output module. The sequence information processing module uses a Transformer encoder; the feature information processing module uses a multi-layer fully connected network; and the result output module uses an MLP network with a sigmoid activation function to output the probability of umami peptides. The feature processing and generation module processes and generates relevant features, providing rich feature data for the model. The input layer of the feature processing and generation module receives a 16-dimensional encoding vector of the peptide sequence and eight physicochemical features. The physicochemical features are standardized and then concatenated with the sequence encoding. The sequence information processing module uses a 12-layer Transformer encoder, with each layer containing eight attention heads, and compensates for positional information through sine-cosine position encoding. A multi-head attention mechanism is used to deeply analyze and efficiently process peptide sequence information, extracting correlations between amino acids. Transformer encoders can simultaneously focus on different parts of a sequence, capturing complex relationships between amino acids, including key local structural features and long-distance associations. Model training module: The Umami-Transformer model is trained using a 5-fold cross-validation method, with the BCELoss function as the loss function, until the classification accuracy of the Umami-Transformer model on the test set exceeds 95%; Umami peptide prediction module: Input the peptide sequence to be tested into the trained Umami-Transformer model to obtain the probability prediction result that it is an umami peptide.
9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the deep learning-based umami peptide prediction method as described in any one of claims 1-7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the deep learning-based umami peptide prediction method as described in any one of claims 1-7.