Targeting peptide recognition method and device, electronic equipment and storage medium

Abstract

The invention discloses a target peptide recognition method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a to-be-recognized target peptide sequence; based on a protein language model, extracting sequence features corresponding to the target peptide sequence to be identified; at least two classification models are adopted to classify the sequence features, and classification results corresponding to the classification models are obtained; and performing comprehensive analysis on the at least two classification results by adopting a trained prediction model to obtain a target peptide recognition result corresponding to the to-be-recognized target peptide sequence. According to the method, the sequence features corresponding to the to-be-recognized target peptide sequence are classified through the at least two trained classification models, and the at least two classification results are comprehensively analyzed, so that the accuracy of the target peptide recognition result obtained based on each prediction classification result is higher.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a method and device for identifying a targeting peptide, an electronic device, and a storage medium. Background Art

[0002] As the incidence of cancer is increasing worldwide, millions of people die from cancer every year, which seriously affects the health and development of human life. With the advancement of medical technology, there are more and more treatment options for tumors, such as targeted tumor therapy. Targeting peptides are a class of small molecule polypeptides that specifically bind to tumor surface markers and play an important role in targeted tumor therapy. Targeting peptide sequences have specific affinity and selectivity and can bind to targets on tumor cells. Therefore, targeting peptides can be used as part of nucleic acid drug delivery to accurately deliver drugs to tumor cells and reduce damage to normal tissues. Tumors can also be diagnosed and imaged by labeling targeting peptides. At the same time, targeting peptides themselves can also bind to targets on tumor cells to kill tumor cells. Therefore, targeting peptides are extremely important for targeted tumor therapy.

[0003] However, there are few targeted peptide databases for tumor markers, so it is necessary to build a targeted peptide database. At the same time, identifying the targeted peptide of a certain tumor in traditional biological experiments is very time-consuming, material-consuming and labor-intensive, so it is very important to develop tools to predict the sequences of targeted peptides for different tumors, which can provide certain technical support for tumor treatment. Summary of the invention

[0004] The present disclosure provides a method and device for identifying a targeted peptide, an electronic device and a storage medium. The main purpose of the present disclosure is to solve the problem that the accuracy of the existing targeted peptide identification method depends on the accuracy of a preset classification algorithm, but the accuracy of the existing classification algorithm is generally low, resulting in a low recognition accuracy of the existing targeted peptide identification method.

[0005] According to a first aspect of the present disclosure, a method for identifying a targeting peptide is provided, comprising:

[0006] Obtaining a target peptide sequence to be identified;

[0007] Extracting sequence features corresponding to the target peptide sequence to be identified based on a protein language model;

[0008] Using at least two classification models to classify the sequence features respectively, and obtaining classification results corresponding to each classification model;

[0009] The trained prediction model is used to comprehensively analyze at least two classification results to obtain a targeting peptide recognition result corresponding to the targeting peptide sequence to be recognized.

[0010] In some embodiments provided in the present disclosure, the prediction model includes target weight values ​​corresponding to each classification prediction model; the comprehensive analysis of at least two classification results using the trained prediction model to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be identified includes:

[0011] The target weight value is used in the prediction model to perform weighted calculation on each of the classification results to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be recognized.

[0012] In some embodiments provided by the present disclosure, before using the trained prediction model to comprehensively analyze at least two classification results to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be identified, the method further includes:

[0013] Constructing training samples based on a pre-built targeting peptide database; the training samples include training targeting peptide sequences;

[0014] Inputting the training sample into a pre-trained protein language model to obtain a training sequence feature corresponding to the training targeting peptide sequence;

[0015] Input the training sequence features into at least two classification models for prediction, and obtain the prediction classification results corresponding to each classification model;

[0016] Inputting the predicted classification results corresponding to each of the classification models into a preset accuracy update model, and setting an initial weight value corresponding to each predicted classification result and an initial accuracy value of the preset accuracy update model;

[0017] Based on the preset accuracy update model, the initial weight value and the initial accuracy value corresponding to each prediction classification result are updated and calculated synchronously, and the target weight value corresponding to each prediction classification result is determined to obtain a trained prediction model.

[0018] In some embodiments provided by the present disclosure, based on the preset accuracy update model, synchronously updating and calculating the initial weight value and the initial accuracy value corresponding to each training classification result, and determining the target weight value corresponding to each training classification result, includes:

[0019] 1) Using a continuous least squares algorithm to optimize the loss function of the prediction model and calculate a first weight value corresponding to each prediction classification result;

[0020] 2) determining a first prediction value of the targeting peptide sequence based on each prediction classification result and a first weight value corresponding to each prediction classification result;

[0021] 3) calculating a first prediction accuracy based on the first prediction value and the true value of the training targeting peptide sequence;

[0022] 4) if the first prediction accuracy value is greater than the initial accuracy value, updating the initial accuracy value to the first prediction accuracy value; and updating the initial weight value corresponding to each of the training classification results to the first weight value;

[0023] 5) Repeat steps 1)-4) until the preset convergence condition is met, and obtain the target weight value corresponding to each of the predicted classification results.

[0024] In some embodiments provided by the present disclosure, after calculating the first prediction accuracy based on the first training prediction value and the true value of the training targeting peptide sequence, the method further comprises:

[0025] If the first prediction accuracy value is less than or equal to the initial accuracy value, the initial accuracy value and the initial weight value are kept unchanged.

[0026] In some embodiments provided in the present disclosure, the training targeting peptide sequence is constructed based on a pre-built targeting peptide database, including:

[0027] Obtaining a target targeting peptide sequence from the pre-built targeting peptide database as a positive sample;

[0028] Obtaining other targeting peptide sequences other than the target targeting peptide sequence from the pre-built targeting peptide database as negative samples;

[0029] Constructing a training targeted peptide sequence data set based on the positive samples and the negative samples;

[0030] The training targeting peptide sequence data set is divided to obtain the training targeting peptide sequence.

[0031] In some embodiments provided by the present disclosure, before constructing a training targeting peptide sequence based on a pre-built targeting peptide database, the method further comprises:

[0032] Integrate public data sets and literature information to obtain targeting peptide information for different tumors;

[0033] A targeting peptide database is constructed based on the targeting peptide information of different tumors, and the targeting peptide database at least contains the names and sequences of tumor targeting peptides corresponding to different tumor markers.

[0034] According to a second aspect of the present disclosure, a targeting peptide recognition device is provided, comprising:

[0035] A first acquisition unit is used to acquire a target peptide sequence to be identified;

[0036] An extraction unit, used for extracting sequence features corresponding to the target peptide sequence to be identified based on a protein language model;

[0037] A classification unit, used to classify the sequence features using at least two classification models, respectively, to obtain classification results corresponding to each classification model;

[0038] The analysis unit is used to use the trained prediction model to comprehensively analyze at least two classification results to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be recognized.

[0039] In some embodiments provided in the present disclosure, the prediction model includes a target weight value corresponding to each classification prediction model; the analysis unit is also used to use the target weight value in the prediction model to perform weighted calculation on each of the classification results to obtain the targeting peptide identification result corresponding to the targeting peptide sequence to be identified.

[0040] In some embodiments, the apparatus further comprises:

[0041] The first construction unit is used to construct a training sample based on a pre-built targeting peptide database before the analysis unit uses the trained prediction model to comprehensively analyze at least two classification results to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be recognized; the training sample includes a training targeting peptide sequence;

[0042] A second acquisition unit is used to input the training sample into a pre-trained protein language model to obtain a training sequence feature corresponding to the training targeting peptide sequence;

[0043] A third acquisition unit is used to input the training sequence features into at least two classification models for prediction, and obtain the prediction classification results corresponding to each classification model;

[0044] A processing unit, configured to input the prediction classification results corresponding to each of the classification models into a preset accuracy update model, and set an initial weight value corresponding to each prediction classification result and an initial accuracy value of the preset accuracy update model;

[0045] An updating unit is used to update the model based on the preset accuracy, synchronously update the initial weight value and the initial accuracy value corresponding to each predicted classification result, determine the target weight value corresponding to each predicted classification result, so as to obtain a trained prediction model.

[0046] The updating unit is further configured to:

[0047] 1) Using a continuous least squares algorithm to optimize the loss function of the prediction model and calculate a first weight value corresponding to each prediction classification result;

[0048] 2) determining a first prediction value of the targeting peptide sequence based on each prediction classification result and a first weight value corresponding to each prediction classification result;

[0049] 3) calculating a first prediction accuracy based on the first prediction value and the true value of the training targeting peptide sequence;

[0050] 4) if the first prediction accuracy value is greater than the initial accuracy value, updating the initial accuracy value to the first prediction accuracy value; and updating the initial weight value corresponding to each of the training classification results to the first weight value;

[0051] 5) Repeat steps 1)-4) until the preset convergence condition is met, and obtain the target weight value corresponding to each of the predicted classification results.

[0052] In some embodiments provided by the present disclosure, the updating unit is further configured to:

[0053] If the first prediction accuracy value is less than or equal to the initial accuracy value, the initial accuracy value and the initial weight value are kept unchanged.

[0054] In some embodiments provided by the present disclosure, the first building unit is further used for:

[0055] Obtaining a target targeting peptide sequence from the pre-built targeting peptide database as a positive sample;

[0056] Obtaining other targeting peptide sequences other than the target targeting peptide sequence from the pre-built targeting peptide database as negative samples;

[0057] Constructing a training targeted peptide sequence data set based on the positive samples and the negative samples;

[0058] The training targeting peptide sequence data set is divided to obtain the training targeting peptide sequence.

[0059] In some embodiments provided by the present disclosure, the device further includes:

[0060] The fourth acquisition unit is used to integrate public data sets and literature information to obtain targeting peptide information for different tumors;

[0061] The second construction unit is also used to construct a targeting peptide database based on the targeting peptide information of different tumors, wherein the targeting peptide database at least includes the names and sequences of tumor targeting peptides corresponding to different tumor markers.

[0062] According to a third aspect of the present disclosure, there is provided an electronic device, including:

[0063] at least one processor; and

[0064] a memory communicatively connected to the at least one processor; wherein,

[0065] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect.

[0066] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the method described in the first aspect.

[0067] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the method as described in the first aspect above.

[0068] The targeted peptide identification method and device, electronic device and storage medium provided by the present disclosure obtain the targeted peptide sequence to be identified; extract the sequence features corresponding to the targeted peptide sequence to be identified based on the protein language model; use at least two classification models to classify the sequence features respectively, and obtain the classification results corresponding to each classification model; use the trained prediction model to comprehensively analyze the at least two classification results to obtain the targeted peptide identification result corresponding to the targeted peptide sequence to be identified. The present disclosure classifies the sequence features corresponding to the targeted peptide sequence to be identified by at least two trained classification models respectively, and comprehensively analyzes the at least two classification results, so that the accuracy of the targeted peptide identification result obtained based on each predicted classification result is higher.

[0069] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0071] Figure 1 A schematic diagram of a method for identifying a targeted peptide provided in an embodiment of the present disclosure;

[0072] Figure 2 A schematic diagram of another method for identifying a targeted peptide provided by an embodiment of the present disclosure;

[0073] Figure 3 A schematic diagram of a ROC curve of a different classification model provided in an embodiment of the present disclosure;

[0074] Figure 4 A schematic diagram of a method for constructing a training targeting peptide sequence provided in an embodiment of the present disclosure;

[0075] Figure 5 A schematic diagram of another method for identifying a targeted peptide provided by an embodiment of the present disclosure;

[0076] Figure 6 A flowchart of a weight value calculation method provided by an embodiment of the present disclosure;

[0077] Figure 7 A schematic diagram of the structure of a targeting peptide recognition device provided in an embodiment of the present disclosure;

[0078] Figure 8 A schematic diagram of the structure of another targeting peptide recognition device provided in an embodiment of the present disclosure;

[0079] Fig. 9 A schematic block diagram of an exemplary electronic device 700 provided for an embodiment of the present disclosure. DETAILED DESCRIPTION

[0080] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0081] The prefixes such as "first" and "second" in the embodiments of the present disclosure are only for distinguishing different description objects and do not constitute any restrictions on the position, order, priority, quantity or content of the description objects. For the statement of the description objects, please refer to the description in the context of the claims or embodiments, and no unnecessary restrictions should be constituted due to the use of prefixes.

[0082] The targeting peptide identification method and device, electronic device and storage medium of the embodiments of the present disclosure are described below with reference to the accompanying drawings.

[0083] Figure 1 A schematic diagram of a method for identifying a targeted peptide provided by an embodiment of the present disclosure. The method is applied in a computer, such as Figure 1 As shown, the method comprises the following steps:

[0084] Step 101, obtaining a target peptide sequence to be identified.

[0085] Step 102: extracting sequence features corresponding to the target peptide sequence to be identified based on a protein language model.

[0086] At present, the traditional peptide sequence feature extraction method mainly extracts peptide features based on the structure and sequence composition of the peptide, while the protein language recognition model (Evolutionary Scale Modeling, ESM)-1b is a model that uses deep learning technology to predict protein structure and function, and is a Transformer pre-training model. The model takes protein sequence or peptide sequence as input, and after hyperparameter optimization training, outputs the protein features of the extracted protein sequence. Since the peptide sequence is a linear amino acid chain formed by the dehydration condensation of multiple amino acids, it has only a primary structure and no spatial structures such as secondary structure and tertiary structure; and protein is a biological macromolecule composed of amino acids as basic units. The sequence of amino acids on the protein molecule and the three-dimensional structure formed thereby constitute the diversity of protein structure. Proteins have primary, secondary, tertiary, and quaternary structures. The ESM-1b model can extract protein features from proteins, and the obtained protein features include not only amino acid sequence features, but also protein secondary structure, tertiary structure, protein function, and physical and chemical properties.

[0087] Therefore, the method provided by the present disclosure for extracting protein features of a target peptide sequence to be identified using a pre-trained protein language recognition model can extract more target peptide features compared to traditional polypeptide sequence feature extraction methods, and can perform identification through more features, thereby making the target peptide identification method more accurate.

[0088] Step 103, using at least two classification models to classify the sequence features respectively, and obtaining classification results corresponding to each classification model;

[0089] In practical applications, the prediction accuracy is not high because the prediction effect of using a single classification model is not good. In order to improve the prediction accuracy, the present disclosure inputs the extracted sequence features into at least two trained classification models for classification, and then obtains the recognition results corresponding to each classification model. Among them, the number of classification models is an empirical value, which can be flexibly adjusted according to the actual situation. For example, when computing resources are tight, in order to reduce the occupation of computing resources, fewer classification models can be selected for prediction, such as Random forest (RF), KNN, and SVM. These three classification models; when the user hopes that the targeted peptide recognition results are more accurate, the number of classification models can be set to be more, for example, 5 classification models can be selected for prediction; the specific types of classification models and the number of classification models are not limited in the embodiments of this application.

[0090] Step 104: Use the trained prediction model to comprehensively analyze at least two classification results to obtain a targeting peptide recognition result corresponding to the targeting peptide sequence to be recognized.

[0091] Since the target weight value corresponding to each classification result is obtained through training, the targeted peptide recognition result calculated based on each predicted classification result and the target weight value corresponding to each predicted classification result is more accurate than the predicted classification result of a single classification model.

[0092] The targeting peptide identification method provided by the present disclosure obtains the targeting peptide sequence to be identified; extracts the sequence features corresponding to the targeting peptide sequence to be identified based on the protein language model; uses at least two classification models to classify the sequence features respectively, and obtains the classification results corresponding to each classification model; uses the trained prediction model to comprehensively analyze the at least two classification results to obtain the targeting peptide identification result corresponding to the targeting peptide sequence to be identified. The present disclosure classifies the sequence features corresponding to the targeting peptide sequence to be identified by at least two trained classification models respectively, and comprehensively analyzes the at least two classification results, so that the accuracy of the targeting peptide identification result obtained based on each predicted classification result is higher.

[0093] As a detailed description of step 104, when using the trained prediction model to comprehensively analyze at least two classification results to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be identified, it can be achieved by but not limited to the following methods, for example: the prediction model includes a target weight value corresponding to each classification prediction model, and the target weight value is used in the prediction model to perform weighted calculation on each of the classification results to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be identified.

[0094] For example, assuming that the predicted classification result obtained by classification model A is y1, and the target weight value of y1 is a; the predicted classification result obtained by classification model B is y2, and the target weight value of y2 is b; the predicted classification result obtained by classification model C is y3, and the target weight value of y3 is c, where a+b+c=1, then a comprehensive analysis is performed in the prediction model, and the comprehensive prediction classification result obtained is (y1×a+y2×b+y3×c) / 3. Specifically, the embodiment of the present application does not limit the number of classification models and target weight values.

[0095] like Figure 2 As shown, the present disclosure also provides another method for identifying a targeting peptide, comprising:

[0096] Step 201, integrating public data sets and literature information to obtain targeting peptide information for different tumors;

[0097] Step 202: construct a targeting peptide database based on the targeting peptide information of different tumors, wherein the targeting peptide database at least includes the names and sequences of tumor targeting peptides corresponding to different tumor markers.

[0098] In practical applications, commonly used peptide databases include antimicrobial peptide databases, signal peptide databases, etc., but there is no targeting peptide database for different tumor markers. Therefore, the disclosed embodiment provides a pre-built targeting peptide database for storing targeting peptides for different tumor markers. Among them, the specific construction method of the targeting peptide database can be adopted but not limited to: integrating public data sets and literature information (targeting peptide related literature), and obtaining targeting peptide information of different tumors to obtain targeting peptide information files, and the targeting peptide information files contain targeting peptide information of different tumors; constructing a targeting peptide database based on the targeting peptide information of different tumors, and the targeting peptide database at least contains the tumor targeting peptide name and targeting peptide sequence corresponding to different tumor markers. In order to facilitate content viewing, the targeting peptide database is visualized.

[0099] Step 203: construct training samples based on a pre-built targeting peptide database; the training samples include training targeting peptide sequences.

[0100] By building a targeting peptide database for different tumor markers and obtaining training samples from the targeting peptide database, the training samples include training targeting peptide sequences. Compared with the method of obtaining targeting peptides from other peptide databases, the process of constructing training samples based on the targeting peptide database takes less time and is more efficient.

[0101] Step 204: input the training sample into a pre-trained protein language model to obtain a training sequence feature corresponding to the training targeting peptide sequence.

[0102] The present disclosure provides a method for extracting training sequence features from training samples (training targeting peptide sequences) using a pre-trained protein language recognition model. Compared with traditional methods for extracting features from polypeptide sequences, more targeting peptide features can be extracted, and identification can be performed through more features, making the targeting peptide identification method more accurate.

[0103] Step 205: Input the training sequence features into at least two classification models for prediction, and obtain the prediction classification results corresponding to each classification model.

[0104] In practical applications, the sample is usually predicted by a decision tree model, a support vector machine, a random forest, a K-nearest neighbor and other classification models, but the effect of using a single classification model for prediction is not good, and the prediction accuracy is not high. In order to improve the prediction accuracy, the present disclosure will be extracted from the training targeting peptide sequence features. The protein features are respectively input into at least two trained classification models for prediction, and then the recognition results corresponding to each classification model are obtained. Among them, the number of classification models is an empirical value, which can be flexibly adjusted according to the actual situation. For example, when the computing resources are relatively tight, in order to reduce the occupation of computing resources, fewer classification models can be selected for prediction, such as Random Forest (RF), KNN, these two classification models, or Random Forest (RF), KNN, SVM, these three classification models can be selected; when the user hopes that the targeting peptide recognition result is more accurate, the number of classification models can be set more, for example, five classification models can be selected for prediction; the specific classification model type, and the number of classification models in this application embodiment are not limited.

[0105] Step 206: input the predicted classification results corresponding to each classification model into a preset accuracy update model, and set an initial weight value corresponding to each predicted classification result and an initial accuracy value of the preset accuracy update model.

[0106] In some embodiments provided by the present disclosure, after the predicted classification results corresponding to each of the classification models are input into the preset accuracy update model, the preset accuracy update model will configure a weight value for each predicted classification result, so that a comprehensive predicted classification result can be obtained based on each predicted classification result and its corresponding weight value. Exemplarily, assuming that the predicted classification result obtained by classification model A is y1, and the weight value of y1 is a; the predicted classification result obtained by classification model B is y2, and the weight value of y2 is b; the predicted classification result obtained by classification model C is y3, and the weight value of y3 is c, where a+b+c=1, then the comprehensive predicted classification result is (y1×a+y2×b+y3×c) / 3. However, it cannot be determined whether the comprehensive predicted classification result obtained by the weight value randomly configured for each predicted classification result is more accurate than the predicted classification result obtained by a single classification model. Therefore, it is necessary to iteratively update the initial accuracy value of the preset accuracy update model.

[0107] Step 207, based on the preset accuracy update model, synchronously update and calculate the initial weight value and the initial accuracy value corresponding to each predicted classification result, determine the target weight value corresponding to each predicted classification result, and obtain the trained prediction model.

[0108] In one embodiment provided by the present disclosure, after determining the initial weight value corresponding to each prediction classification result and the initial accuracy value of the preset accuracy update model, the preset accuracy update model will call its own accuracy update algorithm to iteratively update the accuracy value of the preset accuracy update model until the accuracy value of the preset accuracy update model reaches the maximum, and when updating the accuracy value of the preset accuracy update model, the weight value corresponding to each prediction classification result will also be synchronously updated. When the accuracy value of the preset accuracy update model reaches the maximum, the target weight value corresponding to each prediction classification result is obtained.

[0109] As another implementation method of the embodiment of the present disclosure, in addition to the implementation method in which the accuracy value of the preset accuracy update model reaches the maximum, the accuracy value of the preset accuracy update model can also be set to reach a preset threshold. The preset threshold can be set according to actual needs or a priori, and the specific accuracy value of the preset accuracy update model is not limited.

[0110] In one embodiment provided by the present disclosure, the accuracy value is the accuracy evaluation index of the preset accuracy update model. In the embodiment of the present disclosure, the AUC (Area Under Curve) value can be used as an indicator to measure the performance of the preset accuracy update model, but is not limited to it. The AUC value is the area of ​​the figure surrounded by the receiver operating characteristic curve (Receiver Operating Characteristic curve, ROC curve) and the horizontal axis. The closer the AUC value is to 1, the better the performance of the model, that is, the higher the prediction accuracy value of the model, wherein the ROC curve is a graphical tool for evaluating the performance of a binary classification model. It shows the performance of the model by plotting the true positive rate (True Positive Rate, TPR) and the false positive rate (False Pos itive Rate, FPR) under different classification thresholds. The closer the ROC curve is to the upper left corner, the better the performance of the model, that is, the false alarm rate is low and the correct recognition rate is high.

[0111] To better understand the AUC value, Figure 3 A schematic diagram of a ROC curve of a different classification model provided in an embodiment of the present disclosure. Figure 3 In the figure, the Muti_TAR model represents the preset accuracy update model provided by the present disclosure, GB is the gradient boosting model (Gradient Boosting, GB), RF is the random forest model, LG is the logistic regression (LogisticRegression, LG) model, SVM is the support vector machine model, DT is the decision tree (Decision Tree, DT) model, KNN is the K nearest neighbor model, and AB is the Bayesian model. It can be clearly seen from Figure 3It can be seen that the ROC curve of the preset accuracy update model is closest to the upper left corner, and the AUC value is the largest, which can reach above 0.91, while the AUC values ​​of other classification models are all between 0.70 and 0.83. Obviously, the accuracy value of the preset accuracy update model of the present invention is higher. Therefore, the recognition accuracy of the targeted peptide recognition method provided by the present invention is also higher.

[0112] Step 208, using the target weight value in the prediction model to perform weighted calculation on each of the classification results, to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be recognized.

[0113] Since the target weight value corresponding to each predicted classification result is obtained when the accuracy value of the preset accuracy update model reaches the maximum, the targeted peptide recognition result is calculated based on each predicted classification result and the target weight value corresponding to each predicted classification result, which is more accurate than the predicted classification result of a single classification model. The specific calculation process can refer to the calculation method of the comprehensive predicted classification result mentioned in step 206. Exemplarily, assuming that the predicted classification result obtained by classification model A is y1, and the target weight value of y1 is w1; the predicted classification result obtained by classification model B is y2, and the target weight value of y2 is w2; the predicted classification result obtained by classification model C is y3, and the target weight value of y3 is w3, wherein w1+w2+w3=1, then the targeted peptide recognition result can be calculated by formula (1).

[0114] Y=(y1×w1+ y2×w2+ y3×w3) / 3 Formula (1)

[0115] Wherein, Y represents the target peptide recognition result.

[0116] When there are more than three classification models with their corresponding predicted classification results, the above formula (1) can be further expanded to obtain formula (2):

[0117]

[0118] Among them, Y represents the target peptide recognition result, y i represents the i-th predicted classification result, w i Represents y i The corresponding target weight value, n represents the number of predicted classification results.

[0119] To further explain the above step 203, the construction of training targeting peptide sequences based on the pre-built targeting peptide database may be carried out by, but not limited to, the following methods: Figure 4 As shown, including:

[0120] Step 301: Obtain a target targeting peptide sequence from the pre-built targeting peptide database, and use the target targeting peptide sequence as a positive sample.

[0121] Step 302: acquiring other targeting peptide sequences except the target targeting peptide sequence from the pre-built targeting peptide database, and taking the other targeting peptide sequences as negative samples.

[0122] Taking the targeting peptide sequence for melanocytoma as an example, the targeting peptide sequence for melanocytoma is obtained from the targeting peptide database as a positive sample; and the targeting peptide sequences for other tumors are obtained from the targeting peptide database as negative samples.

[0123] Step 303: construct a training targeting peptide sequence data set based on the positive samples and the negative samples.

[0124] Step 304: divide the training targeting peptide sequence data set to obtain the training targeting peptide sequences.

[0125] In one embodiment provided by the present disclosure, the data set can be divided by, but not limited to, a ten-fold cross validation method. Exemplarily, the training targeting peptide sequence data set is divided into 10 subsets using the ten-fold cross validation method, 2 of which are determined as training targeting peptide sequences, and the remaining 8 subsets are determined as training sets.

[0126] In some embodiments provided by the present disclosure, in order to update the model based on the preset accuracy, synchronously update the model based on the preset accuracy, synchronously update the initial weight value and the initial accuracy value corresponding to each training classification result, determine the target weight value corresponding to each training classification result, and better understand the process of determining the target weight value corresponding to each training classification result. Figure 5 A flowchart of a precision value updating method provided by an embodiment of the present disclosure is shown in FIG. Figure 5 As shown, the method comprises the following steps:

[0127] Step 401, using a continuous least squares algorithm to optimize the loss function of the prediction model, and calculate a first weight value corresponding to each predicted classification result.

[0128] When the first weight value corresponding to each predicted classification result is respectively obtained, it can be implemented in the following ways but not limited to: Figure 6 As shown, including:

[0129] Step 501: Under preset constraints, the continuous least squares algorithm is used to optimize the loss function of the prediction model to obtain an optimized loss function.

[0130] Step 502: Based on the optimized loss function, call the continuous least squares algorithm to optimize the first weight value to obtain the first weight value corresponding to each training classification result.

[0131] In one embodiment provided in the present disclosure, the preset sequential least squares programming (SLSQP) algorithm is a nonlinear constrained optimization algorithm that approximates the original problem by minimizing the quadratic approximation of the objective function, and restricts it according to the constraint conditions to obtain a solution that satisfies the constraint conditions. In the embodiment of the present disclosure, the constraint condition is that the sum of the weight values ​​corresponding to each predicted classification result is a fixed value. For example, the weight value of the predicted classification result y1 of classification model A is w1; the weight value of the predicted classification result y2 of classification model B is w2, and the weight value of the predicted classification result y3 of classification model C is w3, then the constraint condition can be set to w1+w2+w3=1. The objective function is a binary cross entropy loss function, and the first weight value is a set of solutions of the preset continuous least squares algorithm.

[0132] Step 402: determining a first prediction value of the training targeting peptide sequence based on each prediction classification result and a first weight value corresponding to each prediction classification result.

[0133] In order to better understand the process of determining the first prediction value, for example, if the first weight value of the predicted classification result y1 is w1; the first weight value of the predicted classification result y2 is w2, and the first weight value of the predicted classification result y3 is w3, then the first prediction value can be substituted into the above formula (1) for calculation.

[0134] Step 403: Calculate a first prediction accuracy based on the first prediction value and the true value of the training targeting peptide sequence.

[0135] In one embodiment provided by the present disclosure, the preset accuracy value calculation algorithm can adopt but is not limited to the AUC value calculation algorithm. AUC is a performance indicator for measuring the quality of a learner. From the definition, it can be seen that AUC can be obtained by summing the areas of each part under the ROC curve. The calculation principle is: the proportion of positive samples in all sample pairs that are ranked before negative samples.

[0136] You can also use the ROC curve method to determine the AUC value. Please continue to refer to Figure 3The horizontal axis of the ROC curve is FPR, and the vertical axis is TPR. The predicted classification results are sorted according to the probability value of the predicted positive sample. The threshold is gradually reduced from 1. The samples are predicted as positive samples one by one in this order. The current FPR can be calculated each time. The TPR value is plotted with TPR as the vertical axis and FPR as the horizontal axis to obtain the ROC curve.

[0137] Step 404: determine whether the first prediction accuracy value is greater than the initial accuracy value.

[0138] If the first prediction accuracy value is greater than the initial accuracy value, step 4041 is executed; if the first prediction accuracy value is less than or equal to the initial accuracy value, step 4042 is executed;

[0139] Step 4041, updating the initial accuracy value of the preset accuracy update model to the first prediction accuracy value; and updating the initial weight value corresponding to each training classification result to the first weight value.

[0140] Step 4042, keeping the initial accuracy value and the initial weight value unchanged.

[0141] Step 405, repeating steps 401-404 until a preset convergence condition is met, and obtaining the target weight value corresponding to each of the predicted classification results.

[0142] The preset convergence condition includes but is not limited to the number of iterations of the preset accuracy update model reaching the preset number of iterations, or setting an allowable error threshold. After being less than or equal to the allowable error threshold, the preset convergence condition is satisfied. Specifically, the embodiments of the present disclosure do not limit the preset convergence condition.

[0143] After obtaining the target weight value, each predicted classification result and the target weight value corresponding to each predicted classification result can be substituted into the above formula (1) or formula (2) to obtain the targeted peptide recognition result.

[0144] In summary, the embodiments of the present disclosure can achieve the following effects:

[0145] 1. The targeted peptide identification method provided by the present disclosure combines different classification models and has strong robustness.

[0146] 2. The present invention integrates the targeting peptide information for different tumors through a pre-built targeting peptide database. The data comes from a wide range of sources and covers a large number of targeting peptide sequences, providing sufficient data for the development of subsequent algorithms.

[0147] 3. Improved the recognition accuracy of the targeted peptide recognition method.

[0148] Corresponding to the above-mentioned targeting peptide identification method, the present invention also provides a targeting peptide identification device. Since the device embodiment of the present invention corresponds to the above-mentioned method embodiment, details not disclosed in the device embodiment can be referred to the above-mentioned method embodiment, and will not be repeated in the present invention.

[0149] Figure 7 A schematic diagram of the structure of a targeting peptide recognition device provided in an embodiment of the present disclosure is shown in FIG. Figure 7 As shown, including:

[0150] A first acquisition unit 51 is used to acquire a target peptide sequence to be identified;

[0151] An extraction unit 52 is used to extract sequence features corresponding to the target peptide sequence to be identified based on a protein language model;

[0152] A classification unit 53 is used to classify the sequence features using at least two classification models to obtain classification results corresponding to each classification model;

[0153] The analysis unit 54 is used to use the trained prediction model to comprehensively analyze at least two classification results to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be recognized.

[0154] The targeting peptide recognition device provided by the present disclosure obtains the targeting peptide sequence to be identified; extracts the sequence features corresponding to the targeting peptide sequence to be identified based on the protein language model; uses at least two classification models to classify the sequence features respectively, and obtains the classification results corresponding to each classification model; uses the trained prediction model to comprehensively analyze the at least two classification results to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be identified. The present disclosure classifies the sequence features corresponding to the targeting peptide sequence to be identified by at least two trained classification models respectively, and comprehensively analyzes the at least two classification results, so that the accuracy of the targeting peptide recognition result obtained based on each predicted classification result is higher.

[0155] Furthermore, in a possible implementation of this embodiment, as Figure 8 As shown, the prediction model includes target weight values ​​corresponding to each classification prediction model; the analysis unit 54 is also used to use the target weight value in the prediction model to perform weighted calculation on each classification result to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be identified.

[0156] In some embodiments, the apparatus further comprises:

[0157] The first construction unit 55 is used to construct a training sample based on a pre-built targeting peptide database before the analysis unit 54 uses the trained prediction model to comprehensively analyze at least two classification results to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be recognized; the training sample includes a training targeting peptide sequence;

[0158] A second acquisition unit 56 is used to input the training sample into a pre-trained protein language model to obtain a training sequence feature corresponding to the training targeting peptide sequence;

[0159] A third acquisition unit 57 is used to input the training sequence features into at least two classification models for prediction, and obtain the prediction classification results corresponding to each classification model;

[0160] The processing unit 58 is used to input the prediction classification results corresponding to each of the classification models into the preset accuracy update model, and set the initial weight value corresponding to each prediction classification result and the initial accuracy value of the preset accuracy update model;

[0161] The updating unit 59 is used to update the model based on the preset accuracy, synchronously update the initial weight value and the initial accuracy value corresponding to each predicted classification result, determine the target weight value corresponding to each predicted classification result, and obtain the trained prediction model.

[0162] The updating unit 59 is also used for:

[0163] 1) Using a continuous least squares algorithm to optimize the loss function of the prediction model and calculate a first weight value corresponding to each prediction classification result;

[0164] 2) determining a first prediction value of the targeting peptide sequence based on each prediction classification result and a first weight value corresponding to each prediction classification result;

[0165] 3) calculating a first prediction accuracy based on the first prediction value and the true value of the training targeting peptide sequence;

[0166] 4) if the first prediction accuracy value is greater than the initial accuracy value, updating the initial accuracy value to the first prediction accuracy value; and updating the initial weight value corresponding to each of the training classification results to the first weight value;

[0167] 5) Repeat steps 1)-4) until the preset convergence condition is met, and obtain the target weight value corresponding to each of the predicted classification results.

[0168] Furthermore, in a possible implementation of this embodiment, as Figure 8 As shown, the updating unit 59 is also used for:

[0169] If the first prediction accuracy value is less than or equal to the initial accuracy value, the initial accuracy value and the initial weight value are kept unchanged.

[0170] Furthermore, in a possible implementation of this embodiment, as Figure 8 As shown, the first construction unit 55 is also used for:

[0171] Obtaining a target targeting peptide sequence from the pre-built targeting peptide database as a positive sample;

[0172] Obtaining other targeting peptide sequences other than the target targeting peptide sequence from the pre-built targeting peptide database as negative samples;

[0173] Constructing a training targeted peptide sequence data set based on the positive samples and the negative samples;

[0174] The training targeting peptide sequence data set is divided to obtain the training targeting peptide sequence.

[0175] Furthermore, in a possible implementation of this embodiment, as Figure 8 As shown, the device also includes:

[0176] The fourth acquisition unit 510 is used to integrate public data sets and literature information to obtain targeting peptide information of different tumors;

[0177] The second construction unit 511 is further used to construct a targeting peptide database based on the targeting peptide information of different tumors, wherein the targeting peptide database at least includes the names and sequences of tumor targeting peptides corresponding to different tumor markers.

[0178] It should be noted that the above explanation of the method embodiment is also applicable to the device of this embodiment, and the principle is the same, which is not limited in this embodiment.

[0179] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0180] Fig. 9A schematic block diagram of an example electronic device 700 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0181] like Fig. 9 As shown, the device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a ROM (Read-Only Memory) 702 or a computer program loaded from a storage unit 708 to a RAM (Random Access Memory) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An I / O (Input / Output) interface 705 is also connected to the bus 704.

[0182] A number of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0183] The computing unit 701 can be various general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, CPU (Central Processing Unit), GPU (Graphic Processing Units), various dedicated AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSP (Digital Signal Processor), and any appropriate processor, controller, microcontroller, etc. The computing unit 701 performs the various methods and processes described above, such as the targeting peptide identification method. For example, in some embodiments, the targeting peptide identification method can be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 700 via ROM 702 and / or communication unit 709. When the computer program is loaded into RAM 703 and executed by the computing unit 701, one or more steps of the method described above can be executed. Alternatively, in other embodiments, the computing unit 701 may be configured to execute the aforementioned targeting peptide identification method in any other appropriate manner (eg, by means of firmware).

[0184] Various embodiments of the systems and techniques described above herein may be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application Specific Standard Products), SOCs (System On Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that may be executed and / or interpreted on a programmable system including at least one programmable processor that may be a special purpose or general purpose programmable processor that may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0185] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0186] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a RAM, a ROM, an EPROM (Electrically Programmable Read-Only-Memory) or a flash memory, an optical fiber, a CD-ROM (Compact Dis sc Read-Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0187] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball), through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0188] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: LAN (Local Area Network), WAN (Wide Area Network), the Internet, and blockchain networks.

[0189] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0190] It should be noted that artificial intelligence is a discipline that studies how computers can simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.), and includes both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, as well as machine learning / deep learning, big data processing technology, knowledge graph technology, and other major directions.

[0191] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0192] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for identifying a targeting peptide, characterized in that: include: Obtaining a target peptide sequence to be identified; Extracting sequence features corresponding to the target peptide sequence to be identified based on a protein language model; Using at least two classification models to classify the sequence features respectively, and obtaining classification results corresponding to each classification model; The trained prediction model is used to comprehensively analyze at least two classification results to obtain a targeting peptide recognition result corresponding to the targeting peptide sequence to be recognized.

2. The method according to claim 1, characterized in that The prediction model includes target weight values ​​corresponding to each classification prediction model; the comprehensive analysis of at least two classification results using the trained prediction model to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be identified includes: The target weight value is used in the prediction model to perform weighted calculation on each of the classification results to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be recognized.

3. The method according to claim 1, characterized in that Before comprehensively analyzing at least two classification results using the trained prediction model to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be recognized, the method further includes: Constructing training samples based on a pre-built targeting peptide database; the training samples include training targeting peptide sequences; Inputting the training sample into a pre-trained protein language model to obtain a training sequence feature corresponding to the training targeting peptide sequence; Input the training sequence features into at least two classification models for prediction, and obtain the prediction classification results corresponding to each classification model; Inputting the predicted classification results corresponding to each of the classification models into a preset accuracy update model, and setting an initial weight value corresponding to each predicted classification result and an initial accuracy value of the preset accuracy update model; Based on the preset accuracy update model, the initial weight value and the initial accuracy value corresponding to each prediction classification result are updated and calculated synchronously, and the target weight value corresponding to each prediction classification result is determined to obtain a trained prediction model.

4. The method according to claim 3, characterized in that The updating model based on the preset accuracy, synchronously updating and calculating the initial weight value and the initial accuracy value corresponding to each training classification result, and determining the target weight value corresponding to each training classification result, includes: 1) Using a continuous least squares algorithm to optimize the loss function of the prediction model and calculate a first weight value corresponding to each prediction classification result; 2) determining a first prediction value of the training targeting peptide sequence based on each prediction classification result and a first weight value corresponding to each prediction classification result; 3) calculating a first prediction accuracy based on the first prediction value and the true value of the training targeting peptide sequence; 4) if the first prediction accuracy value is greater than the initial accuracy value, updating the initial accuracy value to the first prediction accuracy value; and updating the initial weight value corresponding to each of the training classification results to the first weight value; 5) Repeat steps 1)-4) until the preset convergence condition is met, and obtain the target weight value corresponding to each of the predicted classification results.

5. The method according to claim 4, characterized in that After calculating a first prediction accuracy based on the first training prediction value and the true value of the training targeting peptide sequence, the method further comprises: If the first prediction accuracy value is less than or equal to the initial accuracy value, the initial accuracy value and the initial weight value are kept unchanged.

6. The method according to claim 3, characterized in that The method of constructing a training targeting peptide sequence based on a pre-built targeting peptide database comprises: Obtaining a target targeting peptide sequence from the pre-built targeting peptide database as a positive sample; Obtaining other targeting peptide sequences other than the target targeting peptide sequence from the pre-built targeting peptide database as negative samples; Constructing a training targeted peptide sequence data set based on the positive samples and the negative samples; The training targeting peptide sequence data set is divided to obtain the training targeting peptide sequence.

7. The method according to claim 2, characterized in that Before constructing a training targeting peptide sequence based on a pre-built targeting peptide database, the method further includes: Integrate public data sets and literature information to obtain targeting peptide information for different tumors; A targeting peptide database is constructed based on the targeting peptide information of different tumors, and the targeting peptide database at least contains the names and sequences of tumor targeting peptides corresponding to different tumor markers.

8. A targeting peptide recognition device, characterized in that: include: An acquisition unit, used for acquiring a target peptide sequence to be identified; An extraction unit, used for extracting sequence features corresponding to the target peptide sequence to be identified based on a protein language model; A classification unit, used to classify the sequence features using at least two classification models, respectively, to obtain classification results corresponding to each classification model; The analysis unit is used to use the trained prediction model to comprehensively analyze at least two classification results to obtain the targeting peptide recognition result corresponding to the targeting peptide sequence to be recognized.

9. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-7.

Images