Method, model and device for predicting binding affinity of polypeptides to mhc ii

Abstract

This invention relates to the field of biomedical technology and discloses a method, model, and apparatus for predicting the binding affinity of peptides to MHCII. The method includes: extracting sequence feature information of a target peptide sequence and a target MHCII sequence; obtaining peptide-related attention scores and MHCII-related attention scores based on the target peptide sequence and the target MHCII sequence; obtaining peptide polymerization features based on the peptide-related attention scores and the sequence feature information of the target peptide sequence; obtaining MHCII polymerization features based on the MHCII-related attention scores and the sequence feature information of the target MHCII sequence; and inputting the peptide polymerization features and MHCII polymerization features into an affinity prediction neural network model to predict the affinity between the target peptide sequence and the target MHCII sequence. This invention can improve the efficiency and accuracy of determining the binding affinity of peptides to MHCII.

Description

Technical Field

[0001] This invention relates to the field of biomedical technology, specifically to a method, model, and apparatus for predicting the binding affinity of peptides to MHCII. Background Technology

[0002] Major histocompatibility complex (MHC) molecules play a crucial role in the human immune response. They present antigenic peptides to T cells, thereby activating a specific immune response. There are two main types of MHC molecules: MHC I molecules and MHC II molecules. MHC I molecules present peptides that activate CD8+ T cells, while MHC II molecules present peptides that activate CD4+ T cells.

[0003] MHC class II molecules primarily present exogenous antigenic peptides. They possess open binding grooves capable of accommodating relatively long peptide segments (typically 12-20 amino acids), with their binding core typically consisting of 9 amino acids. MHC class II molecules exhibit high polymorphism, with numerous different allelic forms, resulting in diverse and complex sequence characteristics of their binding peptides. Experimentally determining the binding affinity of a large number of MHC class II molecules for peptides is not only time-consuming and laborious but also costly, and it is difficult to cover all MHC class II alleles and potential peptide sequence spaces. Summary of the Invention

[0004] This invention provides a method, model, and apparatus for predicting the binding affinity of peptides to MHCII, in order to solve the problems of low efficiency and high cost in experimentally determining the binding affinity of a large number of MHCII molecules to peptides, and to ensure the accuracy of determining the binding affinity of peptides to MHCII.

[0005] In a first aspect, the present invention provides a method for predicting the binding affinity of a peptide to MHCII, the method comprising:

[0006] Extract sequence feature information of the target polypeptide sequence and the target MHCII sequence;

[0007] Based on the target peptide sequence and the target MHCII sequence, peptide-related attention scores and MHCII-related attention scores are obtained; the peptide-related attention score is the attention score of an atom in the target peptide sequence relative to the target MHCII sequence, and the MHCII-related attention score is the attention score of an atom in the target MHCII sequence relative to the target peptide sequence.

[0008] Based on the peptide-related attention score and the sequence feature information of the target peptide sequence, peptide aggregation features are obtained;

[0009] Based on the MHCII-related attention scores and the sequence feature information of the target MHCII sequence, MHCII aggregation features are obtained;

[0010] The peptide polymerization features and the MHCII polymerization features are input into an affinity prediction neural network model to predict the affinity of the target peptide sequence and the target MHCII sequence.

[0011] In one optional implementation, obtaining the peptide-related attention score and the MHCII-related attention score based on the target peptide sequence and the target MHCII sequence includes:

[0012] Obtain the polypeptide sequence code corresponding to the target polypeptide sequence and the MHCII sequence code corresponding to the target MHCII sequence;

[0013] Multiply the peptide sequence encoding and the MHCII sequence encoding, and obtain the peptide-MHCII attention score based on the resulting product and the MHCII sequence encoding; obtain the peptide-related attention score based on the peptide-MHCII attention score; and / or,

[0014] The MHCII sequence encoding and the peptide sequence encoding are multiplied together, and the MHCII-peptide attention score is obtained based on the product and the peptide sequence encoding; the MHCII-related attention score is obtained based on the MHCII-peptide attention score.

[0015] In one optional implementation, the extraction of sequence feature information of the target polypeptide sequence and the target MHCII sequence includes:

[0016] The target polypeptide sequence and the target MHCII sequence were respectively transformed using an amino acid numerical dictionary;

[0017] The target polypeptide sequence and the target MHCII sequence after word-based encoding are initially encoded to obtain initial polypeptide sequence encoding and initial MHCII sequence encoding; the initial polypeptide sequence encoding and the initial MHCII sequence encoding include word encoding and position encoding.

[0018] The initial peptide sequence code and the initial MHCII sequence code are re-encoded using a convolutional neural network model to obtain the final peptide sequence code and MHCII sequence code.

[0019] Feature extraction is performed on the final encoded polypeptide sequence and the encoded MHCII sequence respectively to obtain the sequence feature information of the target polypeptide sequence and the sequence feature information of the target MHCII sequence.

[0020] In one optional implementation, the feature extraction from the final polypeptide sequence encoding and the MHCII sequence encoding, respectively, includes:

[0021] The final polypeptide sequence encoding and the MHCII sequence encoding are respectively input into two BERT models for feature extraction;

[0022] The BERT model includes a stacked multi-layer encoder, which sequentially includes a first multi-head attention layer, a first normalization layer, a first residual layer, a first fully connected layer, a second normalization layer, and a second residual layer.

[0023] In one alternative implementation, during model training, the learning rate of the affinity prediction neural network model is 10. -4 -10 -3 The learning rate of the BERT model is 10. -5 -10 -4 .

[0024] In one alternative implementation, the affinity prediction neural network model is a feedforward neural network model.

[0025] Secondly, the present invention provides a model for predicting the binding affinity of a peptide to MHCII, the model comprising:

[0026] The feature extraction module is used to extract sequence feature information of the target polypeptide sequence and the target MHCII sequence;

[0027] The relevant attention score acquisition module is used to acquire a peptide-related attention score and an MHCII-related attention score based on the target peptide sequence and the target MHCII sequence; the peptide-related attention score is the attention score of an atom in the target peptide sequence relative to the target MHCII sequence, and the MHCII-related attention score is the attention score of an atom in the target MHCII sequence relative to the target peptide sequence;

[0028] The feature aggregation module is used to obtain peptide aggregation features based on the peptide-related attention score and the sequence feature information of the target peptide sequence; and to obtain MHCII aggregation features based on the MHCII-related attention score and the sequence feature information of the target MHCII sequence.

[0029] An affinity prediction neural network model is used to predict the affinity of the target polypeptide sequence and the target MHCII sequence based on the polypeptide polymerization characteristics and the MHCII polymerization characteristics.

[0030] Thirdly, the present invention provides a device for predicting the binding affinity of a peptide to MHCII, the device comprising:

[0031] The feature extraction module is used to extract sequence feature information of the target polypeptide sequence and the target MHCII sequence;

[0032] The relevant attention score acquisition module is used to acquire a peptide-related attention score and an MHCII-related attention score based on the target peptide sequence and the target MHCII sequence; the peptide-related attention score is the attention score of an atom in the target peptide sequence relative to the target MHCII sequence, and the MHCII-related attention score is the attention score of an atom in the target MHCII sequence relative to the target peptide sequence;

[0033] The peptide polymerization feature acquisition module is used to obtain peptide polymerization features based on the peptide-related attention score and the sequence feature information of the target peptide sequence;

[0034] The MHCII aggregation feature acquisition module is used to obtain MHCII aggregation features based on the MHCII related attention score and the sequence feature information of the target MHCII sequence;

[0035] The prediction module is used to input the polypeptide polymerization features and the MHCII polymerization features into the affinity prediction neural network model to predict the affinity of the target polypeptide sequence and the target MHCII sequence.

[0036] Fourthly, the present invention provides an electronic device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the method for predicting the binding affinity of a peptide to MHCII as described in the first aspect or any corresponding embodiment.

[0037] Fifthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the method for predicting the binding affinity of a peptide to MHCII according to the first aspect or any corresponding embodiment described above.

[0038] In a sixth aspect, the present invention provides a computer program product, including computer instructions for causing a computer to execute the method for predicting the binding affinity of a peptide to MHCII according to the first aspect or any corresponding embodiment described above.

[0039] The binding affinity prediction method, model, and apparatus for peptides and MHCII provided in this invention utilize a neural network model to predict the affinity between peptide and MHCII sequences, achieving high efficiency and low cost. Furthermore, to ensure prediction accuracy, this invention first obtains peptide-related attention scores and MHCII-related attention scores based on the target peptide and target MHCII sequences. Based on the peptide-related attention scores and sequence feature information of the target peptide sequence, peptide aggregation features are obtained. Based on the MHCII-related attention scores and sequence feature information of the target MHCII sequence, MHCII aggregation features are obtained. Then, the peptide aggregation features and MHCII aggregation features are input into the affinity prediction neural network model for affinity prediction. In other words, this embodiment provides a novel feature extraction method. After extracting the sequence features of the peptide and MHCII, feature aggregation is performed based on the extracted sequence features, achieving sufficient learning of the sequence features of the peptide and MHCII. The affinity prediction neural network model predicts the binding affinity between the peptide and MHCII based on the aggregation features, resulting in higher accuracy. Attached Figure Description

[0040] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0041] Figure 1 This is a schematic diagram of the first process of the method for predicting the binding affinity of a peptide to MHCII according to an embodiment of the present invention.

[0042] Figure 2 This is a schematic diagram of the second process for predicting the binding affinity of a peptide to MHCII according to an embodiment of the present invention.

[0043] Figure 3 This is a schematic diagram of the structure of a CNN model according to an embodiment of the present invention;

[0044] Figure 4 This is a schematic diagram of the structure of the BERT model according to an embodiment of the present invention;

[0045] Figure 5This is a schematic diagram illustrating the process of obtaining peptide-related attention scores according to an embodiment of the present invention;

[0046] Figure 6 This is a structural block diagram of a peptide-MHCII binding affinity prediction device according to an embodiment of the present invention.

[0047] Figure 7 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0050] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0051] According to an embodiment of the present invention, a method for predicting the binding affinity of a peptide to MHCII is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0052] This embodiment provides a method for predicting the binding affinity of a peptide to MHCII, which can be used in various electronic devices, such as mobile phones, tablets, laptops, and desktop computers. Figure 1 This is a flowchart of a method for predicting the binding affinity of a peptide to MHCII according to an embodiment of the present invention, as shown below. Figure 1 As shown, the process includes the following steps:

[0053] Step S101: Extract sequence feature information of the target polypeptide sequence and the target MHCII sequence.

[0054] In this embodiment, the process of extracting sequence feature information of the target polypeptide sequence and the target MHCII sequence includes: first, using an amino acid digital dictionary to perform word transformation on the target polypeptide sequence and the target MHCII sequence respectively; then, encoding the word-transformed target polypeptide sequence and the target MHCII sequence respectively to obtain polypeptide sequence encoding and MHCII sequence encoding; finally, extracting sequence feature information based on polypeptide sequence encoding and MHCII sequence encoding respectively.

[0055] In some optional embodiments, step S101, namely the extraction of sequence feature information of the target polypeptide sequence and the target MHCII sequence, includes:

[0056] Step S1011: The target polypeptide sequence and the target MHCII sequence are transformed into terms using an amino acid digital dictionary.

[0057] In this embodiment, the digital dictionary can be, for example, {'R': 1, 'N': 2, 'D': 3, 'C': 4, 'Q': 5, 'E': 6, 'G': 7, 'H': 8, 'I': 9, 'L': 10, 'K': 11, 'M': 12, 'F': 13, 'P': 14, 'S': 15, 'T': 16, 'W': 17, 'Y': 18, 'V': 19, 'X': 20, 'A': 21}, where 'X' represents an unidentified amino acid in the sequence. The target polypeptide sequence and the target MHCII sequence, converted using the digital dictionary, constitute the lexicalized sequence of the target polypeptide sequence and the lexicalized sequence of the target MHCII sequence.

[0058] Step S1012: Initial encoding is performed on the lexicalized target polypeptide sequence and the target MHCII sequence to obtain initial polypeptide sequence encoding and initial MHCII sequence encoding; the initial polypeptide sequence encoding and the initial MHCII sequence encoding include lexical encoding and position encoding.

[0059] In this embodiment, before initial encoding begins, the length of the lexicalized target polypeptide sequence can be aligned to a predefined length (e.g., 20). If the length of the lexicalized target polypeptide sequence is less than the predefined length, it is padded with 0; if the length of the lexicalized target polypeptide sequence is greater than the predefined length, it is truncated to the predefined length. In this embodiment, the MHC II molecule can be a pre-processed pseudo-sequence with a specific length (e.g., 34), thus eliminating the need for padding or truncation. Lexical encoding and positional encoding of the amino acids in the target polypeptide sequence (lexicalized sequence) and the target MHCII sequence (lexicalized sequence) aligned to the predefined length yield the initial polypeptide sequence code and the initial MHCII sequence code of the predefined length (e.g., 256). The initial polypeptide sequence code and the initial MHCII sequence code are floating-point data.

[0060] Specifically, the initial polypeptide sequence encoding and the initial MHCII sequence encoding include the sum of word encoding and position encoding.

[0061] Step S1013: The initial polypeptide sequence encoding and the initial MHCII sequence encoding are re-encoded using a Convolutional Neural Network (CNN) model (e.g., ...). Figure 2 As shown in the figure, the final polypeptide sequence encoding and MHCII sequence encoding are obtained.

[0062] Specifically, such as Figure 3 As shown, this CNN model can include one-dimensional convolutions with different kernels. After performing convolution calculations on the initial peptide sequence code (or initial MHCII sequence code) using one-dimensional convolutions with different kernels in the CNN model, the convolution results are concatenated along the feature dimension to obtain the final peptide sequence code (or MHCII sequence code). For example, as... Figure 3 As shown, a CNN model can include three parallel one-dimensional convolutions with kernels of 1, 3, and 5 respectively. The output of the CNN model is obtained by concatenating the outputs of the three one-dimensional convolutions.

[0063] Step S1014: Perform feature extraction on the final polypeptide sequence encoding and the MHCII sequence encoding respectively to obtain the sequence feature information of the target polypeptide sequence and the sequence feature information of the target MHCII sequence.

[0064] In some optional embodiments, step S1014, namely, extracting features from the final polypeptide sequence encoding and the MHCII sequence encoding respectively, includes:

[0065] like Figure 2As shown, the final polypeptide sequence encoding and the MHCII sequence encoding are respectively input into two BERT models for feature extraction;

[0066] The BERT model includes stacked multi-layer encoders (e.g., a 4-layer encoder), such as... Figure 4 As shown, the encoder sequentially includes a first multi-head attention layer, a first normalization layer, a first residual layer, a first fully connected layer, a second normalization layer, and a second residual layer.

[0067] Step S102: Based on the target polypeptide sequence and the target MHCII sequence, obtain the polypeptide-related attention score (also known as the polypeptide atom importance score) and the MHCII-related attention score (also known as the MHCII atom importance score); the polypeptide-related attention score is the attention score of the atoms in the target polypeptide sequence relative to the target MHCII sequence, and the MHCII-related attention score is the attention score of the atoms in the target MHCII sequence relative to the target polypeptide sequence.

[0068] Step S103: Based on the peptide-related attention score and the sequence feature information of the target peptide sequence, obtain the peptide polymerization feature.

[0069] Specifically, such as Figure 2 As shown, the peptide-related attention score can be multiplied by the sequence feature information of the target peptide sequence extracted by the BERT model to obtain the peptide aggregation feature.

[0070] Step S104: Based on the MHCII related attention score and the sequence feature information of the target MHCII sequence, obtain the MHCII aggregation feature.

[0071] Specifically, such as Figure 2 As shown, the MHCII-related attention score and the sequence feature information of the target MHCII sequence extracted by the BERT model can be multiplied to obtain the MHCII aggregated feature.

[0072] Step S105: Input the polypeptide polymerization features and the MHCII polymerization features into the affinity prediction neural network model to predict the affinity between the target polypeptide sequence and the target MHCII sequence. Specifically, as shown... Figure 2 As shown, the polypeptide polymerization features and the MHCII polymerization features can be concatenated and input into the affinity prediction neural network model for prediction.

[0073] In some optional embodiments, the affinity prediction neural network model is a feedforward neural network model.

[0074] For example, a feedforward neural network model may include: a second fully connected layer, a third normalized layer, an activation function, a first Dropout layer, and a third fully connected layer.

[0075] Specifically, the feedforward neural network model can be a fully connected neural network model.

[0076] Furthermore, during the training of the entire model (including the CNN model for encoding, the BERT model for feature extraction, and the affinity prediction neural network model), the AdamW optimizer can be used, along with hierarchical learning rates. The learning rate of the BERT model can be up to 10. -5 -10 -4 Between these points, the learning rate of the affinity prediction neural network model can be 10. -4 -10 -3 Furthermore, a larger learning rate can be used at the beginning of model training to accelerate the model's convergence speed, while a smaller learning rate can be used later to fine-tune the model, thus avoiding oscillations around the optimal point.

[0077] The peptide-MHCII binding affinity prediction method provided in this embodiment utilizes a neural network model to predict the affinity between peptide and MHCII sequences, offering high efficiency and low cost. Furthermore, to ensure prediction accuracy, this embodiment first obtains peptide-related attention scores and MHCII-related attention scores based on the target peptide and target MHCII sequences. Then, based on the peptide-related attention scores and sequence feature information of the target peptide sequence, peptide aggregation features are obtained. Similarly, based on the MHCII-related attention scores and sequence feature information of the target MHCII sequence, MHCII aggregation features are obtained. Finally, the peptide aggregation features and MHCII aggregation features are input into the affinity prediction neural network model for affinity prediction. In other words, this embodiment provides a novel feature extraction method. After extracting the sequence features of the peptide and MHCII, feature aggregation is performed based on the extracted sequence features, achieving sufficient learning of the sequence features of the peptide and MHCII. The affinity prediction neural network model predicts the binding affinity between the peptide and MHCII based on the aggregation features, resulting in higher accuracy.

[0078] The method for predicting the binding affinity of peptides to MHCII provided in this embodiment can be applied to immune-related biomedical research, such as for vaccine design, immunotherapy and other related research.

[0079] In some optional embodiments, step S102, namely obtaining the peptide-related attention score based on the target peptide sequence and the target MHCII sequence, includes:

[0080] Step S1021: Obtain the polypeptide sequence code corresponding to the target polypeptide sequence and the MHCII sequence code corresponding to the target MHCII sequence.

[0081] Specifically, the peptide sequence code and MHCII sequence code here can be the sum of term codes and position codes. The process of obtaining the peptide sequence code and MHCII sequence code may include: first, using an amino acid numerical dictionary to perform term conversion on the target peptide sequence and the target MHCII sequence respectively; then, encoding the term-converted target peptide sequence and the target MHCII sequence respectively to obtain the peptide sequence code and the MHCII sequence code.

[0082] Step S1022: Multiply the peptide sequence encoding and the MHCII sequence encoding, and obtain the peptide-MHCII attention score based on the obtained product and the MHCII sequence encoding.

[0083] Specifically, in this embodiment, the peptide-MHCII attention score can be obtained using a peptide-MHCII attention module. That is, the peptide-MHCII attention module is used to multiply the peptide sequence encoding and the MHCII sequence encoding, and obtain the peptide-MHCII attention score based on the obtained product and the MHCII sequence encoding.

[0084] In addition, such as Figure 5 As shown, after multiplying the peptide sequence encoding and the MHCII sequence encoding to obtain the corresponding product, the product can be processed by Softmax. Then, the result of the Softmax processing is multiplied by the MHCII sequence encoding to obtain the peptide-MHCII attention score matrix.

[0085] Step S1023: Based on the peptide-MHCII attention score, obtain the peptide-related attention score.

[0086] In this embodiment, the peptide-MHCII attention scores can be aggregated into peptide-related attention scores through averaging. Specifically, after obtaining the peptide-MHCII attention score matrix, the attention score (i.e., peptide-related attention score, or peptide sequence atom importance score) of each atom in the target peptide sequence relative to the target MHCII sequence can be obtained by averaging the features of each atom in the matrix. That is, after obtaining the peptide-MHCII attention score matrix, the columns of the peptide-MHCII attention score matrix can be averaged to obtain a vector with the same length as the peptide sequence.

[0087] In some optional embodiments, step S102, namely obtaining the MHCII-related attention score based on the target peptide sequence and the target MHCII sequence, includes:

[0088] Step S102a: Obtain the polypeptide sequence code corresponding to the target polypeptide sequence and the MHCII sequence code corresponding to the target MHCII sequence.

[0089] Step S102b: Multiply the MHCII sequence encoding and the peptide sequence encoding, and obtain the MHCII-peptide attention score based on the obtained product and the peptide sequence encoding.

[0090] Specifically, in this embodiment, the MHCII-peptide attention module can be used to obtain the MHCII-peptide attention score. That is, the MHCII-peptide attention module is used to multiply the MHCII sequence encoding and the peptide sequence encoding, and obtain the MHCII-peptide attention score based on the obtained product and the peptide sequence encoding.

[0091] In addition, after multiplying the MHCII sequence encoding and the peptide sequence encoding to obtain the corresponding product, the product can be subjected to Softmax processing. Then, the result of Softmax processing can be multiplied with the peptide sequence encoding to obtain the MHCII-peptide attention score matrix.

[0092] Step S102c: Based on the MHCII-peptide attention score, obtain the MHCII-related attention score.

[0093] In this embodiment, the MHCII-peptide attention scores can be aggregated into MHCII-related attention scores through mean calculation. Specifically, after the matrix of MHCII-peptide attention scores, the attention score of each atom in the target MHCII sequence relative to the target peptide sequence (i.e., the MHCII-related attention score, or MHCII sequence atom importance score) can be obtained by averaging the features of each atom in the matrix.

[0094] This embodiment also provides a peptide-MHCII binding affinity prediction model, which includes: a feature extraction model, a relevant attention score acquisition module, a feature aggregation module, and an affinity prediction neural network model.

[0095] The feature extraction module is used to extract sequence feature information of the target polypeptide sequence and the target MHCII sequence;

[0096] The relevant attention score acquisition module is used to acquire a peptide-related attention score and an MHCII-related attention score based on the target peptide sequence and the target MHCII sequence; the peptide-related attention score is the attention score of an atom in the target peptide sequence relative to the target MHCII sequence, and the MHCII-related attention score is the attention score of an atom in the target MHCII sequence relative to the target peptide sequence;

[0097] The feature aggregation module is used to obtain peptide aggregation features based on the peptide-related attention score and the sequence feature information of the target peptide sequence; and to obtain MHCII aggregation features based on the MHCII-related attention score and the sequence feature information of the target MHCII sequence.

[0098] An affinity prediction neural network model is used to predict the affinity of the target polypeptide sequence and the target MHCII sequence based on the polypeptide polymerization characteristics and the MHCII polymerization characteristics.

[0099] As described in the above example of the method for predicting the binding affinity of peptides to MHCII, the feature extraction module includes a Convolutional Neural Network (CNN) model and a BERT model.

[0100] The peptide-MHCII binding affinity prediction model provided in this embodiment selects an appropriate encoding method and a suitable deep learning neural network to fully learn sequence features, thereby further improving the efficiency and accuracy of peptide-MHCII binding affinity prediction.

[0101] The accuracy of the peptide-MHCII binding affinity prediction model provided in this embodiment has been verified. The specific process is as follows:

[0102] Step 1: Obtain peptide and MHCII binding data from the public database IEDB2016. This database contains 134,281 MHCII-peptide binding affinity data, including 36 HLA-DR molecules, 27 HLA-DQ molecules, 9 HLA-DP molecules, and 8 H-2 molecules among the MHCII molecules. The affinity values ​​from the database... It can be done through the formula Convert the IC50 value to a range of 0-1. If the converted value is greater than or equal to 0.46, it indicates that the polypeptide binds to MHCII; otherwise, it does not.

[0103] Step two involves constructing a sample set based on the peptide and MHCII binding data obtained from the public database IEDB2016. The judgment result from step one (binding or non-binding) is used as the corresponding label, or the affinity value converted to the 0-1 range is directly used as the label. The constructed sample set is then used to train and validate the model. During training and validation, the peptide and MHCII sequences in the samples are first converted into numerical data using an amino acid dictionary. This numerical data serves as the input data for the peptide-MHCII binding affinity prediction model. The length of the peptide lexicalized sequence used as input data is 20, and the length of the MHCII lexicalized sequence used as input data is 34. Five-fold cross-validation is used to train and validate the peptide-MHCII binding affinity prediction model. The results of the five-fold cross-validation are shown in Table 1.

[0104] In addition, the same sample set was used to train and validate the peptide binding affinity prediction model based on DeepSeqPanII and MHCII. The five-fold cross-validation method was also used, and the validation results are shown in Table 1.

[0105] Table 1. Model validation results

[0106]

[0107] As can be seen from the data in Table 1, compared with the peptide-MHCII binding affinity prediction model constructed based on DeepSeqPanII, the peptide-MHCII binding affinity prediction model provided in this embodiment has improved the Pearson Correlation Coefficient (PCC) and the Area Under the ROC Curve (AUC).

[0108] The DeepSeqPanII mentioned above is a deep learning-based model, a recurrent neural network model with an attention mechanism, used to predict the binding of peptides to HLA class II molecules.

[0109] This embodiment also provides a device for predicting the binding affinity of peptides to MHCII. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0110] This embodiment provides a device for predicting the binding affinity of a peptide to MHCII, such as... Figure 6 As shown, it includes:

[0111] The feature extraction module 601 is used to extract sequence feature information of the target polypeptide sequence and the target MHCII sequence;

[0112] The relevant attention score acquisition module 602 is used to acquire a peptide-related attention score and an MHCII-related attention score based on the target peptide sequence and the target MHCII sequence; the peptide-related attention score is the attention score of an atom in the target peptide sequence relative to the target MHCII sequence, and the MHCII-related attention score is the attention score of an atom in the target MHCII sequence relative to the target peptide sequence;

[0113] The polypeptide polymerization feature acquisition module 603 is used to obtain polypeptide polymerization features based on the polypeptide-related attention score and the sequence feature information of the target polypeptide sequence;

[0114] The MHCII aggregation feature acquisition module 604 is used to obtain MHCII aggregation features based on the MHCII related attention score and the sequence feature information of the target MHCII sequence;

[0115] The prediction module 605 is used to input the polypeptide polymerization features and the MHCII polymerization features into the affinity prediction neural network model to predict the affinity of the target polypeptide sequence and the target MHCII sequence.

[0116] In some optional implementations, the relevant attention score acquisition module 602 is specifically used for:

[0117] Obtain the polypeptide sequence code corresponding to the target polypeptide sequence and the MHCII sequence code corresponding to the target MHCII sequence;

[0118] Multiply the peptide sequence encoding and the MHCII sequence encoding, and obtain the peptide-MHCII attention score based on the resulting product and the MHCII sequence encoding; obtain the peptide-related attention score based on the peptide-MHCII attention score; and / or,

[0119] The MHCII sequence encoding and the peptide sequence encoding are multiplied together, and the MHCII-peptide attention score is obtained based on the product and the peptide sequence encoding; the MHCII-related attention score is obtained based on the MHCII-peptide attention score.

[0120] In some optional implementations, the feature extraction module 601 is specifically used for:

[0121] The target polypeptide sequence and the target MHCII sequence were respectively transformed using an amino acid numerical dictionary;

[0122] The target polypeptide sequence and the target MHCII sequence after word-based encoding are initially encoded to obtain initial polypeptide sequence encoding and initial MHCII sequence encoding; the initial polypeptide sequence encoding and the initial MHCII sequence encoding include word encoding and position encoding.

[0123] The initial peptide sequence code and the initial MHCII sequence code are re-encoded using a convolutional neural network model to obtain the final peptide sequence code and MHCII sequence code.

[0124] Feature extraction is performed on the final encoded polypeptide sequence and the encoded MHCII sequence respectively to obtain the sequence feature information of the target polypeptide sequence and the sequence feature information of the target MHCII sequence.

[0125] In some optional implementations, the feature extraction module 601 is specifically used for:

[0126] The final polypeptide sequence encoding and the MHCII sequence encoding are respectively input into two BERT models for feature extraction;

[0127] The BERT model includes stacked multi-layer encoders, which are sequentially comprised of...

[0128] In some optional implementations, the learning rate of the affinity prediction neural network model is 10 during model training. -4 -10 -3 The learning rate of the BERT model is 10. -5 -10 -4 .

[0129] In some alternative implementations, the affinity prediction neural network model is a feedforward neural network model.

[0130] The peptide-MHCII binding affinity prediction device provided in this embodiment of the invention can execute the peptide-MHCII binding affinity prediction method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0131] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0132] The following is a detailed reference. Figure 7 This diagram illustrates a suitable structural schematic for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 701, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 702 or a program loaded from memory 708 into random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the electronic device. The processor 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0133] Typically, the following devices can be connected to I / O interface 705: input devices 706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication device 709 allows electronic devices to exchange data via wireless or wired communication with other devices. Although Figure 7 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0134] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 709, or installed from a memory 708, or installed from a ROM 702. When the computer program is executed by the processor 701, it performs the functions defined in the peptide-MHCII binding affinity prediction method of the embodiments of the present invention.

[0135] Figure 7 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0136] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the method for predicting the binding affinity of peptides to MHCII shown in the above embodiments is implemented.

[0137] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0138] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for predicting the binding affinity of a polypeptide to MHC II, characterized by, The method comprises: extracting sequence feature information of a target polypeptide sequence and a target MHCII sequence; based on the target polypeptide sequence and the target MHCII sequence, obtaining polypeptide-related attention scores and MHCII-related attention scores; the polypeptide-related attention score is the attention score of an atom in the target polypeptide sequence relative to the target MHCII sequence, and the MHCII-related attention score is the attention score of an atom in the target MHCII sequence relative to the target polypeptide sequence; based on the polypeptide-related attention score and the sequence feature information of the target polypeptide sequence, obtaining a polypeptide aggregation feature; based on the MHCII-related attention score and the sequence feature information of the target MHCII sequence, obtaining an MHCII aggregation feature; inputting the polypeptide aggregation feature and the MHCII aggregation feature into an affinity prediction neural network model to predict the affinity of the target polypeptide sequence and the target MHCII sequence; the method comprises: obtaining polypeptide sequence encoding corresponding to the target polypeptide sequence and MHCII sequence encoding corresponding to the target MHCII sequence; multiplying the polypeptide sequence encoding and the MHCII sequence encoding, and based on the obtained product and the MHCII sequence encoding, obtaining polypeptide-MHCII attention scores; based on the polypeptide-MHCII attention scores, obtaining the polypeptide-related attention scores; and / or, multiplying the MHCII sequence encoding and the polypeptide sequence encoding, and based on the obtained product and the polypeptide sequence encoding, obtaining MHCII-polypeptide attention scores; based on the MHCII-polypeptide attention scores, obtaining the MHCII-related attention scores.

2. The method of claim 1, wherein, the method comprises: using a numerical dictionary of amino acids to perform tokenization on the target polypeptide sequence and the target MHCII sequence, respectively; initially encoding the tokenized target polypeptide sequence and the target MHCII sequence, respectively, to obtain initial polypeptide sequence encoding and initial MHCII sequence encoding; the initial polypeptide sequence encoding and the initial MHCII sequence encoding comprise token encoding and position encoding; using a convolutional neural network model to re-encode the initial polypeptide sequence encoding and the initial MHCII sequence encoding, respectively, to obtain final polypeptide sequence encoding and MHCII sequence encoding; extracting features from the final polypeptide sequence encoding and the MHCII sequence encoding, respectively, to obtain sequence feature information of the target polypeptide sequence and sequence feature information of the target MHCII sequence.

3. The method of claim 2, wherein, the method comprises: The final polypeptide sequence encoding and the MHCII sequence encoding are respectively input into two BERT models for feature extraction. The BERT model comprises a stacked multi-layer encoder, and the encoder comprises, in sequence, a first multi-head attention layer, a first normalization layer, a first residual layer, a first full connection layer, a second normalization layer, and a second residual layer.

4. The method of claim 3, wherein, During the model training process, the learning rate of the affinity prediction neural network model is 10 -4 -10 -3 , and the learning rate of the BERT model is 10 -5 -10 -4 .

5. The method of claim 1, wherein, The affinity prediction neural network model is a feedforward neural network model.

6. A polypeptide binding affinity to MHC II prediction apparatus characterized by, The device comprises: a feature extraction module configured to extract sequence feature information of a target polypeptide sequence and a target MHCII sequence; a relevant attention score acquisition module configured to acquire a polypeptide relevant attention score and an MHCII relevant attention score based on the target polypeptide sequence and the target MHCII sequence, the polypeptide relevant attention score being an attention score of an atom in the target polypeptide sequence relative to the target MHCII sequence, and the MHCII relevant attention score being an attention score of an atom in the target MHCII sequence relative to the target polypeptide sequence; a polypeptide aggregation feature acquisition module configured to acquire a polypeptide aggregation feature based on the polypeptide relevant attention score and the sequence feature information of the target polypeptide sequence; an MHCII aggregation feature acquisition module configured to acquire an MHCII aggregation feature based on the MHCII relevant attention score and the sequence feature information of the target MHCII sequence; a prediction module configured to input the polypeptide aggregation feature and the MHCII aggregation feature into an affinity prediction neural network model to predict a binding affinity of the target polypeptide sequence and the target MHCII sequence. The relevant attention score acquisition module is specifically configured to: acquire a polypeptide sequence encoding corresponding to the target polypeptide sequence and an MHCII sequence encoding corresponding to the target MHCII sequence; multiply the polypeptide sequence encoding and the MHCII sequence encoding, and acquire a polypeptide-MHCII attention score based on a product obtained by the multiplication and the MHCII sequence encoding; acquire the polypeptide relevant attention score based on the polypeptide-MHCII attention score; and / or multiply the MHCII sequence encoding and the polypeptide sequence encoding, and acquire an MHCII-polypeptide attention score based on a product obtained by the multiplication and the polypeptide sequence encoding; acquire the MHCII relevant attention score based on the MHCII-polypeptide attention score.

7. An electronic device, comprising: comprise: a memory and a processor, which are communicatively connected to each other, and the memory stores computer instructions, and the processor executes the computer instructions to perform the polypeptide and MHCII binding affinity prediction method in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions, and the computer instructions are used to make a computer execute the polypeptide and MHCII binding affinity prediction method in any one of claims 1 to 5.

Images