Interaction protein generation method and device, electronic equipment and storage medium

Abstract

The invention relates to an interaction protein generation method and device, electronic equipment and a storage medium, a preset reference protein sequence is encoded to obtain a corresponding reference protein code, and the reference protein code comprises a plurality of tag vectors with a connection sequence and is used for representing the amino acid type and the arrangement sequence in the reference protein sequence. The reference protein codes are input into a protein generation model obtained through pre-training, a preset number of candidate protein codes are generated based on the reference protein codes, and the label vectors are sequentially generated in an iteration mode. And screening a target protein code from the candidate protein codes according to the reference protein code. And further decoding the target protein code to obtain a target protein sequence interacting with the reference protein sequence. According to the invention, interacting protein sequences are generated by referring to a known protein sequence structure, so that the controllability of a protein generation process and the success rate of a generation result are improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a method, device, electronic device and storage medium for generating interacting proteins. Background Art

[0002] Language models based on deep learning algorithms, such as ChatGPT, have acquired new capabilities that previous algorithms did not have by training on large-scale data sets (such as social media data, code data, paper data, etc.). These capabilities include natural language understanding, computing power, reasoning ability, etc. Deep learning algorithms enable these models to learn from data autonomously and continuously improve their performance. Among them, ChatGPT is one of the latest natural language models. It uses the Transformer model in deep learning algorithms and is trained through large-scale data sets and self-supervised learning. This allows ChatGPT to generate highly realistic language texts and perform well on a wide range of tasks, such as question answering, text classification, and machine translation. ChatGPT and other natural language models can help achieve more intelligent and automated decision-making and interaction, making people's lives more convenient and efficient. Protein language models are a case of migrating natural language model technology to the field of biology (such as ProGen, ProtGPT2, etc.). By learning massive amounts of protein sequence data, protein language models can generate new protein sequences based on autoregression. However, current protein language models have problems such as poor controllability of protein generation and low success rate of generation results. Summary of the invention

[0003] In view of this, the present disclosure proposes a method, device, electronic device and storage medium for generating interacting proteins, aiming to increase the controllability of the protein language model in the protein generation process and the success rate of the results.

[0004] According to a first aspect of the present disclosure, a method for generating an interacting protein is provided, the method comprising:

[0005] Encoding a preset reference protein sequence to obtain a corresponding reference protein code, wherein the reference protein code includes a plurality of label vectors with a connection order, which are used to characterize the type and arrangement order of amino acids in the reference protein sequence;

[0006] Inputting the reference protein code into a pre-trained protein generation model, the protein generation model generates a preset number of candidate protein codes based on the reference protein code, and the multiple label vectors in each candidate protein code are sequentially generated in an iterative manner;

[0007] Screening a target protein code from the candidate protein codes according to the reference protein code;

[0008] The target protein code is decoded to obtain a target protein sequence that interacts with the reference protein sequence.

[0009] In a possible implementation, encoding a preset reference protein sequence to obtain a corresponding reference protein code includes:

[0010] Generate a character identifier sequence corresponding to the reference protein sequence according to a preset correspondence between amino acid types and character identifiers, and the amino acid types and arrangement order included in the reference protein sequence, wherein the character identifier sequence includes a start tag set before the first character identifier and an end tag set after the last character identifier;

[0011] Performing word segmentation processing on the character identifiers in the character identifier sequence to obtain a tag sequence consisting of a plurality of tags including the start tag and the end tag, each of the tags including at least one character identifier;

[0012] Encode each of the labels in the label sequence to obtain a corresponding label vector;

[0013] The tag vectors are arranged in order according to the position of each tag in the tag sequence to obtain a reference protein code corresponding to the reference protein sequence.

[0014] In a possible implementation, encoding each of the labels in the label sequence to obtain a corresponding label vector includes:

[0015] Performing type encoding, index encoding and chain encoding on each of the tags in the tag sequence to obtain a first vector, a second vector and a third vector;

[0016] The first vector, the second vector and the third vector corresponding to each of the labels are added to obtain a corresponding label vector.

[0017] In a possible implementation, performing type encoding, index encoding, and chain encoding on each of the tags in the tag sequence to obtain a first vector, a second vector, and a third vector includes:

[0018] Directly mapping each of the labels in the label sequence to complete type encoding, and obtaining a corresponding first vector;

[0019] Mapping the position number of each tag in the tag sequence to complete index encoding to obtain a corresponding second vector;

[0020] The reference protein sequence number to which each tag in the tag sequence belongs is mapped to complete the chain encoding, and a corresponding third vector is obtained.

[0021] In a possible implementation, the inputting the reference protein code into a pre-trained protein generation model, and the protein generation model generating a preset number of candidate protein codes based on the reference protein code, includes:

[0022] Inputting the reference protein code into a pre-trained protein generation model, and having the protein generation model randomly sample at least one label vector based on the reference protein code, wherein each label vector has a corresponding first probability value in a first probability distribution corresponding to the reference protein code;

[0023] Iteratively using each generated label vector as input to the protein generation model, and randomly sampling at least one label vector through the reference protein model;

[0024] In response to satisfying a preset stop condition, the iteration process is stopped and a corresponding candidate protein code is determined according to a plurality of label vectors connected in the iteration sequence to obtain a plurality of candidate protein codes.

[0025] In a possible implementation, screening the target protein code from the candidate protein codes according to the reference protein code includes:

[0026] Performing probability value filtering on each of the candidate protein codes based on a preset filtering rule;

[0027] The target protein code is obtained by screening the candidate protein codes after probability value filtering according to the reference protein code.

[0028] In a possible implementation, the probability value filtering of each candidate protein code based on a preset filtering rule includes:

[0029] Deleting candidate protein codes whose first probability value mean corresponding to each of the label vectors is less than a first preset value;

[0030] Determining a second probability value of each of the label vectors in each of the candidate protein codes in a second probability distribution without a corresponding reference protein code;

[0031] Deleting candidate protein codes for which the difference between the mean of the first probability value and the mean of the second probability value corresponding to each of the label vectors is less than a second preset value;

[0032] Delete candidate protein codes in which the proportion of single amino acids in the amino acids corresponding to all the label vectors is greater than a third preset value.

[0033] In a possible implementation, the step of screening the target protein code from the candidate protein codes filtered by the probability value according to the reference protein code includes:

[0034] Inputting the reference protein code and each of the screened candidate protein codes into the trained protein prediction model to obtain corresponding confidences;

[0035] The candidate protein code with the highest confidence is determined as the target protein code.

[0036] In a possible implementation, the training process of the protein prediction model includes:

[0037] Determining a model training set including a plurality of sample protein sequences with annotated protein sequences;

[0038] Co-encoding each of the annotated protein sequences and the sample protein sequence to obtain a first training protein code, wherein the first training protein code includes the sample protein code and the annotated protein code;

[0039] Encoding each of the annotated protein sequences to obtain a second training protein code, wherein the second training protein code includes the annotated protein code;

[0040] The protein prediction model is trained together according to the first training protein code and the second training protein code.

[0041] In a possible implementation, the step of jointly training the protein prediction model according to the first training protein code and the second training protein code includes:

[0042] Inputting the first training protein code into the protein prediction model to determine the predicted protein code based on the sample protein code therein by the protein prediction model;

[0043] The protein prediction model is adjusted based on the difference between the label vector in the predicted protein code and the label vector in the annotated protein code in the first training protein code.

[0044] In a possible implementation, the step of jointly training the protein prediction model according to the first training protein code and the second training protein code further includes:

[0045] Inputting the encoding result of the preset start tag into the protein prediction model, so as to predict the tag multiple times in an iterative manner through the protein prediction model, and determining the predicted protein code according to the prediction order of the multiple tags;

[0046] The protein prediction model is adjusted based on the difference between the label vector in the predicted protein code and the label vector in the annotated protein code in the second training protein code.

[0047] According to a second aspect of the present disclosure, a device for generating an interacting protein is provided, the device comprising:

[0048] A sequence encoding module, used to encode a preset reference protein sequence to obtain a corresponding reference protein code, wherein the reference protein code includes a plurality of label vectors with a connection order, which are used to characterize the type and arrangement order of amino acids in the reference protein sequence;

[0049] A protein generation module, used for inputting the reference protein code into a pre-trained protein generation model, and the protein generation model generates a preset number of candidate protein codes based on the reference protein code, wherein multiple label vectors in each candidate protein code are sequentially generated in an iterative manner;

[0050] A coding screening module, used for screening a target protein coding from the candidate protein coding according to the reference protein coding;

[0051] The sequence generation module is used to decode the target protein code to obtain the target protein sequence that interacts with the reference protein sequence.

[0052] In a possible implementation, the sequence encoding module is further used to:

[0053] Generate a character identifier sequence corresponding to the reference protein sequence according to a preset correspondence between amino acid types and character identifiers, and the amino acid types and arrangement order included in the reference protein sequence, wherein the character identifier sequence includes a start tag set before the first character identifier and an end tag set after the last character identifier;

[0054] Performing word segmentation processing on the character identifiers in the character identifier sequence to obtain a tag sequence consisting of a plurality of tags including the start tag and the end tag, each of the tags including at least one character identifier;

[0055] Encode each of the labels in the label sequence to obtain a corresponding label vector;

[0056] The tag vectors are arranged in order according to the position of each tag in the tag sequence to obtain a reference protein code corresponding to the reference protein sequence.

[0057] In a possible implementation, the sequence encoding module is further used to:

[0058] Performing type encoding, index encoding and chain encoding on each of the tags in the tag sequence to obtain a first vector, a second vector and a third vector;

[0059] The first vector, the second vector and the third vector corresponding to each of the labels are added to obtain a corresponding label vector.

[0060] In a possible implementation, the sequence encoding module is further used to:

[0061] Directly mapping each of the labels in the label sequence to complete type encoding, and obtaining a corresponding first vector;

[0062] Mapping the position number of each tag in the tag sequence to complete index encoding to obtain a corresponding second vector;

[0063] The reference protein sequence number to which each tag in the tag sequence belongs is mapped to complete the chain encoding, and a corresponding third vector is obtained.

[0064] In a possible implementation, the protein production module is further used to:

[0065] Inputting the reference protein code into a pre-trained protein generation model, and having the protein generation model randomly sample at least one label vector based on the reference protein code, wherein each label vector has a corresponding first probability value in a first probability distribution corresponding to the reference protein code;

[0066] Iteratively using each generated label vector as input to the protein generation model, and randomly sampling at least one label vector through the reference protein model;

[0067] In response to satisfying a preset stop condition, the iteration process is stopped and a corresponding candidate protein code is determined according to a plurality of label vectors connected in the iteration sequence to obtain a plurality of candidate protein codes.

[0068] In a possible implementation, the encoding screening module is further used to:

[0069] Performing probability value filtering on each of the candidate protein codes based on a preset filtering rule;

[0070] The target protein code is obtained by screening the candidate protein codes after probability value filtering according to the reference protein code.

[0071] In a possible implementation, the encoding screening module is further used to:

[0072] Deleting candidate protein codes whose first probability value mean corresponding to each of the label vectors is less than a first preset value;

[0073] Determining a second probability value of each of the label vectors in each of the candidate protein codes in a second probability distribution without a corresponding reference protein code;

[0074] Deleting candidate protein codes for which the difference between the mean of the first probability value and the mean of the second probability value corresponding to each of the label vectors is less than a second preset value;

[0075] Delete candidate protein codes in which the proportion of single amino acids in the amino acids corresponding to all the label vectors is greater than a third preset value.

[0076] In a possible implementation, the encoding screening module is further used to:

[0077] Inputting the reference protein code and each of the screened candidate protein codes into the trained protein prediction model to obtain corresponding confidences;

[0078] The candidate protein code with the highest confidence is determined as the target protein code.

[0079] In a possible implementation, the training process of the protein prediction model includes:

[0080] Determining a model training set including a plurality of sample protein sequences with annotated protein sequences;

[0081] Co-encoding each of the annotated protein sequences and the sample protein sequence to obtain a first training protein code, wherein the first training protein code includes the sample protein code and the annotated protein code;

[0082] Encoding each of the annotated protein sequences to obtain a second training protein code, wherein the second training protein code includes the annotated protein code;

[0083] The protein prediction model is trained together according to the first training protein code and the second training protein code.

[0084] In a possible implementation, the step of jointly training the protein prediction model according to the first training protein code and the second training protein code includes:

[0085] Inputting the first training protein code into the protein prediction model to determine the predicted protein code based on the sample protein code therein by the protein prediction model;

[0086] The protein prediction model is adjusted based on the difference between the label vector in the predicted protein code and the label vector in the annotated protein code in the first training protein code.

[0087] In a possible implementation, the step of jointly training the protein prediction model according to the first training protein code and the second training protein code further includes:

[0088] Inputting the encoding result of the preset start tag into the protein prediction model, so as to predict the tag multiple times in an iterative manner through the protein prediction model, and determining the predicted protein code according to the prediction order of the multiple tags;

[0089] The protein prediction model is adjusted based on the difference between the label vector in the predicted protein code and the label vector in the annotated protein code in the second training protein code.

[0090] According to a third aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to implement the above method when executing the instructions stored in the memory.

[0091] According to a fourth aspect of the present disclosure, a non-volatile computer-readable storage medium is provided, on which computer program instructions are stored, wherein the computer program instructions implement the above method when executed by a processor.

[0092] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer-readable code, or a non-volatile computer-readable storage medium carrying the computer-readable code. When the computer-readable code runs in a processor of an electronic device, the processor in the electronic device executes the above method.

[0093] In an embodiment of the present disclosure, a preset reference protein sequence is encoded to obtain a corresponding reference protein code, which includes a plurality of label vectors with a connection order, which are used to characterize the amino acid type and arrangement order in the reference protein sequence. The reference protein code is input into a pre-trained protein generation model to generate a preset number of candidate protein codes based on the reference protein code, wherein the plurality of label vectors included therein are generated sequentially in an iterative manner. The target protein code is screened from the candidate protein codes according to the reference protein code. The target protein code is further decoded to obtain a target protein sequence that interacts with the reference protein sequence. The present disclosure generates interacting protein sequences with reference to known protein sequence structures, thereby improving the controllability of the protein generation process and the success rate of the generation results.

[0094] Further features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate exemplary embodiments, features, and aspects of the disclosure and, together with the description, serve to explain the principles of the disclosure.

[0096] Figure 1 A flow chart showing a method for generating an interacting protein according to an embodiment of the present disclosure;

[0097] Figure 2 A schematic diagram showing a protein production result according to an embodiment of the present disclosure;

[0098] Figure 3 A schematic diagram showing a device for generating an interacting protein according to an embodiment of the present disclosure;

[0099] Figure 4 A schematic diagram showing an electronic device according to an embodiment of the present disclosure;

[0100] Figure 5 A schematic diagram showing another electronic device according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0101] Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0102] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0103] In addition, in order to better illustrate the present disclosure, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present disclosure can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present disclosure.

[0104] The method for generating the interactive protein of the embodiment of the present disclosure can be performed by an electronic device such as a terminal device or a server. Among them, the terminal device can be any fixed or mobile terminal such as a user equipment (User Equipment, UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (Personal Digital Assistant, PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. The server can be a separate server or a server cluster composed of multiple servers. Any electronic device can implement the method for generating the interactive protein of the embodiment of the present disclosure by calling the computer-readable instructions stored in the memory by the processor.

[0105] Figure 1 A flow chart of a method for generating an interacting protein according to an embodiment of the present disclosure is shown. Figure 1 As shown, the method for generating interacting proteins according to the embodiment of the present disclosure may include the following steps S10-S40.

[0106] Step S10: Encode the preset reference protein sequence to obtain the corresponding reference protein code.

[0107] In one possible implementation, the electronic device of the disclosed embodiment can predetermine a reference protein sequence that needs to generate an interactive target protein sequence, and use the reference protein sequence as a reference to generate the target protein sequence, wherein the reference protein sequence includes a plurality of sequentially arranged amino acids. Optionally, the electronic device can represent the reference protein sequence in a preset representation, for example, a protein sequence can be represented by a character identifier sequence having a start tag and an end tag, wherein the start tag and the end tag are represented by preset character identifiers and symbols, and each character identifier in the character identifier sequence except the start tag and the end tag represents an amino acid, respectively. The character identifiers may include identifiers such as letters, numbers, and symbols, and the corresponding relationship with the amino acid type can be preset. Exemplarily, the electronic device can represent the reference protein sequence through a character identifier sequence. <bos>MAFPSTWAQILTWYV <eos>Characterizes the reference protein sequence, where the start tag is <bos>, the end tag is <eos>The reference protein sequence includes 20 amino acids arranged in sequence, and each character represents a class of amino acids.

[0108] Optionally, before encoding the reference protein sequence, the electronic device first characterizes the reference protein sequence through a character identifier sequence, and then encodes the character identifier sequence to obtain the reference protein code. That is, the process of determining the reference protein code may include generating a character identifier sequence corresponding to the reference protein sequence according to the correspondence between the preset amino acid type and the character identifier, as well as the amino acid type and arrangement order included in the reference protein sequence, and the character identifier sequence includes a start tag set before the first character identifier and an end tag set after the last character identifier. Then, the character identifiers in the character identifier sequence are segmented to obtain a label sequence consisting of multiple labels including a start tag and an end tag, and each label includes at least one character identifier. Each label in the label sequence is encoded separately to obtain a corresponding label vector. The label vectors are arranged in order according to the position of each label in the label sequence to obtain the reference protein code corresponding to the reference protein sequence.

[0109] The disclosed embodiment performs word segmentation on the character identifier sequence to avoid memory overflow caused by too long an input sequence, and converts multiple adjacent character identifiers into one label through word segmentation, thereby obtaining a label sequence with a greatly shortened sequence length. For example, when the original character identifier sequence is <bos>MAFPSTWAQILTWYV <eos>Characterizes the reference protein sequence, where the start tag is <bos>, the end tag is <eos>In the case of , you can set the word segmentation method to start with the first character identifier after the start tag, and use every three character identifiers as a tag until the end tag. The result of word segmentation is <bos><h2 style=";text-align:left;direction:ltr">[MAF][PST][WAQ][ILTW][YV]<h2 style=";text-align:left;direction:ltr"> <eos>, thereby converting the character identification sequence of length 17 into a label sequence of length 7.

[0110] Furthermore, after the electronic device obtains the label sequence by word segmentation, each label in the label sequence is encoded respectively to obtain the corresponding label vector and determine the label vectors arranged in sequence as the final encoding result, i.e., the reference protein encoding. Among them, each label can be encoded by performing at least one encoding to obtain the corresponding vector, and then the vectors corresponding to each encoding are added to obtain the label vector. Exemplarily, the disclosed embodiment can perform three types of encoding, namely, type encoding, index encoding, and chain encoding, on each label. That is, the process of encoding the label sequence can be performing type encoding, index encoding, and chain encoding on each label in the label sequence respectively to obtain a first vector, a second vector, and a third vector, and then the first vector, the second vector, and the third vector corresponding to each label are added to obtain the corresponding label vector.

[0111] Optionally, type coding, index coding and chain coding encode the tags based on different dimensions respectively. Among them, type coding is directly encoded based on the content of each tag, position coding is encoded based on the position of each tag in the sequence, and chain coding is encoded based on the reference protein sequence number to which each tag belongs. That is, the process of the electronic device determining the first vector, the second vector and the third vector of each tag in the tag sequence can include: directly mapping each tag in the tag sequence to complete the type coding to obtain the corresponding first vector. Mapping the position number of each tag in the tag sequence to complete the index coding to obtain the corresponding second vector. Mapping the reference protein sequence number to which each tag in the tag sequence belongs to complete the chain coding to obtain the corresponding third vector. The mapping process can be implemented by a preset coding layer. The electronic device can directly input the tag sequence into the preset coding layer to map each tag therein into a first vector. At the same time, the position sequence is determined according to the position of each tag in the tag sequence in the tag sequence, and the position sequence is input into the preset coding layer to map the second vector corresponding to each tag, and the position of each tag in the position sequence is represented by a number between 0-N-1, and N is the length of the tag sequence. Furthermore, the reference protein sequence may be one or more, and when there are multiple reference protein sequences, the corresponding numbers may be determined in advance. The electronic device may determine the numbering sequence according to the number of the reference protein sequence to which each tag in the tag sequence belongs, and input the numbering sequence into a preset coding layer to map and obtain a third vector corresponding to each tag.

[0112] Optionally, the dimensions of the first vector, the second vector and the third vector obtained by the three encoding methods of the embodiments of the present disclosure are the same. After obtaining the first vector, the second vector and the third vector corresponding to each label in the label sequence, the label vector of each label can be obtained by directly adding the first vector, the second vector and the third vector, and the label vectors are arranged according to the label positions corresponding to each label vector to obtain the reference protein encoding.

[0113] Step S20: input the reference protein code into a pre-trained protein generation model, and the protein generation model generates a preset number of candidate protein codes based on the reference protein code.

[0114] In a possible implementation, after the electronic device obtains the reference protein code by encoding the reference protein sequence, the reference protein code can be input into the trained protein generation model to generate protein based on the reference protein code through the protein generation model, and a preset number of candidate protein codes that may be related to the reference protein code are obtained. Among them, each candidate protein code also includes a plurality of label vectors arranged in sequence, and each label vector is generated in sequence in an iterative manner. That is, the process of the protein generation model generating the candidate protein code needs to go through multiple iterations, and only at least one label vector at the same position is generated in each iteration. Optionally, the protein generation model of the embodiment of the present disclosure can be a language model, and the structure is similar to GPT-2, including multiple superimposed Transformer Encoder Layers, each Encoder is composed of a multi-head self-attention mechanism and a feedforward neural network, and different semantic information is captured at different positions of the input reference protein code.

[0115] Optionally, the trained protein generation model includes a probability distribution of multiple label vectors, including a probability distribution with a corresponding reference protein sequence, and a probability distribution without a corresponding reference protein sequence. Wherein, each probability distribution also includes the probability of inputting a plurality of corresponding output results during each iteration. After the reference protein sequence is input into the pre-trained protein generation model, at least one label vector is randomly sampled by the protein generation model based on the reference protein code, and each label vector has a corresponding first probability value in the first probability distribution corresponding to the reference protein code. The label vector generated each time is used as the input of the protein generation model in an iterative manner, and at least one label vector is randomly sampled by the reference protein model. In response to meeting a preset stop condition, the iterative process is stopped and a corresponding candidate protein code is determined according to a plurality of label vectors connected in the iteration order, and a plurality of candidate protein codes are obtained. Wherein, the number of label vectors obtained by random sampling each time and the number of iterations can be determined according to a preset number.

[0116] Further, the training process of the protein generation model in the embodiment of the present disclosure may include two different training methods, including a training method with an input protein sequence and a training method without an input protein sequence. The training process can be completed in the current electronic device or other devices. Optionally, the training process of the protein generation model is to first determine a model training set including multiple sample protein sequences with annotated protein sequences. Then, each annotated protein sequence and the sample protein sequence are jointly encoded to obtain a first training protein code, and the first training protein code includes the sample protein code and the annotated protein code. At the same time, each annotated protein sequence is encoded to obtain a second training protein code, and the second training protein code includes the annotated protein code. The protein prediction model is jointly trained according to the first training protein code and the second training protein code. Among them, in the process of jointly encoding the annotated protein sequence and the sample protein sequence, the annotated protein sequence and the sample protein sequence can be first spliced ​​into a complete character identification sequence, and the input protein sequence includes a start tag set at the start position and an end tag set at the end position, and also includes a spacer tag set between the character identification of the annotated protein and the character identification of the sample protein.

[0117] Exemplarily, in the process of model training according to the first training protein code, the first training protein code can be input into the protein prediction model to determine the predicted protein code based on the sample protein code therein through the protein prediction model. Based on the difference between the label vector in the predicted protein code and the label vector in the annotated protein code in the first training protein code, the protein prediction model is adjusted. In the process of model training according to the second training protein code, the encoding result of the preset start label can be input into the protein prediction model to predict the label multiple times in an iterative manner through the protein prediction model, and determine the predicted protein code according to the prediction order of multiple labels. Based on the difference between the label vector in the predicted protein code and the label vector in the annotated protein code in the second training protein code, the protein prediction model is adjusted. The trained protein prediction model obtains a probability distribution with a corresponding input protein sequence, and a probability distribution without a corresponding input protein sequence.

[0118] Step S30: screening a target protein code from the candidate protein codes according to the reference protein code.

[0119] In a possible implementation, after the electronic device generates a preset number of candidate protein codes through a protein generation model, it can filter multiple candidate protein codes through a preset screening rule to obtain the most matching target protein code. Among them, based on the probability value corresponding to each label vector in the candidate protein code in the probability distribution, each candidate protein code can be firstly filtered by probability value based on the preset filtering rule, and then the target protein code can be obtained by filtering the candidate protein codes after probability value filtering according to the reference protein code.

[0120] Optionally, the probability value filtering process can be based on two different probability values ​​for filtering, including a first probability value determined in a first probability distribution corresponding to a reference protein code in a candidate protein code, and a second probability value determined in a second probability distribution without a corresponding reference protein code. The process of filtering candidate protein codes by an electronic device based on two probability values ​​may include: deleting candidate protein codes whose first probability value mean corresponding to each label vector is less than a first preset value. Then, in the second probability distribution without a corresponding reference protein code, determine the second probability value of each label vector in each candidate protein code. Delete candidate protein codes whose difference between the first probability value mean and the second probability value mean corresponding to each label vector is less than a second preset value. And delete candidate protein codes whose proportion of single amino acids in the amino acids corresponding to all the label vectors is greater than a third preset value.

[0121] Furthermore, the electronic device further screens the candidate protein codes after probability value filtering based on the correlation with the reference protein code to obtain the target protein code. The reference protein code and each of the screened candidate protein codes can be input into the trained protein prediction model to obtain the corresponding confidence. Finally, the candidate protein code with the highest confidence is determined as the target protein code.

[0122] Step S40, decoding the target protein code to obtain a target protein sequence that interacts with the reference protein sequence. In a possible implementation, after determining the target protein code, the electronic device can decode the target protein code to obtain a target protein sequence that interacts with the reference protein sequence. The decoding process corresponds to the encoding process and can be implemented by a pre-trained decoding layer.

[0123] Figure 2 A schematic diagram of a protein production result according to an embodiment of the present disclosure is shown. Figure 2 As shown, through the method for generating interacting proteins disclosed in the present invention, a reference protein sequence can be used as an input protein of a protein generation model for protein generation, so as to accurately generate a target protein sequence that interacts with it as a generated protein through the protein generation model.

[0124] Based on the above technical features, the disclosed embodiment trains the protein generation model in two ways, optimizes the accuracy of the model and increases the controllability of the generated protein. At the same time, the protein generation model can generate protein with a given reference protein sequence as a condition, so as to consider the reference protein sequence in the protein generation process, thereby improving the success rate of the generation result.

[0125] Figure 3 A schematic diagram of a device for generating an interacting protein according to an embodiment of the present disclosure is shown. Figure 3 As shown, the generating device of the interacting protein of the embodiment of the present disclosure may include:

[0126] A sequence encoding module 30, used to encode a preset reference protein sequence to obtain a corresponding reference protein code, wherein the reference protein code includes a plurality of label vectors with a connection order, which are used to characterize the type and arrangement order of amino acids in the reference protein sequence;

[0127] A protein generation module 31 is used to input the reference protein code into a pre-trained protein generation model, and the protein generation model generates a preset number of candidate protein codes based on the reference protein code, and multiple label vectors in each candidate protein code are generated sequentially in an iterative manner;

[0128] A coding screening module 32, for screening a target protein coding from the candidate protein coding according to the reference protein coding;

[0129] The sequence generation module 33 is used to decode the target protein code to obtain the target protein sequence that interacts with the reference protein sequence.

[0130] In a possible implementation, the sequence encoding module 30 is further configured to:

[0131] Generate a character identifier sequence corresponding to the reference protein sequence according to a preset correspondence between amino acid types and character identifiers, and the amino acid types and arrangement order included in the reference protein sequence, wherein the character identifier sequence includes a start tag set before the first character identifier and an end tag set after the last character identifier;

[0132] Performing word segmentation processing on the character identifiers in the character identifier sequence to obtain a tag sequence consisting of a plurality of tags including the start tag and the end tag, each of the tags including at least one character identifier;

[0133] Encode each of the labels in the label sequence to obtain a corresponding label vector;

[0134] The tag vectors are arranged in order according to the position of each tag in the tag sequence to obtain a reference protein code corresponding to the reference protein sequence.

[0135] In a possible implementation, the sequence encoding module 30 is further configured to:

[0136] Performing type encoding, index encoding and chain encoding on each of the tags in the tag sequence to obtain a first vector, a second vector and a third vector;

[0137] The first vector, the second vector and the third vector corresponding to each of the labels are added to obtain a corresponding label vector.

[0138] In a possible implementation, the sequence encoding module 30 is further configured to:

[0139] Directly mapping each of the labels in the label sequence to complete type encoding, and obtaining a corresponding first vector;

[0140] Mapping the position number of each tag in the tag sequence to complete index encoding to obtain a corresponding second vector;

[0141] The reference protein sequence number to which each tag in the tag sequence belongs is mapped to complete the chain encoding, and a corresponding third vector is obtained.

[0142] In a possible implementation, the protein production module 31 is further used to:

[0143] Inputting the reference protein code into a pre-trained protein generation model, and having the protein generation model randomly sample at least one label vector based on the reference protein code, wherein each label vector has a corresponding first probability value in a first probability distribution corresponding to the reference protein code;

[0144] Iteratively using each generated label vector as input to the protein generation model, and randomly sampling at least one label vector through the reference protein model;

[0145] In response to satisfying a preset stop condition, the iteration process is stopped and a corresponding candidate protein code is determined according to a plurality of label vectors connected in the iteration sequence to obtain a plurality of candidate protein codes.

[0146] In a possible implementation, the encoding screening module 32 is further used to:

[0147] Performing probability value filtering on each of the candidate protein codes based on a preset filtering rule;

[0148] The target protein code is obtained by screening the candidate protein codes after probability value filtering according to the reference protein code.

[0149] In a possible implementation, the encoding screening module 32 is further used to:

[0150] Deleting candidate protein codes whose first probability value mean corresponding to each of the label vectors is less than a first preset value;

[0151] Determining a second probability value of each of the label vectors in each of the candidate protein codes in a second probability distribution without a corresponding reference protein code;

[0152] Deleting candidate protein codes for which the difference between the mean of the first probability value and the mean of the second probability value corresponding to each of the label vectors is less than a second preset value;

[0153] Delete candidate protein codes in which the proportion of single amino acids in the amino acids corresponding to all the label vectors is greater than a third preset value.

[0154] In a possible implementation, the encoding screening module 32 is further used to:

[0155] Inputting the reference protein code and each of the screened candidate protein codes into the trained protein prediction model to obtain corresponding confidences;

[0156] The candidate protein code with the highest confidence is determined as the target protein code.

[0157] In a possible implementation, the training process of the protein prediction model includes:

[0158] Determining a model training set including a plurality of sample protein sequences with annotated protein sequences;

[0159] Co-encoding each of the annotated protein sequences and the sample protein sequence to obtain a first training protein code, wherein the first training protein code includes the sample protein code and the annotated protein code;

[0160] Encoding each of the annotated protein sequences to obtain a second training protein code, wherein the second training protein code includes the annotated protein code;

[0161] The protein prediction model is trained together according to the first training protein code and the second training protein code.

[0162] In a possible implementation, the step of jointly training the protein prediction model according to the first training protein code and the second training protein code includes:

[0163] Inputting the first training protein code into the protein prediction model to determine the predicted protein code based on the sample protein code therein by the protein prediction model;

[0164] The protein prediction model is adjusted based on the difference between the label vector in the predicted protein code and the label vector in the annotated protein code in the first training protein code.

[0165] In a possible implementation, the step of jointly training the protein prediction model according to the first training protein code and the second training protein code further includes:

[0166] Inputting the encoding result of the preset start tag into the protein prediction model, so as to predict the tag multiple times in an iterative manner through the protein prediction model, and determining the predicted protein code according to the prediction order of the multiple tags;

[0167] The protein prediction model is adjusted based on the difference between the label vector in the predicted protein code and the label vector in the annotated protein code in the second training protein code.

[0168] In some embodiments, the functions or modules included in the device provided by the embodiments of the present disclosure can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0169] The embodiment of the present disclosure also provides a computer-readable storage medium on which computer program instructions are stored, and the computer program instructions implement the above method when executed by a processor. The computer-readable storage medium can be a volatile or non-volatile computer-readable storage medium.

[0170] An embodiment of the present disclosure further proposes an electronic device, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to implement the above method when executing the instructions stored in the memory.

[0171] The embodiments of the present disclosure also provide a computer program product, including a computer-readable code, or a non-volatile computer-readable storage medium carrying the computer-readable code. When the computer-readable code runs in a processor of an electronic device, the processor in the electronic device executes the above method.

[0172] Figure 4 A schematic diagram of an electronic device 800 according to an embodiment of the present disclosure is shown. For example, the electronic device 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0173] Reference Figure 4 , the electronic device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input / output (I / O) interface 812 , a sensor component 814 , and a communication component 816 .

[0174] The processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.

[0175] The memory 804 is configured to store various types of data to support operations on the electronic device 800. Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0176] The power supply component 806 provides power to the various components of the electronic device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 800.

[0177] The multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the electronic device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0178] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), and when the electronic device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 804 or sent via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0179] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: home button, volume button, start button, and lock button.

[0180] The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the electronic device 800. For example, the sensor assembly 814 can detect the open / closed state of the electronic device 800, the relative positioning of the components, such as the display and keypad of the electronic device 800, and the sensor assembly 814 can also detect the position change of the electronic device 800 or a component of the electronic device 800, the presence or absence of contact between the user and the electronic device 800, the orientation or acceleration / deceleration of the electronic device 800, and the temperature change of the electronic device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of a nearby object without any physical contact. The sensor assembly 814 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0181] The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0182] In an exemplary embodiment, the electronic device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above methods.

[0183] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 804 including computer program instructions, which can be executed by a processor 820 of an electronic device 800 to perform the above method.

[0184] Figure 5 1 is a schematic diagram showing another electronic device 1900 according to an embodiment of the present disclosure. For example, the electronic device 1900 may be provided as a server or a terminal device. Figure 5 , the electronic device 1900 includes a processing component 1922, which further includes one or more processors, and a memory resource represented by a memory 1932 for storing instructions executable by the processing component 1922, such as an application. The application stored in the memory 1932 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 1922 is configured to execute instructions to perform the above method.

[0185] The electronic device 1900 may also include a power supply component 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an input / output (I / O) interface 1958. The electronic device 1900 may operate based on an operating system stored in the memory 1932, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™ or the like.

[0186] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 1932 including computer program instructions, which can be executed by the processing component 1922 of the electronic device 1900 to perform the above method.

[0187] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0188] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0189] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0190] The computer program instructions for performing the disclosed operation may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, executed as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of a computer-readable program instruction, and the electronic circuit may execute a computer-readable program instruction, thereby realizing various aspects of the present disclosure.

[0191] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0192] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0193] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0194] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0195] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.< / eos> < / bos> < / eos> < / bos> < / eos> < / bos> < / eos> < / bos> < / eos> < / bos>

Claims

1. A method for generating an interacting protein, characterized in that: The method comprises: Encoding a preset reference protein sequence to obtain a corresponding reference protein code, wherein the reference protein code includes a plurality of label vectors with a connection order, which are used to characterize the type and arrangement order of amino acids in the reference protein sequence; Inputting the reference protein code into a pre-trained protein generation model, the protein generation model generates a preset number of candidate protein codes based on the reference protein code, and the multiple label vectors in each candidate protein code are sequentially generated in an iterative manner; Screening a target protein code from the candidate protein codes according to the reference protein code; The target protein code is decoded to obtain a target protein sequence that interacts with the reference protein sequence.

2. The method according to claim 1, characterized in that The step of encoding a preset reference protein sequence to obtain a corresponding reference protein code comprises: Generate a character identifier sequence corresponding to the reference protein sequence according to a preset correspondence between amino acid types and character identifiers, and the amino acid types and arrangement order included in the reference protein sequence, wherein the character identifier sequence includes a start tag set before the first character identifier and an end tag set after the last character identifier; Performing word segmentation processing on the character identifiers in the character identifier sequence to obtain a tag sequence consisting of a plurality of tags including the start tag and the end tag, each of the tags including at least one character identifier; Encode each of the labels in the label sequence to obtain a corresponding label vector; The tag vectors are arranged in order according to the position of each tag in the tag sequence to obtain a reference protein code corresponding to the reference protein sequence.

3. The method according to claim 2, characterized in that The step of encoding each of the labels in the label sequence to obtain a corresponding label vector includes: Performing type encoding, index encoding and chain encoding on each of the tags in the tag sequence to obtain a first vector, a second vector and a third vector; The first vector, the second vector and the third vector corresponding to each of the labels are added to obtain a corresponding label vector.

4. The method according to claim 3, characterized in that: The step of performing type encoding, index encoding and chain encoding on each of the tags in the tag sequence to obtain a first vector, a second vector and a third vector comprises: Directly mapping each of the labels in the label sequence to complete type encoding, and obtaining a corresponding first vector; Mapping the position number of each tag in the tag sequence to complete index encoding to obtain a corresponding second vector; The reference protein sequence number to which each tag in the tag sequence belongs is mapped to complete the chain encoding, and a corresponding third vector is obtained.

5. The method according to any one of claims 1 to 4, characterized in that: The step of inputting the reference protein code into a pre-trained protein generation model, and generating a preset number of candidate protein codes based on the reference protein code by the protein generation model, comprises: Inputting the reference protein code into a pre-trained protein generation model, and having the protein generation model randomly sample at least one label vector based on the reference protein code, wherein each label vector has a corresponding first probability value in a first probability distribution corresponding to the reference protein code; Iteratively using each generated label vector as input to the protein generation model, and randomly sampling at least one label vector through the reference protein model; In response to satisfying a preset stop condition, the iteration process is stopped and a corresponding candidate protein code is determined according to a plurality of label vectors connected in the iteration sequence to obtain a plurality of candidate protein codes.

6. The method according to claim 5, characterized in that The step of screening the target protein code from the candidate protein codes according to the reference protein code comprises: Performing probability value filtering on each of the candidate protein codes based on a preset filtering rule; The target protein code is obtained by screening the candidate protein codes after probability value filtering according to the reference protein code.

7. The method according to claim 6, characterized in that The probability value filtering of each candidate protein code based on a preset filtering rule includes: Deleting candidate protein codes whose first probability value mean corresponding to each of the label vectors is less than a first preset value; Determining a second probability value of each of the label vectors in each of the candidate protein codes in a second probability distribution without a corresponding reference protein code; Deleting candidate protein codes for which the difference between the mean of the first probability value and the mean of the second probability value corresponding to each of the label vectors is less than a second preset value; Delete candidate protein codes in which the proportion of single amino acids in the amino acids corresponding to all the label vectors is greater than a third preset value.

8. The method according to claim 6 or 7, characterized in that: The step of screening the target protein code from the candidate protein codes filtered by the probability value according to the reference protein code comprises: Inputting the reference protein code and each of the screened candidate protein codes into the trained protein prediction model to obtain corresponding confidences; The candidate protein code with the highest confidence is determined as the target protein code.

9. The method according to any one of claims 1 to 8, characterized in that: The training process of the protein prediction model includes: Determining a model training set including a plurality of sample protein sequences with annotated protein sequences; Co-encoding each of the annotated protein sequences and the sample protein sequence to obtain a first training protein code, wherein the first training protein code includes the sample protein code and the annotated protein code; Encoding each of the annotated protein sequences to obtain a second training protein code, wherein the second training protein code includes the annotated protein code; The protein prediction model is trained together according to the first training protein code and the second training protein code.

10. The method according to claim 9, characterized in that The step of jointly training the protein prediction model according to the first training protein code and the second training protein code comprises: Inputting the first training protein code into the protein prediction model to determine the predicted protein code based on the sample protein code therein by the protein prediction model; The protein prediction model is adjusted based on the difference between the label vector in the predicted protein code and the label vector in the annotated protein code in the first training protein code.

11. The method according to claim 10, characterized in that The method of jointly training the protein prediction model according to the first training protein code and the second training protein code also includes: Inputting the encoding result of the preset start tag into the protein prediction model, so as to predict the tag multiple times in an iterative manner through the protein prediction model, and determining the predicted protein code according to the prediction order of the multiple tags; The protein prediction model is adjusted based on the difference between the label vector in the predicted protein code and the label vector in the annotated protein code in the second training protein code.

12. A device for generating an interactive protein, characterized in that: The device comprises: A sequence encoding module, used to encode a preset reference protein sequence to obtain a corresponding reference protein code, wherein the reference protein code includes a plurality of label vectors with a connection order, which are used to characterize the type and arrangement order of amino acids in the reference protein sequence; A protein generation module, used for inputting the reference protein code into a pre-trained protein generation model, and the protein generation model generates a preset number of candidate protein codes based on the reference protein code, wherein multiple label vectors in each candidate protein code are sequentially generated in an iterative manner; A coding screening module, used for screening a target protein coding from the candidate protein coding according to the reference protein coding; The sequence generation module is used to decode the target protein code to obtain the target protein sequence that interacts with the reference protein sequence.

13. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to implement the method described in any one of claims 1 to 11 when executing the instructions stored in the memory.

14. A non-volatile computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 11 is implemented.

Images