A method, device and medium for extracting the relationship between a chemical and a protein
By employing a question-and-answer format to extract chemical-protein interaction relationships, and utilizing the BERT model and BiLSTM network to encode protein entity representations, this method addresses the problem of low extraction efficiency in existing technologies and achieves efficient and accurate multiple relationship extraction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SUZHOU UNIV
- Filing Date
- 2022-08-04
- Publication Date
- 2026-04-28
AI Technical Summary
Existing technologies have low efficiency in extracting chemical-protein interaction relationships. With the increasing number of biomedical literatures and the large number of entity pairs, the extraction speed is slow.
Using a question-and-answer approach, we obtain a set of chemical and protein entities and question templates from the text, encode the contextual representation of protein entities using the BERT model and BiLSTM network, and determine the interaction relationships by combining the sigmoid activation function to extract multiple chemical-protein relationships.
It improves the extraction efficiency of chemical-protein interaction relationships, overcomes the problem of a large number of entity pairs, improves extraction accuracy, and avoids the problem of false positives.
Smart Images
Figure CN115269876B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of natural language processing technology, and in particular to a method, apparatus and medium for extracting the interaction relationship between chemicals and proteins. Background Technology
[0002] Automatically identifying the interactions between chemicals and proteins in biomedical literature is of great significance for the development of new drugs, precision medicine, and basic biomedical research. Chemical-protein interaction extraction refers to extracting the interaction between chemical entities and protein entities from the original biomedical literature text. For example, in the sentence "Exposure of LNCaP cells to estramustine for 24hr caused transcriptional inhibition of PSA in a concentration-dependent manner," the chemical "estramustine" has an inhibitory effect on the protein "PSA."
[0003] Currently, the method for extracting chemical-protein interaction relationships from text involves pairing chemical and protein entities in the text to obtain entity pairs, and then determining the interaction relationship between the protein and chemical entities in each entity pair. However, with the rapid increase in the number of biomedical documents in recent years, the number of entity pairs generated by the pairwise chemical-protein entity classification method is large, resulting in a slow extraction speed for chemical-protein interaction relationships.
[0004] Therefore, how to improve the extraction efficiency of the interaction relationship between chemicals and proteins is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0005] The purpose of this application is to provide a method, apparatus, and medium for extracting the interaction relationship between chemicals and proteins, so as to improve the extraction efficiency of the interaction relationship between chemicals and proteins.
[0006] To solve the above-mentioned technical problems, this application provides a method for extracting the interaction relationship between chemicals and proteins, comprising:
[0007] Obtain the set of chemical entities, the set of protein entities, and several question templates from the text; wherein each question template contains an interaction relationship between the chemical and the protein.
[0008] Select one chemical entity from the set of chemical entities as the target chemical entity, and input the target chemical entity into each of the question templates;
[0009] The text is queried using a question template containing the target chemical entity, and target protein entities that interact with the target chemical entity in the text are extracted to obtain a relationship group; wherein, the relationship group includes the target chemical entity, the target protein entity, and the interaction relationship between the target chemical entity and the target chemical entity.
[0010] Preferably, the step of querying the text using a question template containing the target chemical entity and extracting the target protein entity from the text that interacts with the target chemical entity includes:
[0011] Combine the question template containing the target chemical entity with the text to form sentence pairs;
[0012] The sentence pairs are symbolized to convert them into word fragment sequences, and the location information of each protein entity in the text is determined;
[0013] The word segment sequence is input into the BERT model to obtain the contextual representation of each word segment;
[0014] Based on the location information, the context representation sequence corresponding to each protein entity is selected from the context representation;
[0015] The context representation sequence is encoded and fused to obtain a single vector corresponding to each protein entity, and the single vector is concatenated with an indicator vector of the length of the protein entity word segment to obtain the final representation corresponding to each protein entity.
[0016] The final representation is processed through a fully connected layer and a sigmoid activation function to obtain a score for each protein entity;
[0017] Protein entities with scores greater than a threshold are identified as the target protein entities.
[0018] Preferably, the step of forming sentence pairs by combining the question template containing the target chemical entity with the text includes:
[0019] The question template containing the target chemical entity is concatenated with the text using a connector to obtain the sentence pair.
[0020] Preferably, the step of encoding and fusing the context representation sequence to obtain a single vector corresponding to each protein entity includes:
[0021] The context representation sequence is encoded using a BiLSTM network to obtain an output vector sequence for each protein entity;
[0022] The output vector sequence is fused using an attention mechanism to obtain a single vector for each protein entity.
[0023] Preferably, the step of processing the final representation through a fully connected layer and a sigmoid activation function to obtain a score for each protein entity includes:
[0024] The fully connected layer is used to convert the final representation into a single numerical value;
[0025] The single numerical value is normalized using the sigmoid activation function to obtain the score for each protein entity.
[0026] Preferably, before querying the text using a question template containing the target chemical entity, the method further includes:
[0027] The location of the target chemical entity is marked in the text using preset symbols.
[0028] This application also provides an extraction device for the interaction between chemicals and proteins, comprising:
[0029] The acquisition module is used to acquire a set of chemical entities, a set of protein entities, and several question templates from the text; wherein each question template contains an interaction relationship between a chemical and a protein.
[0030] The selection module is used to select a chemical entity from the set of chemical entities as the target chemical entity, and input the target chemical entity into each of the question templates;
[0031] An extraction module is used to query the text using a question template containing the target chemical entity, and extract target protein entities in the text that have an interaction relationship with the target chemical entity to obtain a relationship group; wherein, the relationship group includes the target chemical entity, the target protein entity, and the interaction relationship between the target chemical entity and the target chemical entity.
[0032] Preferably, the extraction module includes:
[0033] A component unit is used to combine a question template containing the target chemical entity with the text to form sentence pairs;
[0034] A conversion unit is used to symbolize the sentence pairs to convert them into word fragment sequences and to determine the location information of each protein entity in the text;
[0035] The input unit is used to input the word segment sequence into the BERT model to obtain the context representation of each word segment;
[0036] A filtering unit is used to filter out the context representation sequence corresponding to each protein entity from the context representation based on the location information;
[0037] The encoding and fusion unit is used to encode and fuse the context representation sequence to obtain a single vector corresponding to each protein entity, and concatenate the single vector with an indicator vector of the length of the protein entity word segment to obtain the final representation corresponding to each protein entity.
[0038] The processing unit is used to process the final representation through a fully connected layer and a sigmoid activation function to obtain a score for each protein entity;
[0039] A determining unit is used to determine the protein entities with scores greater than a threshold as the target protein entities.
[0040] This application also provides an extraction device for the interaction relationship between chemicals and proteins, including a memory for storing computer programs;
[0041] A processor is used to execute the computer program to implement the steps of the method for extracting the interaction relationship between the chemical and the protein.
[0042] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for extracting the interaction relationship between chemicals and proteins.
[0043] This application provides a method for extracting chemical-protein interaction relationships, comprising: obtaining a set of chemical entities, a set of protein entities, and several question templates from a text; wherein each question template contains an interaction relationship between a chemical and a protein; selecting a chemical entity from the chemical entity set as a target chemical entity and inputting the target chemical entity into each question template; querying the text using the question template containing the target chemical entity to extract target protein entities that have interaction relationships with the target chemical entity to obtain a relationship group; wherein the relationship group includes the target chemical entity, the target protein entity, and the interaction relationships between the target chemical entities. This application extracts chemical-protein interaction relationships from text through a question-and-answer format, extracting different types of interaction relationships according to different question templates, and can extract multiple relationships that may exist between the same entity pair. When there are many chemical and protein entities in a text, the question-and-answer extraction method can overcome the disadvantage of a large number of entity pairs caused by pairwise pairing of chemical and protein entities, effectively improving the extraction efficiency of chemical-protein interaction relationships.
[0044] The extraction device and method for the interaction between chemicals and proteins provided in this application have the same beneficial effects as described above. Attached Figure Description
[0045] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0046] Figure 1 A flowchart illustrating a method for extracting the interaction relationship between a chemical and a protein, provided as an embodiment of this application;
[0047] Figure 2 A schematic diagram illustrating a method for extracting the interaction relationship between a chemical and a protein, provided in an embodiment of this application;
[0048] Figure 3 A structural diagram of an extraction device for the interaction between chemicals and proteins provided in an embodiment of this application;
[0049] Figure 4 This is a structural diagram of an extraction device for the interaction between a chemical and a protein, provided in an embodiment of this application. Detailed Implementation
[0050] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.
[0051] Currently, the method for extracting chemical-protein interaction relationships from text involves pairing chemical and protein entities in the text to obtain entity pairs, and then determining the interaction relationship between the protein and chemical entities in each entity pair. For example, if the text contains 10 chemical entities and 10 protein entities, there are 100 entity pairs. The next step is to determine the interaction relationship between the chemical and protein entities in these 100 entity pairs. It is evident that the more chemical and protein entities there are in the text, the larger the number of entity pairs generated by the pairwise classification method, resulting in low efficiency in extracting chemical-protein interaction relationships.
[0052] The core of this application is to provide a method, apparatus, and medium for extracting the interaction relationship between chemicals and proteins.
[0053] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0054] Figure 1 A flowchart illustrating a method for extracting the interaction relationship between a chemical and a protein, as provided in this application embodiment, is shown below. Figure 1 As shown, the methods for extracting the interaction relationship between chemicals and proteins include:
[0055] S10: Obtain the set of chemical entities, the set of protein entities, and several question templates from the text; each question template contains an interaction relationship between the chemical and the protein.
[0056] S11: Select a chemical entity from the set of chemical entities as the target chemical entity, and input the target chemical entity into each question template.
[0057] S12: Use a question template containing the target chemical entity to query the text, and extract the target protein entity that has an interaction relationship with the target chemical entity in the text to obtain the relationship group; wherein, the relationship group includes the target chemical entity, the target protein entity, and the interaction relationship between the target chemical entities.
[0058] This application provides a method for extracting chemical-protein interaction relationships, comprising: obtaining a set of chemical entities, a set of protein entities, and several question templates from a text; wherein each question template contains an interaction relationship between a chemical and a protein; selecting a chemical entity from the chemical entity set as a target chemical entity and inputting the target chemical entity into each question template; querying the text using the question template containing the target chemical entity to extract target protein entities that have interaction relationships with the target chemical entity to obtain a relationship group; wherein the relationship group includes the target chemical entity, the target protein entity, and the interaction relationships between the target chemical entities. This application extracts chemical-protein interaction relationships from text through a question-and-answer format, extracting different types of interaction relationships according to different question templates, and can extract multiple relationships that may exist between the same entity pair. When there are many chemical and protein entities in a text, the question-and-answer extraction method can overcome the disadvantage of a large number of entity pairs caused by pairwise pairing of chemical and protein entities, effectively improving the extraction efficiency of chemical-protein interaction relationships.
[0059] For ease of understanding, each step in the above embodiments will be described in detail below.
[0060] In step S10, the question template is a natural language question constructed based on each type of interaction between chemicals and proteins in the corpus. For example, for an inhibition relationship: "which protein can be inhibited by chemical..." <chemical>?; Activation relationship: "which protein can be activated by chemical..." <chemical>"?"; Reaction substrate: "which protein can be a substrate of chemical..." <chemical>?; Antagonist: "which protein can be affected by antagonist chemicals" <chemical>?; Agonist: "which protein can be affected by agonist chemicals" <chemical>The chemical entity in this embodiment refers to a chemical entity mention, that is, the name of a chemical that appears at a certain position in the text. Similarly, the protein entity refers to a protein entity mention, that is, the name of a protein that appears at a certain position in the text.
[0061] In step S11, a chemical entity is selected from the chemical entity set as the target chemical entity, and the position of the target chemical entity is marked in the text. "@" and "$" are added before and after the target chemical entity to eliminate interference caused by the same chemical entity in different positions.
[0062] In step S12, a chemical entity is filled into a question template of a specific relationship type each time. Then, based on the question template, protein entities with the same type of interaction relationship with the chemical entity are found in the text and formed into a relationship group (triplet). Since different question templates are used for the same chemical entity to extract protein entities with different interaction relationships, multiple interaction relationships can exist between the same pair of chemical entities and protein entities, thus enabling the extraction of multiple interaction relationships between chemicals and proteins.
[0063] The specific steps are as follows:
[0064] (1) Combine the question template containing the target chemical entity with the text to form sentence pairs.
[0065] The question template containing the target chemical entity is concatenated with the text using a connector to obtain sentence pairs.
[0066] (2) The sentence pairs are symbolized to convert them into word fragment sequences, and the location information of each protein entity in the text is determined. The sequence number of each word fragment represents the location information.
[0067] (3) Input the word fragment sequence into the BERT model to obtain the context representation of each word fragment.
[0068] Specifically, sentence pairs are first symbolically converted into word fragment sequences, which are then input into a pre-trained language BERT model. This model consists of multiple iterative modules from bottom to top. Each module uses a multi-head self-attention mechanism to transform the input of the lower layer and then output it to the upper layer. The final output of the model is the context representation corresponding to each word fragment, i.e., the latent vector.
[0069] (4) Based on the location information, the context representation sequence corresponding to each protein entity is selected from the context representation.
[0070] (5) Encode and fuse the context representation sequence to obtain a single vector corresponding to each protein entity, and concatenate the single vector with the indicator vector of the length of the protein entity word segment to obtain the final representation corresponding to each protein entity.
[0071] Specifically, the contextual representation sequence of each protein fragment location is encoded using a bidirectional long short-term memory (BiLSTM) network to obtain an output vector sequence for each protein entity. Then, an attention mechanism is used to fuse the output vector sequences to generate a single vector. Finally, this single vector is concatenated with the protein length embedding vector (based on the word length of the protein fragment) to obtain the final representation of the protein entity.
[0072] (6) The final characterization is processed by a fully connected layer and a sigmoid activation function to obtain the score of each protein entity.
[0073] Specifically, the fully connected layer transforms the final representation of the high-dimensional protein entity into a single numerical value, and then uses the sigmoid activation function to normalize that single numerical value into a score between 0 and 1.
[0074] (7) Protein entities with scores greater than the threshold are identified as target protein entities.
[0075] For example, when the score is greater than a certain threshold (such as 0.5), it is determined that there is a specified interaction relationship between the protein entity and the chemical entity in the question. The target chemical entity in the question template, the target protein entity with a score greater than the threshold, and the interaction relationship between the target chemical entity and the target protein entity constitute a relation triple (relation group).
[0076] This application provides a method for extracting chemical-protein interaction relationships. Utilizing a machine reading comprehension framework, it extracts these relationships from text through a question-and-answer format. Different types of interaction relationships can be extracted based on different question templates, enabling the extraction of multiple interaction relationships that may exist between the same entity pair (chemical entity and protein entity). When there are a large number of chemical and protein entities in a text, this method overcomes the drawback of pairwise classification methods, which result in a large number of entity pairs, effectively improving the extraction efficiency of chemical-protein interaction relationships. Furthermore, protein entity fragment encoding better utilizes known protein entity mention information, avoiding the false positive problem caused by sequence labeling methods in standard machine reading comprehension, thereby improving the accuracy of interaction relationship extraction.
[0077] To make it easier to understand, we will provide further explanation with specific examples below. Figure 2 This is a schematic diagram of a method for extracting the interaction relationship between a chemical and a protein, provided in an embodiment of this application.
[0078] like Figure 2 As shown, the question template containing chemical entities is concatenated with the text to form a sentence pair, which is then symbolically converted into a message sequence. This sequence is then input into the BERT model to obtain the contextual representation of each word in the sentence pair. The contextual representation sequence corresponding to the protein entity fragment is further encoded using a BiLSTM network. The output vector sequence is fused into a single vector by an attention mechanism, and then concatenated with an indicator vector indicating the fragment length to obtain the final representation of the protein entity. The final representation is then passed through a fully connected layer and an activation function to obtain a score for the protein entity. Finally, based on this score, it is determined whether the protein entity has a relationship with the chemical in the question.
[0079] In the sentence "Agonist and antagonist actions of yohimbine as compared to fluparoxan at alpha(2)-adrenergic receptors (AR)s, serotonin(5-HT)(1A), 5-HT(1B), 5-HT(1D) and dopamine D(2) and D(3) receptors," "fluparoxan" is a chemical, while "alpha(2)-adrenergic receptors," "(AR)s," "(5-HT)(1A)," "5-HT(1B)," "5-HT(1D)," and "dopamine D(2) and D(3) receptors" are all proteins.
[0080] The first step is to design a question template for each type of interaction relationship in the corpus. Inhibition relationship: "which protein can be inhibited by chemical..." <chemical>?; Activate relationship:
[0081] "which protein can be activated by chemical <chemical>? ”; Reaction substrate:
[0082] "which protein can be substrate of chemical <chemical>?"; Antagonist: "which protein can be affected by antagonist chemical" <chemical>?"; Agonist: "which protein can be affected by agonist chemical" <chemical>?
[0083] The second step is to ask each chemical in the text five times using the above question template, and extract the corresponding relation triples. Taking the above example and antagonistic relationship as an example, fill the chemical "fluparoxan" in the example sentence into the question template: "which protein can be affected by antagonist chemical fluparoxan?" Then, the question and the sentence are concatenated together with the connecting symbol to obtain the sentence pair "[CLS]which protein can be affected by antagonist chemical fluparoxan? [SEP]Agonist and antagonist actions of yohimbine as compared to fluparoxan at alpha(2)-adrenergic receptors (AR)s, serotonin(5-HT)(1A), 5-HT(1B), 5-HT(1D) and dopamine D(2) and D(3) receptors. [SEP]".
[0084] The third step is to input the sentence pairs into the BERT model to obtain the latent vector of each word segment in the sentence, i.e., the context representation. Then, each protein entity segment is encoded and fused. For example, the word segment sequence (context representation sequence) corresponding to the protein entity "alpha(2)-adrenergic receptors" is 10 word segments: "alpha", "(", "2", "")", "-", "ad", "##rene", "##rg", "##ic", "receptors". After encoding and fusion, these word segment vector sequences (context representation sequences) are converted into a single vector, which is then concatenated with the length vector corresponding to the word segment length of 10 to obtain the final representation of this protein entity.
[0085] Fourth, the final representation of the protein entity is passed through a fully connected layer and a sigmoid activation function to obtain a score for the protein entity. If the score is greater than a threshold, it is determined that the chemical "fluparoxan" has an antagonistic effect on the protein "alpha(2)-adrenergic receptors"; otherwise, it does not. This process is repeated for five rounds of queries on the chemical and the sentence to extract all five potential chemical-protein relationship types.
[0086] In the above embodiments, the method for extracting the interaction relationship between chemicals and proteins has been described in detail. This application also provides embodiments corresponding to the apparatus for extracting the interaction relationship between chemicals and proteins. It should be noted that this application describes the embodiments of the apparatus from two perspectives: one based on functional modules and the other based on hardware.
[0087] Figure 3 A structural diagram of an extraction device for the interaction between a chemical and a protein, provided in an embodiment of this application, is shown below. Figure 3 As shown, the extraction device for the interaction between chemicals and proteins includes:
[0088] The acquisition module 10 is used to acquire the set of chemical entities, the set of protein entities, and several question templates in the text; each question template contains an interaction relationship between the chemical and the protein.
[0089] Module 11 is used to select a chemical entity from the set of chemical entities as the target chemical entity and input the target chemical entity into each question template.
[0090] Extraction module 12 is used to query the text using a question template containing the target chemical entity, and extract the target protein entity in the text that has an interaction relationship with the target chemical entity to obtain a relationship group; wherein, the relationship group includes the target chemical entity, the target protein entity, and the interaction relationship between the target chemical entities.
[0091] Based on the above embodiments, as a preferred embodiment, the extraction module includes:
[0092] The constituent unit is used to combine a question template containing the target chemical entity with text to form sentence pairs;
[0093] The transformation unit is used to symbolize sentence pairs to convert them into word fragment sequences and determine the location information of each protein entity in the text;
[0094] The input unit is used to input the word segment sequence into the BERT model to obtain the context representation of each word segment;
[0095] The filtering unit is used to filter out the context representation sequence corresponding to each protein entity from the context representation based on the location information;
[0096] The encoding and fusion unit is used to encode and fuse the context representation sequence to obtain a single vector corresponding to each protein entity, and concatenate the single vector with the indicator vector of the length of the protein entity word segment to obtain the final representation corresponding to each protein entity.
[0097] The processing unit is used to process the final representation through a fully connected layer and a sigmoid activation function to obtain a score for each protein entity;
[0098] The determination unit is used to identify protein entities with scores greater than a threshold as target protein entities.
[0099] Based on the above embodiments, as a preferred embodiment, the component unit includes:
[0100] The splicing subunit is used to concatenate a question template containing the target chemical entity with text using connector symbols to obtain sentence pairs.
[0101] Based on the above embodiments, as a preferred embodiment, the encoding and fusion unit includes:
[0102] The encoding subunit includes encoding the context representation sequence using a BiLSTM network to obtain an output vector sequence for each protein entity;
[0103] The fusion subunit includes fusing the output vector sequence using an attention mechanism to obtain a single vector for each protein entity.
[0104] Based on the above embodiments, as a preferred embodiment, the processing unit includes:
[0105] A transformation subunit is used to convert the final representation into a single numerical value using the fully connected layer;
[0106] The normalization subunit is used to normalize the single value using the sigmoid activation function to obtain a score for each protein entity.
[0107] Based on the above embodiments, as a preferred embodiment, it further includes:
[0108] The marking module is used to mark the location of target chemical entities in text using preset symbols.
[0109] Since the embodiments of the apparatus and the embodiments of the method correspond to each other, please refer to the description of the embodiments of the method for the embodiments of the apparatus, which will not be repeated here.
[0110] This application provides a device for extracting chemical-protein interaction relationships. It acquires a set of chemical entities, a set of protein entities, and several question templates from text. Each question template contains an interaction relationship between a chemical and a protein. A chemical entity is selected from the chemical entity set as the target chemical entity and input into each question template. The text is then queried using the question template containing the target chemical entity to extract target protein entities that interact with the target chemical entity, resulting in a relationship group. The relationship group includes the target chemical entity, the target protein entity, and the interaction relationships between the target chemical entities. This application extracts chemical-protein interaction relationships from text through a question-and-answer format. Different types of interaction relationships are extracted based on different question templates, enabling the extraction of multiple relationships that may exist between the same entity pair. When there are many chemical and protein entities in a text, the question-and-answer extraction method overcomes the drawback of a large number of entity pairs resulting from pairwise chemical and protein entity pairings, effectively improving the extraction efficiency of chemical-protein interaction relationships.
[0111] Figure 4 A structural diagram of an extraction device for the interaction between a chemical and a protein, as provided in an embodiment of this application, is shown below. Figure 4 As shown, the device for extracting the interaction between chemicals and proteins includes: a memory 20 for storing computer programs;
[0112] The processor 21 is used to execute computer programs to implement the steps of the above embodiment (method for extracting the interaction relationship between chemicals and proteins).
[0113] The device for extracting the interaction between chemicals and proteins provided in this embodiment may include, but is not limited to, smartphones, tablets, laptops, or desktop computers.
[0114] The processor 21 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 21 may be implemented using at least one of the following hardware forms: Digital Signal Processor (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 21 may also include a main processor and a coprocessor. The main processor, also known as the Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 21 may also include an Artificial Intelligence (AI) processor, which is used to handle computational operations related to machine learning.
[0115] The memory 20 may include one or more computer-readable storage media, which may be non-transitory. The memory 20 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 20 is used to store at least the following computer program 201, which, after being loaded and executed by the processor 21, is capable of implementing the relevant steps of the method for extracting the chemical-protein interaction relationship disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 20 may also include an operating system 202 and data 203, and the storage method may be temporary or permanent storage. The operating system 202 may include Windows, Unix, Linux, etc. The data 203 may include, but is not limited to, a set of chemical entities, a set of protein entities, and several question templates.
[0116] In some embodiments, the extraction device for the interaction between chemicals and proteins may further include a display screen 22, an input / output interface 23, a communication interface 24, a power supply 25, and a communication bus 26.
[0117] Those skilled in the art will understand that Figure 4 The structure shown does not constitute a limitation on the extraction device for the interaction between chemicals and proteins, and may include more or fewer components than shown.
[0118] The chemical-protein interaction extraction device provided in this application includes a memory and a processor. When the processor executes a program stored in the memory, it can implement the following method: acquiring a set of chemical entities, a set of protein entities, and several question templates from a text; wherein each question template contains an interaction relationship between a chemical and a protein; selecting a chemical entity from the set of chemical entities as a target chemical entity and inputting the target chemical entity into each question template; querying the text using the question template containing the target chemical entity to extract target protein entities that have an interaction relationship with the target chemical entity from the text to obtain a relationship group; wherein the relationship group includes the target chemical entity, the target protein entity, and the interaction relationship between the target chemical entities.
[0119] Finally, this application also provides an embodiment corresponding to a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps described in the method embodiment for extracting the interaction relationship between chemicals and proteins.
[0120] It is understood that if the methods in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0121] The foregoing has provided a detailed description of a method, apparatus, and medium for extracting the interaction relationship between a chemical and a protein, as provided in this application. The various embodiments in the specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.
[0122] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.< / chemical> < / chemical> < / chemical> < / chemical> < / chemical> < / chemical> < / chemical> < / chemical> < / chemical> < / chemical>
Claims
1. A method for extracting the interaction relationship between a chemical and a protein, characterized in that, include: Obtain the set of chemical entities, the set of protein entities, and several question templates from the text; wherein each question template contains an interaction relationship between the chemical and the protein. Select one chemical entity from the set of chemical entities as the target chemical entity, and input the target chemical entity into each of the question templates; The text is queried using a question template containing the target chemical entity, and target protein entities that interact with the target chemical entity in the text are extracted to obtain a relationship group; wherein, the relationship group includes the target chemical entity, the target protein entity, and the interaction relationship between the target chemical entity and the target chemical entity; The step of querying the text using a question template containing the target chemical entity and extracting target protein entities from the text that interact with the target chemical entity includes: Combine the question template containing the target chemical entity with the text to form sentence pairs; The sentence pairs are symbolized to convert them into word fragment sequences, and the location information of each protein entity in the text is determined; The word segment sequence is input into the BERT model to obtain the contextual representation of each word segment; Based on the location information, the context representation sequence corresponding to each protein entity is selected from the context representation; The context representation sequence is encoded and fused to obtain a single vector corresponding to each protein entity, and the single vector is concatenated with an indicator vector of the length of the protein entity word segment to obtain the final representation corresponding to each protein entity. The final representation is processed through a fully connected layer and a sigmoid activation function to obtain a score for each protein entity; Protein entities with scores greater than a threshold are identified as the target protein entities.
2. The method for extracting the interaction relationship between chemicals and proteins according to claim 1, characterized in that, The step of forming sentence pairs by combining the question template containing the target chemical entity with the text includes: The question template containing the target chemical entity is concatenated with the text using a connector to obtain the sentence pair.
3. The method for extracting the interaction relationship between chemicals and proteins according to claim 1, characterized in that, The process of encoding and fusing the context representation sequence to obtain a single vector corresponding to each protein entity includes: The context representation sequence is encoded using a BiLSTM network to obtain an output vector sequence for each protein entity; The output vector sequence is fused using an attention mechanism to obtain a single vector for each protein entity.
4. The method for extracting the interaction relationship between chemicals and proteins according to claim 1, characterized in that, The final representation is processed through a fully connected layer and a sigmoid activation function to obtain a score for each protein entity, including: The fully connected layer is used to convert the final representation into a single numerical value; The single numerical value is normalized using the sigmoid activation function to obtain the score for each protein entity.
5. The method for extracting the interaction relationship between chemicals and proteins according to claim 1, characterized in that, Before querying the text using a question template containing the target chemical entity, the method further includes: The location of the target chemical entity is marked in the text using preset symbols.
6. An extraction device for the interaction between a chemical and a protein, characterized in that, include: The acquisition module is used to acquire a set of chemical entities, a set of protein entities, and several question templates from the text; wherein each question template contains an interaction relationship between a chemical and a protein. The selection module is used to select a chemical entity from the set of chemical entities as the target chemical entity, and input the target chemical entity into each of the question templates; An extraction module is used to query the text using a question template containing the target chemical entity, and extract target protein entities in the text that have an interaction relationship with the target chemical entity to obtain a relationship group; wherein, the relationship group includes the target chemical entity, the target protein entity, and the interaction relationship between the target chemical entity and the target chemical entity; The extraction module includes: A component unit is used to combine a question template containing the target chemical entity with the text to form sentence pairs; A conversion unit is used to symbolize the sentence pairs to convert them into word fragment sequences and to determine the location information of each protein entity in the text; The input unit is used to input the word segment sequence into the BERT model to obtain the context representation of each word segment; A filtering unit is used to filter out the context representation sequence corresponding to each protein entity from the context representation based on the location information; The encoding and fusion unit is used to encode and fuse the context representation sequence to obtain a single vector corresponding to each protein entity, and concatenate the single vector with an indicator vector of the length of the protein entity word segment to obtain the final representation corresponding to each protein entity. The processing unit is used to process the final representation through a fully connected layer and a sigmoid activation function to obtain a score for each protein entity; A determining unit is used to determine the protein entities with scores greater than a threshold as the target protein entities.
7. An extraction device for the interaction between a chemical and a protein, characterized in that, Includes memory used to store computer programs; A processor, configured to execute the computer program to implement the steps of the method for extracting the interaction relationship between a chemical and a protein as described in any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method for extracting the interaction relationship between chemicals and proteins as described in any one of claims 1 to 5.