A method, device, equipment and medium for predicting drug clinical trial results

Abstract

The present invention belongs to the field of artificial intelligence technology and intelligent medical technology, and in particular, relates to a method, device, equipment and medium for predicting the results of a drug clinical trial. The method includes: obtaining a clinical trial protocol; extracting information about a target drug and a target disease; obtaining information about a first feature item and information about an action relationship; converting a drug feature vector; obtaining information about a second feature item and information about an action relationship; converting a disease feature vector; converting to a clinical trial protocol feature vector; concatenating vectors, and inputting the vectors into a trained deep learning model to obtain a predicted value of the success probability of the clinical trial protocol. The present invention integrates the features of the graph with the features of the clinical trial for the first time to form an integrated feature, and finally predicts the probability of success of the drug clinical trial through deep learning probability prediction.

Description

Technical Field

[0001] The present invention belongs to the fields of artificial intelligence technology and intelligent medical technology, and in particular, relates to a method, device, electronic device and computer-readable storage medium for predicting drug clinical trial results. Background Art

[0002] Drug R&D requires huge human, material and financial resources. The R&D process includes drug target discovery, lead compound screening, lead compound optimization, pre-clinical drug research, clinical trials, and drug approval. Among them, drug clinical trials are divided into Phase I, II, and III clinical trials. According to relevant research, the largest proportion of funds are spent on Phase II and Phase III clinical trials, which usually cost hundreds of millions of dollars and can last up to 3-5 years. At the same time, they face the risk of a high failure rate. Usually, a drug has a high risk of failure in Phase II and Phase III, resulting in the failure of drug R&D, and also affecting the R&D cost and the progress of drug approval.

[0003] The main reasons for clinical trial failure are as follows:

[0004] 1. In the early stages of drug target discovery, the optimal target was not found, resulting in poor clinical therapeutic effects of the final drug.

[0005] 2. A better target was found, but due to the complex medical conditions of the people included in the clinical trial, the side effects of the drug increased, and the main clinical endpoints and goals could not be achieved, resulting in the failure of the clinical trial.

[0006] 3. The clinical trial design is unreasonable, the enrollment population or the trial objectives are set too high or mismatched, resulting in the failure of the clinical trial.

[0007] Predicting the success rate of clinical trial results of drugs before conducting clinical trials can greatly assist clinical trial design and drug development decisions. At this stage, the more common method is to use machine learning to manually formulate some features and perform probability predictions, such as: Feature 1-whether drugs similar to the compound that have been marketed can treat related diseases, Feature 2-whether there are some related research papers on the candidate compound, etc.

[0008] The defects of the existing methods are:

[0009] 1. On the one hand, manually added features are highly empirical, cumbersome, and difficult to expand.

[0010] 2. Some design features of clinical trials are not taken into account.

[0011] 3. For drugs, there is a lot of existing knowledge that has not been incorporated, for example, interactions with existing drugs may affect the prediction of drug effects. Summary of the invention

[0012] The purpose of the present invention is to provide a method, device, electronic device and computer-readable storage medium for predicting drug clinical trial results, so as to solve the problems of cumbersome operation, incomplete feature design and inaccurate prediction effect in the prior art.

[0013] To achieve the above-mentioned and other related purposes, the present invention provides a method for predicting drug clinical trial results, comprising:

[0014] Obtain clinical trial protocols;

[0015] extracting information of the target drug and information of the target disease from the clinical trial protocol;

[0016] According to the information of the target drug, obtaining information of a first characteristic item associated with the target drug and information of an action relationship between the first characteristic item and the target drug from a preset drug atlas;

[0017] Converting the information of the target drug, the information of the first feature item, and the information of the interaction relationship between the target drug and the first feature item into a drug feature vector;

[0018] According to the information of the target disease, obtaining from the preset drug atlas information of a second characteristic item associated with the target disease and information of an action relationship between the second characteristic item and the target disease;

[0019] Converting the information of the target disease, the information of the second feature item, and the information of the interaction relationship between the two into a disease feature vector;

[0020] Converting the clinical trial protocol into a clinical trial protocol feature vector;

[0021] The clinical trial scheme feature vector, the drug feature vector and the disease feature vector are spliced ​​into a complete clinical prediction feature vector, and the clinical prediction feature vector is input into a trained deep learning model to obtain a predicted value of the success probability of the clinical trial scheme; the deep learning model is a classification model.

[0022] To achieve the above-mentioned and other related purposes, the present invention also provides a device for predicting drug clinical trial results, comprising:

[0023] Protocol acquisition module, used to obtain clinical trial protocols;

[0024] An information extraction module, used to extract information about target drugs and target diseases from the clinical trial protocol;

[0025] A first associated information acquisition module, used to acquire information of a first characteristic item associated with the target drug and information of an action relationship between the first characteristic item and the target drug from a preset drug map according to information of the target drug;

[0026] A first conversion module, used to convert the information of the target drug, the information of the first feature item and the information of the interaction relationship between the two into a drug feature vector;

[0027] A second associated information acquisition module is used to acquire information of a second characteristic item associated with the target disease and information of an action relationship between the second characteristic item and the target disease from a preset drug map according to information of the target disease;

[0028] A second conversion module, used to convert the information of the target disease, the information of the second feature item and the information of the interaction relationship between the two into a disease feature vector;

[0029] A third conversion module, used to convert the clinical trial protocol into a clinical trial protocol feature vector;

[0030] A prediction module is used to splice the clinical trial scheme feature vector, the drug feature vector and the disease feature vector into a complete clinical prediction feature vector, and input the clinical prediction feature vector into a trained deep learning model to obtain a predicted value of the success probability of the clinical trial scheme.

[0031] To achieve the above-mentioned purpose and other related purposes, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in any one of claims 1 to 7 when executing the computer program.

[0032] To achieve the above object and other related objects, the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 7 are implemented.

[0033] In the scheme implemented by the above-mentioned drug clinical trial result prediction method, device, equipment and medium, a pre-designed drug map can be used to obtain associated feature items that have an action relationship with the target drug and the target disease. These associated feature items can reflect to a certain extent which genes, proteins and other human elements and symptoms the compounds in the target drug can affect. These effects are directly related to the expected effect of the target drug on the target disease, which helps to determine the optimal target. Using this information as an input item of the deep learning model can greatly improve the accuracy of the prediction. In addition, this scheme converts the entire clinical trial scheme written in natural language into a clinical trial scheme feature vector. This clinical trial scheme feature vector can include most of the feature information of the clinical trial scheme, without the need to manually formulate features, avoiding the problem of incomplete integration of knowledge features and poor accuracy of prediction results due to differences in experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a flow chart of a method for predicting drug clinical trial results provided by an embodiment of the present invention;

[0035] Figure 2 is a flowchart of a first feature item acquisition process provided by an embodiment of the present invention;

[0036] Figure 3 is a flow chart of a process for generating a drug feature vector provided by an embodiment of the present invention;

[0037] Figure 4 is a flowchart of a second feature item acquisition process provided by an embodiment of the present invention;

[0038] Figure 5 is a flow chart of a disease feature vector generation process provided by an embodiment of the present invention;

[0039] Figure 6 is a flow chart of a process for generating a feature vector of a clinical trial protocol provided by an embodiment of the present invention;

[0040] Figure 7 is a structural block diagram of a drug clinical trial result prediction device provided by an embodiment of the present invention;

[0041] Figure 8 It is a structural block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The following describes the embodiments of the present invention through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention.

[0043] See also Figure 1-8 It should be noted that the illustrations provided in this embodiment are only used to illustrate the basic concept of the present invention in a schematic manner, and the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0044] The embodiments of the present application can acquire and process the associated data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application device that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0045] Figure 1 A flow chart showing a preferred embodiment of the method for predicting drug clinical trial results of the present invention.

[0046] The drug clinical trial result prediction method is applied to one or more electronic devices, which are devices that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and whose hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0047] The electronic device may be any electronic product that can perform human-computer interaction with a user, such as a personal computer, a tablet computer, a smart phone, a personal digital assistant (PDA), a game console, an interactive network television (IPTV), a smart wearable device, etc.

[0048] The electronic device may also include a network device and / or a user device, wherein the network device includes, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud consisting of a large number of hosts or network servers based on cloud computing.

[0049] The network where the electronic device is located includes but is not limited to the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (VPN), etc.

[0050] The following will be combined Figure 1 The method for predicting drug clinical trial results of the present invention is described in detail. The method for predicting drug clinical trial results can be applied to predicting drug clinical trial results in the drug development process, for example.

[0051] See also Figure 1 As shown, Figure 1 The flowchart of a method for predicting drug clinical trial results in a specific embodiment includes:

[0052] S1: Obtain clinical trial protocols.

[0053] The drug clinical trial result prediction method of the present invention can be applied to independent terminal electronic devices. For example, the clinical trial plan in natural language form can be input into the electronic device through an input device, or a configuration template of the clinical trial plan can be designed in advance, and the clinical trial plan can be quickly entered by calling the configuration template.

[0054] The drug clinical trial result prediction method of the present invention can also be applied to networked devices. For example, the clinical trial plan can be input on the client, and the client sends the clinical trial plan to the background server through the network. The background server performs subsequent processing and obtains the prediction result, and then returns the prediction result to the client.

[0055] It is understandable that the clinical trial protocol should at least include the component information of the target drug, the type information of the target disease, the condition information of the included population, and the expected treatment effect information. The included population refers to the population participating in the clinical trial. The included population should at least include people suffering from the target disease and may also include healthy people. These healthy people can, for example, be used as reference examples; the expected treatment effect information can, for example, be used as a comparison standard for whether the trial is successful. The clinical trial protocol can also include information such as the trial cycle and trial environment.

[0056] S2: Extract target drug information and target disease information from the clinical trial protocol.

[0057] In a specific embodiment, the present invention can, for example, utilize keyword recognition to automatically extract information about target drugs and target diseases in a clinical trial protocol; or it can also mark each feature in the clinical trial protocol when entering the clinical trial protocol, and then automatically obtain information about the target drugs and target diseases based on the feature markings.

[0058] S3: According to the information of the target drug, information of a first characteristic item associated with the target drug and information of an action relationship between the first characteristic item and the target drug are obtained from a preset drug atlas.

[0059] The drug map described in this step is an existing drug development map based on known knowledge information, which contains association information between features such as compounds, diseases, genes, and proteins, such as the DDKG drug development map, and the map data mainly comes from Ping An's drug knowledge map. It is understandable that the drug map can also use other known open source drug development maps.

[0060] See also Figure 2 As shown, in step S3, i.e., according to the information of the target drug, obtaining the information of the first characteristic item associated with the target drug and the information of the action relationship between the first characteristic item and the target drug from the preset drug map, comprises the following steps:

[0061] S31: Obtain information about the compounds contained in the target drug.

[0062] During the specific implementation process, the compound information of the target drug can be recorded in the clinical trial protocol, and the compound information of the target drug can be automatically obtained from the clinical trial protocol through a natural language processing model; or the target drug and its corresponding compound information can be stored in a preset database, and the name of the target drug involved in the clinical trial protocol is first identified, and then the compound information of the target drug is retrieved from the preset database according to the name of the target drug; or the compound information can be directly used as the name of the target drug and recorded in the clinical trial protocol.

[0063] S32: Compare the information of the compound with the drug map to find the first target node corresponding to the compound in the drug map; for example, keyword search may be used to find the first target node corresponding to the target drug in the drug map.

[0064] S33: Through the first target node, find the first associated node that has an action relationship with the first target node from the drug map; it can be understood that the drug map itself contains the action relationship between each node, so after finding the first target node, the first associated node that has an action relationship with the first target node can be obtained from the drug map.

[0065] S34: Obtain a first feature item corresponding to the first association node, and the interaction relationship between the first feature item and the compound; the first feature item is a disease and / or a gene and / or a protein and / or a compound; in a specific embodiment, the first feature item may be, for example, a gene and a protein, and the interaction relationship between the first feature item and the compound may be, for example, the type of biological or chemical reaction that the gene and protein can produce under the action of the compound.

[0066] Regarding steps S31-S34, it should be noted that the same compound generally corresponds to one target node, and the same target node is often connected to multiple associated nodes. Therefore, the first associated node in the present invention is sometimes more than one. During the specific implementation process, the first characteristic item corresponding to each first associated node should be obtained, as well as the interaction relationship between each first characteristic item and the compound.

[0067] S4: Convert the information of the target drug, the information of the first feature item and the information of the interaction between the two into a drug feature vector; specifically, for example, the Pubmed Bert+RotatE model can be used to convert the information of the target drug, the information of the first feature item and the interaction between the two into a drug feature vector in matrix form, wherein Pubmed Bert is an open source pre-trained natural language model in the field of bio-hospital, which can extract feature information in natural language, and RotatE is an open source knowledge graph embedding model, which can convert feature items in the knowledge graph into vectors in matrix form.

[0068] It is understandable that the drug feature vector includes not only the component information of the target drug, but also the type information of the feature items such as genes, proteins, diseases, etc. related to the target drug, as well as the action relationship information between these feature items and the target drug.

[0069] See also Figure 3 As shown, in step S4, the information of the target drug, the information of the first feature item and the information of the interaction relationship between the target drug and the first feature item are converted into a drug feature vector, which may include the following steps:

[0070] S41: Input the information of the target drug, the information of the first feature item and the information of the interaction relationship between the two into a trained knowledge graph embedding model.

[0071] S42: Convert the target drug information, the first feature item information and the interaction relationship between the target drug information and the first feature item information into a drug feature vector in matrix form through the knowledge graph embedding model.

[0072] For the steps S41-S42, the knowledge graph embedding model adopts a known open source graph embedding algorithm, for example, the Pubmed Bert+RotatE model algorithm can be used; it can be understood that the knowledge graph usually organizes knowledge in the form of a network, each node in the network represents an entity, and the edge represents the relationship between entities. The knowledge graph embedding algorithm can represent entities and relationships as dense low-dimensional vectors, and can efficiently calculate the semantic connection between entities and relationships in a low-dimensional space. The present invention preferably uses the Pubmed Bert+RotatE model to process the knowledge graph data, and can convert the information of the target drug, the information of the first feature item and the information of the action relationship between the two into a drug feature vector.

[0073] S5: According to the information of the target disease, obtain from a preset drug atlas information of a second characteristic item associated with the target disease, and information of an action relationship between the second characteristic item and the target disease.

[0074] The drug map described in this step is an existing drug development map based on known knowledge information, including association information between features such as compounds, diseases, genes, and proteins, such as the DDKG drug development map, and the map data mainly comes from Ping An's drug knowledge map. It is understandable that the drug map can adopt a known open source drug development map.

[0075] See also Figure 4 As shown, in step S5, i.e., according to the information of the target disease, obtaining the information of the second characteristic item associated with the target disease and the information of the action relationship between the second characteristic item and the target disease from the preset drug map, includes the following steps:

[0076] S51: Compare the information of the target disease with the drug map to find a second target node in the drug map corresponding to the target disease; for example, keyword search may be used to find the second target node in the drug map corresponding to the target disease.

[0077] S52: Through the second target node, find the second associated node that has an action relationship with the second target node from the drug map; it can be understood that the drug map itself contains the action relationship between each node, so after finding the second target node, the second associated node that has an action relationship with the second target node can be obtained from the drug map.

[0078] S53: Obtain the second feature item corresponding to the second association node, and the interaction relationship between the second feature item and the target disease; the second feature item is a disease and / or a gene and / or a protein and / or a compound; in a specific embodiment, the second feature item may be, for example, a gene and a protein, and the interaction relationship between the second feature item and the target disease may be, for example, the impact that the target disease can have on the gene and the protein.

[0079] Regarding steps S51-S53, it should be noted that the same disease generally corresponds to one target node, and the same target node is often connected to multiple associated nodes. Therefore, the second associated node in the present invention is sometimes more than one. During the specific implementation process, the second feature item corresponding to each second associated node should be obtained, as well as the functional relationship between each second feature item and the disease.

[0080] S6: Convert the target disease information, the second feature item information and the interaction relationship therebetween into a disease feature vector; specifically, for example, the Pubmed Bert+RotatE model can be used to convert the target disease information, the second feature item information and the interaction relationship therebetween into a disease feature vector in matrix form, wherein Pubmed Bert is an open source pre-trained natural language model in the field of bio-hospital, which can extract feature information in natural language, and RotatE is an open source knowledge graph embedding model, which can convert feature items in the knowledge graph into vectors in matrix form.

[0081] It is understandable that the disease feature vector includes not only the type of the target disease, but also the type information of feature items such as genes, proteins, compounds, etc. related to the target disease, as well as the interaction relationship information between these feature items and the target disease.

[0082] See also Figure 5 As shown, in step S6, the information of the target disease, the information of the second feature item and the information of the interaction relationship between the two are converted into a disease feature vector, which may include the following steps:

[0083] S61: Input the information of the target disease, the information of the second feature item and the information of the interaction relationship between the two into a trained knowledge graph embedding model.

[0084] S62: Convert the target disease information, the second feature item information and the interaction relationship between the target disease information and the second feature item information into a disease feature vector in matrix form through the knowledge graph embedding model.

[0085] For the steps S61-S62, the knowledge graph embedding model adopts a known open source graph embedding algorithm, for example, the Pubmed Bert+RotatE model algorithm can be used; it can be understood that the knowledge graph usually organizes knowledge in the form of a network, each node in the network represents an entity, and the edge represents the relationship between entities. The knowledge graph embedding algorithm can represent entities and relationships as dense low-dimensional vectors, and can efficiently calculate the semantic connection between entities and relationships in low-dimensional space. The present invention preferably uses the Pubmed Bert+RotatE model to process the knowledge graph data, and can convert the information of the target disease, the information of the second feature item and the information of the interaction relationship between the two into a disease feature vector.

[0086] The present invention incorporates the characteristics of drug knowledge graphs, which can potentially capture the action mechanisms between drugs, between drugs and other genes and diseases, comprehensively consider the complex drug action environment that occurs in practice, and can better characterize drug action disease information and improve prediction accuracy.

[0087] This method proposes for the first time to introduce the features of the drug knowledge graph in the clinical trial prediction stage, taking into account the additional information features of the drug knowledge graph. Compared with traditional manual features, it can greatly improve the accuracy of prediction, while guiding drug clinical design and saving R&D cycle costs.

[0088] S7: Convert the clinical trial protocol into a clinical trial protocol feature vector.

[0089] It is understandable that the clinical trial protocol feature vector should contain most or even all of the condition information of the clinical trial protocol, including but not limited to the component information of the target drug, the type information of the target disease, the condition information of the enrolled population, and the expected treatment effect information. The enrolled population refers to the population participating in the clinical trial, and the enrolled population should at least include people suffering from the target disease, and may also include healthy people. These healthy people can be used as reference examples, for example; the expected treatment effect information can be used as a comparison standard for whether the trial is successful. The clinical trial protocol can also include information such as the trial cycle and the trial environment.

[0090] See also Figure 6 As shown, in step S7, the clinical trial scheme is converted into a clinical trial scheme feature vector, including the following steps:

[0091] S71: extracting the condition information contained in the clinical trial protocol; the condition information at least includes the condition information of the enrolled population, the compound component information of the target drug, and the expected therapeutic effect.

[0092] S72: Input the condition information into the natural language processing model.

[0093] S73: Convert the conditional information of the clinical trial protocol into a clinical trial protocol feature vector in matrix form through the natural language processing model.

[0094] For steps S71-S73, the natural language processing model is a known open source language processing model. The natural language processing model may be, for example, the BERT model. BERT is a pre-trained language model that can understand words and sentences in documents. Specifically, it converts words in the text that cannot be directly calculated into computable vectors or matrices, and these digitized vectors should be able to better reflect the meaning of the corresponding words in the sentence.

[0095] The present invention can automatically extract clinical trial protocol features and convert them into feature vectors through a natural language processing model. Using BERT to encode clinical trials, the text feature information can be automatically extracted, saving the tedious process of manual features and improving efficiency and accuracy.

[0096] S8: Concatenate the clinical trial scheme feature vector, the drug feature vector and the disease feature vector into a complete clinical prediction feature vector, and input the clinical prediction feature vector into a trained deep learning model. The deep learning model obtains the predicted value of the success probability of the clinical trial scheme based on the relationship between the action mechanism between drugs, between drugs and other genes, proteins, and diseases and the success probability of clinical trials.

[0097] The deep learning model described in the present invention is a classification model. Specifically, the deep learning model can be, for example, a fully connected neural network model or a Transformer model.

[0098] In a specific embodiment, the training process of the deep learning model is: obtaining historical clinical trial plans and test results; preprocessing the historical clinical trial plans according to the above steps S2-S7; inputting the feature vector obtained after preprocessing into the deep learning model to obtain prediction results; comparing the prediction results with the actual test results, and calculating the loss value; back-propagating the loss value layer by layer through the gradient descent method to update the model parameters; repeating the above process until the loss value is less than a preset range, at which time the model parameters can reflect the relationship between the mechanism of action between drugs and between drugs and other genes, proteins and diseases and the probability of success of clinical trials.

[0099] It can be seen that in the above scheme, the drug clinical trial prediction method of the present invention can use the pre-designed drug map to obtain the associated feature items that have an action relationship with the target drug and the target disease. These associated feature items can reflect to a certain extent which genes, proteins and other human elements and symptoms the compounds in the target drug can affect. These effects are directly related to the expected effect of the target drug on the target disease, which helps to determine the optimal target. Using this information as an input item of the deep learning model can greatly improve the accuracy of the prediction. In addition, this scheme converts the entire clinical trial scheme written in natural language into a clinical trial scheme feature vector. This clinical trial scheme feature vector can include most of the feature information of the clinical trial scheme, without the need to manually formulate features, avoiding the problem of incomplete integration of knowledge features and poor accuracy of prediction results due to differences in experience.

[0100] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0101] See also Figure 7 As shown, in one embodiment, a drug clinical trial result prediction device is provided, which corresponds one-to-one to the drug clinical trial result prediction method in the above embodiment, and includes a scheme acquisition module 10, an information extraction module 20, a first associated information acquisition module 30, a first conversion module 40, a second associated information acquisition module 50, a second conversion module 60, a third conversion module 70, and a prediction module 80. The module referred to in the present invention refers to a series of computer program segments that can be executed by a processor 1 and can complete fixed functions, which are stored in a memory 2. Each functional module is described in detail as follows:

[0102] The protocol acquisition module 10 is used to acquire the clinical trial protocol;

[0103] The information extraction module 20 is used to extract information about target drugs and target diseases from the clinical trial protocol;

[0104] The first associated information acquisition module 30 is used to acquire information of a first characteristic item associated with the target drug and information of an action relationship between the first characteristic item and the target drug from a preset drug map according to information of the target drug;

[0105] The first conversion module 40 is used to convert the information of the target drug, the information of the first feature item and the information of the interaction relationship between the two into a drug feature vector;

[0106] The second associated information acquisition module 50 is used to acquire information of a second characteristic item associated with the target disease and information of an action relationship between the second characteristic item and the target disease from a preset drug map according to information of the target disease;

[0107] The second conversion module 60 is used to convert the information of the target disease, the information of the second feature item and the information of the interaction relationship between the two into a disease feature vector;

[0108] The third conversion module 70 is used to convert the clinical trial protocol into a clinical trial protocol feature vector;

[0109] The prediction module 80 is used to splice the clinical trial scheme feature vector, the drug feature vector and the disease feature vector into a complete clinical prediction feature vector, and input the clinical prediction feature vector into a trained deep learning model to obtain a predicted value of the success probability of the clinical trial scheme.

[0110] In one embodiment, the first association information acquisition module 30 is specifically used for:

[0111] Obtaining information about compounds contained in the target drug;

[0112] Comparing the information of the compound with the drug map, and finding a first target node in the drug map corresponding to the compound;

[0113] Finding a first associated node having an action relationship with the first target node from the drug map through the first target node;

[0114] A first feature item corresponding to the first association node and an action relationship between the first feature item and the compound are obtained; the first feature item is a disease and / or a gene and / or a protein.

[0115] In one embodiment, the first conversion module 40 is specifically used for:

[0116] Inputting information about the target drug, information about the first feature item, and information about the interaction relationship between the target drug and the first feature item into a trained knowledge graph embedding model;

[0117] The information of the target drug, the information of the first feature item and the information of the action relationship between the two are converted into a drug feature vector in matrix form through the knowledge graph embedding model.

[0118] In one embodiment, the second association information acquisition module 50 is specifically used to:

[0119] Comparing the information of the target disease with the drug map, and finding a second target node in the drug map corresponding to the target disease;

[0120] Finding a second associated node having an action relationship with the second target node from the drug map through the second target node;

[0121] A second feature item corresponding to the second associated node and an action relationship between the second feature item and the target disease are obtained; the second feature item is a gene and / or a protein and / or a compound.

[0122] In one embodiment, the second conversion module 60 is specifically used for:

[0123] Inputting information about the target disease, information about the second feature item, and information about the interaction relationship between the target disease and the second feature item into a trained knowledge graph embedding model;

[0124] The information of the target disease, the information of the second feature item and the information of the interaction relationship between the two are converted into a disease feature vector in matrix form through the knowledge graph embedding model.

[0125] In one embodiment, the third conversion module 70 is specifically used for:

[0126] Extracting the condition information contained in the clinical trial protocol; the condition information at least includes the condition information of the enrolled population, the compound component information of the target drug, and the expected therapeutic effect;

[0127] Inputting the condition information into the natural language processing model;

[0128] The condition information of the clinical trial protocol is converted into a clinical trial protocol feature vector in matrix form through the natural language processing model.

[0129] It should be noted that, in the present invention, in order to ensure the security of data, the data involved can be deployed on the blockchain to prevent the data from being maliciously tampered with.

[0130] It should be noted that the step division of the above methods is only for clear description. When implemented, they can be combined into one step or some steps can be split and decomposed into multiple steps. As long as they contain the same logical relationship, they are all within the scope of protection of this patent; adding insignificant modifications to the algorithm or process or introducing insignificant designs without changing the core design of the algorithm and process are all within the scope of protection of this patent.

[0131] It should be noted that the drug clinical trial result prediction device of the present embodiment is a device corresponding to the above-mentioned drug clinical trial result prediction method, and the functional modules in the drug clinical trial result prediction device may correspond to the corresponding steps in the drug clinical trial result prediction method. The drug clinical trial result prediction device of the present embodiment can be implemented in conjunction with the drug clinical trial result prediction method. Accordingly, the relevant technical details mentioned in the drug clinical trial result prediction device of the present embodiment can also be applied in the above-mentioned drug clinical trial result prediction method.

[0132] It should be noted that the above-mentioned functional modules can be fully or partially integrated into one physical entity during actual implementation, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, some or all of the steps of the above method, or the above functional modules can be completed by hardware integrated logic circuits in the processor element or instructions in software form.

[0133] like Figure 8 , which is a schematic diagram of the structure of an electronic device of a preferred embodiment of the present invention for implementing a method for predicting drug clinical trial results.

[0134] The electronic device may include a memory, a processor and a bus, and may also include a computer program stored in the memory and executable on the processor, such as a drug clinical trial result prediction program.

[0135] Among them, the memory includes at least one type of computer-readable storage medium, and the computer-readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. The memory can be an internal storage unit of an electronic device in some embodiments, such as a mobile hard disk of the electronic device. The memory can also be an external storage device of an electronic device in other embodiments, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Further, the memory can also include both an internal storage unit of the electronic device and an external storage device. The memory can not only be used to store application software and various types of data installed in the electronic device, such as the code of a drug clinical trial result prediction program, but also can be used to temporarily store data that has been output or is to be output.

[0136] In some embodiments, the processor may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect the various components of the entire electronic device, and executes or executes the programs or modules stored in the memory (for example, executing the drug clinical trial result prediction program, etc.), and calls the data stored in the memory to execute various functions of the electronic device and process data.

[0137] The processor executes the operating device of the electronic device and various installed application programs. The processor executes the application programs to implement the steps in the above-mentioned various drug clinical trial result prediction method embodiments, such as the steps shown in the figure.

[0138] Exemplarily, the computer program may be divided into one or more modules, and the one or more modules are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device. For example, the computer program may be divided into a solution acquisition module, an information extraction module, a first associated information acquisition module, a first conversion module, a second associated information acquisition module, a second conversion module, a third conversion module, and a prediction module.

[0139] The above-mentioned integrated unit implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, and includes a number of instructions for enabling an electronic device (which can be a personal computer, a computer device, or a network device, etc.) or a processor to perform part of the functions of the drug clinical trial result prediction method described in each embodiment of the present invention.

[0140] The blockchain referred to in this invention is a new application mode of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0141] The bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 The bus is represented by only one straight line, but it does not mean that there is only one bus or one type of bus. The bus is configured to realize the connection and communication between the memory and at least one processor.

[0142] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0143] In summary, the drug clinical trial result prediction method, device, equipment and medium of the present invention can use the pre-designed drug map to obtain the associated feature items that have an action relationship with the target drug and the target disease. These associated feature items can reflect to a certain extent which genes, proteins and other human elements and symptoms the compounds in the target drug can affect. These effects are directly related to the expected effect of the target drug on the target disease, which helps to determine the optimal target. Using this information as an input item of the deep learning model can greatly improve the accuracy of the prediction. In addition, this scheme converts the entire clinical trial scheme written in natural language into a clinical trial scheme feature vector. This clinical trial scheme feature vector can include most of the feature information of the clinical trial scheme, without the need to manually formulate features, avoiding the problem of incomplete integration of knowledge features and poor accuracy of prediction results due to experience differences.

[0144] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0145] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for predicting drug clinical trial results, characterized in that: include: Obtain clinical trial protocols; extracting information of the target drug and information of the target disease from the clinical trial protocol; According to the information of the target drug, obtaining information of a first characteristic item associated with the target drug and information of an action relationship between the first characteristic item and the target drug from a preset drug map, wherein the first characteristic item is a disease and / or a gene and / or a compound; Converting the information of the target drug, the information of the first feature item, and the information of the interaction relationship between the target drug and the first feature item into a drug feature vector; According to the information of the target disease, obtaining information of a second characteristic item associated with the target disease and information of an action relationship between the second characteristic item and the target disease from the preset drug map, wherein the second characteristic item is a disease and / or a gene and / or a compound; Converting the information of the target disease, the information of the second feature item, and the information of the interaction relationship between the two into a disease feature vector; Converting the clinical trial protocol into a clinical trial protocol feature vector; The clinical trial scheme feature vector, the drug feature vector and the disease feature vector are spliced ​​into a complete clinical prediction feature vector, and the clinical prediction feature vector is input into a trained deep learning model to obtain a predicted value of the success probability of the clinical trial scheme; the deep learning model is a classification model.

2. The method for predicting drug clinical trial results according to claim 1, characterized in that: The step of acquiring information of a first characteristic item associated with the target drug and information of an action relationship between the first characteristic item and the target drug from a preset drug atlas according to the information of the target drug includes: Obtaining information about compounds contained in the target drug; Comparing the information of the compound with the drug map, and finding a first target node in the drug map corresponding to the compound; Finding a first associated node having an action relationship with the first target node from the drug map through the first target node; A first characteristic item corresponding to the first associated node and an action relationship between the first characteristic item and the compound are obtained.

3. The method for predicting drug clinical trial results according to claim 1, characterized in that: The step of acquiring information of a second characteristic item associated with the target disease and information of an action relationship between the second characteristic item and the target disease from a preset drug map according to information of the target disease includes: Comparing the information of the target disease with the drug map, and finding a second target node in the drug map corresponding to the target disease; Finding a second associated node having an action relationship with the second target node from the drug map through the second target node; A second feature item corresponding to the second association node and an action relationship between the second feature item and the target disease are obtained.

4. The method for predicting drug clinical trial results according to claim 1, characterized in that: The converting the information of the target drug, the information of the first feature item, and the information of the interaction relationship between the target drug and the first feature item into a drug feature vector includes: Inputting information about the target drug, information about the first feature item, and information about the interaction relationship between the target drug and the first feature item into a trained knowledge graph embedding model; The information of the target drug, the information of the first feature item and the information of the action relationship between the two are converted into a drug feature vector in matrix form through the knowledge graph embedding model.

5. The method for predicting drug clinical trial results according to claim 1, characterized in that: The converting the information of the target disease, the information of the second feature item, and the information of the interaction relationship between the two into a disease feature vector includes: Inputting information about the target disease, information about the second feature item, and information about the interaction relationship between the target disease and the second feature item into a trained knowledge graph embedding model; The information of the target disease, the information of the second feature item and the information of the interaction relationship between the two are converted into a disease feature vector in matrix form through the knowledge graph embedding model.

6. The method for predicting drug clinical trial results according to claim 1, characterized in that: The converting the clinical trial protocol into a clinical trial protocol feature vector comprises: inputting the clinical trial protocol into a trained natural language processing model; The clinical trial protocol is converted into a clinical trial protocol feature vector in matrix form through the natural language processing model.

7. The method for predicting drug clinical trial results according to claim 6, characterized in that: The converting the clinical trial protocol into a clinical trial protocol feature vector comprises: Extracting the condition information contained in the clinical trial protocol; the condition information at least includes the condition information of the enrolled population, the compound component information of the target drug, and the expected therapeutic effect; Inputting the condition information into the natural language processing model; The condition information of the clinical trial protocol is converted into a clinical trial protocol feature vector in matrix form through the natural language processing model.

8. A device for predicting drug clinical trial results, characterized in that: include: Protocol acquisition module, used to obtain clinical trial protocols; An information extraction module, used to extract information about target drugs and target diseases from the clinical trial protocol; A first associated information acquisition module, used to acquire information of a first characteristic item associated with the target drug and information of an action relationship between the first characteristic item and the target drug from a preset drug map according to the information of the target drug, wherein the first characteristic item is a disease and / or a gene and / or a compound; A first conversion module, used to convert the information of the target drug, the information of the first feature item and the information of the interaction relationship between the two into a drug feature vector; A second associated information acquisition module, used to acquire information of a second characteristic item associated with the target disease and information of an action relationship between the second characteristic item and the target disease from a preset drug map according to information of the target disease, wherein the second characteristic item is a disease and / or a gene and / or a compound; A second conversion module, used to convert the information of the target disease, the information of the second feature item and the information of the interaction relationship between the two into a disease feature vector; A third conversion module, used to convert the clinical trial protocol into a clinical trial protocol feature vector; A prediction module is used to splice the clinical trial scheme feature vector, the drug feature vector and the disease feature vector into a complete clinical prediction feature vector, and input the clinical prediction feature vector into a trained deep learning model to obtain a predicted value of the success probability of the clinical trial scheme.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images