Drug target extraction method and system based on large model

Through a drug target extraction system based on a large model, the target information in drug patent documents is automatically identified and matched, and combined with model iteration technology, the problems of low extraction efficiency and poor accuracy in the existing technology are solved, achieving efficient and accurate Target information extraction.

CN119028479APending Publication Date: 2024-11-26SHANDONG LANGCHAO YUNTOU INFORMATION TECH CO LTD

Patent Information

Application Number
CN202411049444.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-01
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

Existing methods for extracting drug target information by manually reading patent documents are inefficient and difficult to accurately understand the context, resulting in a high threshold for extraction of target information and a low accuracy rate, especially when facing massive data.

Method used

Using drug target extraction methods and systems based on large models, the target information in text is automatically identified and extracted through large models, and the existing target library is used for matching. Combined with model iteration technology, the extraction accuracy is continuously optimized, and the extraction threshold is lowered. .

Benefits of technology

It greatly accelerates the speed of target recognition, improves the accuracy of drug target extraction, uses big models to learn implicit patterns and rules, and enhances the predictive ability of drug-target interactions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119028479A_ABST
    Figure CN119028479A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of large models, in particular to a drug target extraction method and system based on a large model, and the method comprises the following steps: extracting drug target information; target spot matching: matching target spot data extracted by the large model by means of an existing target spot library, and distinguishing existing target spot information and unknown target spot information; performing model iteration, and regularly performing model fine adjustment on the large model according to indexes in object storage; the large model-based drug target extraction method and system provided by the invention have the beneficial effects that the target information related to the drug action is automatically identified from the literature, and the target identification speed is greatly accelerated. Implicit modes and laws can be learned from complex data by using a large model, and the modes exceed intuitive understanding of common business personnel.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of large models, and in particular to a drug target extraction method and system based on large models. Background Art

[0002] With the development of drug technology, it is becoming increasingly important to obtain target information from existing drug patents. The existing method of extracting target information by manually reading relevant patent documents requires relevant personnel to have a deep knowledge background and understand the relevant characteristics of the target in order to extract relevant targets. At the same time, faced with the massive and increasing number of patent documents, the manual method is very inefficient. Using the retrieval method, by searching for known targets in the document, it is impossible to understand based on the context, the search effect is not ideal, and the target information is also increasing with drug development.

[0003] Therefore, there is an urgent need for a method that can lower the threshold for extracting target information and improve the accuracy of target extraction. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for extracting drug targets based on a large model to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for extracting drug targets based on a large model, the method comprising the following steps:

[0006] Drug target information extraction;

[0007] Target matching: Match the target data extracted from the large model with the help of the existing target library to distinguish between existing targets and unknown target information;

[0008] Model iteration: Regularly fine-tune large models based on indicators in object storage.

[0009] Preferably, the specific operations of extracting drug target information include:

[0010] System deployment of large model applications;

[0011] The business personnel clicks "Upload File" to upload the drug patent document to the system, and the system immediately starts the document parsing operation;

[0012] The big model automatically identifies and extracts the implicit target information from the parsed text fragments, and the front end displays the target extraction progress information, with the status being extracted;

[0013] The system matches the targets returned by the large model with the target library data;

[0014] After the system matching is completed, the page extraction progress status will be displayed as successful extraction. Select View Results.

[0015] Business personnel check the target information to be confirmed and confirm the accuracy of the data;

[0016] After receiving the notification email, the business expert logs into the system to review the target operations modified by the business personnel.

[0017] Preferably, the specific operations of target matching include:

[0018] Business personnel pull target data;

[0019] According to the full name of the target in the target library, write a prompt and call the large model interface to perform matching operations;

[0020] According to the target abbreviation in the target library, write prompt and call the large model interface to perform matching operations;

[0021] According to the target alias in the target library, write a prompt and call the large model interface to perform matching operations.

[0022] Preferably, the specific operations of model iteration include:

[0023] Personnel with authority can pull target data;

[0024] Pull target data according to requirements and build training sets, validation sets, and test sets;

[0025] Set the training strategy, including but not limited to learning rate and batch size parameters;

[0026] Set the training set to iteratively update the model parameters and perform model fine-tuning training operations;

[0027] Use the test set to evaluate the functionality of the fine-tuned model to ensure that the fine-tuned model achieves the expected results on the new task;

[0028] Deploy the tuned model to the production environment.

[0029] A drug target extraction system based on a large model, the system consists of an information extraction module, a target matching module and a model iteration module;

[0030] Information extraction module, drug target information extraction;

[0031] The target matching module matches the target data extracted by the large model with the help of the existing target library to distinguish the existing target and unknown target information;

[0032] The model iteration module regularly fine-tunes the large model based on the indicators in the object storage.

[0033] Preferably, the information extraction module system deploys a large model application;

[0034] The business personnel clicks "Upload File" to upload the drug patent document to the system, and the system immediately starts the document parsing operation;

[0035] The big model automatically identifies and extracts the implicit target information from the parsed text fragments, and the front end displays the target extraction progress information, with the status being extracted;

[0036] The system matches the targets returned by the large model with the target library data;

[0037] After the system matching is completed, the page extraction progress status will be displayed as successful extraction. Select View Results.

[0038] Business personnel check the target information to be confirmed and confirm the accuracy of the data;

[0039] After receiving the notification email, the business expert logs into the system to review the target operations modified by the business personnel.

[0040] Preferably, in the target matching module, business personnel pull target data;

[0041] According to the full name of the target in the target library, write a prompt and call the large model interface to perform matching operations;

[0042] According to the target abbreviation in the target library, write prompt and call the large model interface to perform matching operations;

[0043] According to the target alias in the target library, write a prompt and call the large model interface to perform matching operations.

[0044] Preferably, in the model iteration module, personnel with authority pull target data;

[0045] Pull target data according to requirements and build training sets, validation sets, and test sets;

[0046] Set the training strategy, including but not limited to learning rate and batch size parameters;

[0047] Set the training set to iteratively update the model parameters and perform model fine-tuning training operations;

[0048] Use the test set to evaluate the functionality of the fine-tuned model to ensure that the fine-tuned model achieves the expected results on the new task;

[0049] Deploy the tuned model to the production environment.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] The drug target extraction method and system based on a large model proposed in the present invention automatically identifies target information related to drug action from literature, greatly accelerating the speed of target identification. Using a large model, implicit patterns and laws can be learned from complex data, and these patterns are beyond the intuitive understanding of ordinary business personnel. With the increase of training data, the generalization ability of the model is continuously enhanced, so that the interaction between drugs and targets can be predicted more accurately, further improving the accuracy of drug target extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a flow chart of drug target information extraction of the present invention;

[0053] Figure 2 This is a target matching flow chart of the present invention;

[0054] Figure 3 It is a flow chart of model iteration of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose and technical solution of the present invention clearly and completely described, and the advantages more clearly understood, the embodiments of the present invention are further described in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are part of the embodiments of the present invention, rather than all of the embodiments, and are only used to explain the embodiments of the present invention, and are not used to limit the embodiments of the present invention. All other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] Example 1: Please refer to Figures 1 to 3 The present invention provides a technical solution: a method for extracting drug targets based on a large model, the method comprising the following steps:

[0057] 1. In order to achieve efficient and convenient drug target information extraction, the following is a detailed operation process:

[0058] 1. The system has deployed a large model application.

[0059] 2. The business personnel clicks "Upload File" to upload the drug patent document to the system. The system then starts the document parsing operation.

[0060] 1) Document parsing failed and the process ends.

[0061] 2) Document parsing is successful, proceed to the next step.

[0062] The big model automatically identifies and extracts the implicit target information based on the parsed text fragments. The front end displays the target extraction progress information, and the status is extracting.

[0063] 3. The system matches the targets returned by the large model with the target library data. For the specific process, see

[0064] 4.2 Target matching process.

[0065] 4. After the system matching is completed, the page extraction progress status will be displayed as successful extraction. Select View Results.

[0066] 1) If the target extracted by the system matches an existing target, the target status is displayed as normal and the process ends.

[0067] 2) If the target extracted by the system does not match the existing target, the target status will be displayed as pending confirmation.

[0068] 5. Business personnel check the target information to be confirmed and confirm the accuracy of the data.

[0069] 1) If the target is extracted correctly, save the target information to the object storage.

[0070] 2) If there is an error in target extraction, modify the target information, submit it to the business expert for review, and notify by email.

[0071] 6. After receiving the notification email, the business expert logs into the system to review the target operations modified by the business personnel.

[0072] 1) If the review is passed, save the target information to the object storage.

[0073] 2) If the review fails, enter the corresponding comments and the process returns to step 5.

[0074] For the overall flow chart, please refer to Figure 1: Overall process description

[0075] 2. Target matching process:

[0076] In order to deal with the problem of inaccurate target information extraction due to large model hallucination, the target data extracted by the large model is matched with the existing target library to distinguish existing targets and unknown target information. The following is the corresponding process information.

[0077] 1. Business personnel pull target data.

[0078] 1) Failed to obtain data in the target library, the process ends

[0079] 2) The target library data is successfully obtained and the process proceeds to the next step.

[0080] 2. According to the full name of the target in the target library, write prompt and call the large model interface for matching

[0081] Operation. The prompt example is as follows:

[0082] Given that "1. The full name and abbreviation of the same target actually refer to the same target; 2. The full name and alias of the same target actually refer to the same target; 3. The full name and alias of the same target actually refer to the same target", please analyze whether XX and YY are the same target. If they are the same target, then strictly answer "yes", otherwise strictly answer "no"

[0083] Among them, XX represents the full name of the target in the target library, and YY is the target returned by the large model.

[0084] 1) If the big model returns the result "yes", the match is successful and the process ends

[0085] 2) If the large model returns a "no" result, the process continues

[0086] 3. According to the abbreviation of the target in the target library, write a prompt and call the large model interface to perform matching operations.

[0087] 1) If the big model returns the result "yes", the match is successful and the process ends

[0088] 2) If the large model returns a "no" result, the process continues

[0089] 4. According to the target alias in the target library, write a prompt and call the large model interface to perform matching operations.

[0090] 1) If the large model returns a "yes" result, the match is successful and the process ends.

[0091] 2) If the large model returns a "no" result, the match fails and the process ends.

[0092] 3. Model iteration:

[0093] To perform adaptive optimization, large models are regularly fine-tuned based on indicators in object storage.

[0094] 1. Authorized personnel pull target data.

[0095] 2. Pull the target data according to the requirements and build the training set, validation set and test set.

[0096] 3. Set the training strategy, including but not limited to parameters such as learning rate and batch size

[0097] 4. Set the training set to iteratively update the model parameters and perform model fine-tuning training operations.

[0098] 5. Use the test set to evaluate the functionality of the fine-tuned model to ensure that the fine-tuned model achieves the expected results on the new task.

[0099] 1) If the model performance meets the expected target, it indicates that the fine-tuning is successful and the process can be continued to the next stage.

[0100] 2) If the model performance does not meet expectations, go to step 3 and perform fine-tuning again until the expected tuning target is met.

[0101] 6. Deploy the tuned model to the production environment.

[0102] Embodiment 2

[0103] Based on the first embodiment, a drug target extraction system based on a large model is proposed, the system is composed of an information extraction module, a target matching module and a model iteration module;

[0104] Information extraction module, drug target information extraction; system deployment large model application;

[0105] The business personnel clicks "Upload File" to upload the drug patent document to the system, and the system immediately starts the document parsing operation;

[0106] The big model automatically identifies and extracts the implicit target information from the parsed text fragments, and the front end displays the target extraction progress information, with the status being extracted;

[0107] The system matches the targets returned by the large model with the target library data;

[0108] After the system matching is completed, the page extraction progress status will be displayed as successful extraction. Select View Results.

[0109] Business personnel check the target information to be confirmed and confirm the accuracy of the data;

[0110] After receiving the notification email, the business expert logs into the system to review the target operations modified by the business personnel.

[0111] The target matching module matches the target data extracted from the large model with the help of the existing target library to distinguish between existing targets and unknown target information; business personnel pull target data;

[0112] According to the full name of the target in the target library, write a prompt and call the large model interface to perform matching operations;

[0113] According to the target abbreviation in the target library, write prompt and call the large model interface to perform matching operations;

[0114] According to the target alias in the target library, write a prompt and call the large model interface to perform matching operations.

[0115] The model iteration module regularly fine-tunes the large model based on the indicators in the object storage; authorized personnel pull target data;

[0116] Pull target data according to requirements and build training sets, validation sets, and test sets;

[0117] Set the training strategy, including but not limited to learning rate and batch size parameters;

[0118] Set the training set to iteratively update the model parameters and perform model fine-tuning training operations;

[0119] Use the test set to evaluate the functionality of the fine-tuned model to ensure that the fine-tuned model achieves the expected results on the new task;

[0120] Deploy the tuned model to the production environment.

[0121] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for extracting drug targets based on a large model, characterized in that: The method comprises the following steps: Drug target information extraction; Target matching: Match the target data extracted from the large model with the help of the existing target library to distinguish between existing targets and unknown target information; Model iteration: Regularly fine-tune large models based on indicators in object storage.

2. The method for extracting drug targets based on a large model according to claim 1, characterized in that: The specific operations of drug target information extraction include: System deployment of large model applications; The business personnel clicks "Upload File" to upload the drug patent document to the system, and the system immediately starts the document parsing operation; The big model automatically identifies and extracts the implicit target information from the parsed text fragments, and the front end displays the target extraction progress information, with the status being extracted; The system matches the targets returned by the large model with the target library data; After the system matching is completed, the page extraction progress status will be displayed as successful extraction. Select View Results. Business personnel check the target information to be confirmed and confirm the accuracy of the data; After receiving the notification email, the business expert logs into the system to review the target operations modified by the business personnel.

3. A method for extracting drug targets based on a large model according to claim 2, characterized in that: The specific operations of target matching include: Business personnel pull target data; According to the full name of the target in the target library, write a prompt and call the large model interface to perform matching operations; According to the target abbreviation in the target library, write prompt and call the large model interface to perform matching operations; According to the target alias in the target library, write a prompt and call the large model interface to perform matching operations.

4. The method for extracting drug targets based on a large model according to claim 1, characterized in that: The specific operations of model iteration include: Personnel with authority can pull target data; Pull target data according to requirements and build training sets, validation sets, and test sets; Set the training strategy, including but not limited to learning rate and batch size parameters; Set the training set to iteratively update the model parameters and perform model fine-tuning training operations; Use the test set to evaluate the functionality of the fine-tuned model to ensure that the fine-tuned model achieves the expected results on the new task; Deploy the tuned model to the production environment.

5. A drug target extraction system based on a large model according to the drug target extraction method based on a large model according to any one of claims 1 to 4, characterized in that: The system consists of an information extraction module, a target matching module and a model iteration module; Information extraction module, drug target information extraction; The target matching module matches the target data extracted by the large model with the help of the existing target library to distinguish the existing target and unknown target information; The model iteration module regularly fine-tunes the large model based on the indicators in the object storage.

6. A drug target extraction system based on a large model according to claim 5, characterized in that: The information extraction module, the system deploys a large model application; The business personnel clicks "Upload File" to upload the drug patent document to the system, and the system immediately starts the document parsing operation; The big model automatically identifies and extracts the implicit target information from the parsed text fragments, and the front end displays the target extraction progress information, with the status being extracted; The system matches the targets returned by the large model with the target library data; After the system matching is completed, the page extraction progress status will be displayed as successful extraction. Select View Results. Business personnel check the target information to be confirmed and confirm the accuracy of the data; After receiving the notification email, the business expert logs into the system to review the target operations modified by the business personnel.

7. A drug target extraction system based on a large model according to claim 6, characterized in that: In the target matching module, business personnel pull target data; According to the full name of the target in the target library, write a prompt and call the large model interface to perform matching operations; According to the target abbreviation in the target library, write prompt and call the large model interface to perform matching operations; According to the target alias in the target library, write a prompt and call the large model interface to perform matching operations.

8. A drug target extraction system based on a large model according to claim 7, characterized in that: The model iteration module allows authorized personnel to pull target data; Pull target data according to requirements and build training sets, validation sets, and test sets; Set the training strategy, including but not limited to learning rate and batch size parameters; Set the training set to iteratively update the model parameters and perform model fine-tuning training operations; Use the test set to evaluate the functionality of the fine-tuned model to ensure that the fine-tuned model achieves the expected results on the new task; Deploy the tuned model to the production environment.

Citation Information

Patent Citations

  • AI-based drug target determination method and device, equipment and storage medium

    CN114596914A

  • Drug target prediction method and device and computer readable storage medium

    CN115132268A

  • Molecular large model based on multi-dimensional molecular information, construction method and application

    CN117524353A

Cited By

  • Target spot extraction method and device based on large model, medium and equipment

    CN119621937A