A Management System and Method for Generating IND Application Materials Based on a Large Model

Abstract

This invention discloses an IND application material generation management system and method based on a large model, relating to the field of application material generation management technology. The system includes: evaluating the reference value of standard application materials in a standard material database for IND application materials; analyzing the functional correlation between functional blocks in IND application materials and standard functional blocks in reference standard application materials; assessing the correlation between different related functional blocks in a related functional block set, generating a target correlation range; analyzing the approval reference value of related functional blocks for functional blocks in IND application materials and related functional blocks in a related functional set under different correlation scales; diagnosing abnormal states of functional blocks, identifying abnormal functional blocks, and marking abnormal functional blocks in IND application materials on the application material generation platform, thereby improving the efficiency of abnormal detection and ensuring the quality of IND application materials.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of declaration material generation management, and particularly relates to an IND declaration material generation management system and method based on a large model. BACKGROUND

[0002] IND is the Chinese meaning of "new drug clinical research application", and IND declaration material refers to the application materials submitted by a drug research and development institution to a drug regulatory agency before the drug is applied to start clinical experiments. The IND declaration material needs to integrate a large amount of research data, and therefore, the generation of the IND declaration material is a very complex and time-consuming task. The traditional declaration material method has the problems of scattered data and low efficiency, and cannot guarantee the consistency and accuracy of the information of the declaration material. The use of the large model technology can effectively integrate data from different sources, thereby quickly generating the declaration material, and the large model technology can effectively guarantee the consistency of the data and avoid the consistency problems caused by the manual generation of the material in the traditional method.

[0003] At present, the technical scheme for generating the IND declaration material using the large model has gradually matured, but the IND declaration material needs to be approved by different approval agencies in different regions in the actual approval process, different approval agencies have different requirements for the IND declaration material, and normal IND declaration material in one approval agency may be judged as abnormal in another approval agency, and the attributes of the drug corresponding to the IND declaration material also have an impact. These will cause problems in the abnormal detection of the IND declaration material, thereby affecting the final declaration progress of the IND declaration material, and even cause the IND declaration material to fail to be successfully declared, resulting in huge losses. SUMMARY

[0004] The present application aims to provide an IND declaration material generation management system and method based on a large model to solve the problems in the prior art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: an IND declaration material generation management method based on a large model, the method comprising:

[0006] Step S1: obtaining a constructed declaration material generation platform, using the declaration material generation platform to generate declaration material for a declared drug, obtaining IND declaration material, obtaining a standard material database of an approval agency for the declaration of the IND declaration material, evaluating the reference value of the standard declaration material in the standard material database to the IND declaration material, and obtaining a reference standard declaration material;

[0007] Step S2: Obtain the reference standard application materials for the IND application materials, analyze the functional correlation between the functional blocks of the IND application materials and the standard functional blocks of the reference standard application materials, and obtain the relevant functional block set;

[0008] Step S3: Obtain the relevant functional block set of the functional block, evaluate the correlation between different relevant functional blocks in the relevant functional block set, generate the target correlation range, and based on the target correlation range, analyze the reference value of the relevant functional blocks for the approval of the functional block in the IND application materials and the relevant functional blocks in the relevant functional set under different correlation scales, and obtain the reference functional block set of the functional block;

[0009] Step S4: Based on the reference function block set and related function block set, diagnose the abnormal status of the function blocks, obtain the abnormal function blocks, and mark the abnormal function blocks in the IND application materials on the application material generation platform to assist the reviewers in the platform to review the IND application materials.

[0010] Furthermore, step S1 includes:

[0011] Step S11: Obtain the pre-built application material generation platform, use the application material generation platform to generate application materials for the drug according to the preset rules, and obtain the IND application materials corresponding to the drug.

[0012] Step S12: Obtain the standard materials database of the approval agency for IND application materials, wherein the standard materials database contains various standard application materials approved by the approval agency;

[0013] Step S13: Evaluate the reference value of the standard application materials in the standard materials database for the IND application materials. The specific evaluation process is as follows:

[0014] Obtain the drug attribute data for the drugs submitted in the standard application materials and the IND application materials, respectively. The drug attribute data includes data for each drug attribute.

[0015] Based on the drug attribute data of the drug submitted in the IND application materials, a drug feature vector A is constructed for the IND application materials, and the drug reference value R of the standard application materials to the IND application materials is calculated. A ;

[0016] Obtain the research type sets F and F' of the IND application materials and standard application materials respectively. The research type set F includes several research types involved in the IND application materials.

[0017] Get the total number of identical research types between research type set F and research type set F'. △sum The calculation of standard application materials has significant reference value for IND application materials. F =(F △ sum / F sum )×(F △ sum / F´ sum ), where F sum Let F' be the total number of research types contained in the research type set F. sum The total number of research types contained in the research type set F';

[0018] Obtain the preset research reference coefficient η F and drug reference coefficient η A Calculate the research reference coefficient η F Multiply by the research reference value R F The value plus the drug reference coefficient η A Multiply by the drug reference value R A The value of α is used to determine the reference value of standard application materials for IND application materials.

[0019] When the reference value α is greater than the preset reference threshold, the standard application materials are determined to have reference value for the IND application materials, and the standard application materials are recorded as reference standard application materials.

[0020] Furthermore, step S2 includes:

[0021] Step S21: Obtain the reference standard application materials for the IND application materials from the standard materials database of the approval agency;

[0022] Use the application material generation platform to divide the IND application materials into various functional blocks;

[0023] Use the application material generation platform to obtain several standard function blocks corresponding to the application materials of each reference standard;

[0024] Step S22: Obtain the structural data of the functional blocks in the IND application materials and several standard functional blocks in the reference standard application materials; extract the standard functional blocks with similar structures to the functional blocks in the IND application materials from the several functional blocks in the reference standard application materials and mark them.

[0025] The correlation between the functional blocks in the IND application materials and several standard functional blocks marked in the reference standard application materials was analyzed. The specific analysis process is as follows:

[0026] The document content of the functional blocks in the IND application materials is obtained, and the document content of the functional blocks is preprocessed. The application material generation platform is used to obtain the standard functional blocks that have been marked after preprocessing and the semantic vectors between the functional blocks.

[0027] Calculate the cosine similarity between the functional block and the standard functional block to obtain the semantic relevance value E between the functional block and the standard functional block;

[0028] Step S23: Use the application material generation platform to obtain the entity sets L and L' of the functional blocks and standard functional blocks;

[0029] Calculate the entity dependency value H between the function block and the standard function block: H = (L∩L´) / (L∪L´);

[0030] Step S24: Use the application material generation platform to obtain the datasets of functional blocks and standard functional blocks;

[0031] Get the total number m of elements in the dataset that have the same index as the standard function block. sum Get the total number M of all elements in the dataset of the function block and the standard function block. sum Get a certain indicator that is present in both the datasets of the function block and the standard function block. Get the values ​​V and V´ of the certain indicator in the datasets of the function block and the standard function block respectively. Calculate the numerical difference U=|VV´| / max{V,V´} between the function block and the standard function block on a certain indicator.

[0032] Get the data difference U between the function block and the standard function block on the same number of metrics. △ Calculate the data correlation value G between the function block and the standard function block = (m sum / M sum )×(1-U);

[0033] Step S25: Normalize the semantic relevance value E, entity relevance value H, and data relevance value G respectively, and calculate the functional relevance value P=ζ between the functional block and the standard functional block. E ×E+ζ H ×H+ζ G ×G, where ζ E ζ H and ζ G These are the preset semantic relevance coefficient, entity relevance coefficient, and data relevance coefficient;

[0034] Obtain the maximum value of the function-related values ​​between the function block and several standard function blocks marked in the reference standard application materials. When the maximum value is greater than the preset maximum threshold, the standard function block corresponding to the maximum value is recorded as the related function block of the function block.

[0035] Step S26: Obtain the reference standard application materials for each IND application, extract the relevant functional blocks from each reference standard application material and aggregate them to obtain the relevant functional block set.

[0036] Furthermore, step S3 includes:

[0037] Step S31: Obtain the relevant functional block set of the functional block in the IND application materials, and use the semantic correlation value, entity correlation value and data correlation value between the functional block and the relevant functional block in the relevant functional block set as various correlation indicators between the functional block and the relevant functional block;

[0038] Step S32: Obtain the values ​​of various related indicators among the various related functional blocks in the relevant functional block set, obtain the maximum and minimum values ​​of a certain related indicator among the various related functional blocks, and construct the numerical range W of a certain related indicator among the various related functional blocks;

[0039] Step S33: Obtain the median μ of the numerical range W, and obtain the standard deviation σ of a certain related index among the relevant functional blocks;

[0040] Centered on the median μ, construct the range Q of the j-th characteristic values ​​of a certain related indicator. j =[μ-j×σ,μ+j×σ], when the value of a certain correlation index between one related functional block and another related functional block is in Q j If it is inside, then determine Q. j There is a one-time inclusion relationship for a certain related indicator;

[0041] Get the total number N of a certain related indicator among the various related functional blocks. sum Get Q j The total number of times N exists that a certain related indicator has an inclusion relationship （j,sum） Calculate Q j The inclusion ratio Z j =N （j,sum） / N sum ;

[0042] When the proportion value Z is included j If Q is the smallest feature data range among several feature data ranges that are greater than a preset inclusion ratio threshold, then Q will be... j The target relevant range Q of a certain related indicator;

[0043] Step S34: Obtain the target relevant range of each relevant indicator. When the values ​​of each relevant indicator between the ε-th relevant functional block in the relevant functional block set and the functional block are all within the target relevant range of each relevant indicator, it is determined that the ε-th relevant functional block has approval reference value for the functional block, and the ε-th relevant functional block is recorded as the reference functional block of the functional block. Obtain each reference functional block of the functional block and aggregate them to obtain the reference functional block set of the functional block.

[0044] The reason for acquiring the target-related data range in the above steps is that the relevant function sets of the function blocks are all standard function blocks that have been approved by the approval agency, and there is a correlation between these standard function blocks and the function blocks in the IND application materials. Therefore, the more similar the correlation between different related function blocks in the relevant function sets is to the correlation between different related function blocks, the lower the probability of anomalies in the function blocks in the IND application materials. Conversely, the lower the correlation, the higher the probability. This scientific mathematical formula can effectively identify anomalies in the function blocks of the IND application materials according to the approval style of the approval agency, thereby ensuring that the style of the function blocks in the IND application materials meets the actual needs of the approval agency.

[0045] Furthermore, step S4 includes:

[0046] Step S41: Obtain the total number S of reference function blocks in the reference function block set of the function block. ▽ sum Get the total number S of related function blocks. sum Calculate the function block outlier τ = 1 - (S) ▽ sum / S sum );

[0047] Step S42: When the abnormal value τ of a function block is greater than the preset abnormal threshold, the function block is determined to be an abnormal function block, and the abnormal function block in the IND application materials is marked on the application material generation platform to assist the reviewers on the platform in reviewing the IND application materials.

[0048] Based on the above method, an IND application material generation and management system based on a large model can also be proposed. The system includes a reference value module, a function-related analysis module, an approval reference value analysis module, and an anomaly detection module.

[0049] The reference value module is used to obtain the standard materials database of the approval agency that submits the IND application materials, evaluate the reference value of the standard application materials in the standard materials database for the IND application materials, and obtain reference standard application materials.

[0050] The Functional Relevance Analysis module is used to obtain the reference standard application materials for IND application materials, analyze the functional relevance between the functional blocks of the IND application materials and the standard functional blocks of the reference standard application materials, and obtain the relevant functional block set;

[0051] The approval reference value analysis module is used to evaluate the correlation between different related functional blocks in the relevant functional block set, generate the target relevant scope, and analyze the approval reference value of the functional blocks in the IND application materials and the related functional blocks in the relevant functional set under different relevant scales, so as to obtain the reference functional block set of the functional blocks.

[0052] The anomaly detection module is used to diagnose the abnormal status of function blocks based on the reference function block set and related function block sets, identify abnormal function blocks, and mark the abnormal function blocks in the IND application materials on the application material generation platform to assist the reviewers on the platform in reviewing the IND application materials.

[0053] Furthermore, the reference value module includes a reference value calculation unit and a reference value evaluation unit;

[0054] The reference value calculation unit is used to obtain the standard materials database of the approval agency that submits the IND application materials, and to calculate the reference value of the standard application materials in the standard materials database to the IND application materials.

[0055] The reference value assessment unit is used to evaluate the reference value of the standard application materials to the IND application materials, and to obtain the reference standard application materials.

[0056] Furthermore, the functional correlation analysis module includes a data correlation value calculation unit and a functional correlation analysis unit;

[0057] The data correlation value calculation unit is used to obtain the reference standard application materials for the IND application materials, and to obtain the function blocks of the IND application materials and the standard function blocks of the reference standard application materials respectively, and calculate the data correlation value between the function blocks and the standard function blocks.

[0058] The function correlation analysis unit is used to analyze the degree of functional correlation between the function block and the standard function block based on the data correlation values ​​between the function block and the standard function block, and to obtain the relevant function block set.

[0059] Furthermore, the approval reference value analysis module includes a target-related scope generation unit and an approval reference value analysis unit;

[0060] The target-related scope generation unit is used to acquire the relevant function block set of the function block, evaluate the correlation between different relevant function blocks in the relevant function block set, and generate the target-related scope.

[0061] The approval reference value analysis unit is used to analyze the approval reference value of functional blocks in the IND application materials and related functional blocks in the relevant functional sets under different relevant scales, based on target-related data, and to obtain the reference functional block set of functional blocks.

[0062] Furthermore, the anomaly detection module includes an anomaly detection unit;

[0063] The anomaly detection unit is used to acquire the reference function block set and related function block set, diagnose the abnormal state of the function blocks, obtain the abnormal function blocks, and mark the abnormal function blocks in the IND application materials on the application material generation platform.

[0064] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention achieves intelligent management of IND application material generation. Compared with traditional IND application material generation methods, this application considers the different approval styles of different approval agencies in different regions for IND application materials. Therefore, starting from the approval agency itself, from the perspective of the drug application and research direction, it obtains standard application materials with reference value from the approval agency's database. Furthermore, considering that the application materials contain a large amount of data, it obtains functional blocks of the application materials according to the standard hierarchical structure and semantics of the application materials, and obtains relevant functional blocks related to the functional blocks of the application materials from the reference standard application materials, thereby obtaining a relevant functional set. Based on the relationship between different relevant functional blocks in the relevant functional set, a specific target relevant range is constructed, thereby filtering the relevant functional blocks in the relevant functional set to obtain reference functional blocks with reference value. Based on the ratio between reference functional blocks and relevant functional blocks, the anomalies of functional blocks are judged, thereby achieving accurate detection of functional block anomalies and greatly improving the efficiency of anomaly detection, ensuring the quality of IND application materials. Attached Figure Description

[0065] Fig. 1 This is a method logic diagram of an IND application material generation and management method based on a large model according to the present invention;

[0066] Fig. 2 This is a flowchart of the modules of an IND application material generation and management system based on a large model, according to the present invention. Detailed Implementation

[0067] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] Example: Figs. 1-2 As shown, this invention provides a technical solution: a method for generating and managing IND application materials based on a large model, the method comprising:

[0069] Step S1: Obtain the constructed application material generation platform, use the application material generation platform to generate application materials for the drug to be applied for, obtain IND application materials, obtain the standard material database of the approval agency for IND application materials, evaluate the reference value of the standard application materials in the standard material database for IND application materials, and obtain reference standard application materials;

[0070] For example, the standard materials database stores standard application materials that have been approved by the approval agency;

[0071] Step S1 includes:

[0072] Step S11: Obtain the pre-built application material generation platform, use the application material generation platform to generate application materials for the drug according to the preset rules, and obtain the IND application materials corresponding to the drug.

[0073] Step S12: Obtain the standard materials database of the approval agency for IND application materials, wherein the standard materials database contains various standard application materials approved by the approval agency;

[0074] Step S13: Evaluate the reference value of the standard application materials in the standard materials database for the IND application materials. The specific evaluation process is as follows:

[0075] Obtain the drug attribute data for the drugs submitted in the standard application materials and the IND application materials, respectively. The drug attribute data includes data for each drug attribute.

[0076] For example, various drug properties include drug molecular fingerprint, oil-water partition coefficient, molecular weight, and hydrogen bond donors;

[0077] Based on the drug attribute data of the drug submitted in the IND application materials, a drug feature vector A is constructed for the IND application materials, and the drug reference value R of the standard application materials to the IND application materials is calculated. A ;

[0078] For example, the drug reference value R is specifically calculated as the cosine similarity between the drug feature vectors in the standard application materials and the IND application materials.

[0079] Obtain the research type sets F and F' of the IND application materials and standard application materials respectively. The research type set F includes several research types involved in the IND application materials.

[0080] Get the total number of identical research types between research type set F and research type set F'. △ sum The calculation of standard application materials has significant reference value for IND application materials. F =(F △ sum / F sum )×(F △ sum / F´ sum ), where F sum Let F' be the total number of research types contained in the research type set F. sum The total number of research types contained in the research type set F';

[0081] For example, several types of research can include pharmacological research, toxicological research, etc.

[0082] Obtain the preset research reference coefficient η F and drug reference coefficient η A Calculate the research reference coefficient η F Multiply by the research reference value R F The value plus the drug reference coefficient η A Multiply by the drug reference value R A The value of α is used to determine the reference value of standard application materials for IND application materials.

[0083] When the reference value α is greater than the preset reference threshold, the standard application materials are determined to have reference value for the IND application materials, and the standard application materials are recorded as reference standard application materials.

[0084] Step S2: Obtain the reference standard application materials for the IND application materials, analyze the functional correlation between the functional blocks of the IND application materials and the standard functional blocks of the reference standard application materials, and obtain the relevant functional block set;

[0085] Step S2 includes:

[0086] Step S21: Obtain the reference standard application materials for the IND application materials from the standard materials database of the approval agency;

[0087] Use the application material generation platform to divide the IND application materials into various functional blocks;

[0088] For example, the application material generation platform can divide the specific functional blocks of IND application materials according to the standard hierarchical structure of CTD (Common Technical Document for Registration of Pharmaceuticals for Human Use), and further divide the IND application material documents based on the different semantics on the basis of the standard hierarchical structure of CTD.

[0089] For example, a functional block can be specifically as follows: Module 2: CTD Overview -> 2.6: Toxicology Summary -> 2.6.6: Genetic Toxicity Study Overview -> Background Introduction;

[0090] Use the application material generation platform to obtain several standard function blocks corresponding to the application materials of each reference standard;

[0091] Step S22: Obtain the structural data of the functional blocks in the IND application materials and several standard functional blocks in the reference standard application materials; extract the standard functional blocks with similar structures to the functional blocks in the IND application materials from the several functional blocks in the reference standard application materials and mark them.

[0092] For example, the structural data of the functional blocks in the IND application materials is as follows: Module 2: CTD Overview -> 2.6: Toxicology Summary -> 2.6.6: Overview of Genetic Toxicity Studies -> Background Introduction;

[0093] The text content of the last path in the result structure data of the functional block in the IND application materials is obtained: the background introduction of the overview of the genotoxicity study. The feature vector of the text corresponding to the background introduction of the genotoxicity study is obtained using a preset sentence embedding model, and the structural feature vector of the functional block in the IND application materials is obtained.

[0094] If the cosine similarity between the structural feature vector of a certain standard functional block in the reference standard application materials and the structural feature vector of a functional block in the IND application materials is less than a preset threshold, then it is determined that the functional block structures of a certain standard functional block are similar to those of the standard functional blocks.

[0095] The correlation between the functional blocks in the IND application materials and several standard functional blocks marked in the reference standard application materials was analyzed. The specific analysis process is as follows:

[0096] The document content of the functional blocks in the IND application materials is obtained, and the document content of the functional blocks is preprocessed. The application material generation platform is used to obtain the standard functional blocks that have been marked after preprocessing and the semantic vectors between the functional blocks.

[0097] For example, preprocessing includes text extraction and text cleaning;

[0098] For example, the specific process by which the application material generation platform obtains the semantic vector of a function block is as follows:

[0099] Use a pre-trained sentence embedding model (such as Sentence-BERT) to process all the text within the function block, thereby generating a high-dimensional semantic vector;

[0100] Calculate the cosine similarity between the functional block and the standard functional block to obtain the semantic relevance value E between the functional block and the standard functional block;

[0101] Step S23: Use the application material generation platform to obtain the entity sets L and L' of the functional blocks and standard functional blocks;

[0102] For example, the specific process of obtaining the entity set L of the function block using the application material generation platform is as follows:

[0103] The application material generation platform uses a NER model specific to the field in which the application materials are located and a preset dictionary to identify key terms in the functional blocks, and then aggregates them to obtain the entity set L of the functional blocks;

[0104] Calculate the entity dependency value H between the function block and the standard function block: H = (L∩L´) / (L∪L´);

[0105] Step S24: Use the application material generation platform to obtain the datasets of functional blocks and standard functional blocks;

[0106] The specific process of obtaining the dataset for the functional blocks using the application material generation platform is as follows:

[0107] The application material generation platform uses a preset Named Entity Recognition (NER) model to automatically identify and extract all numerical data and their context from the text in the functional block according to the format of (numerical, unit, indicator).

[0108] For example, if the text content of the function block is "NOAEL value is 500mg / kg", then (500, "mg / kg", "NOAEL") is extracted, and all extracted data are normalized and indexed, and then aggregated to obtain the dataset of the function block.

[0109] Get the total number m of elements in the dataset that have the same index as the standard function block. sum Get the total number M of all elements in the dataset of the function block and the standard function block. sumGet a certain indicator that is present in both the datasets of the function block and the standard function block. Get the values ​​V and V´ of the certain indicator in the datasets of the function block and the standard function block respectively. Calculate the numerical difference U=|VV´| / max{V,V´} between the function block and the standard function block on a certain indicator.

[0110] Get the data difference U between the function block and the standard function block on the same number of metrics. △ Calculate the data correlation value G between the function block and the standard function block = (m sum / M sum )×(1-U);

[0111] Step S25: Normalize the semantic relevance value E, entity relevance value H, and data relevance value G respectively, and calculate the functional relevance value P=ζ between the functional block and the standard functional block. E ×E+ζ H ×H+ζ G ×G, where ζ E ζ H and ζ G These are the preset semantic relevance coefficient, entity relevance coefficient, and data relevance coefficient;

[0112] Obtain the maximum value of the function-related values ​​between the function block and several standard function blocks marked in the reference standard application materials. When the maximum value is greater than the preset maximum threshold, the standard function block corresponding to the maximum value is recorded as the related function block of the function block.

[0113] Step S26: Obtain the reference standard application materials for each IND application, extract the relevant functional blocks from each reference standard application material and aggregate them to obtain the relevant functional block set.

[0114] Step S3: Obtain the relevant functional block set of the functional block, evaluate the correlation between different relevant functional blocks in the relevant functional block set, generate the target correlation range, and based on the target correlation range, analyze the reference value of the relevant functional blocks for the approval of the functional block in the IND application materials and the relevant functional blocks in the relevant functional set under different correlation scales, and obtain the reference functional block set of the functional block;

[0115] Step S3 includes:

[0116] Step S31: Obtain the relevant functional block set of the functional block in the IND application materials, and use the semantic correlation value, entity correlation value and data correlation value between the functional block and the relevant functional block in the relevant functional block set as various correlation indicators between the functional block and the relevant functional block;

[0117] Step S32: Obtain the values ​​of various related indicators among the various related functional blocks in the relevant functional block set, obtain the maximum and minimum values ​​of a certain related indicator among the various related functional blocks, and construct the numerical range W of a certain related indicator among the various related functional blocks;

[0118] Step S33: Obtain the median μ of the numerical range W, and obtain the standard deviation σ of a certain related index among the relevant functional blocks;

[0119] Centered on the median μ, construct the range Q of the j-th characteristic values ​​of a certain related indicator. j =[μ-j×σ,μ+j×σ], when the value of a certain correlation index between one related functional block and another related functional block is in Q j If it is inside, then determine Q. j There is a one-time inclusion relationship for a certain related indicator;

[0120] Get the total number N of a certain related indicator among the various related functional blocks. sum Get Q j The total number of times N exists that a certain related indicator has an inclusion relationship （j,sum） Calculate Q j The inclusion ratio Z j =N （j,sum） / N sum ;

[0121] When the proportion value Z is included j If Q is the smallest feature data range among several feature data ranges that are greater than a preset inclusion ratio threshold, then Q will be... j The target relevant range Q of a certain related indicator;

[0122] Step S34: Obtain the target relevant range of each relevant indicator. When the values ​​of each relevant indicator between the ε-th relevant functional block in the relevant functional block set and the functional block are all within the target relevant range corresponding to each relevant indicator, it is determined that the ε-th relevant functional block has approval reference value for the functional block, and the ε-th relevant functional block is recorded as the reference functional block of the functional block. Obtain each reference functional block of the functional block and aggregate them to obtain the reference functional block set of the functional block.

[0123] Step S4: Based on the reference function block set and related function block set, diagnose the abnormal status of the function blocks, obtain the abnormal function blocks, and mark the abnormal function blocks in the IND application materials on the application material generation platform to assist the reviewers in the platform to review the IND application materials.

[0124] Step S4 includes:

[0125] Step S41: Obtain the total number S of reference function blocks in the reference function block set of the function block. ▽ sum Get the total number S of related function blocks. sum Calculate the function block outlier τ = 1 - (S) ▽ sum / S sum );

[0126] Step S42: When the abnormal value τ of a function block is greater than the preset abnormal threshold, the function block is determined to be an abnormal function block, and the abnormal function block in the IND application materials is marked on the application material generation platform to assist the reviewers on the platform in reviewing the IND application materials.

[0127] Based on the above method, an IND application material generation and management system based on a large model can also be proposed. The system includes a reference value module, a function-related analysis module, an approval reference value analysis module, and an anomaly detection module.

[0128] The reference value module is used to obtain the standard materials database of the approval agency that submits the IND application materials, evaluate the reference value of the standard application materials in the standard materials database for the IND application materials, and obtain reference standard application materials.

[0129] The Functional Relevance Analysis module is used to obtain the reference standard application materials for IND application materials, analyze the functional relevance between the functional blocks of the IND application materials and the standard functional blocks of the reference standard application materials, and obtain the relevant functional block set;

[0130] The approval reference value analysis module is used to evaluate the correlation between different related functional blocks in the relevant functional block set, generate the target relevant scope, and analyze the approval reference value of the functional blocks in the IND application materials and the related functional blocks in the relevant functional set under different relevant scales, so as to obtain the reference functional block set of the functional blocks.

[0131] The anomaly detection module is used to diagnose the abnormal status of function blocks based on the reference function block set and related function block sets, identify abnormal function blocks, and mark the abnormal function blocks in the IND application materials on the application material generation platform to assist the reviewers on the platform in reviewing the IND application materials.

[0132] The reference value module includes a reference value calculation unit and a reference value evaluation unit.

[0133] The reference value calculation unit is used to obtain the standard materials database of the approval agency that submits the IND application materials, and to calculate the reference value of the standard application materials in the standard materials database to the IND application materials.

[0134] The reference value assessment unit is used to evaluate the reference value of the standard application materials to the IND application materials, and to obtain the reference standard application materials.

[0135] The functional correlation analysis module includes a data correlation value calculation unit and a functional correlation analysis unit.

[0136] The data correlation value calculation unit is used to obtain the reference standard application materials for the IND application materials, and to obtain the function blocks of the IND application materials and the standard function blocks of the reference standard application materials respectively, and calculate the data correlation value between the function blocks and the standard function blocks.

[0137] The function correlation analysis unit is used to analyze the degree of functional correlation between the function block and the standard function block based on the data correlation values ​​between the function block and the standard function block, and to obtain the relevant function block set.

[0138] The approval reference value analysis module includes a target-related scope generation unit and an approval reference value analysis unit.

[0139] The target-related scope generation unit is used to acquire the relevant function block set of the function block, evaluate the correlation between different relevant function blocks in the relevant function block set, and generate the target-related scope.

[0140] The approval reference value analysis unit is used to analyze the approval reference value of functional blocks in the IND application materials and related functional blocks in the relevant functional sets under different relevant scales, based on target-related data, and to obtain the reference functional block set of functional blocks.

[0141] The anomaly detection module includes an anomaly detection unit;

[0142] The anomaly detection unit is used to acquire the reference function block set and related function block set, diagnose the abnormal state of the function blocks, obtain the abnormal function blocks, and mark the abnormal function blocks in the IND application materials on the application material generation platform.

[0143] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for IND declaration material generation management based on a large model, characterized in that, The method comprises: Step S1: obtaining a constructed declaration material generation platform, using the declaration material generation platform to generate declaration materials for declared drugs, obtaining IND declaration materials, obtaining a standard material database of an approval agency for the IND declaration materials, evaluating the reference value of standard declaration materials in the standard material database to the IND declaration materials, and obtaining reference standard declaration materials; Step S2: obtaining the reference standard declaration materials of the IND declaration materials, analyzing the functional correlation between the functional blocks of the IND declaration materials and the standard functional blocks of the reference standard declaration materials, and obtaining a relevant functional block set; Step S3: obtaining the relevant functional block set of the functional blocks, evaluating the correlation between different relevant functional blocks in the relevant functional block set, generating a target correlation range, and analyzing the approval reference value of the relevant functional blocks in the relevant functional block set to the functional blocks under different correlation scales, obtaining a reference functional block set of the functional blocks; Step S4: diagnosing the abnormal state of the functional blocks according to the reference functional block set and the relevant functional block set, obtaining abnormal functional blocks, and marking the abnormal functional blocks in the IND declaration materials on the declaration material generation platform to assist the review personnel in the platform in reviewing the IND declaration materials; The step S2 comprises: Step S21: obtaining each reference standard declaration material of the IND declaration materials from the standard material database of the approval agency; Using the declaration material generation platform, the IND declaration materials are divided into functional blocks; Using the declaration material generation platform, a plurality of standard functional blocks corresponding to each reference standard declaration material are obtained; Step S22: obtaining the structure data of the functional blocks in the IND declaration materials and the plurality of standard functional blocks of the reference standard declaration materials, obtaining standard functional blocks similar in structure to the functional blocks in the IND declaration materials from the plurality of functional blocks in the reference standard declaration materials and marking them; Analyzing the correlation between the functional blocks in the IND declaration materials and the plurality of standard functional blocks marked in the reference standard declaration materials, the specific analysis process being: Obtaining the document content of the functional blocks in the IND declaration materials, pre-processing the document content of the functional blocks, and using the declaration material generation platform to obtain the semantic vectors between the pre-processed marked standard functional blocks and the functional blocks; Calculating the cosine similarity between the functional blocks and the standard functional blocks to obtain the semantic correlation value E between the functional blocks and the standard functional blocks; Step S23: using the declaration material generation platform to obtain the entity sets L and L' of the functional blocks and the standard functional blocks; Calculating the entity correlation value H=(L∩L´) / (L∪L´) between the functional blocks and the standard functional blocks; Step S24: using the declaration material generation platform to obtain the data sets of the functional blocks and the standard functional blocks; acquiring the total number m of elements in the data set of the function block that are identical to elements in the data set of the standard function block sum acquiring the total number M of elements in the data set of the function block and the data set of the standard function block sum acquiring an index that is contained in the data set of the function block and the data set of the standard function block, respectively acquiring the value V and V' of the index in the data set of the function block and the data set of the standard function block, and calculating the numerical difference U of the function block and the standard function block in the index, U = |V-V'| / max{V,V'} obtaining data differences U of the function block and the standard function block on the same number of indicators △ , calculating data correlation values G between the function block and the standard function block G = (m sum / M sum ) × (1-U) Step S25: normalizing the semantic correlation value E, the entity correlation value H and the data correlation value G respectively, and calculating the function correlation value P between the function block and the standard function block, P = ζ E × E + ζ H × H + ζ G × G, wherein ζ E , ζ H and ζ G are respectively preset semantic correlation coefficient, entity correlation coefficient and data correlation coefficient; obtaining a maximum value of the function correlation value between the function block and a plurality of standard function blocks marked in the reference standard declaration material, when the maximum value is greater than a preset maximum threshold value, recording the standard function block corresponding to the maximum value as a relevant function block of the function block; Step S26: obtaining each reference standard declaration material of the IND declaration material, obtaining each relevant function block of the function block from each reference standard declaration material and collecting to obtain a relevant function block set; The step S3 comprises: Step S31: obtaining the relevant function block set of the function block in the IND declaration material, taking the semantic correlation value, the entity correlation value and the data correlation value between the function block and the relevant function block in the relevant function block set as each correlation index between the function block and the relevant function block; Step S32: obtaining the value of each correlation index between each relevant function block in the relevant function block set, obtaining the maximum value and the minimum value of a certain correlation index between each relevant function block, and constructing the numerical range W of a certain correlation index between each relevant function block; Step S33: obtaining the median μ of the numerical range W, and obtaining the standard deviation σ of a certain correlation index between each relevant function block; A jth characteristic value range Q of the certain correlation index is constructed with a median μ as a center j =[μ-j×σ,μ+j×σ], when a value of a certain correlation index between a certain correlation function block and another correlation function block is in Q j , it is determined that Q j There is a one-time inclusion relationship for a certain correlation index; Obtaining the total number N of a certain correlation index between each related function block sum , obtaining Q j The total number N of the containing relationship of a certain correlation index （j,sum） , calculating the containing proportion value Z of Q j j =N （j,sum） / N sum ;​ When the proportion value Z j is the smallest characteristic data range among several characteristic data ranges with a preset proportion threshold value, the Q j target correlation range Q of the certain relevant index; Step S34: obtaining the target correlation range of each correlation index, when the value of each correlation index between the εth relevant function block in the relevant function block set and the function block is within the target correlation range corresponding to each correlation index, it is determined that the εth relevant function block has approval reference value for the function block, and the εth relevant function block is recorded as the reference function block of the function block, obtaining each reference function block of the function block and collecting to obtain the reference function block set of the function block.

2. The IND declaration material generation management method based on a large model according to claim 1, characterized in that, The step S1 comprises: Step S11: obtaining a pre-constructed declaration material generation platform, using the declaration material generation platform to generate declaration materials for declared drugs according to a preset rule, and obtaining the IND declaration material corresponding to the declared drug; Step S12: obtaining the standard material database of the approval agency of the IND declaration material for material declaration, wherein the standard material database comprises each standard declaration material approved by the approval agency; Step S13: evaluating the reference value of the standard declaration material in the standard material database for the IND declaration material, and the specific evaluation process is: obtaining the drug attribute data of the declared drug of the standard declaration material and the IND declaration material respectively, wherein the drug attribute data comprises the data of each drug attribute; According to the drug attribute data of the drug for which the IND declaration material is declared, a drug feature vector A of the IND declaration material is constructed, and a drug reference value R of the standard declaration material to the IND declaration material is calculated A ; obtaining the research type set F and F' of the IND declaration material and the standard declaration material respectively, wherein the research type set F comprises a plurality of research types involved in the IND declaration material; obtaining the total number F of identical research types existing between the research type set F and the research type set F' △ sum , calculating the research reference value R of the standard declaration material to the IND declaration material F = (F △ sum / F sum ) × (F △ sum / F' sum ), wherein F sum is the total number of research types contained in the research type set F, and F' sum is the total number of research types contained in the research type set F' acquiring a preset research reference coefficient η F and a drug reference coefficient η A , calculating a research reference coefficient η F multiplying the value of the research reference value R F and the value of the drug reference coefficient η A multiplying the value of the drug reference value R A to obtain the reference value α of the standard declaration material to the IND declaration material; When the reference value α is greater than a preset reference threshold value, it is determined that the standard declaration material has reference value for the IND declaration material, and the standard declaration material is recorded as a reference standard declaration material.

3. The IND declaration material generation management method based on a large model according to claim 2, characterized in that, The step S4 comprises: Step S41: obtain the total number S of reference function blocks of the reference function block set of the function block ▽ sum , obtain the total number S of related function blocks of the function block sum , calculate the function block outlier τ of the function block τ = 1 - (S ▽ sum / S sum ); Step S42: when the function block abnormal value τ is greater than the preset abnormal threshold value, it is determined that the function block is an abnormal function block, and the abnormal function block in the IND declaration material is marked on the declaration material generation platform to assist the review personnel in the platform to review the IND declaration material.

4. A large model-based IND declaration material generation management system for performing the large model-based IND declaration material generation management method of any one of claims 1-3, characterized in that, The system comprises a reference value module, a function correlation analysis module, an approval reference value analysis module and an abnormality detection module. The reference value module is configured to acquire a standard material database of an approval institution for the IND declaration material, evaluate reference values of standard declaration materials in the standard material database for the IND declaration material, and obtain reference standard declaration materials. The function correlation analysis module is configured to acquire the reference standard declaration materials of the IND declaration material, analyze a function correlation degree between function blocks of the IND declaration material and standard function blocks of the reference standard declaration materials, and obtain a relevant function block set. The approval reference value analysis module is configured to evaluate a correlation condition between different relevant function blocks in the relevant function block set, generate a target correlation range, analyze approval reference values of the relevant function blocks in the relevant function block set for the function blocks of the IND declaration material under different correlation scales according to the target correlation range, and obtain a reference function block set of the function blocks. The abnormality detection module is configured to diagnose an abnormal state of the function blocks according to the reference function block set and the relevant function block set, obtain abnormal function blocks, mark the abnormal function blocks in the IND declaration material on the declaration material generation platform, and assist the review personnel in the platform to review the IND declaration material.

5. The IND declaration material generation management system based on a large model according to claim 4, characterized by, The reference value module comprises a reference value calculation unit and a reference value evaluation unit. The reference value calculation unit is configured to acquire a standard material database of an approval institution for the IND declaration material, and calculate reference values of standard declaration materials in the standard material database for the IND declaration material. The reference value evaluation unit is configured to evaluate the reference values of the standard declaration materials for the IND declaration material according to the reference values of the standard declaration materials for the IND declaration material, and obtain reference standard declaration materials.

6. The IND declaration material generation management system based on a large model according to claim 4, characterized by, The function correlation analysis module comprises a data correlation value calculation unit and a function correlation analysis unit. The data correlation value calculation unit is configured to acquire the reference standard declaration materials of the IND declaration material, acquire the function blocks of the IND declaration material and the standard function blocks of the reference standard declaration materials respectively, and calculate data correlation values between the function blocks and the standard function blocks. The function correlation analysis unit is configured to analyze a function correlation degree between the function blocks and the standard function blocks according to the data correlation values of the function blocks and the standard function blocks, and obtain the relevant function block set.

7. The IND declaration material generation management system based on a large model according to claim 4, characterized by, The approval reference value analysis module comprises a target correlation range generation unit and an approval reference value analysis unit. The target correlation range generation unit is configured to acquire the relevant function block set of the function blocks, evaluate a correlation condition between different relevant function blocks in the relevant function block set, and generate a target correlation range. The approval reference value analysis unit is configured to analyze approval reference values of the relevant function blocks in the relevant function block set for the function blocks of the IND declaration material under different correlation scales according to the target correlation range, and obtain the reference function block set of the function blocks. The approval reference value analysis unit is configured to analyze, according to target related data, approval reference values of related function blocks in a related function set to a function block in the IND declaration material under different related scales, and obtain a reference function block set of the function block.

8. The IND declaration material generation management system based on a large model according to claim 4, characterized by, The abnormality detection module comprises an abnormality detection unit; The abnormality detection unit is configured to acquire the reference function block set and the related function block set, diagnose an abnormal state of the function block, obtain an abnormal function block, and mark the abnormal function block in the IND declaration material on a declaration material generation platform.

Images