A distributed trusted space integrated system for traditional Chinese medicine drug research and development

Abstract

The application relates to the technical field of traditional Chinese medicine research and development, and discloses a distributed trusted space integrated system for traditional Chinese medicine research and development, which comprises the following units: an infrastructure unit, which builds an alliance chain network and generates digital identity certificates for institutions, users and computing devices; a data access unit, which receives medicinal material, prescription, pharmacological and clinical data of each institution after identity verification, and performs extraction conversion, semantic mapping and quality checking according to a traditional Chinese medicine ontology; an intelligent management unit, which performs permission determination and compliance review according to subject attributes, data attributes and task attributes, and generates a task authorization token after the review; a privacy computing unit, which verifies the task authorization token, performs cross-institution computing according to data distribution and computing types of target data nodes to generate a computing certificate; and a trusted application unit, which outputs a result carrying a confidence degree and evidence sources. The application forms a verification mechanism for each link, and ensures the reliability of the trusted space integration.

Description

Technical Field

[0001] This invention relates to the field of traditional Chinese medicine research and development technology, and more specifically, to a distributed trusted spatial integration system for traditional Chinese medicine drug research and development. Background Technology

[0002] Traditional Chinese medicine (TCM) research and development involves various types of data, including medicinal material properties, prescription composition, chemical components, targets, pharmacological experiments, and clinical efficacy. This data is scattered across research institutions, medical institutions, TCM manufacturers, and testing organizations, making it difficult to directly form a unified dataset for correlation analysis. Knowledge graphs, however, can organize TCM research and development data through entities and relationships between them, and, combined with graph neural networks, uncover potential correlations between medicinal materials, components, targets, prescriptions, and disease symptoms.

[0003] Chinese Patent Publication No. CN113764035A discloses a method for predicting targets of traditional Chinese medicine (TCM) compound prescriptions based on knowledge graphs. This method uses a compound prescription-target relationship graph and a TCM compound prescription knowledge graph as input. It employs knowledge graph embedding technology to obtain embedded representations of entities and relationships, then aggregates neighborhood information using a graph neural network, and uses a collaborative filtering algorithm to output matching scores between the compound prescription and the target. It uses semantic relationships and topological structures in the knowledge graph to predict potential compound prescription-target associations. However, centralized knowledge graphs struggle to reliably access, authorize access to, and collaboratively compute heterogeneous TCM R&D data when data from different institutions cannot be directly shared. Furthermore, when data from different institutions cannot be directly aggregated, it is difficult to form a unified relationship graph usable by graph neural networks, and there is a lack of verification mechanisms for participating entities, computation processes, and association results.

[0004] Therefore, it is necessary to design a distributed, trusted spatial integration system for the research and development of traditional Chinese medicine to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a distributed trusted spatial integration system for the research and development of traditional Chinese medicine, which aims to solve the problems of difficulty in trusted access, authorized call and collaborative computing of heterogeneous traditional Chinese medicine research and development data when data from multiple institutions cannot be directly shared, and difficulty in forming a unified relational graph that can be used by graph neural networks when data from different institutions cannot be directly aggregated, and lack of verification mechanism for participating subjects, calculation process and related results.

[0006] This invention proposes a distributed trusted spatial integration system for the research and development of traditional Chinese medicine, comprising: The infrastructure unit is configured to build a consortium blockchain network and generate digital identity credentials for institutions, users, and computing devices; The data access unit is configured to receive medicinal materials, prescriptions, pharmacological and clinical data from various institutions after identity verification, and to extract, transform, semantically map and verify the data according to the ontology of traditional Chinese medicine to determine a standardized dataset. The intelligent governance unit is configured to receive collaborative analysis tasks and the standardized dataset, perform permission determination and compliance review based on subject attributes, data attributes and task attributes, generate a task authorization token after the review is passed, and send the target data node and the task authorization token. The privacy computing unit is configured to verify the task authorization token and perform cross-agency computation to generate computing credentials based on the data distribution and computation type of the target data node. The confidence application unit is configured to construct a cross-institutional Chinese medicine knowledge graph based on the Chinese medicine ontology, and to mine the association between medicinal materials, components, targets, prescriptions and disease symptoms based on graph neural networks and the computational credentials, and output results carrying confidence level and evidence source.

[0007] Furthermore, the infrastructure unit includes an identity registration node and an identity authentication node; The identity registration node binds the organization's qualifications, user roles, or device public keys with digital identities to generate digital identity credentials. The identity authentication node obtains the credential signature, validity period, and issuing node. If the signature verification is successful, the credential is valid, and the issuing node is a licensed node, then the identity verification is successful; otherwise, access is rejected.

[0008] Furthermore, the TCM entity includes medicinal materials, prescriptions, components, targets, pharmacological effects, disease symptoms, and clinical efficacy; The data access unit extracts the fields, data types, and units of measurement of the data, matches the fields with the attributes of traditional Chinese medicine ontology, and performs attribute mapping and unit conversion on the matched fields.

[0009] Furthermore, the quality verification includes field integrity verification, value range verification, and entity relationship constraint verification; if all required fields are not empty, the field values ​​are within the value range, and the entity relationship conforms to the TCM ontology, then the quality verification is deemed to have passed, and the standardized dataset is constructed.

[0010] Furthermore, the intelligent governance unit extracts the research objective, data type, calculation algorithm, result application, and task deadline from the collaborative analysis task; Based on the data type, query the dataset identifier, version number, and source identifier, and determine the data nodes that match the type as candidate data nodes. If the total amount of data of the candidate data nodes reaches the task data amount, then the candidate data nodes are determined as the target data nodes; otherwise, the collaborative analysis task is terminated.

[0011] Furthermore, the permission determination includes: matching the organization type and user role in the subject attributes with the authorized subject of the standardized dataset, and matching the research purpose, calculation algorithm and result usage in the task attributes with the allowed usage and calculation scope in the data attributes; If both matches pass, access rights are granted; otherwise, the collaborative analysis task is terminated.

[0012] Furthermore, the compliance review is a review of the data usage period; If the task period does not exceed the data usage period, the review is deemed successful, and a task authorization token with a valid period is generated; otherwise, the failed item is output.

[0013] Furthermore, the privacy computing unit matches the task identifier and participating data nodes of the collaborative analysis task with the task identifier and target data node of the task authorization token, and verifies the digital signature of the task authorization token. If the digital signature verification is successful, the task identifier is consistent, and all participating data nodes belong to the target data node, then the task authorization token verification is considered successful; otherwise, the task authorization token verification is considered unsuccessful.

[0014] Furthermore, the aforementioned distributed trusted spatial integration system for traditional Chinese medicine drug development includes: When the task authorization token is verified, if the node characteristics are the same but the research objects are different, horizontal federated learning is selected to generate a computational certificate; otherwise, vertical federated learning is selected to generate a computational certificate. If the task authorization token verification fails, cross-agency computation is stopped; The computation credentials include a task identifier, an algorithm version, and a timestamp.

[0015] Furthermore, the confidence application unit constructs a cross-institutional Chinese medicine knowledge graph based on the Chinese medicine ontology, and constructs a graph neural network based on the cross-institutional Chinese medicine knowledge graph; Based on the computational credentials, entities such as medicinal materials, ingredients, targets, prescriptions, and disease syndromes, as well as the relationships between entities, are extracted. The extraction results are then input into the graph neural network to determine the association results between medicinal materials, ingredients, targets, prescriptions, and disease syndromes. The association probability output by the graph neural network is determined as the confidence level, and the association results are bound to the corresponding participating nodes before being output.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: The infrastructure unit utilizes a consortium blockchain network to generate digital identity credentials for institutions, users, and computing devices, enabling multiple institutions participating in traditional Chinese medicine (TCM) research and development to have identifiable identities, thereby mitigating the risk of difficulty in verifying participating entities and data sources. The data access unit receives medicinal materials, prescriptions, pharmacological, and clinical data after successful identity verification. Based on TCM ontology, it performs extraction, transformation, semantic mapping, and quality verification, converting heterogeneous data into standardized datasets and avoiding the risk of inconsistent TCM research and development data. The intelligent governance unit performs permission determination and compliance review based on subject attributes, data attributes, and task attributes. After successful review, it generates a task authorization token and sends it to the target data node, ensuring the adaptability of data access and collaborative analysis tasks and guaranteeing the reliability of authorization constraints for multi-institutional data access. The privacy computing unit verifies the task authorization token and performs cross-institutional computation based on data distribution and computation type, enabling dispersed data to participate in collaborative analysis, while the computation credentials establish a correspondence between the computation process and subsequent applications. The confidence application unit constructs a cross-institutional TCM knowledge graph based on TCM ontology, and combines graph neural networks and computational vouchers to mine the associations between medicinal materials, components, targets, prescriptions and disease symptoms, outputting confidence and evidence sources, thus forming a unified relationship graph and a result verification mechanism, improving the reliability of trusted access, authorized access, collaborative computing and association analysis of TCM R&D data from multiple institutions. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a functional block diagram of a distributed trusted spatial integration system for the research and development of traditional Chinese medicine, provided as an embodiment of the present invention.

[0019] Figure 2 The logical flowchart of the verification task authorization token provided in the embodiment of the present invention. Detailed Implementation

[0020] 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.

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] See Figures 1-2 As shown in some embodiments of this application, a distributed trusted space integration system for traditional Chinese medicine drug development includes: The infrastructure unit is configured to build a consortium blockchain network and generate digital identity credentials for institutions, users, and computing devices; The data access unit is configured to receive medicinal materials, prescriptions, pharmacological and clinical data from various institutions after identity verification, and to extract, transform, semantically map and verify the data according to the ontology of traditional Chinese medicine to determine a standardized dataset. The intelligent governance unit is configured to receive collaborative analysis tasks and standardized datasets, perform permission determination and compliance review based on subject attributes, data attributes and task attributes, generate task authorization tokens after the review is passed, and send the target data node and task authorization tokens. The privacy computing unit is configured to verify task authorization tokens and perform cross-agency computations to generate computing credentials based on the data distribution and computation type of the target data node. The confidence application unit is configured to construct a cross-institutional Chinese medicine knowledge graph based on the Chinese medicine ontology, and to mine the association between medicinal materials, components, targets, prescriptions and disease symptoms based on graph neural networks and computational vouchers, and output results carrying confidence level and evidence source.

[0023] Specifically, the consortium blockchain network built by the infrastructure unit consists of nodes that have obtained access permission. Nodes in the consortium blockchain network possess corresponding digital identities, and nodes not within the permitted scope cannot participate in the issuance of digital identity credentials as valid issuing nodes. The infrastructure unit includes identity registration nodes and identity authentication nodes. Identity registration nodes are used to bind digital identities to institutional qualifications, user roles, or device public keys, while identity authentication nodes are used to verify the corresponding digital identity credentials when institutions, users, or computing devices request access.

[0024] When an organization connects, the identity registration node obtains the organization's qualifications and the digital identity assigned to it, and establishes a mapping between the qualifications and the digital identity. The organization's qualifications indicate its category and access eligibility within the system, while the digital identity uniquely represents the organization during subsequent data access and collaborative analysis. When a user connects, the identity registration node obtains the user's role and the digital identity assigned to them, and establishes a mapping between the role and the digital identity. The user role indicates the user's permitted role in data access or collaborative analysis. When a computing device connects, the identity registration node obtains the device's public key and the digital identity assigned to it, and establishes a mapping between the public key and the digital identity.

[0025] After an organization's qualifications, user role, or device public key are bound to a digital identity, the identity registration node generates a digital identity credential based on the binding. A digital identity credential must at least represent the corresponding digital identity, the content bound to the digital identity, the credential's validity period, and the issuing node that issued the credential. The binding content is the corresponding item from the organization's qualifications, user role, or device public key, enabling the identity authentication node to determine whether the currently requesting access belongs to an organization, user, or computing device, and to determine the access object's identity within the system. When generating a digital identity credential, the identity registration node digitally signs the content to be signed within the digital identity credential, obtaining a credential signature. The content to be signed is the content in the digital identity credential used to indicate the identity and its binding relationship. When the content to be signed changes after the credential is issued, the changed content no longer corresponds to the credential signature. Therefore, the identity authentication node can determine whether the digital identity, bound content, validity period, and issuing node in the credential remain the same as at the time of issuance based on the credential signature.

[0026] Upon receiving an access request, the authentication node retrieves the credential signature, validity period, and issuing node from the digital identity credential carried in the access request, and verifies the credential signature, credential validity, and the issuing node's permission status, respectively. Identity verification is successful when all three verifications pass; otherwise, access is denied to the corresponding organization, user, or computing device.

[0027] During the verification of the credential signature, the authentication node obtains the verification information corresponding to the issuing node and uses this information to verify the correspondence between the credential signature and the content to be signed within the credential. If the verification result shows that the credential signature corresponds to the content to be signed and the issuing node, it indicates that the credential signature was generated by the corresponding issuing node for the current credential content, and the credential content has not undergone any changes after issuance that would cause inconsistencies in the signature; therefore, the signature verification is considered successful. If the credential signature does not correspond to the content to be signed, or does not correspond to the indicated issuing node, the signature verification is considered unsuccessful.

[0028] When verifying the validity of a credential, the identity authentication node reads the validity period from the digital identity credential and obtains the current time when the access request was initiated. The validity period limits the time interval during which the digital identity credential can be used for access verification. If the current time is within the validity period, the credential is considered valid. If the current time is earlier than the start time of the validity period, the credential has not yet entered its usable period. If the current time is later than the end time of the validity period, the credential has exceeded its usage period, and in both cases, the credential is considered invalid. Credential validity does not simply mean that the credential has a validity period field; it means that the verification time meets the time conditions specified by that validity period.

[0029] When verifying whether an issuing node is a licensed node, the identity authentication node reads the issuing node's digital identity from the digital identity credential and matches this digital identity with licensed nodes in the consortium blockchain network. If a node with the same digital identity as the issuing node can be found among the licensed nodes in the consortium blockchain network, it indicates that the credential was generated by a node with issuing qualifications, and the issuing node is determined to be a licensed node. If no matching licensed node can be found, or if the issuing node recorded in the credential does not correspond to any licensed node in the consortium blockchain network, the issuing node is determined not to be a licensed node. Successful identity verification requires that the credential signature corresponds to the credential content, the current time is within the credential's validity period, and the issuing node's digital identity matches with licensed nodes in the consortium blockchain network.

[0030] After successful authentication, the data access unit receives medicinal materials, prescriptions, pharmacological data, and clinical data from various institutions. Data provided by different institutions may differ in field names, data types, and units of measurement. The data access unit extracts, transforms, and semantically maps data based on the TCM ontology, enabling data from different sources to be represented according to the same entities, attributes, and relationships between entities. The TCM ontology includes medicinal materials, prescriptions, components, targets, pharmacological effects, disease syndromes, and clinical efficacy. The TCM ontology can be determined through technical documents, standards, academic dictionaries, etc., related to TCM research and development, but is not specifically limited. Medicinal materials represent the entities of medicinal materials involved in the research and development of TCM drugs; prescriptions represent entities of prescriptions composed of medicinal materials; components represent entities of components corresponding to medicinal materials or prescriptions; targets represent entities of targets related to the effects of components; pharmacological effects represent the pharmacological effects produced by medicinal materials, prescriptions, or components; disease syndromes represent the diseases or syndromes targeted by TCM drug research and development; and clinical efficacy represents the clinical results of prescriptions or medicinal materials for the corresponding disease syndromes.

[0031] When the data access unit extracts data from the received data, it obtains the fields, data type, and unit of measurement for each data item. A field is an item in the data used to represent specific content; the data type indicates whether the field value is recorded as text, numerical value, or a corresponding data format; and the unit of measurement indicates the measurement benchmark corresponding to the numerical field. The data access unit uses the extracted fields as semantic mapping objects and the data type and unit of measurement as conversion conditions.

[0032] When matching fields with attributes of the TCM ontology, the data access unit compares the meaning of the source field with the meaning of the TCM ontology attribute. If a source field corresponds to an attribute in the TCM ontology, the field is considered a match. For example, a source field representing the name of a medicinal material matches the name attribute of the medicinal material entity; a source field representing the medicinal materials that make up a prescription matches the compositional relationship between the prescription and the medicinal materials; and a source field representing the name of an ingredient matches the name attribute of the ingredient entity. Fields with different names but the same corresponding attribute are mapped to the same TCM ontology attribute. Fields whose corresponding attributes cannot be determined are not used as matching fields for attribute mapping.

[0033] For matching fields, the data access unit converts the source field into the corresponding TCM ontology attribute, ensuring that fields with the same meaning but different representations from different institutions are recorded according to the same attribute. The data type after attribute mapping is then consistent with the data type required by the corresponding ontology attribute. If the data type of the source field is consistent with the data type of the ontology attribute, the field value is retained. If the data type of the source field can be converted to the data type required by the ontology attribute, a data type conversion is performed. If it cannot be converted to the corresponding data type, the corresponding field does not meet the quality verification requirements. For matching fields with units of measurement, the data access unit compares the units of measurement of the source data with the units of measurement corresponding to the TCM ontology attribute. If they are the same, the field value and unit of measurement are retained. If they are different but can be converted according to the conversion relationship between corresponding units, the source field value is converted to the unit of measurement used by the TCM ontology attribute.

[0034] After unit conversion, values ​​for the same attribute are expressed using the same unit of measurement, enabling comparison during quality verification and collaborative analysis. Following field extraction, attribute mapping, and unit conversion, the data access unit performs quality verification on the mapped data. Quality verification includes field integrity verification, value range verification, and entity relationship constraint verification; only when all three verifications pass is the quality verification considered successful.

[0035] Field integrity checks are used to determine whether fields necessary to constitute a corresponding ontology entity and entity relationship have values. Required fields are determined based on the expression requirements of corresponding entities and relationships in the TCM ontology; that is, fields whose absence prevents the identification of an entity or the establishment of a corresponding entity relationship are considered required fields. For example, in medicinal material data, the medicinal material name field used to identify the medicinal material entity is a required field; in prescription data, the prescription name field used to identify the prescription entity and the field indicating the medicinal materials contained in the prescription are required fields; in component data, the component name field used to identify the component entity is a required field; in target data, the target name field used to identify the target entity is a required field; in pharmacological data, the field indicating the pharmacological effect and the corresponding entity is a required field; in disease syndrome data, the name field used to identify the disease or syndrome is a required field; and in clinical data, the field indicating the clinical efficacy and its corresponding disease syndrome is a required field.

[0036] Required fields are the minimum field conditions required for an entity and relationship to be valid; they do not change the correspondence between each field and the attributes of the TCM ontology. During integrity verification, the data access unit reads the required fields corresponding to the entity or relationship to which the current data belongs one by one. If all required fields have non-empty values, the field integrity verification is considered successful; if at least one required field is missing, or if the field exists but its value is empty, the field integrity verification is considered unsuccessful.

[0037] Value range validation is used to determine whether a field value falls within the allowed range of the corresponding attribute in the TCM ontology. The value range is determined based on the attribute's data type and the constraints of its unit of measurement within the TCM ontology. For example, for numeric fields, the value range is limited by the minimum and maximum allowed values ​​of the corresponding attribute; the field value is then compared to this range after unit conversion. For text-based name fields, the value range requires that the field value at least meets the corresponding data type and can represent the corresponding entity. Specifically, during validation, the data access unit first determines the TCM ontology attribute corresponding to the field to be validated based on the attribute mapping result, and then reads the value range of that attribute. If the data type of the field value is consistent with the ontology attribute, and the numeric field is between the allowed minimum and maximum values, or the non-numeric field belongs to the corresponding attribute's field value, then the value range validation for that field is considered successful. Therefore, the value range is not temporarily determined by the data access unit based on the received data, nor is it simply based on the maximum and minimum values ​​of data from various institutions; rather, it is determined by the TCM ontology to which the field is mapped and its constraints.

[0038] Entity relation constraint validation is used to determine whether the relationships between entities in the mapped data are permitted by the TCM ontology. The TCM ontology limits the connections that can be established between entities through relation types and the entity types at both ends of the relation. For example, a prescription and medicinal materials can form a relationship where the prescription contains the medicinal materials; a medicinal material and an ingredient can form a relationship where the medicinal material corresponds to the ingredient; an ingredient and a target can form a relationship where the ingredient acts on the target; medicinal materials, prescriptions, or ingredients can form a correspondence with pharmacological effects; prescriptions can form a correspondence with disease syndromes; and clinical efficacy can form a correspondence with corresponding medicinal materials, prescriptions, or disease syndromes.

[0039] When performing entity relationship constraint verification, the data access unit determines whether the entities at both ends of the relationship can be mapped to entity types in the TCM ontology, and then determines whether the relationship represented by the source data can be mapped to the relationship type in the TCM ontology. If the entities at both ends of the relationship exist, the entity types meet the requirements of the corresponding relationship for the entities at both ends, and the relationship used belongs to the inter-entity relationship of the TCM ontology, then the entity relationship is determined to conform to the TCM ontology; otherwise, the entity relationship is determined to not conform to the TCM ontology.

[0040] When all required fields are not empty, all field values ​​in the fields to be verified are within their respective ranges, and all entity relationships conform to the TCM ontology, the data access unit determines that the quality verification has passed and constructs a standardized dataset based on the data that has undergone attribute mapping, unit conversion, and quality verification. The data in the standardized dataset is represented according to the entities, attributes, and relationships between entities as defined by the TCM ontology. If any required field is empty, field values ​​exceed their ranges, or entity relationships do not conform to the TCM ontology, then the three quality verification conditions cannot be met simultaneously. The data access unit will not determine the corresponding data as having passed quality verification. Therefore, the standardized dataset received by the intelligent governance unit consists entirely of data that has undergone semantic mapping and quality verification.

[0041] Collaborative analysis tasks are initiated by authenticated institutions or users based on the needs of traditional Chinese medicine drug research and development. The initiating entity submits the research objective, required data types, computational algorithm, intended use of results, task deadline, and data volume to the intelligent governance unit. The intelligent governance unit generates a task identifier for this task and combines these elements into a collaborative analysis task. The intelligent governance unit receives the collaborative analysis task and standardized dataset, and extracts the research objective, data types, computational algorithm, intended use of results, and task deadline from the collaborative analysis task. The research objective indicates the content to be studied in this collaborative analysis; the data types indicate the categories of data required to complete the task; the computational algorithm indicates the computational method to be performed in the collaborative analysis task; the intended use of results indicates the research and development analysis in which the computational results will be used; and the task deadline indicates the planned execution timeframe for this collaborative analysis task.

[0042] Data type matching indicates that the data type required by the collaborative analysis task is the same as the data type provided by the corresponding standardized dataset. If the collaborative analysis task requires medicinal material data, the data node providing standardized medicinal material data is identified as the data node with matching data type. If the task requires prescription, pharmacological, or clinical data, data nodes that can provide the corresponding standardized data type are queried. The intelligent governance unit identifies the data nodes with matching data type as candidate data nodes and counts the amount of data corresponding to the task data type among all candidate data nodes. The task data volume is the amount of data required to perform this analysis, which can be dynamically determined according to the complexity of the task. The more complex the collaborative analysis task, the higher the corresponding task data volume. This embodiment does not limit the specific task data volume.

[0043] The data volume of each candidate data node is the amount of data in the corresponding node that matches the data type and passes quality verification. The intelligent governance unit adds up the data volumes of each candidate data node to obtain the total data volume of the candidate data nodes, and compares the total data volume with the task data volume. If the total data volume of the candidate data nodes reaches the task data volume, it is determined that the candidate data nodes can collectively meet the data volume requirements of the collaborative analysis task, and the candidate data node is identified as the target data node. If the total data volume of the candidate data nodes does not reach the task data volume, it means that even if all candidate data nodes are called, the required data volume cannot be reached. The intelligent governance unit then terminates the collaborative analysis task to avoid subsequent data deviations.

[0044] The intelligent governance unit determines permissions based on subject attributes, data attributes, and task attributes. Subject attributes include the type of organization initiating the collaborative analysis task and the user role; data attributes include the authorizing entity, permitted uses, and computational scope of the target standardized dataset; and task attributes include the research objective, computational algorithm, and intended use of the results. Permission determination includes two steps: subject matching and task-data matching. Subject matching involves matching the organization type and user role in the subject attributes with the authorizing entity of the standardized dataset. The authorizing entity of the standardized dataset defines which organization types and user roles can participate in accessing the dataset. If the organization initiating the task is of the same type as the standardized dataset, and the user role initiating the task is also of the same type as the standardized dataset, then subject matching succeeds. If the organization type or user role is not within the scope of the authorizing entity, then subject matching fails.

[0045] Task-data matching involves matching the research objective, computational algorithm, and intended use of the results in the task attributes with the permitted uses and computational scope in the data attributes. Specifically, the research objective and intended use of the results should fall within the scope of the standardized dataset, and the computational algorithm should fall within the computational scope of the standardized dataset. If any of the research objective, intended use of the results, or computational algorithm exceeds the scope defined by the corresponding data attribute, the task-data matching fails. If both subject matching and task-data matching pass, the intelligent governance unit determines that the collaborative analysis task has access rights and initiates compliance review. If either subject matching or task-data matching fails, both matches are not simultaneously valid, the intelligent governance unit terminates the collaborative analysis task and does not generate a task authorization token that can be used for cross-institutional computation. The determination of the target data node only indicates that the corresponding node meets the task requirements in terms of data type and volume; it does not directly indicate that the task has permission to access the corresponding data.

[0046] The compliance review refers to the review of the data usage period. The data usage period defines the timeframe within which the corresponding standardized dataset can be used by the collaborative analysis task, while the task period defines the planned execution timeframe for this collaborative analysis task. The intelligent governance unit compares the task period of the collaborative analysis task with the data usage periods of the standardized datasets in each target data node. If the overall task period falls within the data usage period (i.e., the data is already in use when the task starts and the task ends before the end of the data usage period), then the task period is determined not to have exceeded the data usage period; otherwise, it is determined to have exceeded the data usage period. When the collaborative analysis task has access rights and the task period does not exceed the data usage period, the intelligent governance unit determines that the compliance review has passed and generates a task authorization token. The task authorization token corresponds to this collaborative analysis task and at least represents the task identifier, target data node, and validity period. The task identifier in the token is used to limit the token to the corresponding collaborative analysis task, the target data node is used to limit the data nodes that can participate in this cross-institutional computation, and the validity period is used to ensure that the usage time of the task authorization token corresponds to the approved task period.

[0047] The intelligent governance unit digitally signs the task authorization token, establishing a correspondence between the task identifier, target data node, and validity period of the token and the digital signature. The intelligent governance unit then sends the target data node and task authorization token to the privacy computing unit. The privacy computing unit verifies the actual task initiating the computation and the actual data node participating in the computation based on the token's content. If the task's expiration date exceeds the data usage period, the compliance review is deemed unsuccessful, and a failure item is output. This failure item indicates that the collaborative analysis task's expiration date does not meet the data usage period. In this case, no task authorization token is generated for this cross-organizational computation.

[0048] When the privacy computing unit verifies the digital signature of the task authorization token, it obtains the content signed by the intelligent governance unit within the task authorization token, as well as the digital signature carried by the token. The verification of the digital signature includes methods such as developer identity information, hash value, and RFC3161 standard timestamps. The specific verification process is lengthy and well-established, and will not be described in detail here. If the digital signature corresponds to the task identifier, target data node, and validity period in the task authorization token, it indicates that the token content is consistent with the content at the time the signature was generated, and the digital signature verification is deemed successful. If the digital signature does not correspond to the current token content, it indicates that the token content and signature are inconsistent, and the digital signature verification is deemed unsuccessful. Task identifier matching compares the task identifier of the collaborative analysis task actually received by the privacy computing unit with the task identifier recorded in the task authorization token. If they are completely consistent, it means that the current task is the collaborative analysis task authorized by the intelligent governance unit, and the task identifiers are deemed consistent. If they are different, it means that the current task does not correspond to the authorized task, and the task identifiers are deemed inconsistent. Task identifier matching can prevent the task authorization token of a collaborative analysis task from being used for another collaborative analysis task, thus ensuring data credibility.

[0049] The matching of participating data nodes involves comparing each data node actually participating in the cross-agency computation with the target data node recorded in the task authorization token. If every participating data node can find a matching target data node, then all participating data nodes belong to the target data node. The privacy computation unit determines that the task authorization token has passed verification only when the digital signature verification of the task authorization token is successful, the task identifier of the collaborative analysis task matches the task identifier in the token, and all participating data nodes belong to the target data node in the token. If any of the above three conditions is not met, the task authorization token is deemed to have failed verification, and the privacy computation unit can only perform cross-agency computations on tasks and nodes authorized by the intelligent governance unit, further ensuring the credibility of the data.

[0050] When the task authorization token verification fails, the privacy computing unit stops cross-institutional computation and does not generate a computation credential corresponding to this computation. When the task authorization token verification succeeds, the privacy computing unit further selects between horizontal federated learning and vertical federated learning based on the data distribution and computation type of the target data nodes. The data distribution of the target data nodes is represented by node features and research objects; node features are the data attributes used for collaborative analysis in each target data node.

[0051] Taking the "Cross-institutional analysis of the clinical efficacy of Ephedra Decoction and Wind-Cold Exterior Syndrome" as an example, Hospital A's data node stores record A001, with the following fields and values: formula name "Ephedra Decoction", herbal composition "Ephedra, Cinnamon Twig, Bitter Apricot Kernel, Prepared Licorice Root", disease syndrome "Wind-Cold Exterior Syndrome", and clinical efficacy "Effective". Hospital B's data node stores record B001, with the same fields of formula name, herbal composition, disease syndrome, and clinical efficacy, but corresponding to another clinical record generated by Hospital B. In this case, the formula name, herbal composition, disease syndrome, and clinical efficacy are the node features. Since the two nodes use the same feature fields, and records A001 and B001 belong to different clinical research subjects, horizontal federated learning is selected because the node features are the same but the research subjects are different.

[0052] For example, research institution C stores the formula name, medicinal materials, and components for formula F001 "Ma Huang Tang"; pharmacology institution D stores the target and pharmacological effects for the same formula F001; and medical institution E stores the disease syndrome and clinical efficacy for the same formula F001. All three nodes use Ma Huang Tang as their research object, but they have different ontological attribute fields. Therefore, since the research object is the same but the node characteristics are different, vertical federated learning is selected.

[0053] After horizontal or vertical federated learning is completed, the privacy-preserving computation unit generates a computation credential, which includes a task identifier, algorithm version, and timestamp. The task identifier is the same as the task identifier of the collaborative analysis task and is used to establish a correspondence between the computation credential and the collaborative analysis task that has been verified through a task authorization token. The algorithm version indicates the algorithm version used during this horizontal or vertical federated learning process, and the timestamp indicates the computation time corresponding to this computation credential. The trust application unit can distinguish computational content generated by using different algorithm versions for the same task based on the algorithm version, and can also distinguish computational content generated by the same task at different times based on the timestamp. Therefore, although the computation credential does not replace the computation result itself, it can restrict the trust application unit to use the computational content corresponding to the task identifier, algorithm version, and timestamp.

[0054] The confidence application unit connects corresponding entities based on the entity relationships defined by the TCM ontology. For example, it connects prescription entities and medicinal material entities based on the relationship between prescriptions and medicinal materials in standardized data; it connects medicinal material entities and ingredient entities based on the relationship between medicinal materials and components; it connects component entities and target entities based on the relationship between components and targets; and it connects medicinal materials, prescriptions, or components with pharmacological action entities based on relevant data. Furthermore, it forms corresponding entity relationships based on disease syndromes and clinical efficacy data. Since the data from each institution undergoes semantic mapping and quality verification under the same TCM ontology, entities and relationships corresponding to different participating nodes can be added to the cross-institutional TCM knowledge graph according to the same ontology type. The confidence application unit constructs a graph neural network based on the cross-institutional TCM knowledge graph. Entities in the cross-institutional TCM knowledge graph serve as nodes processed by the graph neural network, and entity relationships serve as connections between corresponding nodes. The graph neural network takes entities and entity relationships as input, processes entity information based on other entities and relationships connected to the current entity, and outputs the association probability between corresponding entities.

[0055] After receiving the computational voucher, the confidence application unit reads the task identifier, algorithm version, and timestamp from the voucher. The task identifier identifies the collaborative analysis task requiring association mining, the algorithm version identifies the algorithm version used for the computational content corresponding to the task, and the timestamp identifies the computational batch corresponding to the task. Based on the correspondence among these three elements, the confidence application unit extracts the medicinal materials, components, targets, prescriptions, and disease syndrome entities, as well as the relationships between entities, corresponding to the current computational voucher. The task identifier in the computational voucher determines the collaborative analysis task to which the entities and relationships belong, the algorithm version determines the computational algorithm version corresponding to the entities and relationships, and the timestamp distinguishes the computational content generated by the same task at different times.

[0056] When extracting medicinal material entities, the confidence application unit identifies entities of type medicinal material from the cross-institutional TCM knowledge graph content corresponding to the current calculation voucher. When extracting component entities, it identifies entities of type component; when extracting target entities, it identifies entities of type target; when extracting prescription entities, it identifies entities of type prescription; and when extracting disease syndrome entities, it identifies entities of type disease syndrome. Relationships between entities are extracted based on the relationships established in the TCM ontology. The confidence application unit inputs the extracted medicinal material, component, target, prescription, and disease syndrome entities, along with the relationships between entities, into a graph neural network. The graph neural network processes adjacent entity information based on the connection relationships of entities in the cross-institutional TCM knowledge graph, forming association results between medicinal materials, components, targets, prescriptions, and disease syndromes. These association results represent the potential connections between these entities determined based on the current cross-institutional TCM knowledge graph and the calculation content corresponding to the current calculation voucher.

[0057] The graph neural network outputs the association probability for each association result. The confidence application unit determines the confidence level of the corresponding association result based on the association probability output by the graph neural network. A higher association probability indicates a stronger association between the corresponding entities within the cross-institutional TCM knowledge graph and computational content; a lower association probability indicates a weaker association between the corresponding entities. The confidence level originates from the association probability of the graph neural network. The confidence application unit binds the association result to the corresponding participating nodes. The corresponding participating nodes are the data nodes that actually participate in generating the computational content corresponding to the computational credential after the task authorization token verification is passed. After the association result is bound to the participating nodes, the participating nodes constitute the source of evidence for the association result. When the confidence application unit outputs the association result, it simultaneously outputs the confidence level determined by the graph neural network and the participating nodes bound to the association result. The output result reflects the degree of association between the corresponding medicinal materials, components, targets, prescriptions, and disease symptoms, and also reflects the source of the data nodes corresponding to the association result, thereby ensuring the reliability of the system.

[0058] In summary, the infrastructure unit utilizes a consortium blockchain network to generate digital identity credentials for institutions, users, and computing devices, enabling identifiable identities for multiple institutions participating in traditional Chinese medicine (TCM) research and development, thereby mitigating the risk of difficulty in verifying participating entities and data sources. The data access unit, after successful identity verification, receives medicinal materials, prescriptions, pharmacological, and clinical data. Based on TCM ontology, it performs extraction, transformation, semantic mapping, and quality verification, converting heterogeneous data into standardized datasets and avoiding the risk of inconsistent TCM research and development data. The intelligent governance unit performs permission determination and compliance review based on subject attributes, data attributes, and task attributes. Upon successful review, it generates a task authorization token and sends it to the target data node, ensuring the adaptability of data access and collaborative analysis tasks and guaranteeing the reliability of authorization constraints for multi-institutional data access. The privacy computing unit verifies the task authorization token and performs cross-institutional computations based on data distribution and computation type, enabling distributed data to participate in collaborative analysis, while the computation credentials establish a correspondence between the computation process and subsequent applications. The confidence application unit constructs a cross-institutional TCM knowledge graph based on TCM ontology, and combines graph neural networks and computational vouchers to mine the associations between medicinal materials, components, targets, prescriptions and disease symptoms, outputting confidence and evidence sources, thus forming a unified relationship graph and a result verification mechanism, improving the reliability of trusted access, authorized access, collaborative computing and association analysis of TCM R&D data from multiple institutions.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A distributed, trusted spatial integration system for the research and development of traditional Chinese medicine, characterized in that, include: The infrastructure unit is configured to build a consortium blockchain network and generate digital identity credentials for institutions, users, and computing devices; The data access unit is configured to receive medicinal materials, prescriptions, pharmacological and clinical data from various institutions after identity verification, and to extract, transform, semantically map and verify the data according to the ontology of traditional Chinese medicine to determine a standardized dataset. The intelligent governance unit is configured to receive collaborative analysis tasks and the standardized dataset, perform permission determination and compliance review based on subject attributes, data attributes and task attributes, generate a task authorization token after the review is passed, and send the target data node and the task authorization token. The privacy computing unit is configured to verify the task authorization token and perform cross-agency computation to generate computing credentials based on the data distribution and computation type of the target data node. The confidence application unit is configured to construct a cross-institutional Chinese medicine knowledge graph based on the Chinese medicine ontology, and to mine the association between medicinal materials, components, targets, prescriptions and disease symptoms based on graph neural networks and the computational credentials, and output results carrying confidence level and evidence source.

2. The distributed trusted spatial integration system for traditional Chinese medicine drug development according to claim 1, characterized in that, The infrastructure unit includes an identity registration node and an identity authentication node; The identity registration node binds the organization's qualifications, user roles, or device public keys with digital identities to generate digital identity credentials. The identity authentication node obtains the credential signature, validity period, and issuing node. If the signature verification is successful, the credential is valid, and the issuing node is a licensed node, then the identity verification is successful; otherwise, access is rejected.

3. The distributed trusted spatial integration system for traditional Chinese medicine drug development according to claim 2, characterized in that, The TCM entity includes medicinal materials, prescriptions, components, targets, pharmacological effects, disease symptoms, and clinical efficacy; The data access unit extracts the fields, data types, and units of measurement of the data, matches the fields with the attributes of traditional Chinese medicine ontology, and performs attribute mapping and unit conversion on the matched fields.

4. The distributed trusted spatial integration system for traditional Chinese medicine drug development according to claim 3, characterized in that, The quality verification includes field integrity verification, value range verification, and entity relationship constraint verification. If all required fields are not empty, the field values ​​are within the value range, and the entity relationship conforms to the TCM ontology, then the quality verification is deemed to have passed, and the standardized dataset is constructed.

5. The distributed trusted spatial integration system for traditional Chinese medicine drug development according to claim 4, characterized in that, The intelligent governance unit extracts the research objective, data type, calculation algorithm, result application, and task deadline from the collaborative analysis task; Based on the data type, query the dataset identifier, version number, and source identifier, and determine the data nodes that match the type as candidate data nodes. If the total amount of data of the candidate data nodes reaches the task data amount, then the candidate data nodes are determined as the target data nodes; otherwise, the collaborative analysis task is terminated.

6. The distributed trusted spatial integration system for traditional Chinese medicine drug development according to claim 5, characterized in that, The permission determination includes: matching the organization type and user role in the subject attributes with the authorized subject of the standardized dataset, and matching the research purpose, calculation algorithm and result usage in the task attributes with the allowed usage and calculation scope in the data attributes; If both matches pass, access rights are granted; otherwise, the collaborative analysis task is terminated.

7. The distributed trusted spatial integration system for traditional Chinese medicine drug development according to claim 6, characterized in that, The compliance review refers to a review of the data usage period. If the task period does not exceed the data usage period, the review is deemed successful, and a task authorization token with a valid period is generated; otherwise, the failed item is output.

8. The distributed trusted spatial integration system for traditional Chinese medicine drug development according to claim 7, characterized in that, The privacy computing unit matches the task identifier and participating data nodes of the collaborative analysis task with the task identifier and target data node of the task authorization token, and verifies the digital signature of the task authorization token. If the digital signature verification is successful, the task identifier is consistent, and all participating data nodes belong to the target data node, then the task authorization token verification is considered successful; otherwise, the task authorization token verification is considered unsuccessful.

9. The distributed trusted spatial integration system for traditional Chinese medicine drug development according to claim 8, characterized in that, include: When the task authorization token is verified, if the node characteristics are the same but the research objects are different, horizontal federated learning is selected to generate a computational certificate; otherwise, vertical federated learning is selected to generate a computational certificate. If the task authorization token verification fails, cross-agency computation is stopped; The computation credentials include a task identifier, an algorithm version, and a timestamp.

10. The distributed trusted spatial integration system for traditional Chinese medicine drug development according to claim 9, characterized in that, The confidence application unit constructs a cross-institutional Chinese medicine knowledge graph based on the Chinese medicine ontology, and constructs a graph neural network based on the cross-institutional Chinese medicine knowledge graph; Based on the computational credentials, entities such as medicinal materials, ingredients, targets, prescriptions, and disease syndromes, as well as the relationships between entities, are extracted. The extraction results are then input into the graph neural network to determine the association results between medicinal materials, ingredients, targets, prescriptions, and disease syndromes. The association probability output by the graph neural network is determined as the confidence level, and the association results are bound to the corresponding participating nodes before being output.

Images