CN-122025205-A - Method and device for generating tabu information of medicine

CN122025205ACN 122025205 ACN122025205 ACN 122025205ACN-122025205-A

Abstract

The embodiment of the specification provides a method and a device for generating tabu information of medicines, wherein the method comprises the steps of determining a plurality of pharmacological mechanisms corresponding to a first medicine through a pharmacological mechanism reasoning model based on first medicine information of the first medicine, determining a plurality of first candidate tabu objects of the first medicine through a pharmacological tabu reasoning model based on the pharmacological mechanisms, classifying the first candidate tabu objects of the first medicine into a candidate tabu object set of the first medicine, respectively combining the first medicine with each object in the candidate tabu object set to obtain each medicine object set, wherein each medicine object set comprises the first medicine object set, determining supporting evidence information and anti-evidence information corresponding to the first medicine object set from a designated evidence source, performing countermeasure verification on the first medicine object set by utilizing the supporting evidence information and the anti-evidence information, and determining each tabu information corresponding to each medicine object set according to the countermeasure verification result of each medicine object set so as to generate tabu information of tabu objects with implicit relationship with medicines.

Inventors

ZHOU JING
LIN JINZHEN
YE XIANG
KE XUEJIA
LI CHENGZE
Ying Chenzhe
MENG CHANGHUA
WANG WEIQIANG

Assignees

支付宝(杭州)数字服务技术有限公司

Dates

Publication Date: 20260512
Application Date: 20260105

Claims (16)

1. A method of generating tabu information for a pharmaceutical product, comprising: based on first medicine information of a first medicine, determining a plurality of pharmacological mechanisms corresponding to the first medicine through a pharmacological mechanism reasoning model; Determining a plurality of first candidate tabu objects corresponding to the first medicine through a pharmacological tabu reasoning model based on the pharmacological mechanisms, wherein the first candidate tabu objects belong to a candidate tabu object set of the first medicine; Combining the first medicine with each candidate tabu object in the candidate tabu object set respectively to obtain each medicine object group, wherein the first medicine object group is included; From a specified evidence source, supporting evidence information and opposite evidence information corresponding to a first medicine object group are determined, and the first medicine object group is subjected to countermeasure verification by using the supporting evidence information and the opposite evidence information; And determining each piece of tabu information corresponding to each medicine object group according to the result of the countermeasure verification of each medicine object group, wherein the tabu information at least indicates whether the tabu relation of the corresponding medicine object group is established.
2. The method of claim 1, wherein the candidate contraindicated subject comprises a contraindicated, contraindicated drug, and/or contraindicated population, the contraindicated comprising a disease, pathological condition, and/or symptom in contraindicated relation to the drug.
3. The method of claim 1, wherein the tabu information includes a tabu level and a tabu relationship confidence between the first drug and the respective candidate tabu object.
4. The method of claim 1, further comprising: And determining a plurality of second candidate tabu objects corresponding to the first medicine through a medicine tabu knowledge graph based on the first medicine information, wherein the nodes in the medicine tabu knowledge graph comprise medicine nodes, tabu object nodes and pharmacological mechanism nodes corresponding to pharmacological mechanisms.
5. The method of claim 4, further comprising, prior to said combining said first drug with each candidate tabu object in said candidate tabu object set, respectively, obtaining each drug object set: Calculating an item distance value between every two candidate tabu objects based on object item information corresponding to each first candidate tabu object and object item information of each second candidate tabu object, wherein the object item information at least comprises an object identifier of the corresponding candidate tabu object and a corresponding pharmacological mechanism; merging two candidate tabu objects with the entry distance value lower than the specified distance threshold value into the same candidate tabu object to obtain the candidate tabu object set.
6. The method of claim 4, wherein the determining a number of second candidate tabu objects corresponding to the first drug comprises: Taking a first medicine node corresponding to the first medicine information as a central node, and sampling from the medicine tabu knowledge graph by adopting a random walk mode to obtain a knowledge sub-graph corresponding to the first medicine; determining Node vectors of all nodes in the knowledge sub-graph by adopting a Node2Vec algorithm; Calculating the similarity between the first medicine node and each tabu object node; and determining a tabu object corresponding to the tabu object node with the corresponding similarity exceeding a specified threshold as a second candidate tabu object corresponding to the first medicine.
7. The method of claim 1, wherein the first drug subject group comprises a first drug and a target subject, and wherein the first drug subject group corresponds to a number of pharmacological mechanisms, including a first pharmacological mechanism; The determining supporting evidence information and objecting evidence information corresponding to the first medicine object group comprises the following steps: Determining a plurality of pieces of relevant evidence information corresponding to the first medicine object group from the appointed evidence source; And determining a plurality of first supporting evidence information and a plurality of first countering evidence information from the plurality of related evidence information by a mechanism verification model aiming at the first pharmacological mechanism, wherein each first supporting evidence information indicates that the first pharmacological mechanism leads the target object to be used as a tabu object of the first medicine, and each first countering evidence information indicates that the first pharmacological mechanism is used for weakening the target object to be used as the tabu object of the first medicine.
8. The method of claim 7, wherein the specified evidence source comprises a plurality of evidence sources; the retrieving a plurality of pieces of related evidence information corresponding to the first medicine object group includes: retrieving a plurality of initial evidence information corresponding to the first medicine object group from the appointed evidence source, wherein each initial evidence information comprises a corresponding evidence source identifier, release time corresponding to the evidence text and a correlation score between the evidence text and the first medicine object group; Determining authority weights corresponding to the initial evidence information based on the evidence source identifiers in the initial evidence information; Screening out a plurality of relevant evidence information from the plurality of initial evidence information based on the aging degree, authority weight and relevance score of each initial evidence information, wherein the aging degree of each initial evidence information is determined based on the time difference between the current time and the release time corresponding to the evidence text.
9. The method of claim 8, wherein said screening out a number of relevant evidence information comprises: determining the credible comprehensive score corresponding to each initial evidence information based on the aging degree, authority weight and relevance score of each initial evidence information; And taking the initial evidence information of which the corresponding credible comprehensive score exceeds a preset credible score threshold value as related evidence information.
10. The method of claim 8, wherein said retrieving initial evidence information corresponding to said first drug object group from said specified evidence source comprises: Performing information expansion on the drug information and/or the candidate tabu objects in the first drug object group to obtain corresponding first expansion information; Based on the first extension information, the initial evidence information is retrieved from the plurality of evidence sources.
11. The method of claim 1, wherein said performing challenge verification on said first drug object group using said supporting evidence information and anti-evidence information comprises: Determining a supporting intensity value corresponding to the first pharmacological mechanism based on each first supporting evidence information corresponding to the first pharmacological mechanism aiming at any first pharmacological mechanism in pharmacological mechanisms corresponding to the first medicine object group; And determining evidence conflict scores corresponding to the first medicine object group based on the supporting intensity scores and the objection intensity scores corresponding to the pharmacological mechanisms of the first medicine object group, so as to determine the result of the countermeasure verification.
12. The method of claim 11, wherein the determining a support intensity value for the first pharmacological mechanism comprises: Determining a support intensity value corresponding to the first pharmacological mechanism based on the credible comprehensive score and the authority weight corresponding to each piece of first support evidence information corresponding to the first pharmacological mechanism; the determining an objection intensity value corresponding to the first pharmacological mechanism comprises: And determining an objection intensity value corresponding to the first pharmacological mechanism based on the credible comprehensive score and the authority weight corresponding to each first objection evidence information corresponding to the first pharmacological mechanism.
13. The method of claim 11, wherein the determining each tabu information for each drug object group comprises: aiming at a first medicine object group, determining an authoritative value of an evidence source corresponding to each pharmacological mechanism based on the authoritative weight corresponding to each piece of supporting evidence information corresponding to each pharmacological mechanism; Determining evidence consistency values corresponding to the pharmacological mechanisms based on a plurality of pieces of supporting evidence information corresponding to the pharmacological mechanisms and the number of the supporting evidence information; Determining a contraindication relation confidence coefficient corresponding to the first medicine object group at least based on the evidence source authority value and the evidence consistency value corresponding to each pharmacological mechanism and the evidence conflict score; and determining the tabu grade corresponding to the first medicine object group by designating a large model based on the pharmacological mechanism corresponding to the first medicine object group, the first supporting evidence information and the first objecting evidence information thereof.
14. The method of claim 13, further comprising: and if the evidence conflict score is larger than a specified conflict threshold, punishing the tabu relation confidence based on the specified conflict threshold.
15. An apparatus for generating tabu information of a drug, comprising: The first determining module is configured to determine a plurality of pharmacological mechanisms corresponding to the first medicine through a pharmacological mechanism reasoning model based on first medicine information of the first medicine; A second determining module configured to determine, based on the plurality of pharmacological mechanisms, a plurality of first candidate tabu objects corresponding to the first medicine through a pharmacological tabu reasoning model, the candidate tabu object sets being included in the first medicine; The combination module is configured to respectively combine the first medicine with each candidate tabu object in the candidate tabu object set to obtain each medicine object group, wherein the medicine object group comprises the first medicine object group; The determining and verifying module is configured to determine supporting evidence information and opposite evidence information corresponding to the first medicine object group from a designated evidence source, and conduct countermeasure verification on the first medicine object group by using the supporting evidence information and the opposite evidence information; And a fourth determining module configured to determine each tabu information corresponding to each medicine object group according to a result of the challenge verification of each medicine object group, the tabu information indicating at least whether a tabu relationship of the corresponding medicine object group is established.
16. A computing device comprising a memory and a processor, wherein the memory has executable code stored therein, which when executed by the processor, implements the method of any of claims 1-14.

Description

Method and device for generating tabu information of medicine Technical Field The present disclosure relates to the field of artificial intelligence, and in particular, to a method and apparatus for generating tabu information of a drug. Background At present, in the recommendation scene of the tabu diseases of the medicines, the recommendation of the tabu diseases of the medicines is generally performed based on a manually compiled medicine-disease knowledge graph. The method generally comprises the steps of pre-training a graph neural network based on a manually compiled medicine-disease knowledge graph, wherein the medicine-disease knowledge graph comprises each medicine node, each disease node and a tabu relation or an adaptive relation side between each medicine node and each disease node, the aim of the pre-training the graph neural network is to increase similarity of embedded representation between each medicine node and each disease node with the tabu relation, reduce similarity of embedded representation between each medicine node and each disease node with the adaptive relation, then determining a sub-graph of the first medicine information from the medicine-disease knowledge graph after acquiring the first medicine information required to be subjected to tabu disease recommendation, wherein the sub-graph comprises a first medicine node corresponding to the first medicine information and a one-hop neighbor node of the first medicine node, processing the embedded representation of each node in the sub-graph through the pre-trained graph neural network, and then determining the tabu disease corresponding to the first medicine information based on the similarity between the representation of the first medicine node and the embedded representation of each medicine node in the sub-graph. In the above-mentioned scene, based on the manually compiled medicine-disease knowledge graph, the recommendation of the tabu disease of the medicine is carried out, only the tabu disease indicated by the explicit association path (such as the direct edge of "aspirin→gastric ulcer") in the knowledge graph can be found, and because the medicine-disease knowledge graph is manually compiled, a certain period exists from the marketing of a new medicine to the addition of the tabu disease of the new medicine to the medicine-disease knowledge graph, and hysteresis exists in the addition of the tabu disease of the new medicine in the medicine-disease knowledge graph. Disclosure of Invention One or more embodiments of the present disclosure provide a method and an apparatus for generating tabu information of a drug, so as to generate tabu information of a tabu object having an implicit tabu relationship with the drug, thereby generating relatively more comprehensive tabu information of the tabu object, and generating tabu information of a new drug more friendly. According to a first aspect, there is provided a method of generating tabu information of a pharmaceutical product, comprising: based on first medicine information of a first medicine, determining a plurality of pharmacological mechanisms corresponding to the first medicine through a pharmacological mechanism reasoning model; Determining a plurality of first candidate tabu objects corresponding to the first medicine through a pharmacological tabu reasoning model based on the pharmacological mechanisms, wherein the first candidate tabu objects belong to a candidate tabu object set of the first medicine; Combining the first medicine with each candidate tabu object in the candidate tabu object set respectively to obtain each medicine object group, wherein the first medicine object group is included; From a specified evidence source, supporting evidence information and opposite evidence information corresponding to a first medicine object group are determined, and the first medicine object group is subjected to countermeasure verification by using the supporting evidence information and the opposite evidence information; And determining each piece of tabu information corresponding to each medicine object group according to the result of the countermeasure verification of each medicine object group, wherein the tabu information at least indicates whether the tabu relation of the corresponding medicine object group is established. According to a second aspect, there is provided an apparatus for generating tabu information of a medicine, comprising: The first determining module is configured to determine a plurality of pharmacological mechanisms corresponding to the first medicine through a pharmacological mechanism reasoning model based on first medicine information of the first medicine; A second determining module configured to determine, based on the plurality of pharmacological mechanisms, a plurality of first candidate tabu objects corresponding to the first medicine through a pharmacological tabu reasoning model, the candidate tabu object sets being included in the fi