CN113161013A - Interpretable adverse drug reaction discovery method based on literature knowledge graph - Google Patents
Interpretable adverse drug reaction discovery method based on literature knowledge graph Download PDFInfo
- Publication number
- CN113161013A CN113161013A CN202110489047.9A CN202110489047A CN113161013A CN 113161013 A CN113161013 A CN 113161013A CN 202110489047 A CN202110489047 A CN 202110489047A CN 113161013 A CN113161013 A CN 113161013A
- Authority
- CN
- China
- Prior art keywords
- drug
- adverse
- literature
- biomarker
- reactions
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 29
- 208000030453 Drug-Related Side Effects and Adverse reaction Diseases 0.000 title claims abstract description 25
- 206010061623 Adverse drug reaction Diseases 0.000 title claims abstract description 12
- 206010067484 Adverse reaction Diseases 0.000 claims abstract description 85
- 230000006838 adverse reaction Effects 0.000 claims abstract description 85
- 239000003814 drug Substances 0.000 claims abstract description 78
- 239000000090 biomarker Substances 0.000 claims abstract description 59
- 229940079593 drug Drugs 0.000 claims abstract description 57
- 201000010099 disease Diseases 0.000 claims abstract description 25
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims abstract description 25
- 230000037361 pathway Effects 0.000 claims abstract description 4
- 238000006757 chemical reactions by type Methods 0.000 claims description 5
- 238000010845 search algorithm Methods 0.000 claims description 3
- 238000011160 research Methods 0.000 abstract description 7
- 230000007246 mechanism Effects 0.000 abstract description 5
- 238000004364 calculation method Methods 0.000 abstract description 3
- 206010028980 Neoplasm Diseases 0.000 description 15
- 229960003278 osimertinib Drugs 0.000 description 5
- DUYJMQONPNNFPI-UHFFFAOYSA-N osimertinib Chemical compound COC1=CC(N(C)CCN(C)C)=C(NC(=O)C=C)C=C1NC1=NC=CC(C=2C3=CC=CC=C3N(C)C=2)=N1 DUYJMQONPNNFPI-UHFFFAOYSA-N 0.000 description 5
- 238000010225 co-occurrence analysis Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 201000009925 nephrosclerosis Diseases 0.000 description 4
- 239000002246 antineoplastic agent Substances 0.000 description 3
- 229940041181 antineoplastic drug Drugs 0.000 description 3
- 231100000433 cytotoxic Toxicity 0.000 description 3
- 230000001472 cytotoxic effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 210000002540 macrophage Anatomy 0.000 description 3
- 239000000107 tumor biomarker Substances 0.000 description 3
- 102000001301 EGF receptor Human genes 0.000 description 2
- 108060006698 EGF receptor Proteins 0.000 description 2
- 238000009412 basement excavation Methods 0.000 description 2
- 201000011510 cancer Diseases 0.000 description 2
- 239000008187 granular material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000000144 pharmacologic effect Effects 0.000 description 2
- 102000004169 proteins and genes Human genes 0.000 description 2
- 108090000623 proteins and genes Proteins 0.000 description 2
- 238000011282 treatment Methods 0.000 description 2
- 208000009304 Acute Kidney Injury Diseases 0.000 description 1
- 206010013786 Dry skin Diseases 0.000 description 1
- 102000009465 Growth Factor Receptors Human genes 0.000 description 1
- 108010009202 Growth Factor Receptors Proteins 0.000 description 1
- 208000029523 Interstitial Lung disease Diseases 0.000 description 1
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 208000008636 Neoplastic Processes Diseases 0.000 description 1
- 206010034016 Paronychia Diseases 0.000 description 1
- 241000029132 Paronychia Species 0.000 description 1
- 208000033626 Renal failure acute Diseases 0.000 description 1
- 206010047555 Visual field defect Diseases 0.000 description 1
- 201000011040 acute kidney failure Diseases 0.000 description 1
- 208000012998 acute renal failure Diseases 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 229940121657 clinical drug Drugs 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000034994 death Effects 0.000 description 1
- 231100000517 death Toxicity 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000000502 dialysis Methods 0.000 description 1
- 230000037336 dry skin Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 229940126585 therapeutic drug Drugs 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H70/00—ICT specially adapted for the handling or processing of medical references
- G16H70/40—ICT specially adapted for the handling or processing of medical references relating to drugs, e.g. their side effects or intended usage
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Public Health (AREA)
- Data Mining & Analysis (AREA)
- Medical Informatics (AREA)
- Primary Health Care (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Epidemiology (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Medicinal Chemistry (AREA)
- Pharmacology & Pharmacy (AREA)
- Toxicology (AREA)
- General Physics & Mathematics (AREA)
- Chemical & Material Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Animal Behavior & Ethology (AREA)
- Biomedical Technology (AREA)
- Pathology (AREA)
- Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)
Abstract
The invention discloses an interpretable adverse drug reaction discovery method based on a literature knowledge graph, which comprises the following steps: s1, extracting four entities from medical literature data: diseases, biomarkers, drugs, and adverse reactions; s2, constructing a literature knowledge graph by using the four entities and a naive Bayes model: the literature knowledge graph comprises vertexes and edges, wherein the vertexes comprise vertexes of four entity types, the edges represent relations between the two vertexes, and weights of the edges represent correlation between the two vertexes and are described through importance indexes; s3, based on the adverse drug reactions in the knowledge map of the literature, comparing the instructions with the drug descriptions, finding potential adverse reactions; s4, providing rational biomarker pathway interpretation for the potential adverse reactions based on literature knowledge maps. The invention explores the potential adverse reactions of the medicine by the technical means of calculation medicine and provides power for the mechanism research of the adverse reactions.
Description
Technical Field
The invention relates to the technical field of medicinal informatics and bioinformatics, in particular to an interpretable adverse drug reaction discovery method based on a literature knowledge graph.
Background
Adverse drug reactions are responsible for severe morbidity and mortality in patients and are a source of economic burden to the medical system. Currently, the drug-induced disease is the 5 th death-prone disease, and about 1/3 deaths worldwide are caused by the improper use of therapeutic drugs. Particularly for tumor patients, the patients who use the anti-tumor drugs have relatively high incidence rate of adverse drug reactions, are more likely to experience rare and serious adverse reactions, and can seriously affect the quality of life of the patients. However, because of the limited sample size and the limited generalizability of clinical trials, the identification of rare and severe adverse reactions prior to marketing is limited, so exploring potential adverse reactions is crucial to reducing morbidity. In recent years, with the rapid development and application of computational medicine, more and more researchers think that the computational medicine can be applied to improve the clinical medical aid decision-making service, but the neural network-based reasoning generally adopted is not interpretable.
Disclosure of Invention
Technical problem to be solved
Based on the problems, the invention provides an interpretable adverse drug reaction discovery method based on a literature knowledge graph, aims to search the potential adverse drug reactions through a technical means of computational medicine, and solves the problem that the potential adverse drug reactions are difficult to identify.
(II) technical scheme
Based on the technical problems, the invention provides an interpretable adverse drug reaction discovery method based on a literature knowledge graph, which comprises the following steps:
s1, extracting four entities from medical literature data: diseases, biomarkers, drugs, and adverse reactions;
s2, constructing a literature knowledge graph by using the four entities and a naive Bayes model: the literature knowledge graph comprises vertices and edges, the vertices comprising vertices of four entity types, i.e. disease, biomarker, drug and adverse reaction types, the edges representing a relationship between two vertices, each edge connecting two different types of vertices, the weights on the edges representing the correlation between the two vertices, the weights on the edges being described by an importance index;
s3, based on the adverse drug reactions in the knowledge map of the literature, comparing the instructions with the drug descriptions, finding potential adverse reactions;
s4, providing rational biomarker pathway interpretation for the potential adverse reactions based on literature knowledge maps: and (3) using a depth-first search algorithm to search each path between the drug and the adverse reaction, and extracting the path between the biomarker related to the first 1% of the drug and the potential adverse reaction corresponding to the drug, thereby providing reasonable biomarker path explanation for the potential adverse reaction.
Further, step S1 includes the following steps for the extracted entity: entities with negative meaning and entities whose frequency of occurrence does not exceed 50 times are removed and the remaining entities are considered as entities associated with each summary.
Further, in step S1, determining a disease type entity and a drug type entity according to the classification description, constructing an adverse reaction type entity according to a WHO source dictionary in UMLS, and constructing a biomarker type entity according to the definition of the biomarker.
Further, in step S2, the relationship between each two vertices is described by calculating the importance index between the disease i and the biomarker j, wherein the importance index between the disease i and the biomarker j is:
IMPTNB＝log(p(xi＝1|yj＝1))-log(p(xi＝1|yj＝0))
wherein xiIs 0 or 1, respectively, in the absence or presence of the biomarker i, yiThe value of (a) is 0 or 1, which respectively represents that the disease j does not appear or appears, and p (·) is obtained by a naive Bayes model; and if the importance index is larger than a certain threshold value, the biomarker i and the disease j are considered to have correlation, namely edges exist, the correlation is the obtained importance index, and otherwise, the edges do not exist.
Further, the threshold is zero, the importance index is a regular edge, the weight of the edge is described by the importance index, and if the importance index is negative, the edge does not exist.
Further, step S3 includes collecting all the drugs from the literature knowledge graph and determining the corresponding adverse reactions to form drug adverse reaction pairs, comparing the adverse reactions corresponding to each drug with the content of the drug specification, finding the adverse reactions not reported in the drug specification, and verifying with clinical data to obtain potential adverse reactions.
Further, in step S4, for each drug, extracting the biomarker related to the first 1% of the drug, and comparing the correlations between the biomarkers related to the first 1% of the drug and any potential adverse reaction corresponding to the drug, respectively, wherein the biomarker with the highest correlation is most likely to mediate the occurrence of the potential adverse reaction caused by the drug.
The invention also discloses an interpretable adverse drug reaction discovery system based on the literature knowledge graph, which comprises the following steps:
at least one processor; and at least one memory communicatively coupled to the processor, wherein:
the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the interpretable literature-knowledge-map-based ADR discovery method.
Also disclosed is a non-transitory computer readable storage medium storing computer instructions that cause the computer to perform the interpretable literature-knowledge-map-based method for adverse drug reaction discovery.
(III) advantageous effects
The technical scheme of the invention has the following advantages:
(1) the invention adopts literature knowledge maps to mine the correlation among diseases, biomarkers, medicaments and adverse reactions, fully discovers the potential adverse reactions of the medicaments, provides reasonable biomarker path explanation for the potential adverse reactions, provides a reliable method for discovering the adverse reactions of the medicaments, and provides a basis for the mechanism research of the adverse reactions of the medicaments;
(2) the invention can be explained according to the potential adverse reaction and the biomarker path thereof, more reasonably selects and uses the medicine, has certain reference value for the research of the new field of biomedical literature excavation, and provides power for the mechanism research of the adverse reaction.
Drawings
The features and advantages of the present invention will be more clearly understood by reference to the accompanying drawings, which are illustrative and not to be construed as limiting the invention in any way, and in which:
FIG. 1 is a schematic flow diagram of an interpretable literature-knowledge-based method for adverse drug reaction discovery in accordance with the present invention;
FIG. 2 is a partial schematic representation of a "tumor-biomarker" profile of knowledge in accordance with an embodiment of the present invention;
FIG. 3 is a flow chart of relational modeling of disease and biomarkers according to an embodiment of the invention;
FIG. 4 is an explanatory diagram of the antitumor drug oxcetinic and one of the adverse reactions of the antineoplastic drugs according to the embodiment of the invention;
FIG. 5 is a comparison of the concordance of adverse reaction findings with the co-occurrence analysis method of the examples of the present invention.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
The invention discloses an interpretable method for discovering adverse drug reactions based on a literature knowledge graph, which is shown in figure 1 and comprises the following steps:
s1, extracting four entities from medical literature data: diseases, biomarkers, drugs, and adverse reactions;
in the embodiment of the invention, the disease is cancer, the abstract of a treatise taking cancer treatment as a key word in 1928 and 2020 is downloaded from MEDLINE, namely an international comprehensive biomedical information bibliography database, and four entities are extracted from the abstract: tumors, biomarkers, drugs, and adverse reactions. The UMLS, namely the Meta Thesaurus2020AA version provided by the computerized intelligence retrieval language integration system, is used as a dictionary for entity extraction, and the Apache open source tool CTakes for information extraction of medical free text is used as an entity extraction tool. Entities with negative meaning and entities whose frequency of occurrence does not exceed 50 times are removed and the remaining entities are considered as entities associated with each summary.
Where the classification of T191 is described as "Neoplastic Process" (Table 2), we include it in the tumor type node. The categories of T121 and T200 are described as "pharmacological Substance" and "Clinical Drug", respectively, and we therefore classified them as Drug type nodes. For adverse reaction type nodes, we constructed using the WHO (all called WHO-ART used for coding clinical information related to adverse drug reactions) source dictionary in UMLS, since WHO was used to encode clinical information related to adverse drug reactions (table 1). Finally, for biomarker type nodes, we refer to the definition of biomarkers and include in table 2 categories other than the three types of nodes described above.
TABLE 1 data sources for constructing dictionaries
TABLE 2 concept taxonomy for constructing dictionaries
S2, constructing a literature knowledge graph by using the four entities and a naive Bayes model:
using the entities extracted from the literature to construct a "tumor-biomarker" knowledge map, which consists of vertices and edges, as shown in fig. 2, the vertices comprising vertices of four entity types, i.e., tumor, biomarker, drug, and adverse reaction; the edges are undirected weighted edges and represent the relationship between two vertexes, each edge is connected with two vertexes of different types, the weight of the edges represents the correlation between the two vertexes, and the correlation is obtained by using an importance index calculated according to a naive Bayes model.
A knowledge graph is a data model that represents facts as nodes and relationships between nodes. In a general medical information network, subjects such as diseases, drugs, biomarkers or treatments can be linked together by different types of reference relationships, which enables knowledge discovery at scales and speeds not achievable by traditional pharmacological or clinical trials. The specific construction method comprises the following steps:
s2.1, processing the occurrence condition of the entity in each document abstract, wherein if the occurrence condition is 1, the occurrence condition is zero;
s2.2, respectively calculating the correlation between each two entities by using a naive Bayes model;
the correlation between the nodes is found by utilizing a naive Bayes model, and the naive Bayes model combines the prior probability and the posterior probability at the same time, so that the subjective deviation caused by only using the prior probability is avoided, and the over-fitting phenomenon caused by singly using the sample information is avoided. The calculation method for each relationship is the same, and the principle of the model is mainly illustrated by calculating the relationship between the tumor and the biomarker. By maximum likelihood estimating the learning parameters, we learned the model of each tumor and used an importance index to determine if there is a correlation between the tumor and the biomarker:
IMPTNB＝log(p(xi＝1|yj＝1))-log(p(xi＝1|yj＝0))
wherein xiIs 0 or 1, respectively, in the absence or presence of the biomarker i, yiThe value of (a) is 0 or 1, which respectively represents that the tumor j does not appear or appears, and p (.) is obtained by a naive Bayes model.
S2.3, when the correlation between the two entities is larger than a certain threshold value, determining that an edge exists between the two entities;
if the importance index is greater than a certain threshold, it is determined that there is a correlation between the biomarker i and the tumor j, in this embodiment, the threshold is zero, the importance index is a side that is regularly present, the weight on the side is described by the importance index, and if the importance index is negative, there is no side. The importance measure is chosen because if the presence of a biomarker makes it more likely that a tumor will be observed, there is more confidence that there is an edge between the two. A workflow diagram for relational modeling of disease and biomarkers is shown in figure 3.
S3, finding potential adverse reactions based on the adverse reaction of the drug in the knowledge map of the literature compared with the drug instruction:
collecting all the medicaments from a literature knowledge map, determining corresponding adverse reactions to form medicament adverse reaction pairs, comparing the adverse reactions corresponding to each medicament with the content of the medicament specification, finding the adverse reactions not reported by the medicament specification, and verifying by clinical data to obtain potential adverse reactions.
Table 3 shows the adverse reactions of 8 patients at different times after administration of Osteinib, which are not mentioned in the official manual of Osteinib drugs, but are mentioned in TBKG (Tumor-Biomarker Knowledge Graph Tumor-Tumor factor Knowledge map). For example, acute renal failure requiring dialysis occurs in the patient 8 after one week of oxcetin administration, and the adverse reaction is not reported in the specification, so that the adverse reaction which is not reported in the specification and actually occurs in clinic, namely a potential adverse reaction, is found.
TABLE 3 adverse reactions in 8 patients at different times after taking Oxcetinic
S4, providing reasonable biomarker path explanation for potential adverse reactions based on literature knowledge maps:
and (3) using a depth-first search algorithm to search each path between the drug and the adverse reaction, and extracting the path between the biomarker related to the first 1% of the drug and the potential adverse reaction corresponding to the drug, thereby providing reasonable biomarker path explanation for the potential adverse reaction.
Extracting biomarkers related to the first 1% of the drugs aiming at each drug, and comparing the biomarkers related to the first 1% of the drugs with the relevance of any potential adverse reaction corresponding to the drugs respectively, wherein the biomarker with the highest relevance is most likely to mediate the occurrence of the potential adverse reaction caused by the drugs.
Figure 4 shows the pathway between biomarkers and partial adverse reactions associated with pre-1% of ocitinib. For example, the relevance of the biomarkers of Ositetinib, namely Cytotoxic Granule Protein, Epidermal Growth Factor Receptor and Macrophage activity Factor to 1 percent, namely Cytotoxic Growth Factor Receptor and Macrophage activity Factors is 3.49, 3.64 and 4.59 respectively, and the relevance of the three biomarkers, namely tumor Factors and adverse reaction renal sclerosis is 5.11, 1.44 and 6.25 respectively, so that the Macrophage activity Factors are more likely to mediate the adverse reaction of Ositetinib to the renal sclerosis than the Epidermal Growth Factor Receptor and the Cytotoxic Granule Protein.
We exemplified the biomarkers and adverse reactions associated with the drug oxirtinib "Osimertinib" to show TBKG results. The correlation between ocitinib ' Osimertinib ' and one of the adverse reactions, Nephrosclerosis ', is 4.31. The relevant results mean that the probability of renal sclerosis "nephroschelosis" is 10.4% in the presence of ositinib "Osimertinib" and 0.5% in the absence of ositinib "Osimertinib". The higher the correlation, the greater the likelihood that osetinib "Osimertinib", excluding other factors, will cause adverse reactions.
In the prior art, a co-occurrence analysis method also exists, the method comprises 775 adverse reactions in total, and according to the calculation of a knowledge graph model in our literature, the most important adverse reactions for ocitinib are ranked as follows: dry skin, paronychia, visual field defects, interstitial lung disease, etc. Our model has some agreement with the report in the ocitinib official manual (Kappa 0.68) and is better than the co-occurrence analysis method (Kappa 0.4), as shown in fig. 5 (a). Compared with the co-occurrence analysis method, the model has better specificity, and is more beneficial to analyzing specific contents which are closely related to the medicine and are different from other medicines, as shown in figure 5 (B).
In conclusion, the method for discovering the adverse drug reaction based on the literature knowledge graph has the following advantages that:
(1) the invention adopts literature knowledge maps to mine the correlation among diseases, biomarkers, medicaments and adverse reactions, fully discovers the potential adverse reactions of the medicaments, provides reasonable biomarker path explanation for the potential adverse reactions, provides a reliable method for discovering the adverse reactions of the medicaments, and provides a basis for the mechanism research of the adverse reactions of the medicaments;
(2) the invention can be explained according to the potential adverse reaction and the biomarker path thereof, more reasonably selects and uses the medicine, reduces the occurrence of drug-induced diseases, has certain reference value for the research of the new field of biomedical literature excavation, and provides power for the mechanism research of the adverse reaction.
Finally, it should be noted that the above-described methods may be converted into software program instructions, either implemented by running a system comprising a processor and a memory, or implemented by computer instructions stored in a non-transitory computer readable storage medium. The integrated unit implemented in the form of a software functional unit may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device) or a processor (processor) to execute some steps of the methods according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations fall within the scope defined by the appended claims.
Claims (9)
1. An interpretable adverse drug reaction discovery method based on a literature knowledge graph is characterized by comprising the following steps:
s1, extracting four entities from medical literature data: diseases, biomarkers, drugs, and adverse reactions;
s2, constructing a literature knowledge graph by using the four entities and a naive Bayes model: the literature knowledge graph comprises vertices and edges, the vertices comprising vertices of four entity types, i.e. disease, biomarker, drug and adverse reaction types, the edges representing a relationship between two vertices, each edge connecting two different types of vertices, the weights on the edges representing the correlation between the two vertices, the weights on the edges being described by an importance index;
s3, based on the adverse drug reactions in the knowledge map of the literature, comparing the instructions with the drug descriptions, finding potential adverse reactions;
s4, providing rational biomarker pathway interpretation for the potential adverse reactions based on literature knowledge maps: and (3) using a depth-first search algorithm to search each path between the drug and the adverse reaction, and extracting the path between the biomarker related to the first 1% of the drug and the potential adverse reaction corresponding to the drug, thereby providing reasonable biomarker path explanation for the potential adverse reaction.
2. The interpretable literature-knowledge profile-based method of finding adverse drug reactions according to claim 1, wherein step S1 further comprises subjecting the extracted entities to: entities with negative meaning and entities whose frequency of occurrence does not exceed 50 times are removed and the remaining entities are considered as entities associated with each summary.
3. The interpretable literature-knowledge profile-based ADM discovery method of claim 1, wherein in step S1, disease type entities and drug type entities are determined according to the classification description, adverse reaction type entities are constructed according to a WHO source dictionary in UMLS, and biomarker type entities are constructed according to biomarker definitions.
4. The method of claim 1, wherein in step S2, the relationship between each two vertices is described by calculating the importance index between disease i and biomarker j, wherein the importance index between disease i and biomarker j is:
IMPTNB＝log(p(xi＝1|yj＝1))-log(p(xi＝1|yj＝0))
wherein xiIs 0 or 1, respectively, in the absence or presence of the biomarker i, yiThe value of (a) is 0 or 1, which respectively represents that the disease j does not appear or appears, and p (·) is obtained by a naive Bayes model; and if the importance index is larger than a certain threshold value, the biomarker i and the disease j are considered to have correlation, namely edges exist, the correlation is the obtained importance index, and otherwise, the edges do not exist.
5. The method of claim 4, wherein the threshold is zero, the importance indicator is a regular edge, the weight of the edge is described by the importance indicator, and the negative importance indicator is an edge.
6. The method for discovering the interpretable adverse drug reactions based on the literature knowledge graph as claimed in claim 1, wherein the step S3 comprises collecting all drugs from the literature knowledge graph and determining corresponding adverse reactions to form adverse drug reaction pairs, comparing the corresponding adverse reactions of each drug with the content of the drug specification, discovering the adverse reactions not reported by the drug specification, and verifying the adverse reactions by clinical data to obtain potential adverse reactions.
7. The method for finding the drug adverse reaction according to claim 1, wherein in step S4, the biomarkers related to the first 1% of the drugs are extracted for each drug, and the biomarkers related to the first 1% of the drugs are compared with the correlation between the drug and any one of the potential adverse reactions corresponding to the drug, wherein the biomarker with the highest correlation is most likely to mediate the occurrence of the potential adverse reaction caused by the drug.
8. An interpretable literature-knowledge-based system for finding adverse drug reactions, comprising:
at least one processor; and at least one memory communicatively coupled to the processor, wherein:
the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the method of any of claims 1 to 7.
9. A non-transitory computer-readable storage medium storing computer instructions that cause a computer to perform the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110489047.9A CN113161013A (en) | 2021-04-29 | 2021-04-29 | Interpretable adverse drug reaction discovery method based on literature knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110489047.9A CN113161013A (en) | 2021-04-29 | 2021-04-29 | Interpretable adverse drug reaction discovery method based on literature knowledge graph |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113161013A true CN113161013A (en) | 2021-07-23 |
Family
ID=76873323
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110489047.9A Withdrawn CN113161013A (en) | 2021-04-29 | 2021-04-29 | Interpretable adverse drug reaction discovery method based on literature knowledge graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113161013A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114582459A (en) * | 2022-01-27 | 2022-06-03 | 中南大学湘雅三医院 | Information processing method, device and equipment based on diagnosis and treatment data and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109325131A (en) * | 2018-09-27 | 2019-02-12 | 大连理工大学 | A kind of drug identification method based on biomedical knowledge map reasoning |
-
2021
- 2021-04-29 CN CN202110489047.9A patent/CN113161013A/en not_active Withdrawn
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109325131A (en) * | 2018-09-27 | 2019-02-12 | 大连理工大学 | A kind of drug identification method based on biomedical knowledge map reasoning |
Non-Patent Citations (2)
| Title |
|---|
| DANIEL M. BEAN等: "Knowledge graph prediction of unknown adverse drug reactions and validation in electronic health records", SCIENTIFIC REPORTS, vol. 7, no. 1, pages 16416, XP055656907, DOI: 10.1038/s41598-017-16674-x * |
| MENG WANG等: "Adverse Drug Reaction Discovery Using a Tumor-Biomarker Knowledge Graph", FRONTIERS IN GENETICS, vol. 11, pages 625659 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114582459A (en) * | 2022-01-27 | 2022-06-03 | 中南大学湘雅三医院 | Information processing method, device and equipment based on diagnosis and treatment data and storage medium |
| CN114582459B (en) * | 2022-01-27 | 2023-04-18 | 中南大学湘雅三医院 | Information processing method, device and equipment based on diagnosis and treatment data and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hosseini et al. | Heteromed: Heterogeneous information network for medical diagnosis | |
| CN111292848B (en) | Medical knowledge graph auxiliary reasoning method based on Bayesian estimation | |
| US20200303072A1 (en) | Method and system for supporting medical decision making | |
| US8949079B2 (en) | Patient data mining | |
| Huddar et al. | Predicting complications in critical care using heterogeneous clinical data | |
| Dessi et al. | TF-IDF vs word embeddings for morbidity identification in clinical notes: An initial study | |
| Shen et al. | Constructing node embeddings for human phenotype ontology to assist phenotypic similarity measurement | |
| Omran et al. | Breast cancer identification from patients’ tweet streaming using machine learning solution on spark | |
| Nabi et al. | Machine learning approach: Detecting polycystic ovary syndrome & it's impact on bangladeshi women | |
| CN113161013A (en) | Interpretable adverse drug reaction discovery method based on literature knowledge graph | |
| Lai et al. | BioREx: improving biomedical relation extraction by leveraging heterogeneous datasets | |
| Zhou et al. | Converting semi-structured clinical medical records into information and knowledge | |
| Xu et al. | [Retracted] Using Artificial Intelligence Technology to Solve the Electronic Health Service by Processing the Online Case Information | |
| Juckett et al. | Concept detection using text exemplars aligned with a specialized ontology | |
| Grani et al. | Supporting personalized health care with social media analytics: an application to hypothyroidism | |
| Lin et al. | A top-down binary hierarchical topic model for biomedical literature | |
| Grissette | Drug reaction discriminator within encoder-decoder neural network model: COVID-19 pandemic case study | |
| Liao et al. | ILSHIP: An interpretable and predictive model for hypothyroidism | |
| Altaf et al. | ETL for disease indicators using brute force rule-based NLP algorithm and metadata exploration | |
| Luling et al. | COVID-19 literature mining and analysis research | |
| Egleston et al. | Statistical inference for natural language processing algorithms with a demonstration using type 2 diabetes prediction from electronic health record notes | |
| Riyaz et al. | Improving coronary heart disease prediction by outlier elimination | |
| Rahman et al. | Empowering early detection: A web-based machine learning approach for PCOS prediction | |
| Singh et al. | Pain assessment using intelligent computing systems | |
| Mohammed et al. | Brain Stroke Prediction Model Based SMOTE and Machine Learning Algorithms |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210723 |
|
| WW01 | Invention patent application withdrawn after publication |