Construction method of endocrine disruption effect harmful outcome path network

Abstract

The invention discloses an endocrine disruption effect harmful outcome path network construction method, and relates to the field of endocrine disrupter screening, and the method comprises the steps: obtaining endocrine data; preprocessing the acquired endocrine data; according to the preprocessed endocrine data, constructing an event causal relationship triad reflecting the endocrine disruption effect, wherein the event causal relationship triad comprises a trigger word, a source event and a target event; the data KER1 is used as data KER1; according to the preprocessed endocrine data, acquiring endocrine data of which the confidence coefficient is greater than a preset threshold value from a preset literature database, and taking the acquired endocrine data as data KER2; performing duplicate removal processing on the data KER1 and the data KER2 to obtain final endocrine disruption effect harmful outcome path data; and establishing an endocrine disruption effect harmful outcome path network according to the final endocrine disruption effect harmful outcome path data. In view of insufficient analysis of an action mechanism between an endocrine disrupter and a corresponding bad health outcome in the prior art, the application improves interpretability.

Description

Technical Field

[0001] The present application relates to the field of endocrine disruptor screening, and more specifically, to a method for constructing an adverse outcome pathway network of endocrine disrupting effects. Background Art

[0002] As a class of chemical substances that can interfere with the endocrine system function in organisms, endocrine disruptors pose a serious threat to human health and the ecological environment, and have become a new type of pollutant that has attracted great attention from the global scientific community and the international community. Endocrine disruptors need to meet three conditions simultaneously: having an adverse effect on organisms or their offspring, having an endocrine disruption mechanism of action, and the adverse effect being caused by the endocrine disruption effect. It is difficult to systematically establish the biological connection between the adverse effect and the endocrine disruption effect, which has become the key technical bottleneck in the slow progress of current endocrine disruptor identification work.

[0003] Constructing an adverse outcome pathway (AOP) network of endocrine disrupting effects is an effective way to solve the above problems. AOP establishes a clear causal chain between biological perturbations at the molecular level and adverse outcomes at the organism and population levels, and forms a complete chain of harmful effect mechanisms by integrating molecular initiation events (MIEs), key events (KEs), and adverse outcomes (AOs). However, the existing AOP construction methods have problems such as single data source, cumbersome construction process, and low automation level, and it is difficult to meet the needs of efficient screening and risk assessment of endocrine disruptors.

[0004] Chinese Patent Application, Application No. CN201811597767.1, Publication Date: March 29, 2019, discloses a method for high-throughput screening of endocrine disruptors based on a hierarchical warning structure, which relates to the field of virtual screening and activity prediction of endocrine disruptors. This application uses substructure frequency analysis and substructure ratio analysis to extract the primary warning structures of active data compounds; uses SARpy software to extract the secondary warning structures of compounds that meet the primary warning structures; uses SARpy software to extract the tertiary warning structures; combines the primary warning structures and the secondary warning structures to form an activity prediction module, in which compounds with characteristic structures are first screened out, and then warning compounds with potential endocrine disrupting effects are screened out based on the secondary warning structures; uses the tertiary warning structures as an interference activity prediction module, and then screens the interference activity based on the interference activity prediction module. However, this solution mainly screens endocrine disruptors based on chemical structure characteristics, only focuses on the association between the structure characteristics of compounds and endocrine disrupting activity, and does not establish a causal relationship chain between endocrine disruptors and biological effects. Summary of the Invention

[0005] 1. Technical Problem to be Solved

[0006] In view of the insufficient analysis of the action mechanism between endocrine disruptors and corresponding adverse health outcomes in the prior art, the present application provides a method for constructing an adverse outcome pathway network for endocrine disrupting effects. By constructing a systematic and complete adverse outcome pathway network for endocrine disrupting effects and identifying key nodes and pathways through network topology analysis, the interpretability is improved.

[0007] 2. Technical Solution

[0008] The object of the present application is achieved through the following technical solutions.

[0009] One aspect of the present application provides a method for constructing an adverse outcome pathway network for endocrine disrupting effects, including: obtaining endocrine data; preprocessing the obtained endocrine data; according to the preprocessed endocrine data, constructing an event causality triple <trigger word, source event, target event> reflecting endocrine disrupting effects as data KER1; according to the preprocessed endocrine data, obtaining endocrine data with a confidence level greater than a preset threshold from a preset literature database as data KER2. De-duplicating data KER1 and data KER2 to obtain the final adverse outcome pathway data for endocrine disrupting effects; establishing an adverse outcome pathway network for endocrine disrupting effects according to the final adverse outcome pathway data for endocrine disrupting effects.

[0010] Further, preprocessing the obtained endocrine data includes: obtaining known endocrine disruptors and obtaining corresponding synonyms from a preset literature database of endocrine disruptors, and constructing a list of endocrine disruptor names through data integration; the synonyms represent words with a semantic similarity greater than a preset threshold; obtaining endocrine events from a preset literature database through a predefined list of endocrine words to construct a list of endocrine events; according to the list of endocrine disruptor names and the list of endocrine events, obtaining adverse outcome pathway data for endocrine disrupting effects through text mining; filtering the adverse outcome pathway data for endocrine disrupting effects to obtain filtered literature data.

[0011] Furthermore, obtain the data of the adverse outcome pathways of endocrine disrupting effects through text mining, including: according to the list of endocrine disruptor names and the list of endocrine events, retrieve the preset literature database through text mining to obtain the literature data containing endocrine disruptors and endocrine events; count the co-occurrence frequency of endocrine disruptors and endocrine events in the literature data, and use the literature data with a co-occurrence frequency greater than the preset threshold as the first literature data; according to the first literature data, extract metadata using natural language processing methods and extract endocrine disruptor data using a named entity recognition algorithm; the metadata includes the title, publication date, and abstract of the literature; construct the data of the adverse outcome pathways of endocrine disrupting effects based on the metadata and endocrine disruptor data.

[0012] Furthermore, perform text filtering on the data of the adverse outcome pathways of endocrine disrupting effects to obtain the filtered literature data, including: perform deduplication processing on the data of the adverse outcome pathways of endocrine disrupting effects, delete duplicate data, and obtain the data after the first filtering; according to the data after the first filtering, use the PubMed screening rules to identify and delete review literature, and obtain the data after the second filtering, where the review literature is literature that does not contain experimental data; according to the data after the second filtering, identify the literature language through a language feature dictionary and grammar rules, and obtain Chinese literature data as the data after the third filtering; according to the data after the third filtering, use a named entity recognition algorithm to identify the research object, and obtain the literature data related to mammals as the data after the fourth filtering; count the co-occurrence frequency of endocrine events in the data after the fourth filtering, and obtain the literature data with a co-occurrence frequency greater than the preset threshold as the filtered literature data.

[0013] Specifically, the deduplication process effectively identifies and removes duplicate literature, preventing the same evidence from being calculated multiple times and ensuring the independence of each data point; using the PubMed professional rules to screen the literature type can accurately identify and eliminate review literature without experimental data, retain the original research data, and improve the reliability of the evidence. Identifying the literature language through a language feature dictionary and grammar rules focuses on extracting Chinese literature, integrating local research results, breaking through language barriers, and expanding the knowledge source. Automatically identifying the biological entities involved in the research and focusing on research related to mammals can improve species relevance and enhance the transferability of research results. Counting the co-occurrence frequency of endocrine events and setting a threshold to screen for high-frequency associations can reduce false positives caused by random co-occurrence.

[0014] Furthermore, based on the preprocessed endocrine data, construct an event causal relationship triple <trigger word, source event, target event> reflecting the endocrine disruption effect, as data KER1, including: according to the filtered literature data, use the data structured parsing method to obtain the abstract text of the literature data; adopt a rule-based method, use the predefined causal relationship template, and obtain the data containing causal relationships from the abstract text; based on the data containing causal relationships, construct an event causal relationship triple <trigger word, source event, target event> reflecting the endocrine disruption effect according to the predefined event trigger word dictionary and semantic role annotation rules, as data KER1.

[0015] Furthermore, based on the final adverse outcome pathway data of the endocrine disruption effect, establish an adverse outcome pathway network for the endocrine disruption effect, including: according to the final adverse outcome pathway data of the endocrine disruption effect, extract the source event, target event, and the relationship between events respectively; map the source event and target event to nodes, and map the relationship between events to directed edges; construct an adverse outcome pathway network for the endocrine disruption effect according to the nodes and directed edges.

[0016] Furthermore, after constructing the adverse outcome pathway network for the endocrine disruption effect, it also includes: obtaining the topological structure of the adverse outcome pathway network for the endocrine disruption effect; according to the topological structure, calculate the degree centrality, betweenness centrality, and eccentricity of each event node respectively; the degree centrality represents the connection degree of the node, the betweenness centrality represents the importance of the node as a path mediator, and the eccentricity represents the average distance from the node to other nodes in the network; according to the degree centrality, betweenness centrality, and eccentricity of each event node, and the position of the event node in the network, divide the event nodes into molecular initiating events MIE, key events KE, and adverse outcomes AO; the molecular initiating event MIE represents the initial molecular trigger event of the endocrine disruption effect, the key event KE represents the event with a betweenness centrality greater than the threshold in the endocrine disruption effect, and the adverse outcome AO represents the harmful result caused by the endocrine disruption effect.

[0017] Specifically, the traditional method relies on expert experience to subjectively judge the event type, lacks objective criteria to distinguish molecular initiating events, key events, and adverse outcomes, and there may be differences in the classification of the same event by different experts. In this application, by setting the degree centrality: quantifying the direct connection degree between the node and other nodes to identify hub events with high connectivity; betweenness centrality: evaluating the frequency of the node as a mediator of different paths to identify key transfer nodes; eccentricity: measuring the position characteristics of the node in the network to distinguish the starting and terminal events.

[0018] In the prior art, it is often impossible to objectively evaluate the status of different event nodes in the network. This application automatically identifies based on the characteristics of low in-degree, high out-degree, and the front-end of the network position; quantitatively determines the mediating importance of events through the betweenness centrality threshold; and automatically identifies based on the characteristics of low out-degree, high in-degree, and the end of the network position.

[0019] Another aspect of this application also provides a construction system for an endocrine disruption effect adverse outcome pathway network, including: a data collection module that constructs a list of endocrine disruptor names and a list of endocrine events; a data processing module that, according to the list of endocrine disruptor names and the list of endocrine events, obtains initial endocrine disruption effect adverse outcome pathway data and event relationship data from a preset literature database through text mining; constructs the final endocrine disruption effect adverse outcome pathway data according to the initial endocrine disruption effect adverse outcome pathway data and event relationship data; and a network construction module that establishes an endocrine disruption effect adverse outcome pathway network according to the final endocrine disruption effect adverse outcome pathway data.

[0020] Another aspect of this application also provides a computer-readable storage medium that stores computer instructions, and when the computer instructions are executed by a processor, a method for constructing an endocrine disruption effect adverse outcome pathway network is implemented.

[0021] 3. Beneficial effects

[0022] Compared with the prior art, the advantages of this application are as follows:

[0023] (1) The method for constructing an endocrine disruption effect adverse outcome pathway network provided by this application constructs a complete evidence chain from molecular initiating events to adverse outcomes through systematic integration of multi-source data, realizes the mechanism correlation representation between endocrine disruptors and their biological effects, and provides comprehensive data support for the systematic understanding of endocrine disruption mechanisms and scientific risk control.

[0024] (2) By obtaining a list of synonyms of endocrine disruptors and constructing a list of endocrine events from a preset literature database, this application forms a comprehensive basic data set related to endocrinology, greatly improves the retrieval efficiency and accuracy of subsequent literature information, and effectively solves the problems of incomplete and inefficient collection of literature related to endocrine disruptors in traditional methods.

[0025] (3) By calculating topological parameters such as degree centrality, betweenness centrality, and eccentricity of event nodes in the endocrine disruption effect adverse outcome pathway network, this application realizes the scientific classification of molecular initiating events MIE, key events KE, and adverse outcomes AO, and provides objective and quantitative technical support for identifying key nodes and paths of endocrine disruption effects. Description of the drawings

[0026] Figure 1 Schematic flow chart of a method for constructing an AOP network of endocrine disruptors in the present application;

[0027] Figure 2 Schematic diagram of the AOP network of endocrine disruptors in the present application;

[0028] Figure 3 Schematic diagram of the distribution of Degree values of events included in the AOP network of the present application;

[0029] Figure 4 Schematic diagram of the distribution of Betweenness Centrality values of events included in the AOP network of the present application;

[0030] Figure 5 Schematic diagram of the distribution of Eccentricity values of events included in the AOP network of the present application. Specific implementation manner

[0031] The present application will be described in detail below with reference to the accompanying drawings of the specification and specific embodiments.

[0032] As Figure 1 shown, obtain endocrine data; preprocess the obtained endocrine data; according to the preprocessed endocrine data, construct a triple of event causal relationships reflecting the endocrine disruption effect: trigger word, source event, target event, as data KER1; according to the preprocessed endocrine data, obtain endocrine data with a confidence level greater than a preset threshold from a preset literature database, as data KER2; perform duplicate removal processing on data KER1 and data KER2 to obtain the final harmful outcome path data of the endocrine disruption effect; establish a harmful outcome path network of the endocrine disruption effect according to the final harmful outcome path data of the endocrine disruption effect.

[0033] Among them, in endocrine disruption research, the trigger word clearly defines the nature and intensity of the mechanism of action of the disruptor, and distinguishes between direct and indirect effects. The source event usually represents the initial action point of the endocrine disruptor or the upstream event in the intermediate cascade reaction. The target event can be used as an intermediate effect or the final adverse outcome, and is an important basis for the risk assessment of the disruptor.

[0034] Specifically, S1: Construct a list of endocrine disruptor names: Obtain the substance names and CAS numbers of substances determined as endocrine disruptors at the EU level from List Ⅰ on the official website of the endocrine disruptor list constructed by ECHA, and delete substances without CAS numbers. Batch search for synonyms of the obtained CAS numbers in the Comptox Chemicals Dashboard database, and collect these synonyms to construct a list of endocrine disruptor names.

[0035] A total of 94 endocrine disruptors identified at the EU level were collected, as well as a list of 366 names of endocrine disruptors. For detailed information, please refer to Table 1.

[0036] S2. Construction of a list of endocrine-related events: Retrieve literature on endocrine-related adverse outcome pathways from the WEB OF SCIENCE website according to the following conditions: "Adverse outcome pathway (AOP) (topic) and Endocrine disrupt (topic) and Event (topic)", and extract events related to endocrinology; extract the standard test measurement endpoints from the "OECD Conceptual Framework for the Testing and Assessment of Endocrine Disruptors" document as events related to endocrinology; extract events related to endocrinology from the AOP-Wiki database.

[0037] 130 events related to endocrinology were extracted from the literature, 90 events related to endocrinology were extracted from the document, and 294 events related to endocrinology were extracted from the AOP-Wiki database. The endocrine-related events obtained from the three sources were simplified by their roots. Based on semantic similarity comparison, events with different expressions but the same essence were identified and merged, completely duplicate event entries were deleted, and events referring to the same biological process from different sources were integrated. Finally, an endocrine-related event list containing 278 events was obtained. For detailed information, please refer to Table 2.

[0038] S3. Preliminary locking of literature with potential endocrine AOP: With the help of the AOP-HelpFinder tool, click on the "Search Stressor–Event" module. Enter the list of endocrine disruptor names obtained in step S1 in the "Stressors file" module, and enter the list of endocrine-related events obtained in step S2 in the "Events file" module. The main parameters set include: Reduced search (20%), Output format (With abstracts), Refinement filter (Yes). In this part, 51,999 pieces of literature that may potentially link endocrine disruptors to related events were preliminarily locked, including information such as the PMID (PubMed Unique Identifier), abstract, and publication year of the literature.

[0039] S4. Screening and filtering to determine the final literature with potential AOP: Delete duplicate literature based on the PMIDs of the literature locked in step S3, and use an automated script to compare all literature PMIDs; then input the PMIDs of the literature locked in step S3 into the PUBMED website to filter out reviews and non-English literature; next, screen and filter the abstracts of the literature locked in step S3, and retain the literature whose abstracts contain "Mammal", "Rodent", "Mice", "Mouse", "Rat", "Human", and filter out non-mammal literature; finally, delete the literature irrelevant to endocrine disruptors and the literature with only one event.

[0040] A total of 31,002 duplicate literature, 2,023 reviews, 922 non-English literature, 738 non-mammal literature, 5,142 literature irrelevant to endocrine disruptors, and 7,720 literature with only one event were deleted, and finally 4,578 literature with potential AOP were obtained.

[0041] S5. Mining endocrine-related KER1: Conduct manual review of the abstracts or full texts of the literature with potential AOP determined in step S4. Priority is given to abstract review. For those with obvious conclusions in the abstract part, such as the form of "Event A" leading to "Event B", it is regarded as the endocrine-related KER in the literature and recorded in the form of "Source: Event A; Target: Event B". If there is no obvious conclusion in the abstract, full text review is carried out to extract clear event relationships in the conclusion part. If there is no clear event relationship in the full text, the literature is regarded as having no potential AOP. A total of 1,450 endocrine-related KER1 were mined in this part.

[0042] Especially, the research on endocrine disruptors is scattered in multiple fields such as toxicology, endocrinology, and reproductive science, resulting in highly dispersed relevant knowledge. This solution realizes the conversion from unstructured text to structured data through the triple <trigger word, source event, target event> structure. The triple structure provides a unified framework, facilitating the integration of different research results, representing the full path from molecular initiating events to adverse outcomes, and providing a scientific basis for mechanism-based risk assessment.

[0043] S6. Integrate KER2 in the literature and databases: Search the 294 endocrine-related events identified in the AOP-Wiki database in step S2 in the AOP-Wiki database in sequence to obtain the KERs between the 294 events, and only retain the KERs with high confidence. A total of 268 high-confidence endocrine-related KER2s were obtained through this search in the AOP-Wiki database. Specifically, the construction of the existing adverse outcome pathway network for endocrine disruption effects is inaccurate and has too many false-positive associations due to uneven data quality, insufficient reliability verification, and conflicting research results. Therefore, this application introduces a confidence evaluation mechanism and a threshold screening system to reduce the probability of introducing false-positive associations and reasonably handle conflicting data through confidence weighing.

[0044] Integrate the 1450 endocrine-related KER1 mined from the literature in step S5 with the 268 endocrine-related KER2s in the AOP-Wiki database and delete the duplicates. A total of 101 duplicate KERs were found in this part. After removing the duplicates, a total of 1617 endocrine-related KERs were obtained and recorded in the form of "Source" and "Target".

[0045] S7. Construct the AOP network for endocrine disruptors: Classify the events included in the endocrine-related KERs obtained in step S6. Identify the events that only serve as "Source" as molecular initiating events (MIEs), the events that serve as "Target" as adverse outcomes (AOs), and the remaining events as key events (KEs). Further classify the AOs into reproduction, development, neuro, immune, metabolism, liver, and cancer.

[0046] Import the xlsx file with "Source" as one column and "Target" as one column, the txt file of event classification, and the txt file of AO classification of the endocrine-related KERs obtained in step S6 into the Cyctoscape software to automatically generate a network diagram. Use the tools in the "style" module to perform simple network diagram format settings and distinguish different types of events with different colors. The AOP network for endocrine disruptors is as Figure 2 shown

[0047] S8. AOP network analysis: Use the "Analyze Network" module of the Cyctoscape software to automatically calculate the Degree, Betweenness Centrality, and Eccentricity of the events in the AOP network diagram visualized in step S7 and sort them to obtain the key events.

[0048] The Degree values of the final 64 events are greater than 10. For example, events such as "decreased testosterone synthesis", "increased apoptosis", and "pathological changes in testicular tissue", the larger the Degree value, the more important the function of the event may be in the AOP network. The Degree value distribution of all events is as Figure 3 shown.

[0049] The Betweenness Centrality values of the final 25 events are greater than 1. For example, events such as "altered estradiol levels", "activation of estrogen receptors", and "pathological changes in liver tissue", the larger the Betweenness Centrality value, the more important the function of the event may be in the AOP network. The Betweenness Centrality value distribution of all events is as Figure 4 shown.

[0050] The Eccentricity values of the final 185 events are less than 1. For example, events such as "cryptorchidism in offspring", "damage to male germ cells", and "pathological changes in ovarian tissue", the smaller the Eccentricity value, the more important the function of the event may be in the AOP network. The Eccentricity value distribution of all events is as Figure 5 shown.

[0051] Table 1 List of endocrine disruptor names

[0052]

[0053]

[0054]

[0055]

[0056]

[0057]

[0058]

[0059]

[0060]

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067] Table 2 List of Endocrine Disruptor - related Events

[0068]

[0069]

[0070]

[0071]

[0072]

[0073] The present application and its implementation manners are schematically described above. This description is not restrictive. Without departing from the spirit or basic characteristics of the present application, the present application can be implemented in other specific forms. What is shown in the drawings is only one of the implementation manners of the present application, and the actual structure is not limited thereto. Any reference signs in the claims should not limit the claimed claims. Therefore, if those of ordinary skill in the art are inspired by it and, without departing from the gist of the present creation, design similar structural manners and embodiments to this technical solution without creative efforts, they shall fall within the protection scope of this patent. In addition, the word "comprising" does not exclude other elements or steps, and the word "a" before an element does not exclude including "a plurality of" such elements. The plurality of elements stated in the product claims can also be implemented by one element through software or hardware. The words such as "first" and "second" are used to denote names and do not represent any specific order.

Claims

1. A method for constructing a pathway network of harmful outcomes of endocrine disruption effects, characterized in that: include: Obtain endocrine data; Preprocess the acquired endocrine data; Based on the preprocessed endocrine data, a triplet of event causal relationships reflecting the endocrine disruption effect is constructed: trigger word, source event, target event, as data KER1; According to the preprocessed endocrine data, endocrine data with a confidence level greater than a preset threshold is obtained from a preset literature database as data KER2; The data KER1 and KER2 were deduplicated to obtain the final data of harmful outcome pathways of endocrine disruption effects; Based on the final endocrine disruption effect harmful outcome pathway data, an endocrine disruption effect harmful outcome pathway network was established.

2. The method for constructing a pathway network of harmful outcomes of endocrine disruption effects according to claim 1, characterized in that: Preprocess the acquired endocrine data, including: Acquire known endocrine disrupting substances, and obtain corresponding synonyms from a preset endocrine disrupting substance literature database, and construct a list of endocrine disrupting substance names through data integration; the synonyms represent words whose semantic similarity is greater than a preset threshold; Obtain endocrine events from a preset literature database using a predefined list of endocrine terms and construct an endocrine event list; Based on the list of endocrine disruptors and the list of endocrine events, data on harmful outcome pathways of endocrine disrupting effects were obtained through text mining; Text filtering was performed on the data of harmful outcome pathways of endocrine disruption effects to obtain filtered literature data.

3. The method for constructing a pathway network of harmful outcomes of endocrine disruption effects according to claim 2, characterized in that: Text mining was used to obtain data on harmful outcome pathways of endocrine disruption effects, including: According to the list of endocrine disruptors and the list of endocrine events, the preset literature database is searched through text mining to obtain literature data containing endocrine disruptors and endocrine events; Count the co-occurrence frequencies of endocrine disruptors and endocrine events in the literature data, and take the literature data with a co-occurrence frequency greater than a preset threshold as the first literature data; Extracting metadata based on the first document data using a natural language processing method, and extracting endocrine disruptor data using a named entity recognition algorithm; the metadata includes the title, publication date, and abstract of the document; Based on metadata and endocrine disruptor data, the pathway data of harmful outcomes of endocrine disruption effects were constructed.

4. The method for constructing a pathway network of harmful outcomes of endocrine disruption effects according to claim 2, characterized in that: The text of the harmful outcome pathway data of endocrine disruption effects was filtered to obtain the filtered literature data, including: The data of harmful outcome pathways of endocrine disrupting effects were deduplicated, and duplicate data were deleted to obtain the data after the first filtering; Based on the data after the first filtering, the review literature was identified and deleted using PubMed screening rules to obtain the data after the second filtering, wherein the review literature is the literature that does not contain experimental data; According to the data after the second filtering, the language of the document is identified through the language feature dictionary and grammatical rules, and Chinese document data is obtained as the data after the third filtering; Based on the data after the third filtering, the named entity recognition algorithm was used to identify the research objects and obtain literature data related to mammals as the data after the fourth filtering; The co-occurrence frequency of endocrine events in the data after the fourth filtering is counted, and the literature data with a co-occurrence frequency greater than a preset threshold is obtained as the filtered literature data.

5. The method for constructing a pathway network of harmful outcomes of endocrine disruption effects according to any one of claims 2 to 4, characterized in that: According to the preprocessed endocrine data, a causal relationship triplet <trigger word, source event, target event> reflecting the endocrine disruption effect is constructed as data KER1, including: According to the filtered literature data, the abstract text of the literature data is obtained by using the data structured analysis method; A rule-based approach is adopted to obtain data containing causal relationships from summary text using predefined causal relationship templates; According to the data containing causal relationships, based on the predefined event trigger word dictionary and semantic role labeling rules, an event causal relationship triple <trigger word, source event, target event> reflecting the endocrine disruption effect is constructed as data KER1.

6. The method for constructing a pathway network of harmful outcomes of endocrine disruption effects according to claim 5, characterized in that: Based on the final endocrine disruption effect harmful outcome pathway data, an endocrine disruption effect harmful outcome pathway network was established, including: According to the final endocrine disruption effect harmful outcome pathway data, the source events, target events and the relationships between events were extracted respectively; Map source events and target events as nodes, and map the relationships between events as directed edges; Based on nodes and directed edges, a pathway network of harmful outcomes of endocrine disruption effects was constructed.

7. The method for constructing a pathway network of harmful outcomes of endocrine disruption effects according to claim 6, characterized in that: After constructing the endocrine disruption effect harmful outcome pathway network, it also includes: Obtain the topological structure of the pathway network of harmful outcomes of endocrine disruption effects; According to the topological structure, the degree centrality, betweenness centrality and eccentricity of each event node are calculated respectively; the degree centrality indicates the degree of connection of the node, the betweenness centrality indicates the importance of the node as a path intermediary, and the eccentricity indicates the average distance from the node to other nodes in the network; According to the degree centrality, betweenness centrality and eccentricity of each event node, as well as the position of the event node in the network, the event nodes are divided into molecular initiating events (MIE), key events (KE) and adverse outcomes (AO).

8. The method for constructing a pathway network of harmful outcomes of endocrine disruption effects according to claim 7, characterized in that: The molecular initiating event MIE represents the starting molecular triggering event of the endocrine disrupting effect, the key event KE represents the event whose betweenness centrality in the endocrine disrupting effect is greater than a threshold, and the adverse outcome AO represents the harmful result caused by the endocrine disrupting effect.

9. A system for constructing a pathway network of harmful outcomes of endocrine disruption effects, characterized in that: include: Data collection module, building a list of endocrine disruptors and endocrine events; The data processing module obtains the initial endocrine disrupting effect harmful outcome pathway data and event relationship data from the preset literature database through text mining based on the endocrine disrupting substance name list and the endocrine event list; Based on the initial endocrine disruption effect harmful outcome pathway data and event relationship data, the final endocrine disruption effect harmful outcome pathway data are constructed; The network construction module establishes the endocrine disruption effect harmful outcome pathway network based on the final endocrine disruption effect harmful outcome pathway data.

10. A computer-readable storage medium storing computer instructions, which implement the method according to any one of claims 1 to 8 when executed by a processor.

Images