A method and system for prioritizing thyroid system disruptors

Abstract

The application discloses a kind of thyroid system interferent priority ranking method and system.The method first establishes the test database of the interference effect of 8 kinds of thyroid system target by existing substance, and constructs 8 kinds of thyroid system target interference effect binary classification prediction model based on machine learning algorithm using the data in test database, the interference effect data of target of target substance missing test data is predicted using prediction model, the interference effect prediction data of target of target substance without interference effect test data is obtained, and the interference effect data of 8 kinds of thyroid system target of target compound is integrated, then according to the number of interference effect positive data, the priority level of target compound to thyroid system interference effect is determined.The priority ranking method of the application has clear process, good method scalability, easy to program, and is suitable for determining the priority level of unknown chemicals on potential thyroid system interference effect.

Description

Technical Field

[0001] This invention relates to a method and system for prioritizing thyroid system disruptors, belonging to the technical field of endocrine disruptor screening strategies. Background Technology

[0002] Priority pollutant screening refers to selecting pollutants with significant potential environmental risks from tens of thousands of chemical substances in production and use, based on environmental and health hazards and environmental exposure data, and including them in the scope of priority environmental risk assessment. Currently, priority pollutants of concern both domestically and internationally include endocrine disruptors, persistent, bioaccumulative, toxic (PBTs), persistent, mobile, toxic (PMTs), and persistent organic pollutants (POPs). Experience shows that the implementation of chemical pollutant control requires the support of relatively complete environmental risk assessment and control technologies, chemical hazard databases, and computational toxicology modeling software. However, at present, there is a serious lack of information on chemical hazard and the development of computational toxicology is slow, which has become a significant constraint on chemical pollutant control. Therefore, developing a tool for prioritizing endocrine disruptors is of great significance for supporting the implementation of control measures for new EDCs (endocrine disruptors).

[0003] Studies have shown that endocrine disruptors (EDCs) affect the thyroid and other endocrine systems through mechanisms including: interference with targets related to endogenous hormone regulation, endocrine hormone synthesis, endogenous hormone transporters, endogenous hormone receptors, and targets related to endogenous hormone conversion. Theoretically, by determining whether a target substance can interfere with these endocrine targets, its endocrine disrupting effect can be identified. Alternatively, based on the difference in the number of positive interference effects on endocrine targets, different priority levels can be assigned to substances. Currently, literature (Garcia de Lomana M, et al. In SilicoModels to Predict the Perturbation of Molecular Initiating Events Related to Thyroid Hormone Homeostasis. Chem Res Toxicol. 2021 Feb 15; 34(2):396-411.), patents (CN105893759B, CN110146695B), and software copyrights (2021SR1567339) disclose predictive models and software tools that can predict the potential interference effects of organic chemicals on thyroid system targets. However, no method has been reported on how to prioritize the interference effects of target compounds on the thyroid system based on data on the potential interference effects of organic chemicals on thyroid system targets. Summary of the Invention

[0004] The purpose of this invention is to provide a method for prioritizing thyroid system disruptors, which prioritizes thyroid system disruptors based on experimental or predicted data on the interference effects of thyroid system targets.

[0005] The technical solution for achieving the objective of this invention is as follows:

[0006] A method for prioritizing thyroid system disruptors includes the following steps:

[0007] (1) Construct an experimental database of interference effects of eight thyroid system targets: Collect existing experimental data on the interference effects of substances on targets of the thyroid hormone regulation system, thyroid synthesis system, thyroid transport system, thyroid transformation system, and thyroid receptor and construct a database. The targets of the thyroid hormone regulation system include thyrotropin-releasing hormone receptor (TRHR) and thyroid-stimulating hormone receptor (TSHR). The targets of the thyroid synthesis system include sodium / iodine transporter (NIS) and thyroxine peroxidase (TPO). The target of the thyroid transport system is transthyretin protein (TTR). The target of the thyroid transformation system is deiodinase (DIO). The thyroid receptors include thyroxine alpha receptor (TRα) and thyroxine beta receptor (TRβ).

[0008] (2) Constructing a binary classification prediction model for the interference effect of thyroid system targets: Using the experimental data of the interference effect of 8 thyroid system targets collected in step (1), 8 prediction models for thyroid system targets were constructed based on machine learning algorithms. The optimal models for transthyretin protein, thyroxine alpha receptor and thyroxine beta receptor were based on the k-nearest neighbor algorithm, while the optimal models for thyrotropin-releasing hormone receptor, thyrotropin receptor, sodium / iodine transporter, thyroxine peroxidase and deiodinase were based on the decision tree algorithm.

[0009] (3) Obtain interference effect data of target substance: Based on the experimental database of step (1), obtain the interference effect test data of target substance on targets with existing interference effect test data. For targets without interference effect test data, use the prediction model of related targets in step (2) to predict the interference effect data of targets with missing test data, and obtain the interference effect prediction data of target substance on targets without interference effect test data. Integrate the interference effect test data and interference effect prediction data of target substance to obtain the interference effect data of target substance on 8 thyroid system targets.

[0010] (4) Priority ranking: If the target substance has positive interference effect data for 6-8 thyroid system targets, the substance is set as a high-priority thyroid system interferon; if the target substance has positive interference effect data for 3-5 thyroid system targets, the substance is set as a medium-priority thyroid system interferon; if the target substance has positive interference effect data for 0-2 thyroid system targets, the substance is set as a low-priority thyroid system interferon.

[0011] Furthermore, in step (1), experimental data are collected from academic papers, authoritative software, and databases.

[0012] Furthermore, in step (2), the machine learning algorithm is selected from k-nearest neighbor algorithm, logistic regression algorithm, support vector machine algorithm, decision tree algorithm or random forest algorithm.

[0013] Further, in step (2), the specific method for constructing the binary classification prediction model of thyroid system target interference effect is as follows: The padelpy tool is used to calculate the 1D and 2D molecular structure descriptors and Pubchem molecular fingerprint descriptors of the chemical substances for which thyroid system target interference effect experimental data were obtained in step (1). Then, the machine learning algorithms contained in the scikit-learn module of the Python program are used to construct binary classification models of interference effects for the eight targets of the thyroid system. During modeling, the original datasets of each target are randomly split into training and validation sets. The sensitivity (S) of the training and validation sets is used. n ), specificity (S) p The classification performance is characterized by prediction accuracy (Q), Matthews correlation coefficient (MCC), receiver operating characteristic (ROC) curve, and area under the ROC curve (AUC). The application domain of the classification model is characterized by a method based on Euclidean distance.

[0014] Furthermore, in step (2), the Tanimoto similarity index (T) calculated based on Pubchem molecular fingerprint descriptors is used. s The reliability of the prediction model is evaluated using the following formula:

[0015]

[0016] Among them, A X-i and B X-i These are the i-th molecular structure descriptors of compound A in the validation set and compound B in the training set, respectively. If compound AT S ≥90% indicates that compound A has a similarity of greater than or equal to 90% with at least one compound in the corresponding model training set, meaning the prediction result has "high reliability"; if compound AT SA similarity between 75% and 90% indicates that compound A shares a similarity with at least one compound in the corresponding model training set, meaning the prediction result has "medium reliability"; if compound AT... S If the similarity is ≤75%, it means that compound A has a similarity of less than 75% with the corresponding compounds in the training set of the model, which means that the prediction results have "low reliability".

[0017] A thyroid system disruptor priority ranking system, comprising:

[0018] The target substance input module is used to input the target substance. If the input target substance information is valid, the module for acquiring the target substance interference effect data will be called.

[0019] The module for acquiring interference effect data of target substances is used to obtain interference effect data of target substances on eight thyroid system targets, including an experimental database of interference effects on the eight thyroid system targets and a binary classification prediction model for interference effects on thyroid system targets.

[0020] The aforementioned experimental database of interference effects on eight thyroid system targets stores existing experimental data on the interference effects of substances on targets of the thyroid hormone regulation system, thyroid synthesis system, thyroid transport system, thyroid transformation system, and thyroid receptor. Based on the input target substance, the database retrieves existing experimental data on the interference effects of the target substance on thyroid system targets.

[0021] The binary classification prediction model for the target interference effect of the thyroid system predicts the interference effect data of the target in the test data of the missing target substance, and obtains the predicted data of the interference effect of the target substance on the target in the test data of the test data without interference effect.

[0022] The priority output module is used to determine the priority of a target substance based on the obtained interference effect data of the target substance on 8 thyroid system targets. If the target substance has positive interference effect data for 6-8 thyroid system targets, the substance is set as a high-priority thyroid system interferon; if the target substance has positive interference effect data for 3-5 thyroid system targets, the substance is set as a medium-priority thyroid system interferon; if the target substance has positive interference effect data for 0-2 thyroid system targets, the substance is set as a low-priority thyroid system interferon.

[0023] Furthermore, the input information for the target substance is the SMILES code, CSA number, or structure file, and the input mode is either single substance input mode or batch substance input mode.

[0024] Compared with the prior art, the present invention has the following advantages:

[0025] (1) The prediction model has good fit, robustness and predictive ability, and the priority setting strategy is easy to program.

[0026] (2) The screening method has good scalability, and the classification model of new targets can be easily added to the priority setting system;

[0027] (3) The screening method supports 5 input modes, which can quickly and efficiently set the priority of thyroid system interference effects for a large number of chemicals.

[0028] (4) Screening methods can help identify high-priority thyroid system disruptors that need to be managed, thereby providing strong support for the treatment of new pollutants. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the priority ranking method for thyroid system disruptors.

[0030] Figure 2 It is a binary classification prediction model for the interference effects of 8 thyroid system targets.

[0031] Figure 3 This is a schematic diagram of the input interface for the thyroid system interferon priority ranking system.

[0032] Figure 4 This is a schematic diagram of the results display interface of the thyroid system interferon priority ranking system. Detailed Implementation

[0033] The present invention will now be described in further detail with reference to the embodiments and accompanying drawings.

[0034] The thyroid system disruptor priority ranking method of the present invention mainly includes the following steps: Figure 1 As shown, the core of this strategy includes two aspects: first, obtaining data on the interference effects of the target compound on eight thyroid system targets; and second, determining the priority of the target compound's interference effect on the thyroid system based on the number of positive interference effect data for the eight thyroid system targets.

[0035] To obtain data on the interference effects of the target compound on eight thyroid system targets, an existing experimental database of the interference effects of the substance on the eight thyroid system targets was first established. Then, based on machine learning algorithms, a binary classification prediction model for the interference effects of the target compound on the eight thyroid system targets was constructed using the experimental database. The prediction model was used to predict the interference effect data of the target substance on targets with missing experimental data, and the predicted interference effect data of the target substance on targets without interference effect experimental data was obtained. The integration of the two is the interference effect data of the target compound on the eight thyroid system targets.

[0036] The method for determining the priority of the interference effect of the target compound on the thyroid system is as follows: if the target compound has positive interference effect data for 6-8 thyroid system targets, the substance is set as a high-priority thyroid system interferon; if the target organic chemical has positive interference effect data for 3-5 thyroid system targets, the substance is set as a medium-priority thyroid system interferon; if the target organic chemical has positive interference effect data for 0-2 thyroid system targets, the substance is set as a low-priority thyroid system interferon.

[0037] Example 1

[0038] This embodiment specifically illustrates the implementation of the sorting method of the present invention.

[0039] A method for prioritizing thyroid system disruptors, the specific steps of which are as follows:

[0040] (1) Construct an experimental database of interference effects on eight thyroid system targets:

[0041] Existing experimental data on the interference effects of substances on targets of the thyroid hormone regulatory system, thyroid synthesis system, thyroid transport system, thyroid transformation system, and thyroid receptor were collected and a database was constructed. The targets of the thyroid hormone regulatory system include thyrotropin-releasing hormone receptor and thyroid-stimulating hormone receptor. The targets of the thyroid synthesis system include sodium / iodine transporters and thyroxine peroxidase. The target of the thyroid transport system is transthyretin protein. The target of the thyroid transformation system is deiodinase. The thyroid receptors include thyroxine alpha receptor and thyroxine beta receptor.

[0042] Specifically, experimental data on the interference effects of eight thyroid system targets were collected from academic papers, authoritative software, and databases. These included: 6618 substances from the thyrotropin-releasing hormone receptor (TRH) (70 positive); 6789 substances from the thyroid-stimulating hormone receptor (TSH) (202 positive); 1771 substances from the sodium / iodine transporter (112 positive); 1051 substances from thyroxine peroxidase (255 positive); 445 substances from transthyretin protein (229 positive); 1717 substances from deiodinase (108 positive); 5462 substances from thyroxine alpha (124 positive); and 5480 substances from the beta receptor (143 positive).

[0043] Experimental data on interference effects against eight thyroid system targets were compiled to construct a database containing 8,349 chemical substances. The database contains 29,333 experimental data points, of which 1,243 data points for positive interference effects were found for 970 substances.

[0044] (2) Constructing a binary classification prediction model for thyroid system target interference effects: Using the experimental data of 8 types of thyroid system target interference effects collected in step (1), 8 prediction models for thyroid system targets were constructed based on machine learning algorithms. Specifically:

[0045] The Padelpy tool (Python version PaDEL-Descriptor, https: / / github.com / ecrl / padelpy) was used to calculate the 1D and 2D molecular structure descriptors and Pubchem molecular fingerprint descriptors of the chemical substances with existing thyroid system target interference effect experimental data in step (1). Then, machine learning algorithms such as k-nearest neighbor, logistic regression, support vector machine, decision tree, and random forest contained in the scikit-learn module of the Python program were used to construct binary classification models of interference effects of 8 targets of the thyroid system. During modeling, the original dataset of each target was randomly split into a training set (75%) and a validation set (25%). The sensitivity (S) of the training set and the validation set was used. n ), specificity (S) p The classification performance was characterized by prediction accuracy (Q), Matthews correlation coefficient (MCC), receiver operating characteristic (ROC) curve, and area under the ROC curve (AUC). The application domain of the classification model was characterized using a method based on Euclidean distance. Furthermore, the Tanimoto similarity index (Ti) calculated based on Pubchem molecular fingerprint descriptors was used. s The reliability of the prediction model is evaluated using the following formula:

[0046]

[0047] Among them, A X-i and B X-i These are the i-th molecular structure descriptors of compound A in the validation set and compound B in the training set, respectively. If compound AT S ≥90% indicates that compound A has a similarity of greater than or equal to 90% with at least one compound in the corresponding model training set, meaning the prediction result has "high reliability"; if compound AT S A similarity between 75% and 90% indicates that compound A shares a similarity with at least one compound in the corresponding model training set, meaning the prediction result has "medium reliability"; if compound AT... S If the similarity is ≤75%, it means that compound A has a similarity of less than 75% with the corresponding compounds in the training set of the model, which means that the prediction results have "low reliability".

[0048] The results showed that the optimal models for thyroxine protein, thyroxine alpha receptor, and thyroxine beta receptor were based on the k-nearest neighbor algorithm; while the optimal models for thyrotropin-releasing hormone receptor, thyrotropin receptor, sodium / iodine transporter, thyroxine peroxidase, and deiodinase were based on the decision tree algorithm. The optimal model parameters for the binary classification prediction models of target interference effects in each thyroid system are as follows: Figure 2 As shown.

[0049] (3) Obtain interference effect data of the target substance: Based on the experimental database of step (1), obtain the interference effect test data of the target substance on the target with existing interference effect test data. For the target without interference effect test data, use the prediction model of the relevant target in step (2) to predict the interference effect data of the target with missing test data, and obtain the interference effect prediction data of the target substance on the target without interference effect test data. Integrate the interference effect test data and interference effect prediction data of the target substance to obtain the interference effect data of the target substance on 8 thyroid system targets.

[0050] (4) Priority ranking: If the target substance has positive interference effect data for 6-8 thyroid system targets, the substance is set as a high-priority thyroid system interferon; if the target substance has positive interference effect data for 3-5 thyroid system targets, the substance is set as a medium-priority thyroid system interferon; if the target substance has positive interference effect data for 0-2 thyroid system targets, the substance is set as a low-priority thyroid system interferon.

[0051] Example 2

[0052] This embodiment specifically illustrates the implementation method of the system of the present invention.

[0053] A thyroid system disruptor priority ranking system, comprising:

[0054] The target substance input module is used to input target substances. If the input target substance information is valid, it calls the target substance interference effect data acquisition module. There are five input modes for target substances: single substance input mode based on SMILES code or CSA number; batch input mode based on SMILES code (including CSV, Text, and SMI files containing SMILES code) or CSA number (including CSV and Text files containing CSA number); and substance structure file input mode (SDF and MOL files). Among these, the single substance or batch input mode based on substance CSA number cannot evaluate substances whose CAS number is not in the experimental database of interference effects of eight thyroid system targets. The single substance or batch input mode based on SMILES code or structure file can evaluate substances as long as the SMILES code or structure file is correct. Figure 3The image shown is a schematic diagram of the input interface for the target substance input module.

[0055] The module for acquiring interference effect data of target substances is used to obtain interference effect data of target substances on eight thyroid system targets, including an experimental database of interference effects on the eight thyroid system targets and a binary classification prediction model for interference effects on thyroid system targets.

[0056] The aforementioned experimental database of interference effects on eight thyroid system targets stores existing experimental data on the interference effects of substances on targets of the thyroid hormone regulation system, thyroid synthesis system, thyroid transport system, thyroid transformation system, and thyroid receptor. Based on the input target substance, the database retrieves existing experimental data on the interference effects of the target substance on thyroid system targets.

[0057] The binary classification prediction model for the target interference effect of the thyroid system predicts the interference effect data of the target in the test data of the missing target substance, and obtains the predicted data of the interference effect of the target substance on the target in the test data of the test data without interference effect.

[0058] The priority output module is used to determine the priority of a target substance based on the obtained interference effect data of the target substance on eight thyroid system targets. If the target substance has positive interference effect data for 6-8 thyroid system targets, the substance is set as a high-priority thyroid system interferon; if the target substance has positive interference effect data for 3-5 thyroid system targets, the substance is set as a medium-priority thyroid system interferon; if the target substance has positive interference effect data for 0-2 thyroid system targets, the substance is set as a low-priority thyroid system interferon. Figure 4 The diagram shows the result display interface of the priority output module, which displays the basic information of the target substance (SMILES code, CAS number) and the priority evaluation result.

[0059] Example 3

[0060] 2,4-Di-tert-butylphenol (CAS No.: 96-76-4) is a medium-priority thyroid system disruptor. The steps for evaluating its thyroid system disruptor priority using this invention are as follows:

[0061] First, targets with experimental data showing interference effects were identified. Database queries revealed four targets with experimental data: thyroid-stimulating hormone receptor (TSH), sodium / iodine transporter, thyroxine peroxidase, and deiodinase, all showing positive interference effects. However, no experimental data were available for TSH-releasing hormone receptor (TRH), thyroxine alpha receptor, thyroxine beta receptor, and transthyroxine protein. A predictive model was used to fill in the data gaps. Specifically, the corresponding molecular descriptors for the TSH-releasing hormone receptor, thyroxine alpha receptor, thyroxine beta receptor, and transthyroxine protein interference effect prediction models were calculated. The missing data were then used to evaluate the application domain and reliability. The prediction results showed that 2,4-di-tert-butylphenol (2,4-di-tert-butylphenol) had a positive interference effect on thyroxine alpha receptor, while other targets were predicted to be negative. Based on the experimental and predicted positive interference effect data, 2,4-di-tert-butylphenol showed a positive interference effect on five thyroid system targets, classifying it as a medium-priority thyroid system interferon.

[0062] Example 4

[0063] Furazolidone (CAS No.: 67-45-8) is a medium-priority thyroid system disruptor. The steps for evaluating its thyroid system disruptor priority using this invention are as follows:

[0064] First, targets with experimental data showing interference effects were identified. Database queries revealed two targets with available experimental data: thyroxine alpha receptor and thyroxine beta receptor, both showing negative interference effects. However, no experimental data were available for thyroid-stimulating hormone receptor, thyrotropin-releasing hormone receptor, sodium / iodine transporter, thyroxine peroxidase, deiodinase, and transthyroxine protein. Predictive models were used to fill in the data gaps. Specifically, molecular descriptors for the interference effect prediction models of thyroid-stimulating hormone receptor, thyrotropin-releasing hormone receptor, sodium / iodine transporter, thyroxine peroxidase, deiodinase, and transthyroxine protein were calculated. The missing data were then predicted using the models, and the application domain and reliability were assessed. The prediction results showed that furazolidone had a positive interference effect on thyroid-stimulating hormone receptor, sodium / iodine transporter, transthyroxine protein, and deiodinase, while the other targets were predicted to be negative. Based on the predicted positive interference effect data, furazolidone showed a positive interference effect on four thyroid system targets, classifying it as a medium-priority thyroid system interferon.

[0065] Example 5

[0066] Dexamethasone (CAS No.: 50-02-2) is a low-priority thyroid system disruptor. The steps for evaluating its thyroid system disruptor priority using this invention are as follows:

[0067] First, targets with experimental data showing interference effects were identified. Database queries revealed two targets with experimental data: thyroid-stimulating hormone receptor (TSH) and thyrotropin-releasing hormone receptor (TRH), both showing negative interference effects. However, no experimental data were available for sodium / iodine transporters, thyroxine peroxidase (TPO), deiodinase, thyroxine receptor alpha (TRH), thyroxine receptor beta (TRH), and transthyretin protein. Predictive models were used to fill in the data gaps. Specifically, molecular descriptors for the predicted interference effects of TPO were calculated for TPO, thyroxine peroxidase (TPO), deiodinase (TRH), thyroxine receptor alpha (TRH), thyroxine receptor beta (TRH), and transthyretin protein. The missing data were then predicted using the models, and the application domain and reliability were assessed. The prediction results showed that dexamethasone had negative interference effects on TPO, thyroxine peroxidase (TPO), deiodinase (TRH), thyroxine alpha receptor (TRH), thyroxine beta receptor (TRH), and transthyretin protein. Based on experimental and predicted positive interference effect data, dexamethasone showed negative interference effects on all eight thyroid system targets, classifying it as a low-priority thyroid system disruptor.

Claims

1. A method for prioritizing thyroid system disruptors, characterized in that, Includes the following steps: (1) Construct an experimental database of interference effects of eight thyroid system targets: Collect existing experimental data on interference effects of substances on targets of the thyroid hormone regulation system, thyroid synthesis system, thyroid transport system, thyroid transformation system, and thyroid receptor and construct a database. The targets of the thyroid hormone regulation system include thyrotropin-releasing hormone receptor and thyroid-stimulating hormone receptor. The targets of the thyroid synthesis system include sodium / iodine transporter and thyroxine peroxidase. The targets of the thyroid transport system are transthyretin protein. The targets of the thyroid transformation system are deiodinase. The thyroid receptors include thyroxine alpha receptor and thyroxine beta receptor. (2) Constructing a binary classification prediction model for the interference effect of thyroid system targets: Using the experimental data of the interference effect of 8 thyroid system targets collected in step (1), 8 prediction models for thyroid system targets were constructed based on machine learning algorithms. The optimal models for transthyretin protein, thyroxine alpha receptor and thyroxine beta receptor were based on the k-nearest neighbor algorithm, and the optimal models for thyrotropin-releasing hormone receptor, thyrotropin receptor, sodium / iodine transporter, thyroxine peroxidase and deiodinase were based on the decision tree algorithm. (3) Obtain interference effect data of target substance: Based on the experimental database of step (1), obtain the interference effect test data of target substance on targets with existing interference effect test data. For targets without interference effect test data, use the prediction model of related targets in step (2) to predict the interference effect data of targets with missing test data, and obtain the interference effect prediction data of target substance on targets without interference effect test data. Integrate the interference effect test data and interference effect prediction data of target substance to obtain the interference effect data of target substance on 8 thyroid system targets. (4) Priority ranking: If the target substance has positive interference effect data for 6-8 thyroid system targets, the substance is set as a high-priority thyroid system interferon; if the target substance has positive interference effect data for 3-5 thyroid system targets, the substance is set as a medium-priority thyroid system interferon; if the target substance has positive interference effect data for 0-2 thyroid system targets, the substance is set as a low-priority thyroid system interferon.

2. The method for prioritizing thyroid system disruptors according to claim 1, characterized in that, In step (1), experimental data were collected from academic papers, authoritative software, and databases.

3. The method for prioritizing thyroid system disruptors according to claim 1, characterized in that, In step (2), the machine learning algorithm is selected from k-nearest neighbor algorithm, logistic regression algorithm, support vector machine algorithm, decision tree algorithm or random forest algorithm.

4. The method for prioritizing thyroid system disruptors according to claim 1, characterized in that, In step (2), the specific method for constructing the binary classification prediction model of the interference effect of the thyroid system target is as follows: The padelpy tool is used to calculate the 1D and 2D molecular structure descriptors and Pubchem molecular fingerprint descriptors of the chemical substances in the experimental data of the interference effect of the thyroid system target in step (1). Then, the machine learning algorithm contained in the scikit-learn module in the Python program is used to construct the binary classification model of the interference effect of the 8 targets of the thyroid system. When modeling, the original dataset of each target is randomly split into training set and validation set. The sensitivity, specificity, prediction accuracy, Matthews correlation coefficient, receiver operating characteristic curve and area under ROC curve of the training set and validation set are used to characterize the classification performance. The application domain of the classification model is characterized by the method based on Euclidean distance.

5. The method for prioritizing thyroid system disruptors according to claim 1, characterized in that, In step (2), the Tanimoto similarity index calculated based on Pubchem molecular fingerprint descriptors is used. T s The following formula is used to evaluate the reliability of the prediction model: (1)， in, A X-i and B X-i These are the first two compounds in the validation set (compound A) and the training set (compound B), respectively. i A molecular structure descriptor, if compound A T S ≥ 90% indicates that compound A has a similarity of greater than or equal to 90% with at least one compound in the corresponding model training set, meaning the prediction result has "high reliability"; if compound A T S A similarity between 75% and 90% indicates that compound A shares a similarity with at least one compound in the corresponding model training set, meaning the prediction result has "medium reliability"; if compound A... T S If the similarity is less than 75%, it means that compound A has a similarity of less than 75% with the corresponding model training set compounds, which means that the prediction results have "low reliability".

6. A priority ranking system for thyroid system disruptors, characterized in that, include: The target substance input module is used to input the target substance. If the input target substance information is valid, the module for acquiring the target substance interference effect data will be called. The module for acquiring interference effect data of target substances is used to obtain interference effect data of target substances on eight thyroid system targets, including an experimental database of interference effects on the eight thyroid system targets and a binary classification prediction model for interference effects on thyroid system targets. The aforementioned experimental database of interference effects on eight thyroid system targets stores existing experimental data on the interference effects of substances on targets of the thyroid hormone regulation system, thyroid synthesis system, thyroid transport system, thyroid transformation system, and thyroid receptor. Based on the input target substance, the database retrieves existing experimental data on the interference effects of the target substance on thyroid system targets. The binary classification prediction model for the target interference effect of the thyroid system predicts the interference effect data of the target in the test data of the missing target substance, and obtains the predicted data of the interference effect of the target substance on the target in the test data of the test data without interference effect. The priority output module is used to determine the priority of a target substance based on the obtained interference effect data of the target substance on 8 thyroid system targets. If the target substance has positive interference effect data for 6-8 thyroid system targets, the substance is set as a high-priority thyroid system interferon; if the target substance has positive interference effect data for 3-5 thyroid system targets, the substance is set as a medium-priority thyroid system interferon; if the target substance has positive interference effect data for 0-2 thyroid system targets, the substance is set as a low-priority thyroid system interferon.

7. The thyroid system disruptor priority ranking system according to claim 6, characterized in that, The input information for the target substance is the SMILES code, CSA number, or structure file, and the input mode is either single substance input mode or batch substance input mode.

Images