A graded screening method for chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum

Abstract

This invention discloses a graded screening method for chemically induced lipid metabolism disorders based on nuclear receptor interference spectra, relating to the fields of environmental and health risk assessment and drug screening technology. The method includes: firstly, constructing a multi-dimensional detection system covering core nuclear receptors regulating lipid metabolism; secondly, co-incubating the sample with the system to measure the agonist / antagonist activity of each nuclear receptor, using the obtained activity detection data as a feature, and generating a multidimensional nuclear receptor interference spectrum; thirdly, establishing a quantitative correlation model between the interference spectrum and lipid metabolism disorder phenotypes based on a machine learning model; and finally, graded screening of chemicals based on quantitative indicators of nuclear receptor interference intensity and quantitative indicators of lipid metabolism disorder risk prediction probability. This invention integrates core regulatory nuclear receptors and functional nuclear receptors in the lipid metabolism regulatory network into a single screening system, overcoming the limitations of traditional methods with incomplete target coverage, and achieving a comprehensive and systematic analysis of the metabolic interference effects of chemicals.

Description

Technical Field

[0001] This invention relates to the fields of environmental and health risk assessment and drug screening technology, specifically to a graded screening method for chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum. Background Technology

[0002] Lipid metabolism disorders induced by chemicals (including environmental pollutants, industrial chemicals, pesticides, food additives, and drug candidates) have become a global public health problem. Lipid metabolism disorders not only lead to obesity and non-alcoholic fatty liver disease (NAFLD), but are also closely related to the development of chronic diseases such as cardiovascular disease and diabetes. Traditional chemical risk assessment mainly relies on animal experiments, which has limitations such as long cycles, high costs, low throughput, and species variability. Therefore, establishing a rapid, efficient, and accurate in vitro screening method is of great significance for the risk classification and safety management of chemicals.

[0003] Nuclear receptors (NRs) are a class of ligand-dependent transcription factors that play a central regulatory role in maintaining lipid metabolism homeostasis. For example, peroxisome proliferator-activated receptors (PPARs), liver X receptors (LXRs), and farnesol X receptors (FXRs) directly regulate the expression of genes related to fatty acid synthesis, oxidation, and cholesterol metabolism. Chemicals can disrupt lipid metabolism homeostasis by mimicking or blocking endogenous ligands, interfering with the normal function of these receptors. However, current assessments of the nuclear receptor effects of chemicals often focus on single or a few receptors, neglecting the broad-spectrum interference effects of chemicals on the entire nuclear receptor network, making it difficult to comprehensively predict the complex mechanisms by which they lead to lipid metabolism disorders. Furthermore, how to correlate multidimensional nuclear receptor interference data with the final lipid metabolism phenotype and establish a scientific and objective grading standard remains a pressing technical challenge in this field. Therefore, there is an urgent need to propose a graded screening method for chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum, in order to solve the problem that the low throughput and single target of nuclear receptor screening in the existing technology make it impossible to comprehensively assess the association between nuclear receptor network interference and lipid metabolism phenotype. Summary of the Invention

[0004] The purpose of this invention is to provide a graded screening method for chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum, in order to solve the problem in the prior art that the low throughput of nuclear receptor screening and single target make it impossible to comprehensively assess the association between nuclear receptor network interference and lipid metabolism phenotype.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for graded screening of chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum, comprising the following steps:

[0006] S1. Construct a multi-dimensional detection system for multinuclear receptors covering core nuclear receptors regulating lipid metabolism;

[0007] S2. The sample to be tested is co-incubated with the multi-nuclear receptor multidimensional detection system, and the agonist / antagonist activity of the chemical on each nuclear receptor is determined to obtain activity detection data;

[0008] S3. Based on the activity detection data, after data standardization processing, a multidimensional nuclear acceptor interference spectrum of the chemical is generated.

[0009] S4. Based on a multi-dimensionally validated machine learning model, establish a quantitative correlation model between nuclear receptor interference profiles and lipid metabolism disorder phenotypes.

[0010] S5. Based on the quantitative indicators of nuclear receptor interference intensity and the quantitative indicators of lipid metabolism disorder risk prediction, and combined with the cellular ALT / AST release level indicators, chemicals are screened in a four-level classification: no interference, weak interference, moderate interference, and strong interference.

[0011] Furthermore, the nuclear receptor in S1 includes a core regulatory nuclear receptor and a functional nuclear receptor. The core regulatory nuclear receptor is at least one of PPARα, PPARγ, PPARδ, LXRα, LXRβ, FXR, and RXR; the functional nuclear receptor is at least one of TRβ, RAR, and VDR.

[0012] Furthermore, the multi-dimensional detection system in S1 includes reporter gene detection, fluorescence polarization technology, time-resolved fluorescence resonance energy transfer technology, LC-MS / MS, Western blot protein verification, and flow cytometry analysis; wherein the reporter gene detection uses a dual-luciferase system for internal control normalization.

[0013] Furthermore, the sample to be tested in S2 includes at least one of pure compound, environmental mixture, and food contact material migration; the environmental mixture includes at least one of PM2.5 extract, sewage sample, landfill leachate, and indoor dust extract.

[0014] Furthermore, the sample pretreatment steps are as follows: pure compounds are dissolved in DMSO to prepare a series of concentrations from 0.1 to 100 μM; environmental mixtures are enriched by solid-phase extraction and then re-dissolved in DMSO; food contact material migrants are extracted according to national standard methods and then re-dissolved in DMSO; the cell detection system is incubated for 24 hours, and the cell-free detection system is incubated for 1 to 2 hours.

[0015] Furthermore, the activity detection data in S3 includes the half-maximal effect concentration (EC50) 50 / IC 50The agonist rate / competitive binding inhibition rate at specific concentrations were determined. Data standardization was performed using z-score normalization, and a multidimensional nuclear receptor interference spectrum vector was constructed from the standardized data.

[0016] Furthermore, the machine learning models in S4 include random forest, support vector machine, and ensemble learning XGBoost models; the machine learning models are validated in multiple dimensions, and the validation process includes: dividing the training set and test set in a 7:3 ratio, using 5-fold cross-validation to optimize the model, using LASSO regression and RF-RFE to select features; validating the predictive consistency of the model in different cell lines, animal models, and human sample datasets; and using SHAP values ​​and LIME tools to analyze the model's decision logic.

[0017] Furthermore, the lipid metabolism disorder phenotypes in S4 include lipid accumulation, fatty acid oxidation inhibition, and cholesterol homeostasis imbalance. During the construction of the association model, the Pearson correlation coefficient between the lipid metabolism disorder phenotype and the nuclear receptor interference spectrum was calculated to verify the significance of the association. At the same time, the cellular ALT / AST release level index was combined as a supplementary verification parameter for hepatocellular damage related to lipid metabolism disorder.

[0018] Furthermore, the quantification index of nuclear receptor interference intensity in S5 is based on EC. 50 The core value is the risk probability quantification index for lipid metabolism disorder risk prediction, which is based on the model's output risk probability and combined with cellular ALT / AST release levels. The quantification standard for the four-level screening is as follows:

[0019] No interference: EC of each nuclear receptor 50 >100μM, the model predicted a risk probability <0.3, and the ALT / AST release level was within 1.2 times that of the negative control;

[0020] Weak interference: Single nuclear receptor EC 50 Between 10 and 100 μM, the model predicted a risk probability of 0.3 to 0.5, and the ALT / AST release level was 1.2 to 1.5 times that of the negative control.

[0021] Moderate interference: 2-3 nuclear receptor ECs 50 Between 1 and 10 μM, the model predicted a risk probability of 0.5 to 0.7, and the ALT / AST release level was 1.5 to 2.0 times that of the negative control.

[0022] Strong interference: ≥4 nuclear receptor ECs 50 <10 μM or any nuclear receptor EC 50 <1μM, model predicted risk probability >0.7, and ALT / AST release level exceeded 2.0 times that of the negative control.

[0023] Compared with existing technologies, the present invention provides a graded screening method for chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum, which has the following beneficial effects:

[0024] (1) This invention integrates the core regulatory nuclear receptor and functional nuclear receptor in the lipid metabolism regulatory network into a single screening system, breaking through the limitations of traditional incomplete target coverage and realizing a comprehensive and systematic analysis of the chemical metabolism interference effect.

[0025] (2) This invention uses EC 50 Using model-predicted risk probability and cellular ALT / AST release levels as core molecular indicators, combined with clinical liver enzyme levels, a clear quantitative grading standard of four levels—no interference, weak interference, moderate interference, and strong interference—is established to provide a scientific basis that can be directly applied for chemical risk management.

[0026] (3) This invention introduces the XGBoost ensemble learning model, combines random forest and support vector machine to build a multi-model fusion system, optimizes the model through standardized training / test set partitioning, cross-validation and feature selection, verifies the generalization ability across datasets, and uses SHAP and LIME tools to improve model interpretability and establish a precise quantitative association between nuclear receptor interference spectrum and lipid metabolism disorder phenotype.

[0027] (4) This invention clarifies the specific methods for constructing interference spectra, standardizing data, and verifying correlation models. It verifies the significance of the correlation between spectra and phenotypes through Pearson correlation coefficients, provides graded cases and clinical verification data for specific chemicals, and ensures the effectiveness and reproducibility of the technical solution. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0029] Figure 1 The flowchart of a graded screening method for chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum provided by the present invention is shown. Detailed Implementation

[0030] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0031] Example:

[0032] Please see Figure 1A graded screening method for chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum includes the following steps:

[0033] S1. Construct a multi-dimensional detection system for multiple nuclear receptors covering core nuclear receptors regulating lipid metabolism; nuclear receptors include core regulatory nuclear receptors and functional nuclear receptors. The core regulatory nuclear receptor is at least one of PPARα, PPARγ, PPARδ, LXRα, LXRβ, FXR, and RXR; the functional nuclear receptor is at least one of TRβ, RAR, and VDR; the multi-dimensional detection system includes reporter gene detection, fluorescence polarization technology, time-resolved fluorescence resonance energy transfer technology, LC-MS / MS, Western blot protein verification, and flow cytometry analysis; among them, the reporter gene detection uses a dual-luciferase system for internal control normalization.

[0034] The specific implementation method is as follows: First, expression plasmids for each nuclear receptor are constructed: the coding genes for PPARα, PPARγ, PPARδ, LXRα, LXRβ, FXR, RXR, TRβ, RAR, and VDR are synthesized and cloned into the pcDNA3.1 vector to construct independent expression plasmids for each nuclear receptor; the corresponding response element sequences (PPRE, LXRE, FXRE, etc.) for each nuclear receptor are synthesized and cloned into the pGL3-Basic vector to construct reporter gene plasmids for each nuclear receptor; at the same time, the Renilla luciferase expression plasmid pRL-TK is constructed as an internal control.

[0035] Then, stable cell lines for each nuclear receptor were established: for each nuclear receptor, the corresponding nuclear receptor plasmid, the corresponding reporter gene plasmid and pRL-TK were co-transfected into HEK293T cells at a molar ratio of 9:9:2. After 2 weeks of G418 selection, independent cell lines that stably expressed a single nuclear receptor and its corresponding reporter gene were obtained.

[0036] Next, a cell-free system was constructed: ligand-binding domain proteins of each nuclear receptor were expressed and purified in E. coli, and high-purity proteins were obtained by nickel column affinity chromatography. The concentration was adjusted to 1 μg / μL and stored at -80℃ for later use.

[0037] Finally, a multi-dimensional detection system is established, including:

[0038] Reporter gene detection system: Using the independent and stable cell lines for each nuclear receptor established above, dual-luciferase detection reagents were prepared, detection parameters were set, and a transcriptional activity detection method with internal control normalization was established.

[0039] Fluorescent polarization detection system: Fluorescently labeled nuclear receptor ligands are incubated with purified proteins to establish a competitive binding detection method, with the detection wavelength adjusted according to the fluorescent label;

[0040] TR-FRET detection system: Prepare antibodies labeled with fluorescent donors and receptors to establish a method for detecting co-regulatory factor recruitment;

[0041] Auxiliary validation system: Establish LC-MS / MS lipid metabolite detection method and optimize chromatographic and mass spectrometry parameters; establish Western blot detection method and prepare primary and secondary antibodies for lipid metabolism-related proteins such as SREBP-1c, PPARγ, and LXRα; establish flow cytometry lipid uptake detection method and use Oil Red O fluorescence staining.

[0042] S2. The sample to be tested is co-incubated with a multi-nuclear receptor multi-dimensional detection system to determine the agonist / antagonist activity of the chemical on each nuclear receptor and obtain activity detection data; the sample to be tested includes at least one of pure compound, environmental mixture, and food contact material migration; the environmental mixture includes at least one of PM2.5 extract, sewage sample, landfill leachate, and indoor dust extract.

[0043] The specific implementation method is as follows: First, the test samples are prepared: Bisphenol A (BPA), dibutyl phthalate (DBP), and perfluorooctane sulfonic acid (PFOS) are selected as the pure compounds to be tested, and are dissolved in DMSO to prepare a series of concentrations of 0.1, 1, 10, 50, and 100 μM; effluent from urban sewage treatment plants is collected as an environmental mixture sample, enriched by solid-phase extraction, redissolved in DMSO, and diluted to 1 mL; polypropylene food contact material is selected, and migration tests are conducted according to the national standard GB31604.1-2015. The migration liquid is collected and redissolved in DMSO after solid-phase extraction.

[0044] Then, sample exposure and activity detection were performed to obtain activity detection data. Specifically, the above-mentioned test samples were added to the multidimensional detection system constructed by S1, with three replicates for each concentration. A blank control (DMSO) and a positive control (T0901317 is an LXR agonist and GW9662 is a PPARγ antagonist) were also set up. After incubating the cell system for 24 hours, the fluorescence intensity was detected using a dual-luciferase assay kit, and the agonist / antagonist activity of each nuclear receptor was calculated. After incubating the cell-free system for 1 hour, the fluorescence polarization value was detected, and the competitive binding inhibition rate was calculated. At the same time, the content of lipid metabolites triglycerides and cholesterol was detected by LC-MS / MS, the expression of SREBP-1c protein was detected by Western blot, and the lipid uptake efficiency of HepG2 cells was detected by flow cytometry.

[0045] S3. Based on the activity detection data, after data standardization processing, a multidimensional nuclear acceptor interference spectrum of the chemical is generated; the activity detection data includes the half-maximal effect concentration (EC50). 50 / IC 50The agonist rate / competitive binding inhibition rate at specific concentrations were determined. Data standardization was performed using z-score normalization, and a multidimensional nuclear receptor interference spectrum vector was constructed from the standardized data.

[0046] The specific implementation method involves data processing and interference spectrum generation based on activity detection data: GraphPadPrism software is used to calculate the EC50 of each nuclear receptor. 50 / IC 50 The z-score of all activity data was normalized using SPSS software. Based on the normalized PPARα, PPARγ, PPARδ, LXRα, LXRβ, FXR, RXR, TRβ, RAR, and VDR activity data, a multidimensional nuclear receptor interference spectral vector was constructed for each sample to form its own nuclear receptor interference fingerprint spectrum.

[0047] S4. Based on a multi-dimensionally validated machine learning model, establish a quantitative correlation model between nuclear receptor interference spectrum and lipid metabolism disorder phenotypes. The machine learning model was validated in multiple dimensions, including: dividing the training and test sets in a 7:3 ratio, optimizing the model using 5-fold cross-validation, and performing feature selection through LASSO regression and RF-RFE; verifying the predictive consistency of the model in different cell lines, animal models, and human sample datasets; and using SHAP value and LIME tools to analyze the model's decision logic. Lipid metabolism disorder phenotypes include lipid accumulation, fatty acid oxidation inhibition, and cholesterol homeostasis imbalance.

[0048] The specific implementation method is as follows: First, a dataset is constructed: 50 known chemicals that cause lipid metabolism disorders and 30 chemicals that do not interfere with metabolism are selected as research objects, their nuclear receptor interference spectra are measured and lipid metabolism disorder phenotypes are detected to construct a dataset; the dataset is divided into a training set (56 types) and a test set (24 types) in a 7:3 ratio.

[0049] The 50 known chemicals that cause lipid metabolism disorders include:

[0050] Endocrine disruptors (20 types): Bisphenol A, Bisphenol F, Bisphenol S, Dibutyl Phthalate, Dioctyl Phthalate, Butylbenzyl Phthalate, Diethyl Phthalate, Dimethyl Phthalate, Perfluorooctane Sulfonate, Perfluorooctanoic Acid, Perfluorobutane Sulfonate, Perfluorohexane Carboxylic Acid, BDE-47, BDE-99, BDE-209, BDE-153, Tetrabromobisphenol A, Hexabromocyclododecane, Polychlorinated Biphenyls-153, Polychlorinated Biphenyls-138;

[0051] Classical nuclear receptor drug / modulator (15 types):

[0052] T0901317, GW3965, GW1929, WY14643, GW501516, GW9662, GSK864, Z-Gynecosterone, Resmetirom, Bexarotin, Rosiglitazone, Pioglitazone, Fenbufen, Fenofibrate, Benzafibrate;

[0053] High-risk chemicals in the industrial / food sector (15 types):

[0054] Benzo[a]pyrene, fluoranthene, pyrene, anthracene, glyphosate, malathion, cypermethrin, deltamethrin, Sudan I, Sudan IV, tartrazine, sunset yellow, methyl trans oleate, methyl trans linoleate, benzene[a]anthene.

[0055] The 30 chemicals that do not interfere with metabolism include:

[0056] Non-biologically active solvents / inert compounds (10 types):

[0057] Dimethyl sulfoxide (≤0.1%), anhydrous ethanol (low concentration), polyethylene glycol 400, sucrose, mannitol, sodium hyaluronate, sodium carboxymethyl cellulose, glycerol, sorbitol, polypropylene glycol;

[0058] 10 industrial / food chemicals proven to have no lipid metabolism interference:

[0059] Vitamin C, citric acid, potassium sorbate, erythritol, steviol glycosides, tributyl citrate, polysorbate 80, sodium chloride, potassium chloride, potassium dihydrogen phosphate;

[0060] Pharmacological / experimental tools with no binding activity to nuclear receptors (10 types):

[0061] Bovine serum albumin, sodium fluorescein, sodium azide (low concentration), ibuprofen (low concentration), metronidazole, rifampin (low concentration), glucose, lactose, trehalose, gelatin;

[0062] Then, model training and optimization were performed: Random Forest, Support Vector Machine, and XGBoost models were trained with nuclear receptor interference spectrum vector as input and lipid accumulation rate as output; Feature selection was performed using RF-RFE to select PPARγ, LXRα, TRβ, and FXR as core features; 5-fold cross-validation was used to optimize model hyperparameters, with the learning rate of the XGBoost model set to 0.1 and the tree depth set to 6, and the number of decision trees in the Random Forest model set to 100.

[0063] Next, model validation was performed: On the test set, the XGBoost model achieved a prediction accuracy of 91.67%, outperforming Random Forest (83.33%) and Support Vector Machine (79.17%). Applying the model to mouse primary hepatocytes and human hepatocellular carcinoma Huh7 cells datasets, the prediction accuracies were 89.29% and 87.50%, respectively, indicating good generalization ability. SHAP analysis of the XGBoost model showed that PPARγ contributed the most to the prediction results (35%), followed by LXRα (28%) and TRβ (18%).

[0064] Finally, a correlation model was established: the XGBoost model was selected as the quantitative correlation model between nuclear receptor interference spectrum and lipid metabolism disorder phenotype. The Pearson correlation coefficient between the risk probability of lipid metabolism disorder predicted by the model and the actual lipid accumulation rate was calculated to be 0.89 (P < 0.001), indicating that the two have a significant positive correlation.

[0065] S5. Based on the quantitative indicators of nuclear receptor interference intensity and the quantitative indicators of lipid metabolism disorder risk prediction probability, chemicals are screened in a four-level classification: no interference, weak interference, moderate interference, and strong interference; the quantitative indicator of nuclear receptor interference intensity is based on EC50. 50 The core value is the risk probability quantification index for lipid metabolism disorder risk prediction, which is based on the model's output risk probability and combined with cellular ALT / AST release levels. The quantification standard for the four-level screening is as follows:

[0066] No interference: EC of each nuclear receptor 50 >100μM, the model predicted a risk probability <0.3, and the ALT / AST release level was within 1.2 times that of the negative control;

[0067] Weak interference: Single nuclear receptor EC 50 Between 10 and 100 μM, the model predicted a risk probability of 0.3 to 0.5, and the ALT / AST release level was 1.2 to 1.5 times that of the negative control.

[0068] Moderate interference: 2-3 nuclear receptor ECs 50 Between 1 and 10 μM, the model predicted a risk probability of 0.5 to 0.7, and the ALT / AST release level was 1.5 to 2.0 times that of the negative control.

[0069] Strong interference: ≥4 nuclear receptor ECs 50 <10 μM or any nuclear receptor EC 50 <1μM, model predicted risk probability >0.7, and ALT / AST release level exceeded 2.0 times that of the negative control.

[0070] The specific implementation method involves inputting the nuclear receptor interference spectra of bisphenol A, dibutyl phthalate, perfluorooctane sulfonic acid, and wastewater effluent samples from S2 into the XGBoost association model established in S4, outputting the risk prediction probability, and simultaneously detecting the levels of liver enzymes (ALT, AST) in HepG2 cells corresponding to each sample, combined with EC 50 The values ​​were categorized and screened, and the results are as follows:

[0071] Bisphenol A (BPA): Two nuclear receptors, PPARγ and LXRα, are EC. 50 The concentrations were 5.2 μM and 6.8 μM, respectively. The model predicted a risk probability of 0.62, and the ALT / AST release levels were 1.6 times that of the negative control, indicating moderate interference.

[0072] Dibutyl phthalate (DBP): Only PPARδ nuclear acceptor EC 50 The concentration was 45.3 μM, the model predicted a risk probability of 0.38, and the ALT / AST release level was 1.1 times that of the negative control, which was determined to be a weak interference.

[0073] Perfluorooctane sulfonic acid (PFOS): five nuclear acceptors, PPARα, PPARγ, LXRα, LXRβ, and FXR. 50 All <5μM, including PPARγEC 50 The concentration was 0.8 μM, the model predicted a risk probability of 0.89, and the ALT / AST release level was 3.0 times that of the negative control, indicating strong interference.

[0074] Wastewater effluent samples from wastewater treatment plants: EC of various nuclear receptors 50 All values ​​were >100 μM, the model predicted a risk probability of 0.21, and the ALT / AST release level was 1.0 times that of the negative control, indicating no interference.

[0075] In summary, the chemical-induced lipid metabolism disorder screening method based on nuclear receptor interference spectrum provided by this invention constructs a standardized, high-throughput, multi-dimensional detection system, establishes a highly reliable machine learning correlation model, and formulates quantitative screening standards. It can be directly applied to environmental monitoring agencies for screening metabolic interference risks of environmental pollutants, chemical enterprises for safety evaluation of industrial chemicals and food additives, and pharmaceutical companies for screening off-target effects of drug candidate compounds. At the same time, it can provide technical support for regulatory authorities in chemical risk management, screening of green alternatives, and regulatory compliance evaluation, and has broad industrial application prospects and market value.

[0076] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A graded screening method for chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum, characterized in that, Includes the following steps: S1. Construct a multi-dimensional detection system for multinuclear receptors covering core nuclear receptors regulating lipid metabolism; S2. The sample to be tested is co-incubated with the multi-nuclear receptor multidimensional detection system, and the agonist / antagonist activity of the chemical on each nuclear receptor is determined to obtain activity detection data; S3. Based on the activity detection data, after data standardization processing, a multidimensional nuclear acceptor interference spectrum of the chemical is generated. S4. Based on a multi-dimensionally validated machine learning model, establish a quantitative correlation model between nuclear receptor interference profiles and lipid metabolism disorder phenotypes. S5. Based on the quantitative indicators of nuclear receptor interference intensity and the quantitative indicators of lipid metabolism disorder risk prediction, and combined with the cellular ALT / AST release level indicators, chemicals are screened in a four-level classification: no interference, weak interference, moderate interference, and strong interference.

2. The method for graded screening of chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum according to claim 1, characterized in that, The nuclear receptor in S1 includes a core regulatory nuclear receptor and a functional nuclear receptor. The core regulatory nuclear receptor is at least one of PPARα, PPARγ, PPARδ, LXRα, LXRβ, FXR, and RXR; the functional nuclear receptor is at least one of TRβ, RAR, and VDR.

3. The method for graded screening of chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum according to claim 1, characterized in that, The multi-dimensional detection system in S1 includes reporter gene detection, fluorescence polarization technology, time-resolved fluorescence resonance energy transfer technology, LC-MS / MS, Western blot protein verification, and flow cytometry analysis; wherein the reporter gene detection uses a dual-luciferase system for internal control normalization.

4. The method for graded screening of chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum according to claim 1, characterized in that, The sample to be tested in S2 includes at least one of pure compound, environmental mixture, and food contact material migration; the environmental mixture includes at least one of PM2.5 extract, sewage sample, landfill leachate, and indoor dust extract.

5. The method for graded screening of chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum according to claim 1, characterized in that, The activity detection data in S3 includes the half-maximal effect concentration (EC5). 50 / IC 50 The agonist rate / competitive binding inhibition rate at specific concentrations were determined. Data standardization was performed using z-score normalization, and a multidimensional nuclear receptor interference spectrum vector was constructed from the standardized data.

6. The method for graded screening of chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum according to claim 1, characterized in that, The machine learning models in S4 include random forest, support vector machine, and ensemble learning XGBoost models. The machine learning models are validated in multiple dimensions. The validation process includes: dividing the training set and test set in a 7:3 ratio, using 5-fold cross-validation to optimize the model, and using LASSO regression and RF-RFE to select features; verifying the predictive consistency of the model in different cell lines, animal models, and human sample datasets; and using SHAP value and LIME tools to analyze the model's decision logic.

7. The method for graded screening of chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum according to claim 1, characterized in that, The lipid metabolism disorder phenotypes in S4 include lipid accumulation, fatty acid oxidation inhibition, and cholesterol homeostasis imbalance. During the construction of the association model, the Pearson correlation coefficient between the lipid metabolism disorder phenotype and the nuclear receptor interference spectrum was calculated to verify the significance of the association. At the same time, the cellular ALT / AST release level index was combined as a supplementary verification parameter for hepatocellular damage related to lipid metabolism disorder.

8. The method for graded screening of chemically induced lipid metabolism disorders based on nuclear receptor interference spectrum according to claim 1, characterized in that, The quantitative index of nuclear receptor interference intensity in S5 is based on EC. 50 The core value is the risk probability quantification index for lipid metabolism disorder risk prediction, which is based on the model's output risk probability and combined with cellular ALT / AST release levels. The quantification standard for the four-level screening is as follows: No interference: EC of each nuclear receptor 50 >100μM, the model predicted a risk probability <0.3, and the ALT / AST release level was within 1.2 times that of the negative control; Weak interference: Single nuclear receptor EC 50 Between 10 and 100 μM, the model predicted a risk probability of 0.3 to 0.5, and the ALT / AST release level was 1.2 to 1.5 times that of the negative control. Moderate interference: 2-3 nuclear receptor ECs 50 Between 1 and 10 μM, the model predicted a risk probability of 0.5 to 0.7, and the ALT / AST release level was 1.5 to 2.0 times that of the negative control. Strong interference: ≥4 nuclear receptor ECs 50 <10 μM or any nuclear receptor EC 50 <1μM, model predicted risk probability >0.7, and ALT / AST release level exceeded 2.0 times that of the negative control.

Images