Method for calculating combination of perfluoroalkyl and polyfluoroalkyl compounds and PPAR alpha based on pharmacophore model

Abstract

The invention relates to the technical field of computational toxicology, and discloses a method for calculating combination of perfluorinated and polyfluorinated alkyl compounds and PPARalpha based on a pharmacophore model, which comprises the following steps: acquiring a PFAS initial data set, acquiring a PPARalpha ligand binding domain crystal structure, and optimizing the structure; selecting representative PFAS to carry out molecular docking, and carrying out simulated annealing optimization to obtain an optimal binding posture; then generating a pharmacophore model; the models are screened; performing rigid fitting screening by using a pharmacophore model; carrying out molecular dynamics simulation on the representative hit object, and verifying the stability; protein is extracted by culturing cells in vitro, a thermal displacement experiment is carried out, and the thermal stability effect of PFAS on PPARalpha is verified. According to the method, comprehensive virtual screening of PFAS-PPAR alpha interaction based on pharmacophores is provided, a calculation framework based on protein and compound structures is established, and the method is suitable for environmental pollutant screening and safer chemical design.

Description

Technical Field

[0001] This invention relates to the field of computational toxicology, and more specifically, to a method for calculating the binding of perfluorinated and polyfluoroalkyl compounds to PPARα based on a pharmacophore model. Background Technology

[0002] Perfluorinated and polyfluoroalkyl substances (PFAS) are widely used in non-stick cookware, food packaging, and fire-fighting foams due to the high stability of their carbon-fluorine bonds. However, their "permanent chemical" nature means they persist in the environment and bioaccumulate, causing global environmental and public health crises. Currently, over 12,000 PFAS have been identified, with new variants constantly emerging. However, due to limitations in experimental conditions, the vast majority of PFAS lack toxicological evaluation, leading to delays in regulatory decisions.

[0003] Peroxisome proliferator-activated receptor α (PPARα) is a core mediating target for PFAS-induced hepatotoxicity and lipid metabolism interference. PFAS activates its signaling pathway by binding to the PPARα ligand-binding domain. Current technologies for predicting PFAS-PPARα binding have the following limitations: The assessment is fragmented: it focuses on a single PFAS (such as PFOA, PFOS) or a limited set of compounds, which cannot cover the entire chemical space of PFAS and is difficult to support risk warnings for emerging PFAS in the "replacement cycle". Method limitations: Existing computational methods either rely on single-molecule docking / machine learning (such as models based on 206 molecular descriptors) or only focus on global molecular properties (such as size and electrostatics), failing to fully capture the unique 3D molecular recognition patterns (such as halogen bonds) of PFAS and PPARα binding. Insufficient validation: Most studies lack multi-dimensional validation combining computation and experiment, and static docking results cannot reflect the stability of the binding complex under physiological conditions; Poor regulatory adaptability: The compound-specific results output by existing methods are insufficient to meet the grouped risk assessment needs of thousands of PFAS, resulting in low regulatory efficiency.

[0004] Therefore, there is an urgent need to develop a systematic, mechanistic, and high-confidence framework for assessing the PFAS-PPARα binding potential in order to address the aforementioned technical challenges. Summary of the Invention

[0005] In view of this, the present invention proposes a method for calculating the binding of perfluorinated and polyfluoroalkyl compounds to PPARα based on the pharmacophore model, aiming to solve the problems of narrow coverage, unclear mechanism, insufficient validation and poor regulatory compatibility in the current PFAS-PPARα binding assessment method.

[0006] This invention proposes a method for calculating the binding of perfluorinated and polyfluoroalkyl compounds to PPARα based on a pharmacophore model, comprising the following steps: 1) Obtain the initial dataset of perfluorinated and polyfluoroalkyl compounds from the US EPA CompTox database, remove compounds without usable 3D structures and complex mixtures, and obtain 10,946 PFAS; obtain the crystal structure of PPARα ligand binding domain from the RCSB PDB database, remove co-crystallized ligands, water molecules and heteroatoms, add hydrogen atoms under physiological pH conditions, complete the amino acid side chains and perform local energy minimization. 2) Four representative perfluorinated and polyfluoroalkyl compounds were selected: perfluorobutyric acid, perfluorooctanoic acid, perfluorododecanoic acid, and perfluorooctane sulfonic acid. Molecular docking was performed using the CDOCKER algorithm of Discovery Studio 2021 based on the CHARMm force field. Each compound was run 10 times independently to generate 10 initial conformations. The optimal binding posture was obtained through simulated annealing optimization. 3) Based on the docking results of step 2), generate a pharmacophore model containing 3 to 6 chemical features, with a minimum distance of 2 Å between features and a shape similarity threshold of 0.5; 4) The PPARα binding activity was experimentally confirmed using 12 positive perfluorinated and polyfluoroalkyl compounds; a validation dataset was constructed using 600 negative decoy compounds, and four high-performance models with AUC≥0.85 and MCC∈[0.15,0.35] were selected. 5) For the 10946 perfluorinated and polyfluoroalkyl compounds obtained in step 1), the FAST algorithm was used to generate ≤255 conformations with an energy threshold of 10 kcal / mol; the four pharmacophore models screened in step 4) were used for rigid fitting screening, allowing up to two features to be omitted, and the binding potential results were obtained by stratification according to the number of matching models. 6) Perform 300ns molecular dynamics simulations on representative hits to verify the stability of the complex; The thermal stabilizing effect of perfluorinated and polyfluoroalkyl compounds on PPARα was experimentally verified by analyzing the thermal displacement of MH-S and HL-1 cells.

[0007] Preferably, the removal of co-crystallized ligands, water molecules and heteroatoms in step 1) is performed using UCSF Chimera software; The addition of hydrogen atoms under physiological pH conditions and the completion of amino acid side chains were performed using Chimera software. The local energy minimization employs the GROMOS96 force field.

[0008] Preferably, the simulated annealing parameters in step 2) are: heating to 700K in 2000 steps, cooling to 300K in 5000 steps, calculating the partial charge of the ligand using the Momany-Rone method, and minimizing the total potential energy of the docking attitude to a gradient tolerance of 0.

[0009] Preferably, the chemical features described in step 3) include the hydrogen bond acceptor features of PFAS carboxylate / sulfonate groups, the hydrophobic features of the perfluoroalkyl chain, and the halogen bond site features formed by the fluorine atom and the PPARα cysteine ​​residue.

[0010] Preferably, the four high-performance models mentioned in step 4) are as follows: PFBA_01: AUC is 0.982, MCC is 0.583; PFDoA_08: AUC is 0.907, MCC is 0.606; PFOS_07: AUC is 0.894, MCC is 0.459; PFOS_08: AUC is 0.915, MCC is 0.379.

[0011] Preferably, the results of obtaining the binding potential by stratifying according to the number of matching models in step 5) are as follows: Matching 4 models: 942; Matching 3 models: 2498; Matching 2 models: 4286; Matching 1 model: 2012; No match: 1069.

[0012] Preferably, the molecular dynamics simulation in step 6) uses AMBER99SB-ILDN force field to treat PPARα, and general AMBER force field GAFF to treat the ligands. The simulation system includes a TIP3P water model and 150mM NaCl. The equilibrium phase includes 200ps constant volume and 500ps constant pressure.

[0013] Preferably, in step 6), the PFAS concentration for the thermal displacement analysis of the cells is 100 μM, the thermal intervention temperature is 50℃, 55℃, 60℃, and 65℃, the soluble PPARα is detected by Western blot, and the primary antibody is an anti-PPARα antibody.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: Mechanistic breakthrough: For the first time, the unique halogen bond mode of PFAS binding with PPARα was captured systematically. Combined with hydrogen bonding and hydrophobic interactions, a mechanistic evaluation framework based on 3D molecular recognition was constructed, which makes up for the shortcomings of existing methods that "emphasize properties and neglect structure". Balancing coverage and accuracy: Achieving large-scale screening of 10,946 PFAS, while outputting 942 high-confidence binding candidates through consensus prediction of 4 high-performance pharmacophore models, thus balancing "breadth" and "accuracy". Regulatory compatibility: It demonstrates that PPARα interference is a "class-wide property" of PFAS (90% of screened compounds have binding potential), providing data support for group-based risk assessment of PFAS and enabling rapid screening of the toxicity of emerging alternative PFAS. Reliability assurance: Through multi-dimensional verification of "molecular dynamics (300ns) + cellular thermal displacement", the stability of the prediction results under physiological conditions and the reproducibility of the experiment are ensured, reducing the risk of false positives. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 Molecular docking diagrams of PFOA, PFOS, PFBA, and PFDoA with PPARα; Figure 2 This is a representative pharmacophore model diagram of the PFAS-PPARα interaction; where green spheres represent hydrogen bond acceptors; cyan spheres represent hydrophobic features; magenta spheres represent halogen bond sites; and gray boxes represent exclusion volumes. Figure 3 The figures show receiver operating characteristic (ROC) curves for six validated pharmacophore models; the blue curve represents model performance, and the red dashed line represents random classification. Figure 4 The distribution map of PFAS-PPARα binding potential determined by virtual screening. Figure 4 In the diagram, A represents a bar chart showing the distribution of compounds based on the number of pharmacophore-matched compounds. Figure 4 B in the diagram represents the Venn diagram of the overlap patterns among the four validation pharmacophore models. Figure 5 This is a diagram verifying the binding of PFHxS within the PPARα ligand binding domain; Figure 6 The figure shows the results of molecular dynamics simulation analysis of the PFAS-PPARα complex over 300 ns. Figure 6 In this context, A represents the RMSD of the PPARα backbone after least-squares fitting of the protein. Figure 6 B and C in the figure are the root mean square wave (RMSF) curves calculated by 300 ns molecular dynamics simulation of PFOS and PFHxS. Figure 6 In this context, D represents the radius of gyration. Figure 6 E and F in the figure show the Gibbs energy landscape of the stable conformation basin of the PFOS and PFHxS complex; Figure 7 The figure shows the results of the differential thermostability test of PPARα induced by PFHxS in different cell types. Figure 7 In the figure, A represents the PPARα expression results in the lysates of MH-S and HL-1 cells treated with the vector (-) or 100 μM PFHxS (+) at the specified temperature. Figure 7 In the figure, B represents the thermal displacement curve of the MH-S cell. Figure 7 C in the figure represents the thermal displacement curve of HL-1 cells. Detailed Implementation

[0016] Various exemplary embodiments of the present invention will now be described in detail. This detailed description should not be considered as a limitation of the present invention, but rather as a more detailed description of certain aspects, features, and embodiments of the present invention. It should be understood that the terminology used in this invention is merely for describing particular embodiments and is not intended to limit the present invention.

[0017] Furthermore, regarding the numerical ranges in this invention, it should be understood that each intermediate value between the upper and lower limits of the range is also specifically disclosed. Every smaller range between any stated value or intermediate value within a stated range, and any other stated value or intermediate value within said range, is also included within this invention. The upper and lower limits of these smaller ranges may be independently included or excluded from the range.

[0018] Unless otherwise stated, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. While only preferred methods and materials have been described herein, any methods and materials similar or equivalent to those described herein may be used in the implementation or testing of this invention. All references to this specification are incorporated by way of citation to disclose and describe methods and / or materials associated with those references. In the event of any conflict with any incorporated reference, the content of this specification shall prevail.

[0019] Various modifications and variations can be made to the specific embodiments described in this specification without departing from the scope or spirit of the invention, as will be apparent to those skilled in the art. Other embodiments derived from this specification will also be apparent to those skilled in the art. This specification and embodiments are merely exemplary.

[0020] The terms “include,” “including,” “have,” “contain,” etc., used in this article are all open-ended terms, meaning that they include but are not limited to.

[0021] This invention proposes a method for calculating the binding of perfluorinated and polyfluoroalkyl compounds to PPARα based on a pharmacophore model, comprising the following steps: 1) Obtain the initial dataset of perfluorinated and polyfluoroalkyl compounds from the US EPA CompTox database, remove compounds without usable 3D structures and complex mixtures, and obtain 10,946 PFAS; obtain the crystal structure of PPARα ligand binding domain from the RCSB PDB database, remove co-crystallized ligands, water molecules and heteroatoms, add hydrogen atoms under physiological pH conditions, complete the amino acid side chains and perform local energy minimization. 2) Four representative perfluorinated and polyfluoroalkyl compounds were selected: perfluorobutyric acid, perfluorooctanoic acid, perfluorododecanoic acid, and perfluorooctane sulfonic acid. Molecular docking was performed using the CDOCKER algorithm of Discovery Studio 2021 based on the CHARMm force field. Each compound was run 10 times independently to generate 10 initial conformations. The optimal binding posture was obtained through simulated annealing optimization. 3) Based on the docking results of step 2), generate a pharmacophore model containing 3 to 6 chemical features, with a minimum distance of 2 Å between features and a shape similarity threshold of 0.5; 4) The PPARα binding activity was experimentally confirmed using 12 positive perfluorinated and polyfluoroalkyl compounds; a validation dataset was constructed using 600 negative decoy compounds, and four high-performance models with AUC≥0.85 and MCC∈[0.15,0.35] were selected. 5) For the 10946 perfluorinated and polyfluoroalkyl compounds obtained in step 1), the FAST algorithm was used to generate ≤255 conformations with an energy threshold of 10 kcal / mol; the four pharmacophore models screened in step 4) were used for rigid fitting screening, allowing up to two features to be omitted, and the binding potential results were obtained by stratification according to the number of matching models. 6) Perform 300ns molecular dynamics simulations on representative hits to verify the stability of the complex; The thermal stabilizing effect of perfluorinated and polyfluoroalkyl compounds on PPARα was experimentally verified by analyzing the thermal displacement of MH-S and HL-1 cells.

[0022] In this invention, the removal of co-crystallized ligands, water molecules and heteroatoms in step 1) is performed using UCSF Chimera software; The addition of hydrogen atoms under physiological pH conditions and the completion of amino acid side chains were performed using ChimeraChimera software. The local energy minimization employs the GROMOS96 force field.

[0023] In this invention, the parameters for simulated annealing in step 2) are: heating to 700K in 2000 steps, cooling to 300K in 5000 steps, calculating the partial charge of the ligand using the Momany-Rone method, and minimizing the total potential energy of the docking attitude to a gradient tolerance of 0.

[0024] In this invention, the chemical features described in step 3) include the hydrogen bond acceptor features of PFAS carboxylate / sulfonate groups, the hydrophobic features of the perfluoroalkyl chain, and the halogen bond site features formed by the fluorine atom and the PPARα cysteine ​​residue.

[0025] In this invention, the four high-performance models mentioned in step 4) are as follows: PFBA_01: AUC is 0.982, MCC is 0.583; PFDoA_08: AUC is 0.907, MCC is 0.606; PFOS_07: AUC is 0.894, MCC is 0.459; PFOS_08: AUC is 0.915, MCC is 0.379.

[0026] In this invention, the results of obtaining the binding potential by stratifying according to the number of matching models in step 5) are as follows: Matching 4 models: 942; Matching 3 models: 2498; Matching 2 models: 4286; Matching 1 model: 2012; No match: 1069.

[0027] In this invention, the molecular dynamics simulation described in step 6) uses AMBER99SB-ILDN force field to treat PPARα, and general AMBER force field GAFF to treat the ligands. The simulation system includes a TIP3P water model and 150mM NaCl. The equilibrium phase includes 200ps constant volume and 500ps constant pressure.

[0028] In this invention, the PFAS concentration for the thermal displacement analysis of cells in step 6) is 100 μM, the thermal intervention temperatures are 50℃, 55℃, 60℃, and 65℃, soluble PPARα is detected by Western blot, and the primary antibody is an anti-PPARα antibody. Example 1 1. Data Collection and Preprocessing PFAS structures were systematically sourced from the U.S. Environmental Protection Agency's CompTox Chemicals Database. The initial dataset contained 12,039 individual PFAS compounds, representing different structural categories, including perfluorocarboxylic acids (PFCAs), perfluorosulfonic acids (PFSAs), fluoropolymers (FTOH), and polyfluoroalkyl phosphates (PAPs). All molecular structures were preprocessed using a standardized workflow, systematically identifying and removing compounds lacking usable 3D structural representations, and eliminating 1,087 structures that could not be processed by computational modeling software. Six complex chemical mixtures and entries containing multiple independent molecular entities were also excluded. The final dataset contained 10,946 individual PFAS compounds suitable for computational analysis.

[0029] The 3D crystal structure of the PPARα ligand-binding domain was retrieved from the Research Consortium for Structural Bioinformatics Protein Database (RCSB PDB) with accession code 2ZMN. This structure represents the highest resolution available, at 2.9 Å. Obtained in the form of a complex with a synthetic agonist ligand, this provides valuable information for understanding the structure of the natural ligand-binding cavity. A comprehensive protein preparation procedure was performed using the UCSF Chimera molecular modeling software to optimize the structure for computational studies. All co-crystallized ligands, water molecules, and heteroatoms were systematically removed from the protein structure to create an empty binding site suitable for docking calculations. Missing hydrogen atoms were added using standard protonation states suitable for physiological pH conditions, with particular attention paid to ionizable residues within the binding cavity. The prepared protein structure was further optimized using Chimera software to complete any missing atoms in the amino acid side chains and to minimize local energy to mitigate potential spatial conflicts introduced during preparation.

[0030] 2. Reference Compound Selection and Structural Classification To maximize the structural diversity captured by the pharmacophore model while maintaining computational tractability, a systematic structural analysis was conducted, classifying the twelve compounds into different chemical categories based on functional groups and carbon chain length. Perfluorocarboxylic acids (PFCAs) were further subdivided into short-chain, medium-chain, and long-chain representatives according to chain length, while perfluorosulfonic acids (PFSAs) were analyzed separately due to their unique head group chemistry. Based on the above analysis, four compounds were selected as the best representatives of the broader PFAS chemical space: perfluorobutyric acid (PFBA) representing short-chain carboxylic acids with 4 carbon atoms, perfluorooctanoic acid (PFOA) representing medium-chain carboxylic acids with 8 carbon atoms, perfluorododecanic acid (PFDoA) representing long-chain carboxylic acids with 12 carbon atoms, and perfluorooctanesulfonic acid (PFOS) representing the sulfonic acid functional group category with 8 carbon atoms. This selection strategy ensures coverage of the major PFAS head group types and the full range of carbon chain lengths common in environmental and biological systems.

[0031] 3. Molecular docking scheme and characterization of binding sites Molecular docking calculations were performed using the CHARMm-based CDOCKER algorithm in the Discovery Studio 2021 software suite. This algorithm provides a robust platform for protein-ligand interaction prediction through systematic conformational sampling and energy minimization. The docking scheme was designed to identify the optimal binding posture while thoroughly sampling the conformational space available for each PFAS ligand within the PPARα binding cavity.

[0032] For each target compound, ten independent docking runs were performed to ensure statistical robustness and capture the variability of potential binding modes. Initial ligand conformations were generated using a systematic conformational search algorithm, creating 10 random initial conformations for each compound to avoid bias towards specific conformational states. These initial conformations were subjected to molecular dynamics simulation protocols within the binding site, utilizing 1000 kinetic steps at a target temperature of 1000 K to allow for sufficient conformational sampling while maintaining computational efficiency.

[0033] The binding site was defined by analyzing the location of co-crystallized ligands in the original protein structure. The search space extended 8 Å in all directions from the geometric center of the native ligand to encompass potential alternative binding modes. Simulated annealing optimization was used to refine the docking attitude, starting with a 2000-step heating phase at a target temperature of 700 K, followed by a 5000-step cooling phase at a final temperature of 300 K. This temperature cycling scheme effectively explored the local energy landscape while avoiding trapping in high-energy conformations.

[0034] All energy calculations employed the CHARMm molecular mechanics force field, providing an accurate representation of protein-ligand interactions through a well-validated parameter set. The Momany-Rone method was used to calculate the ligand partial charge, providing reliable electrostatic parameters for organofluorine compounds. All docking postures underwent full potential minimization with zero gradient tolerance to ensure optimal geometry and eliminate residual strain during docking. Acceptor preparation protocols included automatic protonation state allocation and optimized hydrogen bond networks to ensure accurate protein conformation during docking calculations.

[0035] 4. Development and Feature Definition of Pharmacophore Model The pharmacophore model was generated using the interaction pharmacophore generation protocol available in Discovery Studio. This protocol analyzes protein-ligand interactions in molecular docking results to identify the essential chemical features required for biological activity. This ensures that the resulting pharmacophore model reflects the actual molecular recognition patterns observed at the PPARα binding site.

[0036] The pharmacophore generation process is configured to identify 3 to 6 chemical features per model, representing an optimal balance between specificity and versatility in virtual screening applications. A minimum feature distance of 2 Å is set to prevent spatial overlap of pharmacophore elements while allowing for reasonable conformational flexibility in the matched compounds. Based on the observed interaction patterns between the docked PFAS compounds and the PPARα binding site residues, the chemical features are automatically classified as hydrogen bond acceptors, hydrogen bond donors, hydrophobic regions, aromatic centers, and halogen sites.

[0037] Advanced geometric constraints were implemented to ensure pharmacophore quality and screening reliability. A minimum shape similarity threshold was set to 0.5 to maintain meaningful structural relationships while allowing for chemical diversity among screened hits. An exclusion volume was automatically generated to represent regions within the binding site that could not accommodate additional chemical groups, with a maximum exclusion volume distance of 5 Å to capture the local spatial environment.

[0038] A comprehensive feature set was employed to capture all molecular interactions observed in PFAS-PPARα binding. Conventional hydrogen bonding patterns were identified and encoded as directional features with appropriate geometric constraints. Halogen interactions were specifically identified and parameterized using a specialized algorithm designed for organofluorine chemistry. Hydrophobic features were mapped based on the lipophilic contact surfaces between the PFAS alkyl chain and nonpolar protein residues. Electrostatic interactions were captured via ionic features, representing charge-charge interactions between the PFAS head group and charged amino acid residues within the binding site.

[0039] 5. Validation and performance evaluation of the pharmacophore model The pharmacophore model was validated using a rigorous statistical framework to assess sensitivity and specificity under realistic screening conditions. First, a balanced validation dataset was constructed, containing twelve experimentally confirmed PPARα-binding PFAS compounds as positive controls and 600 carefully selected negative control compounds. The negative control set was generated using the web-based LIDEB-LUDE v2 platform, which creates chemical decoys with similar physicochemical properties to the active compounds but without known biological activity. It maintained a 1:50 ratio of active to inactive compounds, closely simulating the typical hit rates encountered in large-scale virtual screening.

[0040] Multiple performance metrics were calculated to comprehensively evaluate model quality and screening utility. The area under the receiver operating characteristic (ROC) curve (AUC) was calculated to assess the model's overall discriminative ability, with a minimum acceptable threshold of 0.85. The Matthews correlation coefficient (MCC) was calculated to account for dataset imbalance, with an acceptable range between 0.15 and 0.35, reflecting the inherent challenges of highly imbalanced screening datasets. Sensitivity and specificity values ​​were determined to assess the ability to correctly identify active compounds while minimizing the false positive rate.

[0041] Additional validation criteria included evaluating the minimum fit value of the active compound, which needed to reach a threshold of 2.0 to ensure meaningful pharmacophore matching. Statistical significance tests were performed using bootstrapping procedures to assess the robustness of performance metrics across multiple randomized dataset partitions. Only pharmacophore models that simultaneously met all validation criteria were selected for subsequent virtual screening applications, ensuring high confidence in the screening results.

[0042] 6. Virtual screening scheme and target identification Large-scale virtual screening was performed using the Ligand Profiler protocol implemented in Discovery Studio. This protocol provides efficient algorithms that can simultaneously match compound databases across multiple pharmacophore models. The screening workflow was optimized to maintain the accuracy and reliability of hit identification while processing large chemical databases.

[0043] The FAST conformational sampling algorithm was used to generate a three-dimensional conformational ensemble for all screened compounds, striking a balance between computational efficiency and adequate coverage of the accessible conformational space. A maximum of 255 conformations were generated for each compound to capture conformational flexibility while maintaining reasonable computational requirements. The energy threshold for conformational generation was set 10 kcal / mol above the global minimum to include energy-accessible conformations while excluding high-energy states unlikely to occur under biological conditions.

[0044] Virtual screening employs a rigid fitting method, matching compound conformations to pharmacophore models. This provides computational efficiency suitable for large-scale screening while maintaining accuracy for well-defined pharmacophore features. Considering potential conformational limitations, up to two pharmacophore features can be omitted during the fitting process, offering flexibility for compounds that nearly match pharmacophore requirements but may differ in fine structural details. During screening, the minimum feature distance constraint is maintained at 0.5 Å to ensure geometric consistency with the original pharmacophore model.

[0045] Target compounds were identified based on pharmacophore matching scores and geometric fit quality, and detailed analyses of binding and interaction modes were performed on all identified targets. Compounds matching multiple pharmacophore models were given higher priority for further analysis, as consensus predictions typically demonstrate higher reliability than single-model predictions. Screening results were compiled and annotated to facilitate downstream analysis and experimental validation planning.

[0046] 7. Molecular dynamics simulation Molecular dynamics (MDS) simulations lasting 300 nanoseconds (ns) were performed to validate the stability of the predicted PFAS-PPARα binding interaction, using PFOS and perfluorohexanesulfonic acid (PFHxS) as representative compounds. The protein-ligand complex was placed in a rectangular simulation box with a minimum distance of 12 Å from the box boundary and solvated using the TIP3P water model. Sodium (Na₂SO₄) was added. + ) and chlorine (Cl - The ions reached a physiological ionic strength of 150 mM.

[0047] Energy minimization was performed using the steepest descent method and the conjugate gradient algorithm, with a convergence criterion of 10 kJ / mol / root mean square force. Equilibrium consisted of a 200 ps constant-volume phase at 310 K using a velocity rescaling thermostat, followed by a 500 ps constant-pressure phase at 1.0 bar using a Parrinello-Rahman barostat. During equilibrium, 1000 kJ / mol / nm was applied to the heavy atoms. 2 Position constraints.

[0048] Production simulations used an AMBER99SB-ILDN force field to treat the protein, a universal AMBER force field (GAFF) to treat the ligands, a time step of 2 fs, and LINCS constraints to handle hydrogen bonds. Electrostatic interactions were calculated using the particle mesh Ewald method with a cutoff of 1.2 nm. Trajectory analyses included calculations of the root mean square deviation (RMSD) of the protein backbone and ligand atoms, analysis of the root mean square fluctuation (RMSF) of binding site flexibility, and monitoring of hydrogen bond, hydrophobic contact, and halogen bond interactions throughout the simulation.

[0049] 8. Cell thermal displacement analysis Macrophage line MH-S and cardiomyocyte line HL-1 were maintained under standard culture conditions (37°C, 5% CO2 atmosphere) using Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum, 100 units / mL penicillin, and 100 μg / mL streptomycin. During the logarithmic growth phase, cells were harvested and counted to obtain 1 million cells for lysis buffer preparation. Cell lysis buffer was obtained by liquid nitrogen freezing followed by three freeze-thaw cycles to ensure complete cell lysis and protein extraction. Subsequently, the lysis buffer was incubated with PFHxS at a final concentration of 100 μmol at room temperature for 1 hour to allow protein-ligand equilibrium. After compound incubation, aliquots of the lysis buffer-compound mixture were subjected to thermal intervention at specific temperature points (50°C, 55°C, 60°C, and 65°C) for 3 minutes each.

[0050] Following thermal intervention, samples were rapidly transferred to ice for 5 minutes to halt protein denaturation. Cell lysates were clarified by centrifugation at 16000g for 15 minutes at 4°C to separate soluble proteins from aggregates. The PPARα content of soluble protein fractions was analyzed using Western blot analysis with a specific anti-PPARα primary antibody (1:2000 dilution, Abcolonal, catalog number A25296, Wuhan, China) and an appropriate horseradish peroxidase-conjugated secondary antibody. Thermal shift curves were generated by plotting the relationship between soluble PPARα protein fractions and temperature, and statistical analysis was performed using ANOVA (followed by Tukey) with multiple comparisons to identify significant thermostability effects.

[0051] 9. Statistical Analysis All computational and experimental data underwent comprehensive statistical analysis to ensure robust interpretation of the results. The pharmacophore model performance was evaluated using sensitivity, specificity, positive predictive value, negative predictive value, and MCC calculated from the confusion matrix, with 95% confidence intervals determined using a bootstrap procedure. AUC values ​​were calculated using the trapezoidal rule. Descriptive statistics were used to perform virtual screening to characterize the distribution of pharmacophore fit values, and compounds were ranked based on composite scores from multiple models. Molecular dynamics simulation data were processed using time-series analysis to assess system stability, and RMSD and RMSF values ​​were calculated using a block averaging procedure. Melting temperatures were determined using nonlinear regression and 4-parameter logistic curve fitting for cellular thermal displacement analysis data. Statistical comparisons were performed using Student's t-tests to handle differences between two groups, and one-way ANOVA followed by Tukey post-hoc tests to handle differences between three or more groups. All analyses were performed using R statistical software version 4.4.1, and statistical significance was defined as p < 0.05.

[0052] The experimental results are as follows: 1. Molecular docking reveals different PFAS-PPARα binding modes Molecular docking analysis determined the stable binding postures of all four representative PFAS compounds within the PPARα ligand binding domain. These compounds exhibited different interaction spectra. Molecular docking of PFOA, PFOS, PFBA, and PFDoA with PPARα was shown in the figure. Figure 1As shown, these spectra correlate with structural features, including chain length and head group functionality. Perfluorobutyric acid (PFBA) establishes four favorable interactions with the acceptor, forming hydrogen bonds with MET330 and ALA333 residues, while simultaneously engaging in halogen interactions with CYS275 and ALA333. This compound positions the carboxylate group within the polar binding region, while the short perfluorinated chain occupies the entrance to the hydrophobic cavity. In contrast, perfluorooctanoic acid (PFOA) exhibits three favorable interactions and enhanced cavity penetration; the carboxylate group forms a hydrogen bond with CYS276, and the C8 perfluoroalkyl chain enables halogen interactions with ILE272, CYS275, and CYS276, extending deeper into the binding cavity compared to shorter-chain analogs.

[0053] Furthermore, perfluorododecanoic acid (PFDoA) exhibits the most extensive interaction network among carboxylic acids, establishing five favorable contacts across the entire binding cavity length. Hydrogen bonding occurs with GLU251, ILE272, CYS276, and HIS440, while hydrophobic interactions involve CYS276, ILE354, and HIS440, demonstrating how the extended chain length enables comprehensive acceptor occupancy. Notably, PFOS displays the highest number of favorable interactions, with six distinct molecular contacts, representing a significant enhancement relative to carboxylic acid analogs. The sulfonate group forms hydrogen bonds with GLU251, ILE272, CYS276, and THR279, while simultaneously engaging in hydrophobic interactions with VAL255 and CYS275, demonstrating superior coordination ability compared to carboxylate group functionality.

[0054] The docking results revealed that the binding complexity gradually increases with chain length, from four interactions in C4 compounds to five interactions in C12 derivatives, with the binding spectrum conferred by the sulfonic acid root head group being superior to that of carboxylic acid derivatives. All compounds successfully occupied the PPARα binding cavity through a combination of hydrogen bonds, halogen bonds, and hydrophobic interactions, laying the molecular foundation for subsequent pharmacophore model development.

[0055] 2. The pharmacophore model development captured the basic PFAS binding features. The pharmacophore generation protocol yielded 32 distinct pharmacophore models from molecular docking results, with the model distribution varying according to compound structural complexity. Perfluorobutyric acid (PFBA) generated two pharmacophore models due to its limited interaction diversity, while perfluorooctanoic acid (PFOA), perfluorododecanic acid (PFDoA), and PFOS each generated 10 models, reflecting their broader binding modes. Each of these pharmacophore models contains 3 to 6 chemical features, capturing hydrogen bond acceptors, hydrogen bond donors, hydrophobic regions, and halogen bond sites identified from the PFAS-PPARα docking interaction.

[0056] Systematic analysis of the generative model revealed unique characteristic patterns corresponding to different binding modes and molecular orientations within the PPARα cavity, such as... Figure 2 As shown, representative pharmacophore models from PFAS-PPARα interactions are presented (green spheres: hydrogen bond acceptors; cyan spheres: hydrophobic features; magenta spheres: halogen bond sites; gray wireframes: exclusion volumes). These models consistently identify polar head base regions as key for acceptor recognition, and hydrogen bond acceptor features are localized to interact with key residues, including GLU251, CYS276, and THR279. Furthermore, hydrophobic features map perfluoroalkyl chain binding regions, while halogen bond features specifically capture the unique fluorine-protein interactions of PFAS compounds. Additionally, exclusion volumes are automatically generated around each pharmacophore to represent protein regions that cannot accommodate additional chemical groups, thereby enhancing the selectivity of subsequent virtual screening procedures.

[0057] Characteristic composition analysis revealed a clear relationship between the structural features of PFAS and the complexity of the pharmacophore. Short-chain models primarily contained hydrogen bond acceptors and halogen bonds, exhibiting minimal hydrophobic components, while long-chain models contained a wide range of hydrophobic features across the binding cavity length. Notably, sulfonate-derived models consistently demonstrated enhanced hydrogen bond acceptor capacity compared to carboxylate analogs, corresponding to the superior binding performance observed in molecular docking analysis. The systematic capture of these different interaction modes provides comprehensive coverage of PFAS binding modes, laying the foundation for rigorous pharmacophore validation and subsequent virtual screening applications.

[0058] 3. Performance evaluation of the pharmacophore model Comprehensive validation of 32 generated pharmacophore models using a balanced test set revealed significant performance differences, with 6 models meeting the acceptance criteria established for virtual screening applications. This validation employed 12 experimentally confirmed PPARα-PFAS combinations as positive controls (Table 1) and 600 carefully selected decoy compounds as negative controls, maintaining a 1:50 ratio to reflect realistic virtual screening conditions. Performance evaluation utilized multiple statistical metrics, including AUC, MCC, sensitivity, and specificity, to ensure robust model evaluation under imbalanced dataset conditions. Results are shown below. Figure 3 As shown, the receiver operating characteristic (ROC) curves of six validated pharmacophore models demonstrate the distinguishing performance between active PFAS and decoy compounds. The blue curve represents model performance; the red dashed line represents random classification.

[0059] In the validated models, PFBA_01 demonstrated excellent discriminative ability, with an AUC of 0.982 and an MCC of 0.583, achieving a perfect sensitivity of 1.000 while maintaining a high specificity of 0.963. This model successfully identified all twelve positive control compounds and correctly classified 578 of the 600 negative controls, resulting in 22 false positive predictions. In contrast, the perfluorododecanoic acid-derived model PFDoA_08 showed robust but slightly lower performance, with an AUC of 0.907 and an MCC of 0.606, maintaining a sensitivity of 0.833 and a specificity of 0.980. Similarly, PFOA_09 achieved robust discriminative ability, with an AUC of 0.942 and an MCC of 0.568, exhibiting a sensitivity of 0.917 and a specificity of 0.968, correctly identifying eleven of the twelve active compounds with an extremely low false positive rate.

[0060] Furthermore, the PFOS-derived models PFOS_07 and PFOS_08 showed acceptable performance levels with AUC values ​​of 0.894 and 0.915, respectively, but their lower MCC values ​​reflected the challenges associated with sulfonate pharmacophores. PFOS_07 achieved an MCC of 0.459, with a sensitivity of 0.833 and a specificity of 0.955, while PFOS_08 exhibited an MCC of 0.379, with a sensitivity of 0.917 and a specificity of 0.913. Validation results confirmed that four pharmacophore models possessed the optimal combination of sensitivity, specificity, and statistical robustness required for large-scale virtual screening applications. Based on complementary structural features and validated performance metrics, PFBA_01, PFDoA_08, PFOA_09, and PFOS_08 were selected for subsequent compound library screening.

[0061] 4. Virtual screening reveals broad PFAS-PPARα binding potential A large-scale virtual screening of 10,807 PFAS compounds was conducted using four validated pharmacophore models, combined with potential distributions such as... Figure 4 As shown, Figure 4 In the diagram, A represents a bar chart showing the distribution of compounds based on the number of pharmacophore-matched compounds. Figure 4B in the diagram represents the Venn diagram of the overlap patterns among the four validating pharmacophore models, with the central intersection representing the highest confidence prediction used for experimental validation. This reveals a vast binding potential across the entire chemical library, with 9738 compounds exhibiting compatibility with at least one pharmacophore model. The screening results show a clear stratification pattern based on the number of pharmacophore matches, indicating varying predicted binding affinity and structural complementarity with PPARα binding sites. Significant differences exist in the performance of individual pharmacophores, with PFOS_08 identifying the most potential binders (7495 compounds), followed by PFOA_09 with 6789 compounds, while PFDoA_08 and PFBA_01 identified 3666 and 3896 compounds, respectively.

[0062] Distribution analysis revealed a structured pattern of pharmacophore matching in the screening library, with 942 compounds matching all four pharmacophore models, representing the highest confidence predictions for PPARα binding activity. Furthermore, 2498 compounds matched three pharmacophores, 4286 matched two pharmacophores, and 2012 matched only one pharmacophore model. Conversely, 1069 compounds showed no pharmacophore matching, representing approximately 10% of the screening library and exhibiting the lowest predicted binding potential. This comprehensive matching pattern indicates that the vast majority of PFAS compounds possess structural features compatible with PPARα binding, and the consensus predictions from multiple models provide higher confidence for the binding potential assessment.

[0063] Furthermore, the screening results revealed unique selectivity patterns among the validated pharmacophore models. The sulfonate-derived PFOS_08 exhibited the broadest recognition spectrum, identifying nearly 70% of the compound library as potential binders. In contrast, the short-chain carboxylate model PFBA_01 showed greater selectivity, identifying approximately 36% of the compounds as potential matchers. The systematic application of multiple complementary pharmacophore models successfully prioritized 942 high-confidence compounds for experimental validation, providing comprehensive coverage of PFAS structural diversity and establishing a robust framework for targeted experimental evaluation of PPARα binding potential across a broader PFAS chemical space.

[0064] 5. Experimental and computational verification confirms PPARα binding prediction. 5.1 Molecular docking verification using representative compounds Molecular docking validation using PFHxS as an independent test compound confirmed the predictive accuracy of the computational framework and demonstrated a binding mode consistent with the established pharmacophore model. Figure 5 The binding verification of PFHxS within the PPARα ligand binding domain is as follows: Figure 5As shown, five favorable interactions are established between the two compounds: hydrogen bonds with ILE272, CYS276, and THR279 residues, and hydrophobic interactions with VAL255 and CYS275. This binding spectrum is very similar to the patterns observed in PFOS, particularly in the coordination of the sulfonate group with polar binding site residues. Furthermore, the compound exhibits halogen-bonded interactions with ILE272 and CYS275, confirming the importance of fluorine-protein contact in PFAS receptor recognition.

[0065] The validation docking results showed excellent agreement with the pharmacophore model predictions. PFHxS successfully matched three of the four validation models, and the fit values ​​exceeded the predetermined threshold. The binding mode of this compound occupies the same spatial region identified in the original pharmacophore development, with the sulfonate group located within the hydrogen bond acceptor feature, and the perfluoroalkyl chain aligned with hydrophobic and halogen bond elements. Notably, this binding posture demonstrates optimal complementarity with PFOS-derived pharmacophores while showing partial compatibility with carboxylate group models, consistent with the structural similarity among sulfonate-containing compounds, and validates the discriminative power of the developed screening framework.

[0066] 5.2 Molecular dynamics simulations confirm binding stability Molecular dynamics simulations exceeding 300 ns provided comprehensive validation of the binding stability of PFOS and PFHxS complexes with PPARα, as shown in the results. Figure 6 As shown, Figure 6 In this context, A represents the RMSD of the PPARα backbone after least-squares fitting of the protein. Figure 6 B and C in the figure are the root mean square wave (RMSF) curves calculated by 300 ns molecular dynamics simulation of PFOS and PFHxS. Figure 6 In this context, D represents the radius of gyration. Figure 6 E and F in the figure show the Gibbs energy landscape of the stable conformational basin of the PFOS and PFHxS complexes. Blue line / surface: PFHxS; Red line / surface: PFOS. This demonstrates a persistent interaction mode throughout the extended simulation. RMSD analysis reveals stable ligand conformations; the RMSD values ​​of PFOS and PFHxS stabilize at around 4 Å after initial equilibrium, indicating minimal positional drift within the binding cavity. Furthermore, protein backbone RMSD calculations show that the values ​​for both complexes remain around 2.5 Å, confirming that the receptor structure remains stable and that no significant conformational perturbation occurs after PFAS binding.

[0067] RMSF analysis revealed that both ligands exhibited restricted mobility within the binding site, with RMSF values ​​remaining below 2.0 Å throughout the simulated trajectory. This analysis uncovered specific regions of enhanced stability corresponding to key interaction sites identified in the initial docking study. Furthermore, radius of gyration measurements showed that the values ​​for both complexes remained stable at approximately 1.90 ns, indicating a compact protein conformation maintained without unfolding or significant structural expansion during the simulation.

[0068] Gibbs energy landscape analysis revealed that both PFOS and PFHxS complexes possess unique conformational basins, with energy minimums located at similar coordinates in principal component space. Both compounds occupy stable energy traps with energy barriers exceeding 10 kJ / mol, confirming thermodynamically favorable binding states that persist under physiological conditions.

[0069] 5.3 Cell thermal displacement analysis demonstrates the cell type-specific protein stabilizing effect. Cellular thermal displacement analysis using MH-S macrophage and HL-1 cardiomyocyte lysates provided experimental validation of PFAS-PPARα binding by directly measuring changes in protein thermal stability, revealing cell type-specific response patterns. Figure 7 The figure shows the results of the differential thermostability test of PPARα induced by PFHxS in different cell types. Figure 7 In the figure, A represents the PPARα expression results in the lysates of MH-S and HL-1 cells treated with the vector (-) or 100 μM PFHxS (+) at the specified temperature. Figure 7 In the figure, B represents the thermal displacement curve of the MH-S cell. Figure 7 C in the figure represents the thermal displacement curve of HL-1 cells; Figure 7 Medium black circle: DMSO control; Brown square: PFHxS treatment. In MH-S cell lysates, PFHxS treatment resulted in enhanced thermal stability at intermediate temperatures, with the compound-treated samples showing increased PPARα protein abundance at 55°C compared to the carrier control. However, the thermal protective effect failed to be maintained at higher temperatures, with treated and control samples showing similar protein degradation patterns at 60°C and 65°C. The thermal shift pattern indicates the presence of a moderate stabilizing effect that is overcome with increasing temperature.

[0070] In contrast, HL-1 cell lysates exhibited unique thermal characteristics, namely enhanced PPARα protein stability over a wider temperature range. PFHxS treatment produced a significant thermal stabilizing effect, with peak protection observed at 55–60°C, and the treated sample maintained approximately 1.6-fold higher relative protein abundance compared to the DMSO control. This protective effect persisted at multiple temperature points, indicating more robust protein-ligand interactions in cardiomyocyte lysates compared to macrophage-derived samples.

[0071] Differential thermal shift responses between cell types may reflect differences in protein expression levels, cellular cofactor availability, or the presence of competitive endogenous ligands, all of which can affect PFAS-PPARα binding affinity. Western blot analysis confirmed consistent PPARα protein detection across temperature points prior to thermal denaturation, validating that the observed shift represents genuine protein-ligand stabilization rather than an artifact arising from differential protein expression or sample preparation.

[0072] This study presents a comprehensive virtual screening of PFAS-PPARα interactions based on pharmacophores, identifying 9,738 compounds with potential binding activity from nearly 11,000 screened structures. The validated pharmacophore model successfully captured fundamental molecular recognition patterns involving hydrogen bonding, hydrophobic interactions, and unique halogen bonding mechanisms. Molecular dynamics simulations confirmed binding stability over extended timescales, while the systematic identification of 942 high-confidence compounds provided practical prioritization for experimental validation. These findings support the necessity of group-based PFAS regulation and establish a mechanism-based computational framework applicable to broader environmental pollutant screening and safer chemical design.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for calculating the binding of perfluoro- and polyfluoroalkyl compounds to PPARa based on a pharmacophore model, characterized in that, Comprising the following steps: 1) Obtain the initial dataset of perfluoroalkyl and polyfluoroalkyl substances from the US EPA CompTox database, remove compounds without available 3D structure and complex mixtures, and obtain 10946 PFAS; Obtain PPARα ligand binding domain crystal structure from RCSB PDB database, remove co-crystallized ligands, water molecules and heteroatoms, add hydrogen atoms under physiological pH conditions, complete amino acid side chains and perform local energy minimization; 2) Select 4 representative perfluoroalkyl and polyfluoroalkyl compounds: perfluorobutyric acid, perfluorooctanoic acid, perfluorododecanoic acid, perfluorooctanesulfonic acid; Use the CDOCKER algorithm of Discovery Studio 2021, based on the CHARMm force field to perform molecular docking, 10 independent runs for each compound, generate 10 initial conformations, and obtain the optimal binding pose by simulated annealing optimization; 3) Based on the docking results of step 2), generate a pharmacophore model containing 3-6 chemical features, with a minimum distance of 2Å between features and a shape similarity threshold of 0.5; 4) Use 12 positive perfluoroalkyl and polyfluoroalkyl compounds to confirm the PPARα binding activity, and use 600 negative decoy compounds to construct a validation dataset, and select 4 high-performance models with AUC≥0.85 and MCC∈[0.15,0.35]; 5) For the 10946 perfluoroalkyl and polyfluoroalkyl compounds obtained in step 1), use the FAST algorithm to generate ≤255 conformations with an energy threshold of 10 kcal / mol; Use the 4 pharmacophore models selected in step 4) for rigid fitting screening, allowing up to 2 features to be omitted, and layer by matching model number to obtain binding potential results; 6) Perform 300ns molecular dynamics simulation on representative hits to verify the stability of the complex; The thermal shift analysis of MH-S and HL-1 cells verifies the thermal stability of perfluoroalkyl and polyfluoroalkyl compounds on PPARα.

2. A method of calculating the binding of perfluoro- and polyfluoroalkyl compounds to PPARa based on a pharmacophore model according to claim 1, characterized in that, The removal of co-crystallized ligands, water molecules and heteroatoms in step 1) is performed using UCSF Chimera software; The addition of hydrogen atoms under physiological pH conditions and the completion of amino acid side chains are performed using ChimeraChimeraX software; The local energy minimization uses the GROMOS96 force field.

3. A method of calculating the binding of perfluoro- and polyfluoroalkyl compounds to PPARa based on a pharmacophore model according to claim 2, characterized in that, The parameters of simulated annealing in step 2) are 2000 steps of heating to 700K, 5000 steps of cooling to 300K, using the Momany-Rone method to calculate the partial charge of the ligand, and minimizing the total potential energy of the docking pose to a gradient tolerance of 0.

4. The method for calculating the binding of perfluoro- and polyfluoroalkyl compounds to PPARa based on pharmacophore models according to any one of claims 1 to 3, characterized in that, The chemical features in step 3) include the hydrogen bond acceptor feature of PFAS carboxylate / sulfonate, the hydrophobic feature of perfluoroalkyl chain, and the halogen bond site feature formed by fluorine atoms and PPARα cysteine residues.

5. A method of calculating the binding of perfluoro- and polyfluoroalkyl compounds to PPARa based on a pharmacophore model according to claim 4, characterized in that, The 4 high-performance models in step 4) are: PFBA_01: AUC is 0.982, MCC is 0.583; PFDoA_08: AUC is 0.907, MCC is 0.606; PFOS_07: AUC is 0.894, MCC is 0.459; PFOS_08: AUC is 0.915, and MCC is 0.

379.

6. The method for calculating the binding of perfluorinated and polyfluoroalkyl compounds to PPARα based on a pharmacophore model according to claim 5, characterized in that, The binding potential results obtained by the hierarchical division according to the number of matched models in step 5) are as follows: Match 4 models: 942 kinds, match 3 models: 2498 kinds, match 2 models: 4286 kinds, match 1 model: 2012 kinds, and no match: 1069 kinds.

7. A method of calculating the binding of perfluoro- and polyfluoroalkyl compounds to PPARa based on a pharmacophore model according to claim 6, characterized in that, In step 6), the molecular dynamics simulation adopts the AMBER99SB-ILDN force field to process PPARα, and the general AMBER force field GAFF to process the ligand. The simulation system contains a TIP3P water model and 150 mM NaCl. The equilibrium stage includes 200 ps of constant volume and 500 ps of constant pressure.

8. The method of claim 7, wherein the method is based on a pharmacophore model for calculating the binding of perfluoro- and polyfluoroalkyl compounds to PPARa, characterized in that, In step 6), the PFAS concentration for the heat shift analysis of the cells is 100 μM, the heat intervention temperature is 50°C, 55°C, 60°C, and 65°C, and the soluble PPARα is detected by Western Blot. The primary antibody is an anti-PPARα antibody.

Images