Methods for screening fxr modulators for anti-hbv activity based on machine learning

Abstract

The application discloses a method for screening FXR modulators with anti-HBV activity based on machine learning. FXR modulator compounds are collected, a classification prediction model is constructed based on a machine learning method, and the prediction effects of SVM and RF prediction models are calculated and evaluated respectively by using molecular fingerprint descriptors; the RF prediction model with better prediction performance is used to predict the FXR modulating activity of molecules in a drug library (FDA_HY-L022 Library); molecular docking analysis is performed on compounds with higher predicted activity probability; and the FXR modulating activity is detected by experiment. The AUC value of the FXR modulator virtual screening model constructed by the application reaches 0.93, which shows good prediction ability for the data set, and the accuracy of the model is verified by a cell experiment. In addition, anti-HBV activity tests using a HepG2.2.15 cell model find that six molecules have both strong FXR modulating activity and certain HBV replication inhibiting activity.

Description

Technical Field

[0001] This invention belongs to the field of drug discovery and relates to a method for screening FXR modulators with anti-HBV activity based on machine learning. Background Technology

[0002] FXR, a member of the nuclear receptor superfamily of NRs, is highly expressed in the gut-hepatic axis. Based on its role in regulating bile acid homeostasis, glucose and lipid metabolism, and inflammatory responses, FXR has become a highly anticipated target in the development of drugs for liver diseases such as PBC, NAFLD, NASH, and liver fibrosis. In recent years, studies have shown that exogenous FXR modulators have a significant inhibitory effect on HBV replication, and two synthetic non-steroidal FXR modulators are currently undergoing clinical trials for their anti-HBV effects. Therefore, FXR modulators have significant application potential in the treatment of liver-related diseases and in the fight against HBV. However, current research on the anti-HBV effects of FXR modulators is still limited, and there are no relevant machine learning models for screening FXR modulators with anti-HBV activity. With the continuous development of machine learning technology, machine learning models can be used to extract useful features from large amounts of compound and biological information to screen for active compounds. This invention constructs an efficient machine learning model that can accurately screen potential FXR modulators with anti-HBV activity, providing strong support for the application research of FXR modulators in the fight against HBV. Summary of the Invention

[0003] The purpose of this invention is to propose a method for screening FXR modulators with anti-HBV activity based on machine learning, aiming to assist drug developers in efficiently and accurately discovering potential FXR modulators with anti-HBV activity. This method improves prediction accuracy by integrating a random forest screening model and in vitro cell activity validation, thereby enhancing the efficiency and accuracy of drug screening.

[0004] To achieve the above objectives, the technical solution of the present invention is as follows: As a first aspect, a method for screening FXR modulators with anti-HBV activity based on machine learning is provided, comprising the following steps:

[0005] Step S1: Obtain active and inactive FXR modulator compounds from the ChEMBL database, and collect FXR modulator binding ability data for these compounds, including: ligand affinity for FXR binding, and half-activation concentration (EC) after ligand binding to FXR. 50 IC50 50 And the SMILES string. Preferably, the collected data is preprocessed, including data cleaning, data deduplication, standardization of concentration units, and label extraction. The label is set to indicate whether it has FXR modulating activity.

[0006] Step S2: Use molecular fingerprinting to perform cluster analysis on the database compounds to construct an FXR modulator compound library.

[0007] Preferably, the data in the FXR regulator compound library is divided into a training set and a test set, and the balance of the proportion of compounds labeled as having FXR regulatory activity and not having FXR regulatory activity in the training set is adjusted.

[0008] Step S3 involves feature calculation on the preprocessed data. Using PaDEL-Descriptor, the following 12 molecular fingerprint descriptors were selected as training features: AtomPairs2D-FingerprintCount, AtomPairs2D-Fingerprinter, Estate-Fingerprinter, Extended-Fingerprinter, Fingerprinter, GraphOnly-Fingerprinter, KlekotaRoth-FingerprintCount, KlekotaRothFingerprinter, MACCS-Fingerprinter, Pubchem-Fingerprinter, Substructure-FingerprintCount, and SubstructureFingerprinter. These 12 molecular fingerprint descriptors were then used individually and in various combinations.

[0009] Step S4: Build a machine learning model and train it to obtain the FXR regulatory activity prediction model.

[0010] Specifically, various machine learning methods were employed to establish FXR regulator activity prediction models, such as Random Forest (RF) and Support Vector Machine (SVM) algorithms, and the overall prediction performance of the two algorithms was compared. One or more combinations of molecular fingerprint descriptors were used as training features to establish FXR regulator activity prediction models, and their effects were compared. The FXR regulatory activity prediction models obtained by calculating Accuracy, Precision, Recall, F1 score, and AUC were used to optimize the training features and machine learning methods.

[0011] Step S5: Train the FXR regulator activity prediction model using preferred training features and machine learning methods (random forest is used in this invention). Input the data of the test compound (activity unknown), perform activity prediction, and select the top-ranking test compounds with the highest FXR regulation activity prediction results as candidate compounds.

[0012] Specifically, the data of the compounds to be tested are from the FDA_HY-L022 Library compound database.

[0013] Step S6, as a preferred step, involves further analysis of compounds with high predicted activity by the RF model to screen for potential FXR modulators. Crystal structures of three FXR-ligand complexes were selected, and molecular docking techniques were used to score these candidate compounds obtained in step S5. The binding energy fractions of the compounds to the FXR receptors were calculated, and the scoring order was determined. Several top-ranked compounds with potential FXR regulatory activity were then selected. This invention yielded 252 potential modulators.

[0014] Step S7, FXR regulator activity screening. To screen for potential FXR regulators, compounds from a 252-drug library (FDA_HY-L022 Library) were screened in HEK-293T-FXR-Luc cells for activity validation. Specifically, cell lines containing FXR-responsive elements and stably expressing luciferase were constructed. The selected compounds were screened at a concentration of 10 μM. Compounds with a regulation fold greater than 130% were considered agonists; compounds with a regulation fold less than 50% were considered antagonists or inverse agonists. As a control, DMSO was added to achieve a regulation fold of 100%.

[0015] Step S8: Phenotypic Validation of Anti-HBV Activity. To screen for the anti-HBV activity of potential FXR modulators, the anti-HBV activity of 95 compounds screened in S7 was validated in HepG2.2.15 cells. Specifically, the effect of FXR agonist compounds on HBeAg and HBsAg antigen secretion was determined using HepG2.2.15 cell ELISA, and cell viability was tested using the MTT assay to exclude the influence of cytotoxicity on the anti-HBV activity test.

[0016] Step S9: Finally, potential novel FXR modulator compounds are screened.

[0017] Furthermore, in step S1, data preprocessing includes:

[0018] Data cleaning: Remove FXR modulator binding capacity data that is not active and molecules with failed 3D conformation;

[0019] Data deduplication: Duplicates are removed by ID, retaining only unique IDs; the numerical values ​​of different FXR regulator binding capacities (ECs) produced when the same inhibitor molecule is acted upon by different types of FXR mutants. 50 (Value), keep the lowest value;

[0020] Standardized concentration units: The concentration units of compounds in the database are not consistent. To facilitate subsequent screening, the data units will be uniformly converted to the concentration unit nM (nmol / L).

[0021] Tag extraction: using FXR modulator binding capacity values ​​(EC) 50 The activity threshold (NM) is set at 2000 nM. Molecules with an NM value less than the activity threshold are labeled as category "1" and are considered as activity regulators of FXR. Otherwise, they are considered as weak or inactive regulators and are labeled as category "0". The category is the activity label.

[0022] Furthermore, in step S2, the cluster analysis uses the MACCS molecular fingerprint in RDKIT to calculate the Tanimoto coefficient, thereby obtaining molecular fingerprint clusters.

[0023] Specifically, with the threshold set to 80%, the database showed 192 compound clusters.

[0024] Furthermore, in step S4, the selection of the classification prediction model is achieved by calculating accuracy, precision, recall, F1 score, and AUC. The training features and machine learning methods are optimized by comprehensively comparing the scores of each indicator. The formulas for calculating each indicator are as follows:

[0025] (1) Accuracy: Accuracy is the most common performance evaluation metric for classification problems. It measures the proportion of samples that the model correctly predicts. However, in the case of an imbalanced class distribution, accuracy may be misleading, so it is necessary to combine it with other metrics to evaluate model performance.

[0026]

[0027] (2) Precision and Recall: Precision and recall are important metrics for imbalanced class distribution problems. Precision measures the proportion of true positives among the predicted positives, while recall measures the proportion of correct positives predicted by the model out of the actual positives.

[0028]

[0029] (3) F1 score: The F1 score is the harmonic mean of precision and recall, which takes into account the performance of both. It is a comprehensive performance metric that can provide a more comprehensive evaluation when dealing with imbalanced datasets.

[0030]

[0031] (4) AUC is calculated based on the area under the ROC curve (Receiver Operating Characteristic Curve), also known as the AUC-ROC curve (Area Under the Receiver Operating Characteristic Curve), and is an important tool for evaluating the performance of binary classification models. This curve plots the relationship between the True Positive Rate and the False Positive Rate at different thresholds, providing a visual understanding of the model's performance in classification tasks. The size of the area under the AUC-ROC curve reflects the model's ability to distinguish between positive and negative examples, thus evaluating the model's quality. The ROC curve is plotted with FPR (False Positive Rate) on the horizontal axis and TPR (True Positive Rate, or Recall) on the vertical axis. The AUC value ranges from 0 to 1; a larger value indicates better model performance.

[0032] FPR = FP / (TN + FP)

[0033] TPR = Recall = TP / (TP + FN)

[0034] Where TP represents true positives, FP represents false positives, FN represents false negatives, and TN (True Negatives) represents true negatives, which is the number of samples correctly predicted as negative.

[0035] Furthermore, in step S4, after comparing SVM and RF machine learning algorithms, the optimal model is selected by the average AUC of 10 cross-validations performed under different hyperparameter settings.

[0036] Furthermore, in step S4, RF is implemented by calling the RandomForest-Classifier module through the SKLEARN library. The model uses 10 cross-validations, 500 decision trees, a maximum depth of 30 for each tree, and a minimum number of samples required for node splitting of 2. A random forest classifier with a random seed of 42 is also set to correct the model instability caused by overfitting of a single decision tree. The SVM model is constructed using the SVM module of the SKLEARN library. The optimal hyperparameters of the algorithm are determined through grid search and 10 cross-validations. In order to comprehensively and rigorously evaluate the model performance, the average values ​​of each classification index need to be calculated and summarized under the optimal parameter settings.

[0037] Furthermore, in step S5, the RF parameters are set as follows: total sample size 1781, training set to test set ratio 8:2, training set compounds 1425, test set compounds 356, two classifications: 'active', 'inactive', and RF and SVM modeling are performed.

[0038] Furthermore, in step S6, the top 10% of compounds with the highest activity prediction probability are selected for molecular docking, and a binding energy fraction of ≤-10kcal / mol is set as the threshold for screening. A total of 252 compounds are identified as compounds with potential FXR regulating activity.

[0039] Furthermore, in step S7, compounds from the 95 drug library (FDA_HY-L022 Drug Library) with a regulatory activity of 130% or higher at a concentration of 10 μM are considered agonists; compounds with a regulatory fold of less than 50% are considered antagonists or inverse agonists.

[0040] Furthermore, in step S8, the anti-HBV activity of 37 FXR agonists was analyzed. Among them, 2 compounds showed an HBeAg inhibition rate >50%, and 6 small molecules showed an HBsAg inhibition rate >50%. MTT assay was performed on cell viability, and 36 compounds showed a viability rate greater than 70%, indicating that 36 FXR agonists have anti-HBV activity and do not affect cell viability.

[0041] As a second aspect, we provide FXR modulators of anti-HBV activity obtained through machine learning screening, which are any compounds with the structural formulas shown in FL113, FL116, FL889, FL917, FL1387, FL1578, and FL1855.

[0042]

[0043] The beneficial effects of this invention include: This invention employs advanced machine learning technology to provide an efficient method for screening FXR modulators, which is of great significance in the field of anti-HBV therapeutics. This invention develops an innovative screening process: First, molecular fingerprinting is used to process compounds in the database; second, advanced support vector machine and random forest models are used to train the database, improving the accuracy and stability of the model in predicting FXR modulator activity; then, molecular docking is used to assist in the screening of FXR modulators; finally, in vitro activity verification effectively tests whether the model has learned key structural features from the training dataset, ensuring the model's screening capability. Experimental results show that the model exhibits good performance on the test set, with an average AUC value of 0.93, demonstrating its predictive accuracy in FXR modulator screening. Attached Figure Description

[0044] Figure 1 This is a flowchart of the machine learning-based FXR modulator discovery method for anti-HBV activity described in this invention.

[0045] Figure 2 This is the HBV inhibitor structure with FXR-modulating activity described in this invention. Detailed Implementation

[0046] This invention preprocesses the collected compound data, including data cleaning, deduplication, and tag extraction; constructs a molecular fingerprint database using molecular fingerprint extraction; builds a classification and prediction model for FXR regulator active compounds using random forest (RF) and support vector machine (SVM) algorithms, determines the optimal parameters of the algorithms through 10 cross-validations, and selects the optimal model from the two models through multidimensional comparison. Finally, the RF (random forest) model is selected to predict the activity of molecules in the drug library (FDA_HY-L022 Library); molecular docking analysis is performed on compounds with high predicted activity probabilities for further screening; and the FXR regulatory activity is experimentally detected using the HEK-293T-FXR-Luc stable cell line containing FXR response elements (FXRE) and stably expressing luciferase, with GW4064 as a positive control.

[0047] The following will combine Figure 1 The technical solution and beneficial effects of the present invention will be described in detail below.

[0048] This invention provides a machine learning-based method for predicting the activity of FXR modulators, specifically including the following steps:

[0049] S1. Collect FXR regulator data from the ChEMBL database, and then preprocess the collected data, including data cleaning, data deduplication, and label extraction.

[0050] Data Acquisition: FXR regulator data, including affinity values ​​and EC50 values, were collected from the ChEMBL database. 50 Value, IC 50 The values ​​and SMILES strings were used to collect 1781 data points related to FXR regulators;

[0051] Table 1. Statistical table of sample size for each dataset.

[0052] Dataset Positive samples negative samples total training set 711 714 1425 test set 178 178 356 total 889 892 1781

[0053] Data preprocessing mainly includes data cleaning, data deduplication, and label extraction.

[0054] Data cleaning mainly involves removing active and unusable data, as well as useless molecules such as metals and metal oxides.

[0055] Data deduplication primarily involves standardizing the SMILES strings and removing duplicates, retaining only the unique SMILES as the molecular representation; this also addresses the different EC values ​​produced when the same inhibitor molecule is interacted with different types of FXR mutants. 50 Values, keep the lowest value;

[0056] The affinity values ​​of the ligands extracted from the tags for binding to FXR, and the half-activation concentration (EC) values ​​generated after the ligands bind to FXR. 50 IC50 50 The activity threshold, set at 2000 nM, was determined to be a numerical value that reflects the binding affinity of the ligand to FXR. 50 Molecules with values ​​less than the activity threshold are labeled as category "1" and considered as active regulators of FXR; otherwise, they are considered as weak or inactive inhibitors and labeled as category "0".

[0057] S2. Use relevant tools to calculate features and filter the calculated descriptors. Then, divide the dataset into training and test sets.

[0058] Feature calculations used 12 molecular fingerprint descriptors from PaDEL-Descriptor: AtomPairs2D-FingerprintCount, AtomPairs2D-Fingerprinter, Estate-Fingerprinter, Extended-Fingerprinter, Fingerprinter, GraphOnly-Fingerprinter, KlekotaRoth-FingerprintCount, KlekotaRothFingerprinter, MACCS-Fingerprinter, Pubchem-Fingerprinter, Substructure-FingerprintCount, and SubstructureFingerprinter.

[0059] The fingerprint descriptor is the binary value corresponding to the feature. For example, each feature of the MACCS fingerprint corresponds to a specific chemical substructure, such as hydroxyl, benzene ring or nitrogen atom. If the structure exists, the value of the binary bit corresponding to the feature is 1, otherwise it is 0.

[0060] The selection of a classification prediction model involves calculating accuracy, precision, recall, F1 score, and AUC. The model is chosen by comprehensively comparing the scores of each metric. The formulas for each metric are as follows:

[0061] (1) Accuracy: Accuracy is the most common performance evaluation metric for classification problems. It measures the proportion of samples that the model correctly predicts. However, in the case of an imbalanced class distribution, accuracy may be misleading, so it is necessary to combine it with other metrics to evaluate model performance.

[0062]

[0063] (2) Precision and Recall: Precision and recall are important metrics for imbalanced class distribution problems. Precision measures the proportion of true positives among the predicted positives, while recall measures the proportion of correct positives predicted by the model out of the actual positives.

[0064]

[0065] (3) F1 score: The F1 score is the harmonic mean of precision and recall, which takes into account the performance of both. It is a comprehensive performance metric that can provide a more comprehensive evaluation when dealing with imbalanced datasets.

[0066]

[0067] (4) AUC is calculated based on the area under the ROC curve (Receiver Operating Characteristic Curve), also known as the AUC-ROC curve (Area Under the Receiver Operating Characteristic Curve). It is an important tool for evaluating the performance of binary classification models. This curve plots the relationship between the True Positive Rate and the False Positive Rate at different thresholds, providing a visual understanding of the model's performance in classification tasks. The size of the area under the AUC-ROC curve reflects the model's ability to distinguish between positive and negative examples, thus evaluating the model's quality. The ROC curve is plotted with FPR (False Positive Rate) on the horizontal axis and TPR (True Positive Rate, or Recall) on the vertical axis. The AUC value ranges from 0 to 1; a larger value indicates better model performance.

[0068] FPR = FP / (TN + FP)

[0069] TPR = Recall = TP / (TP + FN)

[0070] Where TP represents true positives, FP represents false positives, FN represents false negatives, and TN (True Negatives) represents true negatives, which is the number of samples correctly predicted as negative.

[0071] Table 2 compares the performance of two machine learning models, Random Forest (RF) and Support Vector Machine (SVM), and also shows the performance comparison of the two models under different feature selection methods, including accuracy, AUC, precision, recall, and F1 score. Considering all factors, the Random Forest model of this invention exhibits the best performance. Table 3 compares the results calculated by Random Forest for different combinations of molecular fingerprint descriptors, and the results show that the Substructure Fingerprint Count fingerprint achieves very good results in all indicators.

[0072] Table 2 compares the computation of twelve molecular fingerprint descriptors using Random Forest (RF) and Support Vector Machine (SVM).

[0073]

[0074]

[0075] Table 3 Results of RF Model Calculation Based on Molecular Fingerprint Descriptor Combination

[0076]

[0077] S3 applies the classification model to an external validation set for further validation and evaluation, predicts the activity of unknown molecules, and outputs the prediction results of whether each molecule has FXR regulatory activity.

[0078] S4. By combining the docking model, molecular docking was performed on the top 10% of compounds with the highest predicted activity. The binding energy was scored, and the threshold for the binding energy score was set as ≤-10 kcal / mol. A total of 252 compounds were identified as compounds with potential FXR regulating activity.

[0079] S5, FXR-regulated activity screening:

[0080] (1) Construction of the HEK-293T-FXR-Luc cell line: First, high-purity, endotoxin-free lentiviral vectors and their helper packaging vector plasmids were extracted and co-transfected into HEK-293T cells using HG transgene reagent. After a series of culture and treatments, including adding enhancing buffer after transfection, replacing fresh culture medium, collecting and concentrating cell supernatant rich in lentiviral particles, and finally measuring and labeling the viral titer in HEK-293T cells to ensure that the quality and quantity of the virus met the experimental requirements. Subsequently, HEK-293T cells were infected with packaged CMV-Luc-PGK lentivirus. After cell seeding, virus infection, replacement of culture medium, detection of infection efficiency, and resistance screening, a stable HEK-293T-FXR-Luc cell line was successfully constructed.

[0081] (2) Validation of HEK-293T-FXR-Luc cells: To validate the stability and functionality of the constructed HEK-293T-FXR-Luc cell line, we conducted a series of experiments. First, we observed the cell morphology under an inverted microscope to ensure that the cell confluence was suitable for the experiment. Then, through steps such as cell resuspension, seeding in 96-well plates, and changing the culture medium, we used the GMOne-Step luciferase reporter gene assay kit to detect the luciferase activity of the cells.

[0082] (3) Determination of the initial concentration of DMSO in drug screening: Before drug screening, the toxicity of DMSO to HEK-293T-FXR-Luc cells was first determined. The results were verified by setting up ten concentration gradients of DMSO from 0‰ to 9‰, and the growth status and luciferase activity of cells under different concentrations of DMSO were observed to determine the appropriate range of DMSO concentration.

[0083] (4) Determination of the effective concentration of the positive control drug GW4064 and EC 50 Determination: By comparing the effects of different concentrations of GW4064 on the luciferase activity of HEK-293T-FXR-Luc cells, 8 μM GW4064 was ultimately determined to be the optimal positive drug concentration. This result provides an important reference for our subsequent drug screening.

[0084] (5) Activity analysis of the FXR regulation effect of the test molecules at the cellular level: After determining the concentrations of DMSO and GW4064, the screening of the physical compounds began. Through cell detection, the compounds involved in the patent have good FXR regulation activity.

[0085] Table 4. Regulatory activities of compounds on intracellular FXR

[0086]

[0087]

[0088] Table 5. Structure of FXR modifier compounds

[0089]

[0090]

[0091]

[0092]

[0093]

[0094]

[0095]

[0096]

[0097]

[0098] S6, In vitro anti-HBV activity assay of FXR modulators: The sample solution was prepared by dissolving the aforementioned 95 FXR modulator compounds in DMSO at a concentration of 30 μM. The effects of the 95 FXR modulators on HBeAg and HBsAg antigen secretion were determined using ELISA. The specific experimental procedures are as follows: First, HepG2.2.15 cells in suitable condition were washed with PBS, and trypsin was added to disperse the cells. Culture medium (MEM + 10% FBS + 380 mg / mL G418) was added and pipetteted to form a single-cell suspension. Then, the suspension was transferred to an EP tube (15 mL), centrifuged to remove the supernatant, and 2 mL of complete culture medium (MEM + 10% FBS + 380 μg / mL G418) was added and pipetteted to form a single-cell suspension. The cells were then re-seeded in 48-well plates with approximately 3000 cells per well and incubated at 37°C and 5% CO2 for 24 hours. Finally, the culture medium was aspirated, and the cells were treated with the prepared sample solution (final concentration 30 μM). Lamivudine was used as a positive control, and cells without sample treatment were used as a blank control. The cells were incubated at 37°C and 5% CO2 for 72 hours, and the culture medium was collected. The HBeAg detection kit (Shanghai Kehua Bioengineering Co., Ltd., National Medical Device Registration Certificate No. 20163400144) and the HBsAg detection kit (Shanghai Kehua Bioengineering Co., Ltd., National Drug Approval No. S10910113) were used to detect the samples. The absorbance (OD value) was measured using an ELISA reader at a detection wavelength of 450 nm and a reference wavelength of 630 nm. The inhibition rates of the compounds against HBsAg and HBeAg were calculated according to the following formula:

[0099]

[0100] S7, FXR regulator in vitro MTT assay for cell viability of the compound

[0101] (1) Wash HepG2.2.15 cells in suitable condition with PBS, remove the cell culture medium, add 0.25% trypsin to digest the cells, add culture medium (MEM + 10% FBS + 380 mg / mL G418) and pipette to form a single-cell suspension; then, transfer the suspension to an EP tube (15 mL), centrifuge to remove the supernatant, add 2 mL of complete culture medium (MEM + 10% FBS + 380 mg / mL G418) and pipette to form a single-cell suspension, and re-seed in a 96-well plate with approximately 3000 cells per well;

[0102] (2) After 12 hours of adhesion, remove the culture medium and add 100 μL of culture medium (MEM+10%EBS+380mg / mLG418) containing different concentrations of the test sample (serially diluted from below 200μM to 1.56μM). Also set up blank wells (containing only culture medium) and control wells (without adding solvent of the test sample).

[0103] (3) After the cells and drugs were co-incubated at 37°C and 5% CO2 for 24 hours, 10 μL of MTT solution (10 mg / mL) was added to each well and co-incubated for 4 hours. Then, 100 μL of MTT solution was added and the cells were incubated at 37°C overnight. The absorbance (A value) of each well was measured at 550 nm.

[0104] (4) Calculate cell viability. Viability = (experimental wells - blank wells) / (control wells - blank wells) × 100%. The experiment was repeated three times.

[0105] The analysis of the anti-HBV activity of 37 FXR agonists, namely FXR regulators with a regulatory activity greater than 130% (Table 6), showed that 6 small molecules inhibited HBsAg secretion by cells by >50%, and 2 small molecules inhibited HBeAg secretion by >50%. Figure 2 The structures of compounds with regulatory activity greater than 130% in HEK-293T-FXR-Luc cell assays and inhibition rates of HBsAg secretion and HBeAg inhibition rates greater than 50% in HepG 2.2.15 cell assays were presented. MTT assays showed that 36 compounds exhibited cell viability greater than 70%, indicating that these 36 FXR agonists possessed anti-HBV activity without affecting cell viability.

[0106] Table 6. Screening results of anti-HBV activity

[0107]

[0108] Therefore, this invention proposes a highly efficient and accurate machine learning-based method for screening FXR modulators. It not only successfully constructed a virtual screening model for FXR modulators with an AUC value as high as 0.93, demonstrating excellent dataset prediction capabilities, but also verified the reliability of the model through cell experiments. Furthermore, in tests targeting anti-HBV activity, this method successfully screened six active molecules that possess both significant FXR agonist activity and the ability to inhibit HBsAg and HBeAg secretion, providing new candidate compounds for the development of FXR-related drugs and demonstrating the enormous potential and application value of this invention in the field of drug discovery.

Claims

1. The application of an FXR modulator in the preparation of anti-HBV drugs, characterized in that, The FXR modifier is any one of the compounds shown in FL113, FL116, and FL1855. 。

Images