Method for screening toxic components of traditional Chinese medicine by using MACCS structural similarity model constructed based on Vibrio fischeri
The screening of Chinese medicine ingredients through the MACCS structural similarity model based on Vibrio ferriformis solves the problem of low data processing and screening efficiency of QSAR model, and achieves rapid, cost-effective detection and evaluation of Chinese medicine toxic components.
Patent Information
- Application Number
- CN202411128218.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-16
- Publication Date
- 2024-12-13
AI Technical Summary
The existing QSAR model based on Vibrio Fischer has the problem that the data cannot be included in the training set and test set, which leads to differences in model evaluation and selection, and the construction and screening process takes a long time, reducing the efficiency of drug ingredients screening.
Using the MACCS structural similarity model based on Vitiligo Fischer, by mining toxic component data, using OpenBabel and MACCS algorithms to perform intermolecular similarity matching, screening mutually matched molecules with Tanimoto coefficient ≥0.5, and analyzing the Degree value in Cytoscape to construct Toxic molecular similarity model, similarity matching screening of Chinese medicine ingredients.
It reduces data loss, improves the efficiency of screening of drug toxic ingredients, and achieves rapid, repeatable and cost-effective detection of Chinese medicine toxic ingredients and toxicity assessment of compatibility.
Smart Images
Figure CN119132447A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of toxicity evaluation, and specifically relates to a method for screening toxic components of traditional Chinese medicine by utilizing a MACCS structural similarity model constructed based on Vibrio fischeri. Background Art
[0002] Vibrio fischeri is a common marine luminescent bacterium. When exposed to toxic substances in the external environment, its luminescence intensity will be affected. The change in its luminescence intensity can be used as an indicator to determine the biological toxicity of harmful and toxic substances in the environment. Vibrio fischeri has a higher sensitivity to toxins than general biological cells, so it is often used for the rapid detection and evaluation of tiny toxic substances.
[0003] Models built with Vibrio fischeri toxicity data combined with quantitative structure-activity relationship (QSAR) are widely used in the screening of toxic substances. However, there are still some problems with models built based on conventional QSAR, such as the failure to include all data when building training sets and test sets, and differences in the evaluation and selection of models after the model is built, which will result in partial data loss. In addition, it takes a lot of time to build training sets, test sets, and evaluate and screen after model construction, which greatly reduces the efficiency of drug ingredient screening.
[0004] Screening of toxic components of traditional Chinese medicine is of great significance for safe medication, disease prevention and treatment, and drug development. However, due to the diversity and complexity of traditional Chinese medicine components, toxic substances are difficult to accurately determine, which makes it difficult to screen toxic components of traditional Chinese medicine. In addition, the existing methods of screening toxic components of traditional Chinese medicine through conventional cell and animal experiments have the disadvantages of being time-consuming, expensive, and inefficient. Therefore, it would be of great significance if the existing model constructed by combining the toxicity data of Vibrio fischeri with quantitative structure-activity relationship could be improved and used for screening toxic components of traditional Chinese medicine. Summary of the invention
[0005] In order to solve the technical problems that when screening drug toxic components using a model based on conventional QSAR, the data cannot be fully included when constructing training sets and test sets, and there are differences in the evaluation and selection of the model after the model is built, resulting in partial data loss; and it takes a lot of time to construct the training set, the test set and the evaluation and screening after the model is built, which greatly reduces the efficiency of drug toxic component screening, the present invention improves the existing model constructed with Vibrio fischeri toxicity data, obtains the MACCS structural similarity model based on Vibrio fischeri, and applies the model to the screening of toxic components of traditional Chinese medicine, solving the problems of data loss and low screening efficiency.
[0006] In order to solve the above technical problems and achieve corresponding technical effects, the present invention specifically provides the following technical solutions:
[0007] The first object of the present invention is to provide a method for screening toxic components of traditional Chinese medicine using a MACCS structural similarity model constructed based on Vibrio fischeri, the method comprising the following steps:
[0008] S1. Construction of MACCS similarity model of toxic molecules: Toxic molecules were mined from the dataset of toxic components that Vibrio fischeri could sense, OpenBabel was installed, the MACCS algorithm was called, the mined toxic molecules were formatted, and the MACCS characterization script was run to match the molecular similarity, and the molecules with Tanimoto coefficient ≥ 0.5 that matched each other were screened to obtain the interaction points. The interaction points were imported into CytoNCA in Cytoscape, and the Degree values of the molecules with Tanimoto coefficient ≥ 0.5 that matched each other were analyzed. 90, half of the maximum value of Degree 180, was taken as the screening threshold, and finally the molecules with Degree value greater than 90 were screened out as a toxic molecule similarity dataset for constructing the toxic molecule similarity model;
[0009] S2. Screening of toxic substances in traditional Chinese medicine: Obtain traditional Chinese medicine ingredients and convert them into SMILES format. Use the similarity model constructed in ① to perform similarity matching between the above-mentioned Chinese medicine ingredients and the toxic ingredients obtained in ①. If the Chinese medicine ingredients satisfy the Tanimoto coefficient ≥ 0.5 and Degree ≥ 90, the Chinese medicine ingredients are judged to be toxic ingredients.
[0010] In one embodiment of the present invention, S1 runs the MACCS characterization script using Python 3.12.3.
[0011] In one embodiment of the present invention, S2 obtains traditional Chinese medicine ingredients from literature or from the TCMSP database.
[0012] The second object of the present invention is to provide a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for screening toxic components of traditional Chinese medicine using the MACCS structural similarity model constructed based on Vibrio fischeri is implemented.
[0013] The third object of the present invention is to provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for screening toxic components of traditional Chinese medicine using the MACCS structural similarity model constructed based on Vibrio fischeri.
[0014] Beneficial effects of the present invention:
[0015] The present invention constructs a MACCS structural similarity model based on the toxicity data of Vibrio fischeri, which can directly match the screened traditional Chinese medicine components with the toxic components in the model, thereby reducing data loss in the process of model construction, data conversion, etc. The model is characterized by high speed and strong operability, and can be applied to similarity matching between multiple compound structures, and then used for toxic substance screening.
[0016] The method provided by the present invention for screening toxic components of traditional Chinese medicine using the MACCS structural similarity model constructed based on Vibrio fischeri can be used for detecting trace toxicity of traditional Chinese medicine and evaluating changes in toxicity of traditional Chinese medicine combinations, and has the advantages of being rapid, repeatable, economical and efficient. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 This is a cluster analysis result diagram of the docking results of traditional Chinese medicine toxic molecules and proteins LuxA_A, LuxA_C, LuxB_B, LuxB_D, LuxC, LuxD, LuxE, and LuxG; among them, Figure 1 a in the figure is the cluster analysis result of the docking results of EF, ZJW and ED toxic molecules with proteins LuxA_A, LuxA_C, LuxB_B, LuxB_D, LuxC, LuxD, LuxE and LuxG. Figure 1 b in the figure is the cluster analysis result diagram of the docking results of PF, ESP and SSP toxic molecules with proteins LuxA_A, LuxA_C, LuxB_B, LuxB_D, LuxC, LuxD, LuxE and LuxG;
[0018] Figure 2 Docking analysis and visualization results of B-101_LuxE;
[0019] Figure 3 The docking analysis and visualization results of R-637542_LuxG;
[0020] Figure 4 Docking analysis and visualization results of W-8655_LuxC;
[0021] Figure 5 Docking analysis and visualization results of Z-8468_LuxG;
[0022] Figure 6 This is the docking analysis and visualization result diagram of T-88944_LuxG;
[0023] Figure 7The results of using RMSF, RMSD and Rg to evaluate the fluctuation of protein amino acid residues, the stability of molecule-protein binding and the folding state of protein; Figure 7 The a in the figure is the result of evaluating the fluctuation of protein amino acid residues using RMSF. Figure 7 The result of b in the figure is the evaluation of the stability of the binding between the molecule and the protein using RMSD. Figure 7 c in the figure is the result of evaluating the protein folding state using Rg;
[0024] Figure 8 is the free energy profile of the binding between molecules and proteins; Figure 8 a in the figure is the free energy morphology diagram of B-101_LuxE. Figure 8 b in the figure is the free energy morphology diagram of R-637542_LuxG. Figure 8 c in is the free energy morphology diagram of Z-8468_LuxG, Figure 8 The d in is the free energy morphology diagram of W-8655_LuxC, Figure 8 The e in FIG. is the free energy morphology diagram of T-88944_LuxG. DETAILED DESCRIPTION
[0025] The present invention is further described in detail below in conjunction with specific examples and drawings. The following examples are convenient for better understanding of the present invention, but are not intended to limit the present invention. The experimental methods in the following examples are conventional methods unless otherwise specified. The drugs, reagents, and materials used in the following examples can be purchased through commercial channels unless otherwise specified.
[0026] Example 1: A method for screening toxic components of traditional Chinese medicine using the MACCS structural similarity model constructed based on Vibrio fischeri
[0027] (1) Constructing the MACCS similarity model of toxic molecules based on the toxic components of Vibrio fischeri to screen the toxic components of traditional Chinese medicine
[0028] ①Construction of MACCS similarity model of toxic molecules
[0029] The present invention first excavates 1277 toxic molecules from the published public data set of toxic components that Vibrio fischeri can perceive (https: / / github.com / hhaootian / toxicity / blob / main / src / clean.py), installs OpenBabel (https: / / github.com / openbabel / openbabel / releases), calls the MACCS algorithm, and formats the 1277 toxic molecules. Then, Python 3.12.3 is used to run the MACCS characterization script (https: / / github.com / ZhangChengCADEN / MACCS-Model) to perform intermolecular similarity matching. If the Tanimoto coefficient is greater than or equal to 0.5, it is judged to be a molecule with strong toxicity. Mutually matching compounds with a Tanimoto coefficient ≥ 0.5 are screened to obtain 73,710 interaction points. Finally, the obtained interaction points were imported into CytoNCA in Cytoscape (https: / / cytoscape.org / ), and the Degree values of molecules with Tanimoto coefficients greater than or equal to 0.5 were analyzed. 90, half of the maximum Degree value of 180, was taken as the screening threshold (the first 50% of the Degree value data). Finally, 304 compounds with Degree values greater than 90 were screened out as a toxic compound similarity dataset for constructing a toxic molecule similarity model.
[0030] ②Use the MACCS similarity model of toxic molecules to screen toxic substances in traditional Chinese medicine
[0031] Firstly, the components of traditional Chinese medicines Evodiae Fructus (EF), Zuojin Wan (ZJW), Evodia Decoction (ED), Psoraleae Fructus (PF), Ershen Pills (ESP), Sishen Pills (SSP) were obtained. Zuojin Wan is composed of Evodiae Fructus and Coptis chinensis, Evodia Decoction is composed of Evodiae Fructus, ginger, ginseng and jujube, Ershen Pills is composed of salt psoraleae, stewed nutmeg, jujube and ginger, and Sishen Pills is composed of salt psoraleae, stewed nutmeg, vinegar-processed Schisandrae Fructus, processed Evodiae Fructus, jujube and ginger. The components of PF were obtained from the literature, with a total of 102 components; the components of EF, ZJW, ED, ESP and SSP were obtained from the TCMSP database (https: / / www.tcmsp-e.com / ), and 176 components of EF, 48 components of Coptis chinensis in ZJW, 133 components of jujube in ED, 265 components of ginger, 190 components of ginseng, 64 components of nutmeg in ESP, and 50 components of Schisandra chinensis in SSP were obtained through the TCMSP database. The above components were converted into SMILES format, and the similarity model constructed in ① was used to perform similarity matching between the above Chinese medicine components and 304 toxic components. When the Chinese medicine component meets the Tanimoto coefficient greater than or equal to 0.5 and the Degree is greater than or equal to half of the maximum value, the Chinese medicine component is judged to be a toxic component of Chinese medicine. According to the above method, 7 toxic components were screened in EF, 13 toxic components in ZJW, 20 toxic components in ED, 4 toxic components in PF, 16 toxic components in ESP, and 22 toxic components in SSP.
[0032] (2) Verification of the screening results of the MACCS similarity model for toxic molecules
[0033] ① Docking analysis of Vibrio fischeri luciferase and toxic molecules
[0034] The luciferases LuxA, LuxB, LuxC, LuxD, LuxE, and LuxG of Vibrio fischeri are involved in the luminescence regulation of Vibrio fischeri. The LuxA gene and LuxB gene encode the α subunit and β subunit of the luciferase, respectively. The heterodimer they form is called LuxAB, which has the ability to emit light in bacteria. Based on published articles, the species of Vibrio fischeri, and the analysis of protein structures, the best LuxAB (protein structure ID is 3FGC) and LuxG (protein structure ID is 1BKJ) (https: / / www.rcsb.org / pages / about-us / index) (RCSB PBD) were selected, while LuxC (protein structure ID is A0A510UCI8), LuxD (protein structure ID is A7MAR5), and LuxE (protein structure ID is P24272) were obtained through the AlphaFold protein structure database (https: / / alphafold.com / ). Schro dinger Maestro 2021 software (https: / / www.schrodinger.com / release-download / ) was used for docking: First, the 304 toxic components obtained in Example 1 were processed with the OPLS4 field using LigPrep, the ligand size was limited to less than 500 atoms, and ionization was performed using Epik. Protein Preparation Wizard was then used for preprocessing before protein docking. When assigning H-bonds, PROPKA was used for optimization, and the OP LS4 field was selected to minimize all atoms. The docking frame was then constructed, and Receptor Grid Generation was used to construct docking frames for LuxA, LuxB, and LuxG in a “picking to identify the ligand” manner. At the same time, blind docking was used for LuxC, Lux D, and LuxE. Finally, in ligand docking, standard precision (SP) was used for molecule-protein docking. After the docking is completed, the scores of molecule and protein binding are exported, and the molecules and docking scores are visualized and clustered using the microbial information platform (clustering method: complete, distance method: Euclidean, callback function: pheatmap).
[0035] Since docking software is used to perform force field processing on molecular structures and batch docking of molecules and proteins, even if the same molecules in different traditional Chinese medicines are docked with the same protein, the docking results may not be the same. Therefore, we performed cluster analysis on all docking results to see whether the same molecules and proteins in different traditional Chinese medicines are clustered together to verify the accuracy of the docking results. LuxAB is a large complex. During the docking process, the docking operation of the molecule to the protein cannot be completed. Based on the protein structure, we split the protein into LuxA_A, LuxA_C, LuxB_B, and LuxB_D, so that the molecules can fully perform computational docking and matching on the protein, avoiding the local influence of molecular action on the protein, and taking into account the diversity of the binding between molecules and proteins. Cluster analysis found that most of the components with the same number were classified into one category, indicating that the binding difference between the same components in different traditional Chinese medicines and the same protein is not much (see Figure 1 ).
[0036] Through the analysis of cluster diagrams and docking results, we selected the component and protein complexes according to the lowest docking score for visualization and molecule-protein interaction analysis. The main binding mode of the molecules is hydrogen bond interaction. The main binding amino acid residues of Psoralea corylifolia and its compatibility Ershen Pills and Sishen Pills are PHE, VAL, SHR, ARG. The main binding amino acid residues of Evodia rutaecarpa and its compatibility Zuojin Pills and Evodia rutaecarpa decoction are THR, TYH, GLY, SER, HIP, LYS, ARG, PHE, HIS, TYR, and VAL. Due to the large number of dockings, we selected the toxic molecules with the lowest docking score, the largest number of docking bonds, and the best ligand-receptor binding in each traditional Chinese medicine for further analysis. Among Psoralea corylifolia, B-101_LuxE was selected as the strongest (docking score of -6.842), and the binding sites are VAL162, PHE280 (see Figure 2 ), R-637542_LuxG (docking score of -7.487) was selected in Ershen Pills, and the binding sites were LYS_A:167, ARG_A:169, HIP_A:11, SER_A:13 (see Figure 3 ), Sishen Pills are the same as Ershen Pills. W-8655_LuxC (docking score of -5.54) was selected from Evodia rutaecarpa, and the binding sites are THR28, TYR92, GLY153 (see Figure 4 ), Z-8468_LuxG (docking score of -7.115) was selected in Zuojin Pills, and the binding sites were SER_A:13, HIP_A:11, ARG_A169, LYS_A:167, SER_B:38 (see Figure 5 ), T-88944_LuxG (docking score of -7.161) was selected in Wuzhuyu Decoction, and the binding sites were LYS_A:167, ARG_A:169, HIP_A:11, ARG_A:15 (see Figure 6 ). (B-101: 3-hydroxybenzaldehyde, R-637542: (E)-3-(4-hydroxyphenyl)prop-2-enoic acid, W-8655: 4-hydroxy-3,5-dimethoxybenzaldehyde, Z-8468: 4-hydroxy-3-methoxybenzoic acid, T-88944: 1-(2-hydroxy-4-methoxyphenyl)ethanone).
[0037] ②Molecular dynamics analysis
[0038] Molecular dynamics simulation uses GROMACS2020 (https: / / manual.gromacs.org / ), which needs to be compiled before use. First, use Pymol (https: / / pymol.org / ) to separate proteins and molecular complexes. Use ATB (https: / / atb.uq.edu.au / index.py) to add force fields to molecules and install the corresponding force field packages into GROMACS. When processing proteins, make sure that the ligand molecule force field is consistent with the protein force field. After adding the force field to the molecule, check whether hydrogen has been added. If not, it needs to be added. Then, call the GROMACS command to convert the molecular format of the ligand to the gro format, build the topological structure of the protein, make sure that the water molecules have been removed from the protein in advance, and then build the complex topology file of the ligand and protein. Determine the size of the molecular dynamics simulation box according to the size of the complex, and add solvent water to the box. The next step is to model the biological system, add ions to the complex, energy minimization, NVT thermal equilibrium, NPT pressure equilibrium, density analysis steps, and use DuIvyTools (https: / / github.com / CharlesHahn / DuIvyTools / tree / dev / DuIvyTools) for data visualization. Finally, the GPU is called for calculation to perform molecular dynamics simulations on the constructed biological system. The analysis of the simulation results mainly focuses on the calculation of the steady-state root mean square deviation (RMSD), root mean square fluctuation (RMSF), and the calculation of the radius of gyration (Rg) and the drawing of a 3D free energy landscape using Origin 2021 based on RMSD and Rg.
[0039] ③Molecular dynamics simulation
[0040] The present invention uses Schrodinger software to carry out the docking of molecule and protein, is with semi-flexible docking mode, and receptor is in dynamic change in real environment, adopts molecular dynamics simulation to judge the dynamic interaction of molecule and protein.Therefore, in order to further verify the binding degree and stability of Chinese medicine component docking protein, molecular dynamics simulation greater than 10ns is carried out.We select the ligand and receptor binding body (B-101_LuxE, R-637542_LuxG, W-8655_LuxC, Z-8468_LuxG, T-88944_LuxG) screened in molecular docking result to carry out molecular dynamics simulation.RMSF is used to evaluate the fluctuation mobility of protein amino acid residues, and the smaller the value, the smaller the fluctuation degree. Between 100 and 4000 atoms, the RMSF values of B-101_LuxE, R-637542_LuxG, W-8655_LuxC, Z-8468_LuxG, and T-88944_LuxG fluctuated within the range of 0.4 nm, indicating that the Chinese medicinal ingredients B-101, R-637542, W-8655, Z-8468, and T-88944 had little effect on the overall structure of the protein (see Figure 7 a in the figure). RMSD is used to further evaluate the stability of the molecule-protein binding. The smoother the RMSD curve, the more stable the molecule-protein binding. Within 0-15ns, B-101_LuxE, R-637542_LuxG, W-8655_LuxC, Z-8468_LuxG, and T-88944_LuxG fluctuate within 0.5nm, indicating that the molecule and protein can be stably bound ( Figure 7 b). The protein folding state is further evaluated by Rg. As time goes by, the larger the Rg value, the more likely the protein is in the process of unfolding, and vice versa, the more likely it is in the process of folding ( Figure 7 c) in.
[0041] We make free energy topography, plotting it with RMSD, Rg and Gibbs relative free energy as the X, Y and Z axes, to simulate the lowest energy conformation of the receptor during the entire binding process with the molecule, and judge the strength of the interaction between the ligand and the protein. If the interaction between the ligand and the protein is strong, the peak of the free energy topography is obvious and smooth, and there are fewer multiple minimum energy clusters with undulating surfaces. The free energy topography of W-8655_LuxC and T-88944_LuxG is almost a single and smooth minimum energy cluster ( Figure 8 d and e), indicating that the ligand binds stably to the receptor. The free energy topography of B-101_LuxE and R-637542_LuxG has fewer dispersed energy clusters ( Figure 8 a and b), indicating that the ligand-receptor binding is relatively stable. Z-8468_LuxG is a narrow, rough, and dispersed energy cluster ( Figure 8c), indicating that the stability of these complexes is poor. According to the results of molecular dynamics simulation of various indicators, W-8655 has stable effects on LuxC and T-88944 has stable effects on LuxG, B-101 has relatively stable effects on LuxE and R-637542 has relatively stable effects on LuxG, and Z-8468 has poor stability on LuxG.
[0042] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
Claims
1. A method for screening toxic components of traditional Chinese medicine using the MACCS structural similarity model constructed based on Vibrio fischeri, characterized in that: The steps include: S1. Construction of MACCS similarity model of toxic molecules: Toxic molecules were mined from the dataset of toxic components that Vibrio fischeri could sense, OpenBabel was installed, the MACCS algorithm was called, the mined toxic molecules were formatted, and the MACCS characterization script was run to match the molecular similarity, and the molecules with Tanimoto coefficient ≥ 0.5 that matched each other were screened to obtain the interaction points. The interaction points were imported into CytoNCA in Cytoscape, and the Degree values of the molecules with Tanimoto coefficient ≥ 0.5 that matched each other were analyzed. 90, half of the maximum value of Degree 180, was taken as the screening threshold, and finally the molecules with Degree value greater than 90 were screened out as a toxic molecule similarity dataset for constructing the toxic molecule similarity model; S2. Screening of toxic substances in traditional Chinese medicine: Obtain traditional Chinese medicine ingredients and convert them into SMILES format. Use the similarity model constructed in ① to perform similarity matching between the above-mentioned Chinese medicine ingredients and the toxic ingredients obtained in ①. If the Chinese medicine ingredients satisfy the Tanimoto coefficient ≥ 0.5 and Degree ≥ 90, the Chinese medicine ingredients are judged to be toxic ingredients.
2. The method according to claim 1, characterized in that S1 ran the MACCS characterization script using Python 3.12.
3.
3. The method according to claim 1, characterized in that S2 The way to obtain traditional Chinese medicine ingredients is to obtain them from literature or from the TCMSP database.
4. A computer device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for screening toxic components of traditional Chinese medicine using the MACCS structural similarity model constructed based on Vibrio fischeri as described in any one of claims 1 to 3 is implemented.
5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the method for screening toxic components of traditional Chinese medicine using the MACCS structural similarity model constructed based on Vibrio fischeri as described in any one of claims 1 to 3 is implemented.