Molecular docking method of E3 ligase target protein

Abstract

The invention discloses a molecular docking method of E3 ligase target protein, which comprises the following steps: determining an E3 enzyme required to be searched by a user in response to E3 enzyme basic information input by the user; according to the determined E3 enzyme, a PDB structure needing to be docked is selected; and in response to the coordinate of the docking box entered by the user, uploading the small molecules needing to be docked, and after docking is completed, sending a docking result according to the mailbox reserved by the user. The method is applied to molecular docking prediction of E3 ligase target protein, helps to understand the protein structure of E3 enzyme and analyze the interaction mechanism of E3 enzyme and substrate protein, and can provide advanced layout for design of protein degradation related drugs and even provide important guidance for development of broad-spectrum protein degradation agents. The invention provides important support for efficient development of basic research and drug development related to E3 enzyme, and meanwhile, helps to reveal the action mechanism of E3 enzyme in the fields of tumors, neurodegenerative diseases, autoimmune diseases and the like, and provides theoretical basis and technical support for precise medical treatment.

Description

Technical Field

[0001] The invention relates to a molecular docking method for an E3 ligase target protein. Background Art

[0002] E3 ligases are key catalytic enzymes in the ubiquitin-proteasome pathway, responsible for recognizing specific substrates and transferring ubiquitin molecules to substrate proteins, thereby regulating the stability, activity and subcellular localization of proteins. There are many types of E3 enzymes, and their specific recognition functions determine that they play an important role in regulating biological processes such as cell growth, apoptosis, and immune response. Studying the structure and function of E3 enzymes and their targets is of great significance for understanding the mechanism of disease occurrence and developing targeted treatment strategies.

[0003] At present, there are hundreds of known E3 enzyme family members, which show high specificity in substrate recognition and ubiquitin transfer mechanisms. In order to deeply understand the molecular mechanism of E3 enzymes, analyzing their high-resolution three-dimensional structures has become one of the key research directions. However, traditional experimental methods such as X-ray crystallography and cryo-electron microscopy face certain challenges in E3 enzyme research, including harsh experimental conditions and the difficulty of crystallizing some E3 enzymes. Summary of the invention

[0004] The present invention provides a molecular docking method for E3 ligase target protein in order to solve the problems existing in the above-mentioned prior art.

[0005] The present invention proposes to be deployed on a CentOS server, and the underlying language of the front-end interface is developed based on the HTML5 standard. NGL Viewer is used as a 3D molecular viewer to intuitively display the interaction between the E3 enzyme target and the ligand. The recording of the website access source is implemented through the RevolverMaps control. The underlying service architecture uses Docker for containerized deployment to ensure the flexibility and scalability of the system. NGINX is responsible for receiving HTTP requests from ports 80 and 443, and the back end is developed using PHP (Hypertext Preprocessor) to provide efficient data processing functions. User data storage is based on the MySQL database, and OpenBabel is used as a tool for molecular format conversion and 3D coordinate generation. The technical solution adopted by the present invention is:

[0006] A molecular docking method for E3 ligase target protein, comprising:

[0007] In response to the basic information of the E3 enzyme entered by the user, determining the E3 enzyme that the user needs to find;

[0008] According to the determined E3 enzyme, select the PDB structure to be docked;

[0009] In response to the coordinates of the docking box entered by the user, the small molecule to be docked is uploaded. After the docking is completed, the docking result is sent according to the email address left by the user.

[0010] Furthermore, the inputted basic information of the E3 enzyme includes: the function and category of the E3 enzyme and the protein sequence of the E3 enzyme.

[0011] Furthermore, the format of the small molecule to be docked is mol2 format or sdf format.

[0012] Furthermore, the E3 enzyme is docked with the small molecule using CoDock-Ligand as a docking engine.

[0013] Further, after the E3 enzyme and the small molecule are docked, a scoring function is used to evaluate the binding free energy of the E3 enzyme and the small molecule. The scoring function is:

[0014] ΔG binding =ΔG intermolecular +ΔG torsional +ΔG solvent

[0015] Where: ΔG intermolrcular is the intermolecular interaction energy, ΔG torsional Contribution to the conformational freedom of the ligand molecule, ΔG solvent It refers to the energy changes of receptor and ligand desolvation;

[0016] ΔG intermolecular =E vdW +E hydrophobic +E electrostatic

[0017] E vdW This is estimated using the Lennard-Jones potential energy function:

[0018]

[0019] Where r: atomic distance, ∈: potential energy depth, σ: equilibrium distance between two atoms;

[0020] E hydrophoboc It is related to the solvent accessibility and the number of hydrophobic atoms on the molecular surface;

[0021] E electrostatic Estimate using the Coulomb potential:

[0022]

[0023] where q 1 q 2 : the charge of the interacting atoms, ∈ r : dielectric constant;

[0024] ΔG torsional =-RTln(N)

[0025] Where R is the gas constant, T is the temperature, and N is the number of rotatable bonds in the ligand.

[0026] Furthermore, a scoring function is added to the scoring function, and the scoring function uses a random forest algorithm to associate the interaction characteristics of the receptor and the ligand with the experimentally measured binding free energy. The calculation process is as follows:

[0027] S1: Feature extraction:

[0028] Extract a set of features describing the ligand-receptor interaction from the molecular structure, including: the type of atom pairs, the distance between atom pairs, and the physical quantity of the interaction energy;

[0029] S2: Random Forest Modeling:

[0030]

[0031] in: Predicted binding free energy, N trees : The number of decision trees in the random forest, T i (x): prediction of feature vector x by the i-th tree, x: feature vector describing ligand-receptor interaction;

[0032] The features used by the scoring function are based on the distance and type between atom pairs, and are formulated as:

[0033] x=[f 11 ,f 12 ,…,f ij ,…,f nn ]

[0034] where f ij : the interaction frequency between receptor atom type i and ligand atom type j within a certain distance range, n: the total number of atom types of receptor and ligand;

[0035] The loss function for supervised learning training of experimental binding free energy data by random forest is as follows:

[0036]

[0037] Where N: the number of samples in the dataset, Model-predicted binding free energy, y i : Experimentally measured binding free energy;

[0038] S3: Binding free energy prediction: input molecular features, output binding free energy ΔGbinding .

[0039] The present invention has the following beneficial effects:

[0040] The present invention is applied to the molecular docking prediction of E3 ligase target proteins, helps to understand the protein structure of E3 enzymes, analyzes their interaction mechanism with substrate proteins, can be used to plan ahead for the design of protein degradation-related drugs, and even provide important guidance for the development of broad-spectrum protein degraders. The present invention provides important support for the efficient development of basic research and drug development related to E3 enzymes, while helping to reveal the mechanism of action of E3 enzymes in the fields of tumors, neurodegenerative diseases, autoimmune diseases, etc., and provides theoretical basis and technical support for precision medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flow chart of the present invention.

[0042] Figure 2 For the search page.

[0043] Figure 3 For the search results page.

[0044] Figure 4 Select the interface for the PDB.

[0045] Figure 5 Select the interface for the docking box.

[0046] Figure 6 Upload interface for small molecules.

[0047] Figure 7 This is the interface for submitting tasks.

[0048] Figure 8 Query page for results.

[0049] Fig. 9 This is the result interface of the crystal structure docking example.

[0050] Fig.10 This is the result interface of the Alphafold predicted structure docking example. DETAILED DESCRIPTION

[0051] The present invention will be further described below in conjunction with the accompanying drawings.

[0052] like Figure 1 The present invention provides a molecular docking method for E3 ligase target protein, which first prepares an E3 enzyme target information data set (1059), and each target provides the corresponding crystal structure and Alphafold predicted structure data of the known E3 enzyme target.

[0053] For small molecule docking, the present invention uses CoDock-Ligand as the docking engine. Users only need to enter the search box ( Figure 2 Enter the relevant information of E3 enzyme, such as UniProt, gene name and protein name, and the search results page ( Figure 3 The search results page provides a list of proteins that match the searched E3 enzyme information, as well as related functional information, E3 enzyme category, and protein sequence.

[0054] After selecting the corresponding E3 enzyme, you can select the required PDB structure ( Figure 4 For AlphaFold predicted structures, an average confidence score is also provided to confirm the reliability of AlphaFold predicted structures. The binding pocket of each PDB structure is predicted by DeepPocket, and users can select different pockets for docking ( Figure 5 Select interface for docking box), docking pocket is set to 28×28×28 After selecting the docking pocket, upload the small molecule file in mol2 or sdf format ( Figure 6 For small molecule upload interface). On the result display page, CoDock-Ligand is used to sort the docking scores. The top 10 models with the best scores will be displayed on the Results page, and 3D visualization can be performed through NGL. The corresponding docking results are also available for download. ( Figure 7 To submit the task interface, Figure 8 Result query page)

[0055] The present invention is applied to the molecular docking prediction of E3 ligase target proteins, which helps to understand the protein structure of E3 enzymes and analyze their interaction mechanism with substrate proteins. It can provide early layout for the design of protein degradation-related drugs and even provide important guidance for the development of broad-spectrum protein degraders.

[0056] After the E3 enzyme and the small molecule are docked, a scoring function is used to evaluate the binding free energy of the E3 enzyme and the small molecule. The scoring function is:

[0057] ΔG binding =ΔG intermolecular +ΔG torsional +ΔG solvent

[0058] Where: ΔG intermolecular is the intermolecular interaction energy, ΔG torsional Contribution to the conformational freedom of the ligand molecule, ΔG solvent It refers to the energy changes of receptor and ligand desolvation;

[0059] ΔG intermolecular =E vdW +E hydrophobic +E electrostatic

[0060] E vdW This is estimated using the Lennard-Jones potential energy function:

[0061]

[0062] Where r: atomic distance, ∈: potential energy depth, σ: equilibrium distance between two atoms;

[0063] E hydrophobic It is related to the solvent accessibility and the number of hydrophobic atoms on the molecular surface;

[0064] E electrostatic Estimate using the Coulomb potential:

[0065]

[0066] where q 1 q 2 : the charge of the interacting atoms, ∈ r : dielectric constant;

[0067] ΔG torsional =-RTln(N)

[0068] Where R is the gas constant, T is the temperature, and N is the number of rotatable bonds in the ligand.

[0069] In addition, a scoring function is added to the scoring function. The scoring function uses a random forest algorithm to associate the interaction characteristics of the receptor and the ligand with the experimentally measured binding free energy. The calculation process is as follows:

[0070] S1: Feature extraction:

[0071] Extract a set of features describing the ligand-receptor interaction from the molecular structure, including: the type of atom pairs, the distance between atom pairs, and the physical quantity of the interaction energy;

[0072] S2: Random Forest Modeling:

[0073]

[0074] in: Predicted binding free energy, N trees : The number of decision trees in the random forest, T i (x): prediction of feature vector x by the ith tree, x: feature vector describing ligand-receptor interaction;

[0075] The features used by the scoring function are based on the distance and type between atom pairs, and are formulated as:

[0076] x=[f 11 ,f 12 ,…,f ij ,…,f nn ]

[0077] where f ij : the interaction frequency between receptor atom type i and ligand atom type j within a certain distance range, n: the total number of atom types of receptor and ligand;

[0078] The loss function for supervised learning training of experimental binding free energy data by random forest is as follows:

[0079]

[0080] Where N: the number of samples in the dataset, Model-predicted binding free energy, y i : Experimentally measured binding free energy;

[0081] S3: Binding free energy prediction: input molecular features, output binding free energy ΔG binding .

[0082] The C chain of the crystal structure 5v3o of CRBN was selected as the target protein for molecular docking. The sdf file of the small molecule was uploaded. The docking results are as follows: Fig. 9 .

[0083] Select the Alphafold predicted structure of Uniprot accession number: A2RRH5 as the target protein for molecular docking, upload the mol2 file of the small molecule, and the docking results are as follows: Fig.10 .

[0084] The above description is only a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements without departing from the principle of the present invention, and these improvements should also be regarded as within the protection scope of the present invention.

Claims

1. A molecular docking method for an E3 ligase target protein, characterized in that: include In response to the basic information of the E3 enzyme entered by the user, determining the E3 enzyme that the user needs to find; According to the determined E3 enzyme, select the PDB structure to be docked; In response to the coordinates of the docking box entered by the user, the small molecule to be docked is uploaded. After the docking is completed, the docking result is sent according to the email address left by the user.

2. The molecular docking method of the E3 ligase target protein according to claim 1, characterized in that: The input E3 enzyme basic information includes: E3 enzyme function, category and protein sequence of E3 enzyme.

3. The molecular docking method of the E3 ligase target protein according to claim 1, characterized in that: The format of the small molecule to be docked is mol2 format or sdf format.

4. The molecular docking method of the E3 ligase target protein according to claim 1, characterized in that: The E3 enzyme is docked with a small molecule using CoDock-Ligand as a docking engine.

5. The molecular docking method of the E3 ligase target protein according to claim 1, characterized in that: After the E3 enzyme and the small molecule are docked, a scoring function is used to evaluate the binding free energy of the E3 enzyme and the small molecule. The scoring function is: ΔG binding =ΔG intermolecular +ΔG torsional +ΔG solvent Where: ΔG intermolecular is the intermolecular interaction energy, ΔG torsional Contribution to the conformational freedom of the ligand molecule, ΔG solvent It refers to the energy changes of receptor and ligand desolvation; ΔG intermolecular =And vdW +E hydrophobic +E electrostatic E vdW This is estimated using the Lennard-Jones potential energy function: Where r: atomic distance, ∈: potential energy depth, σ: equilibrium distance between two atoms; E hydrophobic It is related to the solvent accessibility and the number of hydrophobic atoms on the molecular surface; E electrostatic Estimate using the Coulomb potential: Where q1q2: the charges of the interacting atoms, ∈ r : dielectric constant; ΔG torsional =-RTln(N) Where R is the gas constant, T is the temperature, and N is the number of rotatable bonds in the ligand.

6. The molecular docking method of the E3 ligase target protein according to claim 5, characterized in that: A scoring function is added to the scoring function. The scoring function uses a random forest algorithm to associate the interaction characteristics of the receptor and the ligand with the experimentally measured binding free energy. The calculation process is as follows: S1: Feature extraction: Extract a set of features describing the ligand-receptor interaction from the molecular structure, including: the type of atom pairs, the distance between atom pairs, and the physical quantity of the interaction energy; S2: Random Forest Modeling: in: Predicted binding free energy, N trees : The number of decision trees in the random forest, T i (x): prediction of feature vector x by the ith tree, x: feature vector describing ligand-receptor interaction; The features used by the scoring function are based on the distance and type between atom pairs, and are formulated as: x=[f 11 ,f 12 ,…,f ij ,…,f nn ] where f ij : the interaction frequency between receptor atom type i and ligand atom type j within a certain distance range, n: the total number of atom types of receptor and ligand; The loss function for supervised learning training of experimental binding free energy data by random forest is as follows: Where N: the number of samples in the dataset, Model-predicted binding free energy, y i : Experimentally measured binding free energy; S3: Binding free energy prediction: input molecular features, output binding free energy ΔG binding .

Images