A protac molecular drug design method based on a del platform

Abstract

This invention discloses a PROTAC molecular drug design method based on the DEL platform, including generating bifunctional molecules, generating a PROTAC-like DEL bifunctional molecule library, and evaluating the bifunctional molecules in the PROTAC-like DEL bifunctional molecule library using a protein degradation-targeting chimera system mediated by bifunctional molecules. Based on the DEL platform, this invention rapidly generates and evaluates bifunctional molecules through molecular generation models and computer-aided drug design, accelerating the construction of PROTAC-like DEL bifunctional molecule libraries. Furthermore, by modeling and simulating the protein-bifunctional molecule ternary complex system mediated by bifunctional molecules, the binding modes of the bifunctional molecules in the PROTAC-like DEL bifunctional molecule library are predicted, further optimizing the bifunctional molecules and making the designed bifunctional molecules more rational, significantly reducing the difficulty of molecule synthesis.

Description

Technical Field

[0001] This invention relates to the field of computer-aided drug molecule design, and in particular to a PROTAC molecular drug design method based on the DEL platform. Background Technology

[0002] To address some of the problems associated with traditional "undruggable" targets and overcome drug resistance, protein degradation-targeting chimeras (PROTACs) technology has attracted significant attention from the pharmaceutical industry. On October 6, 2004, Israeli scientists Aaron Ciechanover and Avram Hershko, along with American scientist Irwin Rose, were awarded the Nobel Prize in Chemistry for their discovery of ubiquitin (Ub)-regulated protein degradation: when ubiquitin is linked to a protein, it causes the protein to be transported to the proteasome for degradation. This discovery laid the theoretical foundation for PROTAC protein degradation technology. A PROTAC molecule is a bifunctional hybrid molecule composed of an E3 ubiquitin ligase ligand, a target protein ligand, and a linker connecting the two. This ubiquitinates the target protein, which is then recognized by the proteasome, ultimately leading to its degradation. However, the design of PROTAC molecules remains a technical challenge.

[0003] DNA-encoded libraries (DELs) are a cutting-edge method for discovering lead compounds. By linking a specific DNA sequence to each small molecule, a structural unit of the compound corresponds one-to-one with the DNA sequence. A DEL molecule mainly consists of three parts: the small molecule, the DNA sequence, and an intermediate linker. Its structure is similar to that of PROTAC molecules, which also use linkers to connect E3 and target protein ligands. Therefore, DEL technology has inherent experimental advantages in both finding novel structures of PROTAC ligands and optimizing the linker portion of PROTACs. By modifying current DEL screening techniques and constructing specialized bifunctional DEL molecules based on the structural characteristics of PROTAC molecules, the application of DEL technology can be expanded to "undruggable" targets, while also providing new ideas for the discovery of PROTAC drugs.

[0004] The challenge of developing PROTAC molecules using the DEL technology platform lies in the fact that DEL is an affinity screening technique. While it can generate rich structure-activity relationship (SAR) information during the screening process, allowing for some inference of small molecule-protein interactions, it cannot directly generate information on the binding modes of small molecules and proteins. PROTAC systems, on the other hand, are ternary complexes formed between target proteins and E3 ligases mediated by bifunctional molecules. These complexes involve both protein-small molecule interactions and complex protein-protein interactions. Therefore, understanding and predicting the conformation of the complex and protein interactions is of paramount importance for the design and optimization of PROTAC molecules. How to combine the prediction of complex conformation and protein interactions with the DEL technology platform to develop PROTAC molecular drug design methods has become a technical problem that needs to be solved in this field. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a PROTAC molecular drug design method based on the DEL platform, which combines computer-aided drug design to realize the design of PROTAC molecules on the DEL platform.

[0006] To address the aforementioned technical problems, the present invention provides a PROTAC molecular drug design method based on the DEL platform, comprising:

[0007] Step S1: Generate bifunctional molecules: Based on warhead molecules and E3 ligase ligands, generate multiple sets of bifunctional molecules of warhead molecules-linkers-E3 ligase ligands using a molecular generation model.

[0008] Step S2: Generate PROTAC-like DEL bifunctional molecular library: Screen and evaluate the bifunctional molecules generated in step S1, add the bifunctional molecules that meet the specifications to the DEL library to generate PROTAC-like DEL bifunctional molecular library, and evaluate its KL divergence.

[0009] Step S3: Evaluate the bifunctional molecules in the PROTAC-like DEL bifunctional molecule library using a bifunctional molecule-mediated protein degradation targeting chimerism system.

[0010] In this invention, the molecular generation model can be, for example, the PROTAC-DEL Model, which can generate a linker between the warhead molecule and the E3 ligase ligand based on the characteristics of the linking sites of the warhead molecule and the E3 ligase ligand using artificial intelligence (AI) methods.

[0011] Specifically, in step S1, the steps of generating bifunctional molecules using the molecular generation model include:

[0012] S11. Define the linker site between the warhead molecule and the E3 ligase ligand, and modify the corresponding SMILES string.

[0013] S12. Using the warhead molecule SMILES string and the E3 ligase ligand SMILES string as input, multiple different bifunctional molecules are generated.

[0014] The warhead molecules are obtained based on DEL screening.

[0015] Specifically, step S2, the steps for screening and evaluating bifunctional molecules, include:

[0016] Step S21: The bifunctional molecules in step S1 are initially screened based on the linker length and the number of aromatic rings to obtain bifunctional molecules that meet the criteria. Generally, the number of heavy atoms (non-hydrogen atoms) in the linker is around 12, and the number of aromatic rings does not exceed 2. Therefore, the initial screening is carried out based on the criteria in that order.

[0017] Step S22: Further screen bifunctional molecules that meet the screening criteria of step S21 using their physicochemical properties. The physicochemical properties of the bifunctional molecules include molecular weight, number of hydrogen bond acceptors, number of hydrogen bond donors, and lipid-water partition coefficient. It should be noted that the order of S21 and S22 can be changed. Among them, the number of hydrogen bond acceptors does not exceed 10, the number of hydrogen bond donors does not exceed 5, and the logarithm of the lipid-water partition coefficient is between -2 and 5 (these parameters are based on three of the Ribinsky five rules). The molecular weight can be determined according to the molecules reported for different targets. For example, for molecules reported for the CRBN target, the molecular weight can be limited to less than 1000 Daltons.

[0018] Step S23: Calculate the feasibility score for the synthesis of bifunctional molecules;

[0019] Step S24: Evaluate the membrane permeability, solubility, and metabolic sites of the bifunctional molecule;

[0020] Step S25: Based on the scores from step S23 and the evaluation results from step S24, the medicinal chemist selects bifunctional molecules and adds those that meet the selection criteria to the DEL library to generate a PROTAC-like DEL bifunctional molecule library.

[0021] Step S26: Based on the known physicochemical properties of PROTAC molecules, perform KL divergence evaluation on the PROTAC-like DEL bifunctional molecular library obtained in step S25.

[0022] In step S22, during the screening process based on the physicochemical properties of bifunctional molecules, the physicochemical properties of PROTAC molecules targeting relevant targets already included in the PROTAC-DB 2.0 database can also be referenced. Currently, the PROTAC-DB 2.0 database includes numerous PROTAC molecules and summarizes, classifies, and categorizes the physicochemical properties of these PROTAC molecules (such as molecular weight, number of hydrogen bond acceptors, number of hydrogen bond donors, and lipid-water partition coefficient), providing valuable reference. In step S23, the feasibility score of the bifunctional molecule is calculated using the SA score in RDKit.

[0023] In step S24, the membrane permeability, solubility, and metabolic sites of the bifunctional molecule were evaluated using the ADMET property model.

[0024] In step S25, medicinal chemists mainly select based on the ranking in steps S23 and S24, such as selecting the top 100 bifunctional molecules in the overall ranking. In addition, medicinal chemists can also combine their own experience to screen the selected bifunctional molecules that rank high in the overall ranking.

[0025] In step S26, the main purpose of performing the KL divergence assessment is to evaluate whether the physicochemical property distribution of the PROTAC-like DEL bifunctional molecular library constructed in this application is consistent with that of the reported PROTAC molecules.

[0026] Specifically, step S3, the evaluation of the PROTAC-like DEL bifunctional molecular library, includes:

[0027] Step S31: Simulate protein-protein docking between the target protein and the E3 degrading enzyme to determine the conformation of the starting protein;

[0028] Step S32: Using LeDock software, bind the molecular building blocks of the warhead molecules corresponding to the bifunctional molecules in the PROTAC-like DEL bifunctional molecular library to the target protein and select an appropriate binding mode; bind the molecular building blocks corresponding to the E3 ligase ligand to the E3 ligase and select an appropriate binding mode.

[0029] Step S33: Using LeDock software, with molecular building blocks corresponding to linkers in the PROTAC-like DEL bifunctional molecular library, in a state of no steric hindrance or minimum steric hindrance, the selected warhead molecule-target protein structure and E3 ligase ligand-E3 ligase structure with reasonable binding modes are linked to form a ternary complex simulated structure of target protein-warhead molecule-linker-E3 ligase ligand-E3 ligase. Then, the warhead molecule and E3 ligase ligand are constrained to be in the corresponding coordinate positions (applying a certain constraint force to keep them in a stable position). The linker is optimized for energy to be in the lowest energy conformation (generally by energy gradient descent). The ternary complex simulated structure of target protein-warhead molecule-linker-E3 ligase ligand-E3 ligase is obtained by screening, that is, the protein-bifunctional molecular conformation.

[0030] Step S34: For the protein-bifunctional molecule conformation in step S33, use molecular dynamics simulation to screen and obtain stable protein-bifunctional molecule conformations;

[0031] Step S35: Perform molecular dynamics trajectory analysis on the stable protein-bifunctional molecular conformation obtained in step S34 to obtain a stable and reasonable protein-bifunctional molecular conformation.

[0032] Step S36: For the protein-bifunctional molecule conformation obtained in step S35, calculate the binding free energy of the bifunctional molecule and the protein using the molecular mechanics / Poisson-Boltzmann surface area method, and evaluate the binding strength between the bifunctional molecule and the protein.

[0033] In step S31, Rosetta software is used for protein-protein docking. The indicators considered in determining the initial protein conformation include docking score and key residue interaction analysis. Generally, a lower docking score indicates lower energy and greater stability. Key residue interaction analysis mainly examines whether hydrogen bonds have formed at key sites. Additionally, the software requires analysis of the protein-protein interface, including factors such as embedding surface area, shape complementarity, number of polar atoms, number of hydrogen bonds, and hydrogen bond energy. For example, the embedding surface area can be selected according to the software's ranking. Shape complementarity is generally recommended to exceed 0.65. The analysis of the number of polar atoms, hydrogen bonds, and hydrogen bond energy is as follows: Generally, too few polar atoms and too few hydrogen bonds are detrimental to binding, while stronger hydrogen bond interactions are beneficial.

[0034] Specifically, in step S34, the molecular dynamics simulation is performed in AMBER20 software, using a model of a protein-bifunctional molecule in aqueous solution. The simulation process includes:

[0035] Step S341: Use the steepest descent method to minimize energy. During the minimization process, apply constraint forces to all heavy atoms in the protein-bifunctional molecule to optimize their positions.

[0036] Step S342: After energy minimization, the system is heated from 297K to 310K within 50ps. During this process, a confinement force is continuously applied to the heavy atoms in the protein-bifunctional molecule.

[0037] After steps S343 and 50ps, the entire system, under the constraint of heavy atoms, continues to undergo the equilibrium process of the isothermal and isovolume ensemble for 100ps and then the isothermal and isobaric ensemble for another 200ps.

[0038] After steps S344 and S343 are completed, the entire system undergoes molecular dynamics simulation for no less than 100 ns under an unconstrained isothermal and isobaric ensemble.

[0039] When executed in AMBER20 software, the force field of the bifunctional molecule was generated using the gaff2 small molecule force field file, and the protein force field was generated using the FF19SB force field file. The TIP3P water model was also used, and metal cations and corresponding anions were added to neutralize the charge with counterions, thus establishing a model of the protein-bifunctional molecule (ternary complex simulated structure) in aqueous solution.

[0040] Specifically, in step S35, the molecular dynamics trajectory analysis includes:

[0041] In step S351 and step S344, during the molecular dynamics simulation, a simulation trajectory of 50ns-100ns is selected, and the molecular conformation of the protein-bifunctional molecule is collected every 100ps.

[0042] Step S352: Using the k-means algorithm, the molecular conformations collected in step S351 are divided into multiple different clusters, and the cluster center of each cluster is found.

[0043] This invention, based on the DEL platform, rapidly generates and evaluates bifunctional molecules through molecular generation models and computer-aided drug design, accelerating the construction of PROTAC-like DEL bifunctional molecule libraries. Furthermore, by modeling and simulating the protein-bifunctional molecule ternary complex system mediated by bifunctional molecules, the binding modes of bifunctional molecules in the PROTAC-like DEL bifunctional molecule library are predicted, further optimizing the bifunctional molecules and making the designed bifunctional molecules more reasonable, greatly reducing the difficulty of molecule synthesis. Attached Figure Description

[0044] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 These are three bifunctional molecules generated by the molecular generation model in the embodiments.

[0046] Figure 2 This is one of the conformation diagrams of the ternary complex formed by the protein-bifunctional molecule as ultimately predicted in the examples. Detailed Implementation

[0047] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0048] Example

[0049] Based on the DEL platform, bifunctional molecules were designed using the target proteins bromodomain protein 4 (BRD4) and E3 degrading enzyme (CRBN).

[0050] Step S1: Generate bifunctional molecules

[0051] Step S11: Use the DEL platform to screen out warhead molecules that have a signal with BRD4, define the linkage sites in conjunction with the reported E3 degrading enzymes targeting BRD4, and mark and modify the corresponding SMILES strings.

[0052] Step S12: Using the warhead molecule SMILES string and the E3 ligase ligand SMILES string as input, 16566 bifunctional molecular structures with different linkers were generated through the PROTAC-DEL Model.

[0053] Step S2: Generate a PROTAC-like DEL bifunctional molecular library

[0054] Step S21: Referring to the definitions of the number of hydrogen bond acceptors, the number of hydrogen bond donors, and the lipid-water partition coefficient in the Ribinsky five principles, and combining the molecular weight, number of hydrogen bond acceptors, number of hydrogen bond donors, and lipid-water partition coefficient of PORTAC molecules already included in the PROTAC-DB 2.0 database, 16,566 bifunctional molecules were screened, and 15,701 bifunctional molecules were obtained after filtering.

[0055] Step S22: Select a linker skeleton with approximately 12 heavy atoms (non-hydrogen atoms) (9-15), remove linkers that are too short or too long, and obtain 15677 bifunctional molecules;

[0056] Steps S23-S25: ​​The synthesis feasibility score is calculated using the SA score in RDKit, and the membrane permeability, solubility, and metabolic sites of bifunctional molecules are evaluated using the ADMET property model. Chemists further screen bifunctional molecules based on a comprehensive consideration of the synthesis feasibility score and the evaluation results of membrane permeability, solubility, and metabolic sites, selecting those with higher synthesis feasibility scores and rankings. Combining this with group stability factors, structures that might be PAINS are removed, resulting in 100 structurally reasonable bifunctional molecules. These are added to the DEL library to generate a PROTAC-like DEL bifunctional molecule library (the PROTAC-like DEL bifunctional molecule library construction process is the same as the DEL library construction process, including library construction, library encoding, screening, decoding, verification, and optimization steps, which will not be elaborated here). Figure 1 Three of its reasonable structures are shown;

[0057] Step S26: Refer to the physicochemical properties of PROTAC molecules already included in the PROTAC-DB 2.0 database, evaluate the KL divergence of the generated PROTAC-like DEL bifunctional molecular library with it. The evaluation results are consistent, and the established PROTAC-like DEL bifunctional molecular library meets the requirements.

[0058] Step S27: Based on the designed PROTAC-like DEL library, library production and subsequent DEL screening are performed to obtain the corresponding screening data. DEL data experts analyze and screen the data based on the comprehensive ranking of enrichment and copy data, combined with molecular structure and corresponding family information, to obtain 5 candidate bifunctional molecules. Binding mode simulation is then performed (it should be noted that in other embodiments, the number of candidate bifunctional molecules is not limited to 5; other numbers of bifunctional molecules can be used for subsequent evaluation, such as 3, 4, 6, 8, 10, etc.).

[0059] Step S3: Evaluate the bifunctional molecules in the PROTAC-like DEL bifunctional molecule library using a bifunctional molecule-mediated protein degradation targeting chimera system.

[0060] Step S31: In Rosetta software, protein-protein docking is performed on the target protein BRD4 and the E3 degrading enzyme CRBN. Since there are many crystal structures for BRD4-CRBN, this embodiment uses the crystal structure 6BN from the RCSB PDB database for modeling. First, a certain distance is created between BRD4 and CRBN, so that the binding pockets in the two proteins face each other. Using Rosetta software for local sampling, 50 protein-protein binding conformations are generated for each candidate bifunctional molecule. The top 10 binding conformations are ranked by docking score for further analysis. Then, based on the known linker length (9-15), calculated embedding surface area, shape complementarity (greater than 0.65), and the assessed number of polar atoms, number of hydrogen bonds, and hydrogen bond energy, one protein conformation is selected as the starting conformation for each candidate bifunctional molecule.

[0061] Step S32: In LeDock software, the molecular building blocks of the warhead molecules of the five candidate bifunctional molecules finally obtained in step S2 are respectively docked to the BRD4 receptor protein, and the molecular building blocks of the E3 ligase ligand are bound to the CRBN receptor protein. The docking results are further selected by combining docking scoring and reported small molecule protein binding modes, and reliable docking conformations are selected.

[0062] Step S33: In LeDock software, connect the two ends of the molecular building blocks of the linkers of the 5 bifunctional molecules to the molecular building blocks of the warhead molecule and the molecular building blocks of the E3 ligase ligand molecule, respectively. For each bifunctional molecule, sample 100 different three-dimensional conformations and select the conformation with no steric hindrance or the lowest steric hindrance to obtain 5 protein-bifunctional molecule conformations corresponding to the 5 bifunctional molecules.

[0063] Step S34: In AMBER20 software, the force field of bifunctional molecules is represented by the gaff2 small molecule force field file, and the force field of proteins is represented by the FF19SB force field file. The TIP3P water model is also used, and the metal cation Na is added. + and the corresponding anion Cl - The counterions neutralize the charge, keeping the initial environment neutral, thus obtaining a model of the molecular dynamics initiation protein-bifunctional molecule (ternary complex) in aqueous solution;

[0064] Step S341: Apply constraint forces to all heavy atoms of the protein-bifunctional molecule, perform energy minimization processing on the steepest gradient descent of each system, and optimize the position of water molecules.

[0065] Step S342: For the system after minimizing the capability in step S341, the temperature is increased from 297K to 310K within 50ps. During this process, a confinement force is continuously applied to the heavy atoms.

[0066] Step S343: The system after heating in step S342 is subjected to a 100ps isothermal constant volume (NVT) ensemble (temperature 310K) and a 200ps isothermal constant pressure (NPT) ensemble (temperature 310K, pressure 1.0bar) to pre-equilibrate the entire system. During this process, a constraint force is continuously applied to the heavy atoms.

[0067] After steps S344 and S343 are completed, the entire system is subjected to a 100ns molecular dynamics simulation under an unconstrained isothermal and isobaric ensemble (temperature 310K, pressure 1.0bar) to obtain the 100ns small molecule dynamics trajectory.

[0068] In steps S35 and S344, the small molecule dynamics trajectory obtained in the first 50 ns shows a certain optimization process for the small molecule conformation, but it is still unstable. Therefore, molecular dynamics trajectory analysis is performed on the latter 50 ns.

[0069] Step S351: Extract the molecular dynamics trajectory after 50 ns, and output a protein-bifunctional molecule conformation every 100 ps, ​​for a total of 500 conformations.

[0070] Step S352: The k-means algorithm is used to cluster the 500 conformations into ten different clusters, and the cluster center conformations of these ten clusters are identified, resulting in ten different conformations. Combining structure-activity relationship analysis and molecular dynamics simulation correlation analysis, a reasonable conformation is selected as the predicted binding mode for the corresponding bifunctional molecule, resulting in five predicted binding modes for bifunctional molecules, i.e., protein-bifunctional molecule binding modes. The combination of structure-activity relationship analysis and molecular dynamics simulation correlation analysis mainly refers to software scoring and basic common sense considerations such as the orientation of hydrophobic and hydrophilic groups of the molecule, whether hydrogen bonds and other interactions are formed with the protein, and hydrophobic interactions between small molecules and proteins, which will not be elaborated here.

[0071] Step S36: Based on the predicted binding modes of the five bifunctional molecules obtained in step S352, and combined with existing reports on PROTAC molecule binding modes (e.g., whether the binding modes of similar groups are consistent), the binding free energy between the bifunctional molecules and proteins is calculated using molecular mechanics / Poisson-Boltzmann surface area method. The binding strength between the bifunctional molecules and proteins is evaluated, and the five bifunctional molecules are ranked. Simultaneously, considering the key interactions between the bifunctional molecules and proteins (e.g., whether key protein sites interact with the bifunctional molecules), the two optimal bifunctional molecules are selected, exhibiting high enrichment signals, indicating that these two bifunctional molecules are rationally designed. Figure 2 The final predicted conformation of one of the molecules is shown.

[0072] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A PROTAC molecular drug design method based on the DEL platform, characterized in that, Includes the following steps: Step S1: Generate bifunctional molecules: Based on warhead molecules and E3 ligase ligands, generate multiple sets of bifunctional molecules of warhead molecules-linkers-E3 ligase ligands using a molecular generation model. Step S2: Generate PROTAC-like DEL bifunctional molecular library: Screen and evaluate the bifunctional molecules generated in step S1, add the bifunctional molecules that meet the specifications to the DEL library to generate PROTAC-like DEL bifunctional molecular library, and evaluate its KL divergence. Step S3: Evaluate the bifunctional molecules in the PROTAC-like DEL bifunctional molecule library using a bifunctional molecule-mediated protein degradation targeting chimera system, specifically including: Step S31: Simulate protein-protein docking between the target protein and the E3 degrading enzyme to determine the conformation of the starting protein; Step S32: Using LeDock software, bind the molecular building blocks of the warhead molecules of the bifunctional molecules in the PROTAC-like DEL bifunctional molecular library to the target protein and select an appropriate binding mode; bind the molecular building blocks of the E3 ligase ligand to the E3 ligase and select an appropriate binding mode. Step S33: Using LeDock software, molecular building blocks of linkers from the PROTAC-like DEL bifunctional molecular library are used to link the selected warhead molecule-target protein structure and E3 ligase ligand-E3 ligase structure with reasonable binding modes under conditions of no steric hindrance or minimal steric hindrance to form a target protein-warhead molecule-linker-E3 ligase ligand-E3 ligase structure. Then, the warhead molecule and E3 ligase ligand are constrained, and the linker is energy optimized to screen and obtain the target protein-warhead molecule-linker-E3 ligase ligand-E3 ligase structure, i.e., the protein-bifunctional molecular conformation. Step S34: For the protein-bifunctional molecule conformation in step S33, use molecular dynamics simulation to screen and obtain stable protein-bifunctional molecule conformations; Step S35: Perform molecular dynamics trajectory analysis on the stable protein-bifunctional molecular conformation obtained in step S34 to obtain a stable and reasonable protein-bifunctional molecular conformation. Step S36: For the protein-bifunctional molecule conformation obtained in step S35, calculate the binding free energy of the bifunctional molecule and the protein using the molecular mechanics / Poisson-Boltzmann surface area method, and evaluate the binding strength between the bifunctional molecule and the protein. The warhead molecules are obtained based on DEL screening.

2. The design method as described in claim 1, characterized in that, In step S1, the steps of generating bifunctional molecules using the molecular generation model include: S11. Define the linker site between the warhead molecule and the E3 ligase ligand, and modify the corresponding SMILES string. S12. Using the warhead molecule SMILES string and the E3 ligase ligand SMILES string as input, multiple different bifunctional molecules are generated.

3. The design method as described in claim 1, characterized in that, Step S2, the screening and evaluation steps for bifunctional molecules, includes: Step S21: Use the linker length and the number of aromatic rings to perform preliminary screening of the bifunctional molecules in step S1 to obtain bifunctional molecules that meet the criteria. Step S22: Further screen bifunctional molecules that meet the screening criteria in step S21 using their physicochemical properties; the physicochemical properties of the bifunctional molecules include molecular weight, number of hydrogen bond acceptors, number of hydrogen bond donors, and lipid-water partition coefficient. Step S23: Calculate the feasibility score for the synthesis of bifunctional molecules; Step S24: Evaluate the membrane permeability, solubility, and metabolic sites of the bifunctional molecule; Step S25: Based on the scores from step S23 and the evaluation results from step S24, the medicinal chemist selects bifunctional molecules and adds those that meet the selection criteria to the DEL library to generate a PROTAC-like DEL bifunctional molecule library. Step S26: Based on the known physicochemical properties of PROTAC molecules, perform KL divergence evaluation on the PROTAC-like DEL bifunctional molecular library obtained in step S25.

4. The design method as described in claim 3, characterized in that, In step S22, the physicochemical properties of the bifunctional molecule are obtained using the PROTAC-DB 2.0 database; In step S23, the feasibility score for the synthesis of bifunctional molecules was calculated using evaluation software. The membrane permeability, solubility, and metabolic sites of the bifunctional molecule in step S24 were evaluated using the ADMET property model.

5. The design method as described in claim 1, characterized in that, In step S31, the software used for protein-protein docking is Rosetta software.

6. The design method as described in claim 1, characterized in that, In step S34, the molecular dynamics simulation is performed in AMBER20 software, using a model of a protein-bifunctional molecule in aqueous solution. The simulation process includes: Step S341: Minimize energy using the steepest descent method; Step S342: After minimization, the system is heated from 297K to 310K within 50ps, while simultaneously applying constraint forces to the heavy atoms in the protein-bifunctional molecule. After steps S343 and 50ps, the entire system, under the constraint of heavy atoms, continues to undergo the equilibrium process of the isothermal and isovolume ensemble for 100ps and then the isothermal and isobaric ensemble for another 200ps. After steps S344 and S343 are completed, the entire system undergoes molecular dynamics simulation for no less than 100 ns under an unconstrained isothermal and isobaric ensemble.

7. The design method as described in claim 6, characterized in that, In step S35, the molecular dynamics trajectory analysis includes: In step S351 and step S344, during the molecular dynamics simulation, a simulation trajectory of 50ns-100ns is selected, and the molecular conformation of the protein-bifunctional molecule is collected every 100ps. Step S352: Using the k-means algorithm, the molecular conformations collected in step S351 are divided into multiple different clusters, and the cluster center of each cluster is found.

Images