Generating targeted aav9 antibody epitope affinity peptide ligands using protein language models and screening methods and applications thereof

Abstract

The application relates to a method for generating a target AAV9 antibody epitope affinity peptide ligand by using a protein language model and a screening method and application thereof, and relates to an affinity peptide candidate sequence construction and screening method, an affinity peptide ligand, an affinity medium and application thereof for targeting adeno-associated virus type 9 (AAV9). The AAV9 capsid region recognized by Fab2-4 is taken as a target, structure analysis, natural amino acid positioning, key residue extraction and affinity peptide template construction are carried out, a candidate peptide library is generated by using a protein language model, and a candidate affinity peptide ligand is screened by combining molecular docking, binding free energy evaluation, refined docking and molecular dynamics simulation. The affinity peptide ligand is any one of SEQ ID NO. 1, SEQ ID NO. 2, SEQ ID NO. 3 or SEQ ID NO. 4, can be immobilized on a solid phase matrix to form an affinity medium, is used for AAV9 recognition, capture and purification, and provides a technical scheme for AAV9 purification process development.

Description

Technical Field

[0001] This invention relates to the fields of viral vector isolation and purification, computational-aided molecular design, and artificial intelligence-aided peptide sequence design. Specifically, it relates to a method for generating and screening affinity peptide candidate ligands for adeno-associated virus type 9 (AAV9) using a protein language model, the affinity peptide candidate sequences obtained by this method, and their application in the isolation and purification of AAV9. Background Technology

[0002] In recent years, adeno-associated virus (AAV) vectors have become one of the most important delivery platforms for in vivo gene therapy due to their low pathogenicity, low immunogenicity, ability to maintain exogenous gene expression in vivo for a long time, and good tissue specificity. Among the many AAV vectors, AAV9 has shown outstanding application value in neuromuscular diseases, inherited metabolic diseases, and systemic drug delivery scenarios due to its good in vivo delivery ability to tissues such as the central nervous system, myocardium, skeletal muscle, and liver. It has become one of the important capsid serotypes for many gene therapy products under development and on the market (Nature Reviews Drug Discovery, 2021, 20(3): 173-174.; Signal Transduction and Targeted Therapy, 2024, 9: 76.). With the development and industrialization of AAV9-related gene therapy products, the downstream purification process of viral vectors has put forward higher requirements for recovery rate, activity retention, impurity removal efficiency, and process cost control. AAV feedstock typically contains a large amount of host cell protein, residual nucleic acid, empty capsids, and process-related impurities. If the purification efficiency is insufficient, it will not only affect the quality properties of the virus particles, but also significantly increase production costs and increase the pressure of subsequent quality control (Nature Reviews Drug Discovery, 2019, 18(5): 358-378.).

[0003] Currently, affinity chromatography has become an important technique for the capture and purification of AAV vectors. Existing commercial AAV affinity media mostly employ antibody fragments, single-domain antibodies, or other protein ligands, which can achieve highly selective capture of specific serotypes of AAV. For example, Cytiva has disclosed a single-domain antibody (WO2025083267A1) that specifically binds to AAV9 and can be used for its separation and purification, but its elution is usually performed at low pH. However, these affinity ligands still have several limitations. First, many protein affinity mediators typically rely on acidic conditions for elution, which can negatively impact the integrity and bioactivity of the AAV capsid. Second, these ligands often lack resistance to alkali washing, oxidation, and multiple cycles of reuse, affecting the stability and economics of industrial scale-up processes. Furthermore, the high cost of protein ligand preparation and their strong specificity to different serotypes limit their rapid migration applications in novel or engineered AAV capsids (Molecular Therapy Methods & Clinical Development, 2020, 19: 362-373). Therefore, developing novel AAV9 affinity ligands that can achieve binding and dissociation under milder conditions while possessing good chemical stability and scalable synthesis capabilities is of significant practical importance.

[0004] Compared to protein ligands, peptide affinity ligands offer advantages such as well-defined sequences, chemical synthesis capabilities, lower cost, ease of modification, better physicochemical stability, and suitability for alkali-resistant washing and gentle elution, making them a key development direction for next-generation AAV affinity purification media. For example, Avitide has disclosed a cyclic peptide affinity ligand (US20220213447A1) that specifically binds to AAV9, but it still faces the challenge of low-pH elution. Current methods for obtaining small peptide ligands largely rely on phage display, random peptide library screening, or empirical modification. While these methods can yield sequences that can bind to the target, they often suffer from long screening cycles, insufficient structural interpretation, poor sequence migration, and difficulties in cross-serotype adaptation. Especially for capsid systems like AAV9 with complex surface topology and well-defined immune-related epitopes, relying solely on random screening without high-resolution structural epitope information often fails to efficiently obtain candidate ligands that possess affinity, specificity, and engineering potential (Journal of Chromatography A, 2024, 1734: 465320.; Molecular TherapyMethods & Clinical Development, 2020, 19: 362-373.). Therefore, rational design based on well-defined structural epitopes has become an important direction for the development of AAV affinity peptide ligands.

[0005] Recent structural biology studies have provided new target sources for AAV9 ligand design. Mietzsch et al. systematically resolved the structures of multiple human anti-AAV9 neutralizing antibodies derived from patients treated with Zolgensma. The structure of the complex formed by Fab2-4 and the AAV9 capsid has been resolved and included in PDB 9B7N (Nature Communications, 2025, 16: 3731). Related studies show that Fab2-4 mainly recognizes neutralizing epitopes near the 2-fold depression of the AAV9 capsid. Its binding region covers the 2-fold depression and the adjacent 3-fold protrusion side, representing a representative antibody recognition hotspot with sufficient exposure on the AAV9 capsid surface and well-defined spatial features (Nature Communications, 2025, 16: 3731). This type of epitope not only reflects the key interface of AAV9 in vivo immune recognition but also provides a high-resolution structural template that can be directly utilized for affinity peptide ligand design. At the same time, the development of artificial intelligence methods such as protein language models has made it possible to generate, screen and optimize candidate peptide libraries based on key residue templates, providing new technical means for constructing a rational design process for specific epitopes of the AAV9 capsid.

[0006] This invention, based on the structural features of AAV9 identified by Fab2-4, and combining key residue extraction, affinity peptide template construction, candidate sequence library construction, multi-round virtual screening, and molecular dynamics stability evaluation, establishes a method for constructing and screening AAV9 affinity peptide candidate sequences. Compared to existing techniques that rely primarily on random screening and empirical optimization, this invention uses the well-defined AAV9 capsid PDB structure as its foundation, providing a clear structural basis for obtaining candidate sequences and improving the targeting and efficiency of candidate ligand screening. This method addresses, to some extent, the problems of high randomness, long screening cycles, lack of clear epitope support for candidate sequences, and insufficient direct connection with subsequent affinity medium development in existing technologies. It facilitates the acquisition of candidate affinity peptide ligands that can be further used for immobilization verification and purification process research, thus providing technical support for the development of AAV9 affinity purification materials. Summary of the Invention

[0007] The purpose of this invention is to address the problems of existing methods for obtaining AAV9 affinity ligands, which rely heavily on random screening, lack clear structural epitope guidance, have low screening efficiency, and are difficult to quickly obtain candidate ligands that can be used for purification process development. This invention provides a method for constructing and screening candidate sequences of AAV9 affinity peptides, and further provides AAV9 affinity peptide ligands obtained by the method, affinity media containing the affinity peptide ligands, and their application in AAV9 recognition, capture, separation, and purification.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A protein language model was used to generate an affinity peptide ligand targeting the AAV9 antibody epitope, wherein the amino acid sequence of the affinity peptide ligand is any one of SEQ ID NO.1, SEQ ID NO.2, SEQ ID NO.3 and SEQ ID NO.4.

[0010] A method for screening AAV9 antibody epitope affinity peptide ligands using protein language models includes the following steps:

[0011] (1) Obtain the structure of the AAV9-Fab2-4 complex and perform structural pretreatment;

[0012] (2) The AAV9 capsid epitope regions identified by Fab2-4 were divided, and the binding site preferences of different amino acids in each sub-region were analyzed using natural amino acid localization analysis.

[0013] (3) Extract key residues based on the localization results and construct an affinity peptide template;

[0014] (4) Based on the affinity peptide template, the variable sites in the template are conditionally generated, amplified, and deduplicated using a protein language model to construct candidate affinity ligand peptides.

[0015] (5) Perform molecular docking, binding free energy calculation, refined docking verification and molecular dynamics simulation analysis on the candidate affinity peptide library to screen out candidate affinity peptide ligands.

[0016] In step (1), the structural file numbered 9B7N in the public database PDB is used as the target source; the chain and antibody fragments involved in binding with the AAV9 capsid are retained for interface reference; missing hydrogen atoms are added and the protonation state is adjusted, followed by local energy minimization to eliminate potential structural inconsistencies, and then converted into the PDBQT software format suitable for docking and simulation.

[0017] In step (2), based on the interface between Fab2-4 and AAV9 surfaces, combined with spatial location and local physicochemical characteristics, the binding region is divided into sub-regions; by performing molecular docking between each residue and the epitope sub-region, comparing scores and spatial locations, the amino acid that performs best in a specific region is selected as the key residue, providing a basis for template construction.

[0018] In step (3), the key residues include Arg, Val, Lys, His, Gln and Ser; the overall structure of the affinity peptide template constructed from the key residues is RV-X1-X2-KHQ-X3-S, where X1, X2 and X3 are variable sites.

[0019] In step (4), the affinity peptide template composed of key residues is input into the protein language model. Under the condition of fixing the key residues R, V, K, H, Q and S, amino acid combinations are generated for the variable sites X1, X2 and X3 so that the generated sequence simultaneously meets the requirements of key residue retention, sequence context rationality and the feasibility of peptide chemical synthesis. After the generated candidate sequence is deduplicated, length consistency checked and sequence rationality screened, the initial candidate peptide sequence for virtual screening is obtained.

[0020] In step (5), ADCP is used to perform preliminary docking of candidate peptides with AAV9 epitopes. Due to its high precision in peptide-protein complex docking, reasonable binding conformations are quickly screened. Subsequently, MMGBSA is used to calculate the binding free energy and evaluate the thermodynamic stability of the complex. The preliminary screening results are then refined and verified. Rosetta and HADDOCK are used to optimize the interface conformation and energy to further confirm the reliability of the binding mode. Finally, molecular dynamics simulation is used to analyze the dynamic stability of the complex in the solution environment, and MMPBSA residue-by-residue energy decomposition is used to determine the contribution of key residues, thereby screening out the thermodynamically and kinetically optimal candidate affinity peptides.

[0021] A method for separating and purifying AAV9 using protein language models to generate affinity peptide ligands targeting AAV9 antibody epitopes, characterized in that the affinity medium comprises a solid matrix and affinity peptide ligands immobilized on the matrix, comprising the following steps:

[0022] (1) Load the sample containing AAV9 onto the affinity medium to allow AAV9 to bind to the immobilized affinity peptide ligand;

[0023] (2) The affinity medium is cleaned to remove unbound impurities;

[0024] (3) Elute AAV9 bound to the affinity medium to achieve AAV9 capture and recovery.

[0025] The adsorption buffer is 20 mM pH 7.5 phosphate buffer (PB).

[0026] The elution buffer is a pH 3.0, 0.1M Gly-HCl buffer.

[0027] The specific explanation is as follows:

[0028] This invention provides a method for constructing and screening candidate affinity peptide sequences for AAV9 (SEQ ID NO.5). The method is based on the complex structure (PDB 9B7N) formed by the human neutralizing antibody Fab2-4 and AAV9, and determines the target epitope by referencing the binding site of the antibody to the AAV9 capsid protein. By analyzing the complex structure and combining it with the localization results of natural amino acids, key residues related to binding to the target epitope are screened, and an affinity peptide template is constructed based on this. Subsequently, a protein language model is used to conditionally generate and amplify variable sites in the template to construct a candidate affinity peptide library. Further, molecular docking, binding free energy assessment, refined docking verification, and molecular dynamics simulation analysis are combined to screen the candidate peptide library, thereby obtaining candidate affinity peptide ligands with binding ability to the AAV9 target region.

[0029] The present invention also provides an AAV9 affinity peptide ligand obtained by the method, wherein the amino acid sequence of the affinity peptide ligand is any one of SEQ ID NO.1, SEQ ID NO.2, SEQ ID NO.3, and SEQ ID NO.4. The amino acid sequence of the AAV9 capsid protein is SEQ ID NO.5, which serves as a reference sequence for the target recognition of the affinity peptide ligand.

[0030] For obtaining the complex structure and performing preprocessing in step (1), this invention uses the structure file numbered 9B7N in the publicly available PDB database as the target source. After obtaining the structure, water molecules and irrelevant small molecules are removed, and the chain involved in binding with the AAV9 capsid and the antibody fragment are retained for interface reference. Missing hydrogen atoms are added and the protonation state is adjusted. Subsequently, local energy minimization is performed to eliminate potential structural inconsistencies, and the structure is converted into the PDBQT software format suitable for docking and simulation.

[0031] For the epitope region division and natural amino acid localization analysis in step (2), this invention divides the binding region into sub-regions based on the Fab2-4 and AAV9 surface binding interface, combined with spatial location and local physicochemical characteristics. The reason for using natural amino acids for localization analysis is that natural amino acids, as basic building blocks of peptides, have high chemical feasibility, are convenient for subsequent synthesis and immobilization, and can represent side chains with different physicochemical properties (positively charged, negatively charged, polar, hydrophobic, or aromatic), facilitating a systematic evaluation of the binding preferences of different residue types at epitopes. Different amino acids are derived from a standard set of twenty natural amino acids. By performing molecular docking between each residue and the epitope sub-region, comparing scores and spatial locations, the amino acid exhibiting the best performance in a specific region is selected as the key residue, providing a basis for template construction.

[0032] In step (3), the key residues include Arg, Val, Lys, His, Gln, and Ser; the overall structure of the affinity peptide template constructed based on the key residues is RV-X1-X2-KHQ-X3-S, where X1, X2, and X3 are variable sites; a candidate affinity ligand peptide library is constructed based on the template, and the final affinity peptide ligands are obtained through subsequent screening.

[0033] In step (4), a conditional sequence amplification of the template variable sites is performed using a protein language model. Specifically, an affinity peptide template composed of key residues is input into the protein language model. Under the condition of fixing key residues such as R, V, K, H, Q, and S, amino acid combinations are generated for the variable sites X1, X2, and X3, so that the generated sequence simultaneously meets the requirements of key residue preservation, sequence context rationality, and feasibility of peptide chemical synthesis. After deduplication, length consistency checks, and sequence rationality screening, the generated candidate sequences are used to obtain the initial candidate peptide sequences for virtual screening.

[0034] In step (5), the candidate peptide library is virtually screened and its stability is evaluated. First, ADCP is used to perform preliminary docking of candidate peptides with AAV9 epitopes, leveraging its high precision in peptide-protein complex docking to rapidly screen for suitable binding conformations. Subsequently, MMGBSA is used to calculate the binding free energy and assess the thermodynamic stability of the complex. The preliminary screening results are then refined and validated using Rosetta and HADDOCK to optimize the interface conformation and energy, further confirming the reliability of the binding mode. Finally, molecular dynamics simulations are used to analyze the dynamic stability of the complex in solution, and MMPBSA residue-by-residue energy decomposition is combined to determine the contribution of key residues, thereby screening out the thermodynamically and kinetically optimal candidate affinity peptides. This process, from initial screening, energy evaluation, conformation optimization to dynamic validation, forms a complete systematic method to systematically evaluate peptide ligand binding ability and complex stability.

[0035] This invention further provides an affinity medium containing the aforementioned affinity peptide ligand, the affinity medium comprising a solid matrix and an affinity peptide ligand coupled to the solid matrix. The solid matrix is ​​preferably an activated porous carrier capable of forming a stable covalent coupling with the peptide ligand. In the preparation of the affinity medium, the solid matrix is ​​first pretreated and activated to provide reaction sites for peptide coupling; then, the affinity peptide ligand is dissolved in a suitable buffer solution, and DMSO may be added to improve peptide solubility; by controlling the coupling time, reaction pH, peptide concentration, and reaction temperature, the peptide ligand is ensured to fully contact and immobilize with the solid matrix; finally, the coupled affinity medium is blocked and washed to remove uncoupled peptides and reaction byproducts, thereby obtaining a stable and reusable affinity medium.

[0036] In one specific embodiment, the affinity peptide ligand is SEQ ID NO.2; the corresponding affinity medium is Sep4FF-P5 formed by coupling SEQ ID NO.2 to the Sepharose 4 Fast Flow matrix.

[0037] This invention also provides a method for separating and purifying AAV9 using the aforementioned affinity medium, wherein the affinity medium comprises a solid matrix and an affinity peptide ligand coupled to the solid matrix. The method includes the following steps:

[0038] (1) Equilibrate the affinity medium with an adsorption buffer and load the sample containing AAV9 into the affinity medium to allow AAV9 to bind to the immobilized affinity peptide ligand;

[0039] (2) Use an adsorption buffer solution to wash away unbound impurities;

[0040] (3) Use elution buffer to elute AAV9 bound to the affinity medium, thereby achieving the capture and recovery of AAV9.

[0041] Preferably, the adsorption buffer condition is pH 7.5 and no NaCl is added; preferably, the elution buffer is a Gly-HCl buffer with a pH of 3.0.

[0042] Furthermore, the affinity medium prepared by the above method exhibits good adsorption capacity for AAV9 and can achieve effective elution and recovery; it has weak adsorption on non-target protein BSA, but can still selectively capture AAV9 in complex sample systems containing Vero cell supernatant. Chromatographic separation experiments and SDS-PAGE analysis show that the affinity medium can stably and reproducibly achieve the adsorption, elution, and recovery of AAV9, demonstrating the application potential of the affinity peptide ligand in the separation and purification of AAV9.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. This invention targets the Fab2-4 site that recognizes AAV9. The selected target site is located in the antibody recognition region on the AAV9 capsid surface, which is well-exposed and has a clearly defined spatial structure, providing a clear structural basis for the construction and screening of AAV9 affinity peptide ligands. Compared with random peptide library screening or empirical modification methods, this invention has the advantages of a well-defined target and a clear screening path.

[0045] 2. This invention establishes a method for screening AAV9 affinity peptide candidate sequences, which combines structural analysis, natural amino acid localization, key residue template construction, protein language model candidate peptide library generation, and multi-round binding evaluation. By utilizing protein language models to generate variable site sequences under the condition of fixed key residues, this invention can expand the candidate peptide sequence space while maintaining the target binding structure, and can be used for screening and evaluating AAV9 specific epitope affinity peptide ligands.

[0046] 3. The candidate affinity peptide ligands shown in SEQ ID NO.1 to SEQ ID NO.4 obtained by this invention all showed a tendency to form stable complexes with the AAV9 target region in the calculation evaluation. Among them, SEQ ID NO.2, as a representative sequence, was further verified by immobilization and chromatographic experiments to illustrate the feasibility of the application of the candidate affinity peptide ligands in the construction of AAV9 affinity media.

[0047] 4. The affinity medium provided by this invention can be used for the adsorption and elution of AAV9; the example affinity medium Sep4FF-P5 can achieve the capture and recovery of AAV9 under the adsorption conditions and exhibits certain separation selectivity in complex sample systems, indicating that the affinity peptide ligand has application value for AAV9 recognition, capture material development and separation and purification process construction. Attached Figure Description

[0048] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0049] Figure 1 This is a schematic diagram of the AAV9 affinity peptide ligand design pattern in Example 1 of the present invention;

[0050] The process involves first analyzing the Fab2-4 antibody binding site and identifying the target region, then identifying key residues through natural amino acid docking, constructing affinity peptide templates, and generating an initial peptide library using the ESM3 model. Subsequently, a series of computer-based virtual screening and validation methods, including ADCP, MMGBSA, HADDOCK, Rosetta, and molecular dynamics simulations, are used to finally obtain candidate peptide ligands and conduct binding mechanism analysis.

[0051] Figure 2 This is a schematic diagram of antibody Fab2-4 binding to AAV9 in Example 1 of the present invention;

[0052] This figure illustrates the binding site of Fab2-4 on the surface of the AAV9 capsid and its corresponding region. Different colored background areas represent the VP3 subunit of the AAV9 capsid, while the red portion represents the Fab2-4 antibody. The figure shows that the main binding interface of Fab2-4 covers the 2-fold depression and its adjacent 3-fold protrusion sides, providing a structural basis for target determination in this invention.

[0053] Figure 3 This is a diagram showing the docking of natural amino acids and the determination of key residues in Example 1 of the present invention;

[0054] The figure shows that after the natural amino acids were located and docked using AutoDock Vina, the amino acids with higher scores at corresponding sites in each region were selected as key residues. Region 1 consists of R and V, and Region 2 consists of K, H, Q, and S.

[0055] Figure 4 This is a sequence identification diagram of affinity peptide template construction and initial peptide library;

[0056] (a) shows the affinity peptide template RV-X1-X2-KHQ-X3-S constructed based on key residues; (b) shows the initial candidate peptide library and its sequence identification generated by the ESM3 model.

[0057] Figure 5 This is a virtual screening result diagram of ADCP and MMGBSA in Embodiment 1 of the present invention;

[0058] Among them, (a) is the ADCP docking screening result diagram, and (b) is the result of MMGBSA binding free energy calculation.

[0059] Figure 6 This is a detailed docking verification diagram of Rosetta and HADDOCK in Embodiment 1 of the present invention;

[0060] Among them, (a) is the Rosetta fine docking verification result diagram and (b) is the HADDOCK fine docking verification result diagram, which are used to further evaluate the binding mode and binding ability of candidate peptides to the AAV9 capsid target region.

[0061] Figure 7 This is a molecular dynamics simulation analysis diagram from Embodiment 1 of the present invention;

[0062] Wherein, (a) is the RMSD variation graph, and (b) is the R g The change diagram, (c) is d min Variation diagram, (d) represents N cont The variation diagram is used to evaluate the structural stability and interfacial contact of the candidate peptide-AAV9 complex during the simulation process.

[0063] Figure 8 This is a comparison diagram of the polypeptide conformation before and after a 100 ns molecular dynamics simulation in Example 1 of this invention;

[0064] In the figures, (a) is P2 (SEQ ID NO.1), (b) is P5 (SEQ ID NO.2), (c) is P11 (SEQ ID NO.3), and (d) is P13 (SEQ ID NO.4). In each group of figures, the yellow conformation represents the candidate peptide conformation before molecular dynamics simulation, and the blue conformation represents the candidate peptide conformation after 100 ns molecular dynamics simulation, which is used to show the conformational changes of the candidate peptide before and after the simulation.

[0065] Figure 9 This is a residue-by-residue binding free energy analysis diagram of gmx_MMPBSA in Example 1 of the present invention;

[0066] Among them, (a)–(d) are residue-by-residue binding free energy analysis diagrams of the complexes formed by SEQ ID NO.1, SEQ ID NO.2, SEQ ID NO.3 and SEQ ID NO.4 with AAV9, respectively, which are used to analyze the contribution of AAV9 capsid residues and candidate peptide residues to the binding free energy of the complexes.

[0067] Figure 10 This is a detailed analysis diagram of the binding mechanism of the complex formed by SEQ ID NO.2 (P5) and SEQ ID NO.5 (AAV9) in Embodiment 1 of the present invention;

[0068] Among them, (a) is the paired RMSD matrix diagram; (b) is the free energy morphology diagram; (c) is the electrostatic interaction analysis diagram; (d) is the hydrophobic interaction analysis diagram; and (e) is the hydrogen bond interaction analysis diagram.

[0069] Figure 11This is a chromatographic diagram exploring the chromatographic conditions of Sep4FF-P5 against AAV9 in Example 2 of this invention;

[0070] Among them, (a) shows the adsorption and elution results of Sep4FF-P5 on AAV9 under different pH conditions, which is used to investigate the effect of pH on the adsorption and elution behavior of affinity media; (b) shows the adsorption results of Sep4FF-P5 on AAV9 under different salt concentration conditions, which is used to investigate the effect of ionic strength on the adsorption effect of affinity media.

[0071] Figure 12 This is a chromatographic diagram showing the chromatographic purification results of AAV9 by Sep4FF-P5 under optimal adsorption conditions in Example 2 of this invention;

[0072] In the figure, (a) is the chromatographic result and (b) is the SDS-PAGE analysis chromatogram of the corresponding components. The meaning of each lane is as follows: M, protein molecular weight standard; F, sample loading material; W, flow-through component; E, elution component.

[0073] Figure 13 This is a graph showing the evaluation results of the adsorption selectivity of Sep4FF-P5 for AAV9 in Example 2 of the present invention;

[0074] In the figure, (a) is the adsorption chromatogram of BSA solution; (b) is the adsorption chromatogram of Vero cell supernatant and AAV9 mixed sample; (c) is the SDS-PAGE analysis of the corresponding components in (b). The meanings of each lane are as follows: M, protein molecular weight standard; Vero, Vero cell supernatant; AAV, AAV9 raw material; F, loading mixture; W, flow-through component; E, elution component. Detailed Implementation

[0075] Example 1

[0076] A method for constructing and screening affinity peptide candidate sequences targeting AAV9. Figure 1The overall technical route of this embodiment is illustrated. First, based on the structure of the AAV9-Fab2-4 complex, the binding region of Fab2-4 on the surface of the AAV9 capsid is analyzed to determine the target epitope region for subsequent screening. Then, natural amino acid localization analysis is performed on the target epitope region, and key residues are extracted based on the binding positions and scoring results of different amino acids in each sub-region. Next, an affinity peptide template is constructed based on the extracted key residues, and conditional sequence generation and amplification are performed around the variable sites in the template using a protein language model to construct a candidate affinity ligand peptide library. Further, the constructed candidate peptide library is subjected to molecular docking screening, binding free energy assessment, refined docking verification, and molecular dynamics simulation analysis in sequence to gradually narrow down the candidate sequence range and screen out candidate affinity peptide ligands that have binding ability to the AAV9 target region. Finally, the selected candidate sequences are analyzed for interfacial interactions and binding mechanisms to obtain candidate affinity peptide ligands that can be further used for affinity media construction and separation and purification studies. Specifically, as follows:

[0077] (1) Generation of candidate affinity peptide library

[0078] To construct a candidate affinity peptide library targeting the AAV9 capsid and screen for preferred affinity peptides, this embodiment uses the AAV9 site recognized by the Fab2-4 antibody as the design target to conduct structure-driven peptide library generation and screening. Fab2-4 is a human neutralizing antibody derived from patients treated with AAV9 gene therapy. The complex structure formed by Fab2-4 with the AAV9 capsid (PDB: 9B7N) shows that Fab2-4 primarily recognizes neutralizing epitopes near the 2-fold depression of the AAV9 capsid, and the binding region covers the 2-fold depression and its adjacent 3-fold protrusion sides, as shown in the figure. Figure 2 As shown, this region has ample surface exposure and clear spatial features, making it suitable as a design target for affinity peptide ligands.

[0079] In the specific implementation process, the AAV9-Fab2-4 complex structure file was first obtained, and the structure was preprocessed, including removing irrelevant molecules, adding missing hydrogen atoms, and minimizing energy, to obtain an acceptor structure suitable for subsequent docking analysis. Subsequently, based on the binding mode of Fab2-4 on the AAV9 capsid surface, the epitope region was divided and used as the target region for subsequent natural amino acid docking and template construction.

[0080] This embodiment employs a natural amino acid localization method, performing molecular docking of natural amino acids with the AAV9 target region. The binding site preferences and affinity trends of different amino acids in each sub-region are compared to screen suitable amino acid residues as key template sites. The results are as follows: Figure 3As shown. Based on the docking results, amino acids with better scores and good spatial matching in each region were selected as key residues, specifically Arg, Val, Lys, His, Gln, and Ser. Based on these key residues, an affinity peptide template was constructed, as shown. Figure 4 As shown in a, the overall template structure is: RV-X1-X2-KHQ-X3-S. X1, X2, and X3 are variable sites used for subsequent candidate sequence generation.

[0081] After obtaining the aforementioned key residue templates, the templates are input into the protein language model ESM3 to generate sequences from variable sites in the templates, thus constructing a candidate polypeptide sequence library, such as... Figure 4 As shown in b. During sequence generation, the R, V, K, H, Q, and S sites in the template are fixed, and mutations are only performed on the X1, X2, and X3 sites. The generated results are then deduplicated to obtain an initial peptide library containing 180 candidate polypeptide sequences for subsequent virtual screening.

[0082] (2) Virtual screening and evaluation of candidate affinity peptide libraries

[0083] To screen candidate sequences from the initial peptide library that can stably bind to the AAV9 target region, this embodiment uses ADCP and MMGBSA for virtual screening of the initial peptide library. First, ADCP molecular docking was performed on the generated 180 candidate peptides with the AAV9 target region, and initial screening was conducted based on docking scores and the rationality of the binding conformation. The docking results are as follows: Figure 5 As shown in (a), the ADCP docking scores of the candidate peptides ranged from -19.1 to -24.3 kcal / mol. To improve the specificity of subsequent analyses, peptides with docking scores below -23 kcal / mol that could stably locate in the preset target region were included in further evaluation, resulting in a total of 16 candidate peptides.

[0084] Subsequently, the MMGBSA binding free energy of the aforementioned 16 candidate peptides was further calculated to obtain a quantitative evaluation of the binding strength between the candidate peptides and the AAV9 complex. The relevant results are as follows: Figure 5 As shown in (b). The results showed that four peptides performed better, namely P2, P5, P11 and P13, and were designated as SEQ ID NO.1, SEQ ID NO.2, SEQ ID NO.3 and SEQ ID NO.4, respectively.

[0085] Building upon this, to further improve the reliability of the candidate peptide screening results, this embodiment employs two independent refined docking methods, Rosetta and HADDOCK, to further validate the candidate peptide-AAV9 complexes after preliminary screening using ADCP and MMGBSA. The results are as follows: Figure 6 (a) and Figure 6 As shown in (b). Generally, a lower score indicates a more favorable conformation of the corresponding complex, and better relative binding performance between the candidate peptide and the AAV9 target region. Figure 6 It is evident that different candidate peptides exhibited varying performance in the two refined docking evaluation systems. However, the four candidate peptides, P2, P5, P11, and P13, generally scored lower, indicating a more reasonable binding conformation. The specific refined docking scores of the aforementioned candidate peptides with the AAV9 complex are shown in Table 1. Table 1 indicates that P2, P5, P11, and P13 all possessed low Rosetta and HADDOCK scores, providing a basis for further molecular dynamics simulation verification.

[0086] Table 1

[0087]

[0088] Based on the initial screening, combined with free energy calculation and refined docking results, this embodiment screened out 4 candidate affinity peptide ligands from the generated candidate affinity peptide library, which are denoted as SEQ ID NO.1, SEQ ID NO.2, SEQ ID NO.3 and SEQ ID NO.4, respectively.

[0089] (3) Molecular dynamics simulation verification

[0090] To further confirm the dynamic stability between the aforementioned candidate peptides and the AAV9 target region, this embodiment conducted 100 ns molecular dynamics simulations on the complexes formed by SEQ ID NO.1, SEQ ID NO.2, SEQ ID NO.3, and SEQ ID NO.4 with AAV9. The simulations were performed using GROMACS 2023.2 software, employing the AMBER 99SB-ILDN force field and TIP3P water model to construct the systems. Electroneutrality was maintained by adding counterions, and the salt concentration was adjusted to physiological conditions. The simulated trajectories were used for subsequent root mean square deviation (RMSD) and radius of gyration (R²). g ), minimum interface distance (d) min ) and atomic contact number (N cont Analysis, results as follows Figure 7 As shown. By Figure 7 (a) As can be seen, the RMSD of all four candidate peptide-AAV9 complexes increased to some extent in the early stage of the simulation, and then entered a relatively stable fluctuation state without a continuous increasing trend, indicating that the overall conformation of the complex did not undergo significant instability or dissociation during the simulation. Figure 7 (b) It can be seen that the Rg values ​​of each system remained within a relatively narrow range during the 100 ns simulation, indicating that the overall compactness of the complex remained relatively stable and no significant loosening occurred. Figure 7 (c) As can be seen, the minimum interfacial distance between the candidate peptide and the AAV9 target region remained at a low level overall, suggesting that the candidate peptide maintained close contact with the AAV9 surface throughout the simulation. Figure 7 (d) As can be seen, all systems maintained a high number of interfacial atomic contacts. Although there were some differences in the number and fluctuation of contacts between different candidate peptides, there was no situation where the number of contacts continuously decreased to a low level, indicating that there was a continuous and effective interfacial interaction between the candidate peptides and the AAV9 target region. (Based on RMSD and R...) g d min and N cont The analysis results show that the complexes formed by SEQ ID NO.1, SEQ ID NO.2, SEQ ID NO.3, and SEQ ID NO.4 with AAV9 maintained relatively stable overall conformation and interfacial contact state during the 100 ns simulation, indicating that the four candidate peptides all possess good interfacial binding stability. Further comparison of representative conformations before and after the 100 ns simulation yielded the following results: Figure 8 As shown, the overall binding posture of the four candidate peptides did not change much before and after the simulation. None of them underwent significant dissociation or large conformational shift. They only underwent some adaptive adjustments in local side chain orientation and interface adhesion. This indicates that while maintaining stable binding with the AAV9 target region, the candidate peptides still possess a certain degree of interface flexibility, which is conducive to forming a more stable complex conformation.

[0091] To further quantitatively evaluate the binding strength between the candidate peptide and AAV9, this embodiment calculates the MMPBSA binding free energy of the simulated trajectory and further performs residue-by-residue energy decomposition. The results are as follows: Figure 9 As shown in the figure; the results of the total binding free energy analysis are shown in Table 2. Meanwhile, based on ΔG... bind The theoretical dissociation constant K is obtained through thermodynamic calculations. d This is used to compare the binding ability of candidate peptides.

[0092] Table 2

[0093]

[0094] Total binding free energy analysis results showed that the complexes formed by SEQ ID NO.1, SEQ ID NO.2, SEQ ID NO.3 and SEQ ID NO.4 with AAV9 all had negative ΔG. bind The values ​​ranged from -7.73 to -8.88 kcal / mol, indicating that all four candidate peptides could form thermodynamically favorable complexes with the AAV9 target region. Based on ΔG... bind Theoretical K obtained by conversiond The range is 3.05 × 10 -7 Up to 2.13×10 -6 The concentration of mol / L indicates that all four candidate peptides possess a certain binding capacity. The energy components show that the four candidate peptide-AAV9 complexes all exhibit favorable energy contributions from van der Waals and electrostatic interactions, suggesting multiple non-covalent interactions between the candidate peptides and the AAV9 target region. The combined MMPBSA binding free energy and theoretical KAAV9... d The analysis results show that SEQ ID NO.1, SEQ ID NO.2, SEQ ID NO.3 and SEQ ID NO.4 all have the potential to be used as AAV9 affinity peptide ligands.

[0095] Residue-by-residue energy decomposition analysis, such as Figure 9 As shown, the positively charged Arg residues at the N-terminus of the candidate peptides, along with other charged or polar residues, contribute the most to the binding of the complex, indicating that electrostatic interactions play a dominant role in interfacial stability. Simultaneously, hydrophobic sites also contribute to the local binding energy, aiding in interface stability. SEQ ID NO.1 and SEQ ID NO.4 exhibit multi-conformation low-energy regions, suggesting a certain degree of interface flexibility, which is beneficial for adapting to local structural differences at the AAV9 target site. SEQ ID NO.2 and SEQ ID NO.3 show concentrated low-energy regions, better interfacial convergence, and exhibit more stable binding conformations.

[0096] (4) Analysis of the binding mechanism of preferred ligand P5

[0097] To further elucidate the interaction mechanism between the candidate affinity peptide and the AAV9 target region, this embodiment selected SEQ ID NO.2 (P5) as a representative candidate ligand for interfacial interaction analysis. The results are as follows: Figure 10 As shown. It should be noted that SEQ ID NO.2 (P5) in this embodiment is used to representatively demonstrate the binding mechanism between the candidate affinity peptide and AAV9, and does not exclude the application of SEQ ID NO.1, SEQ ID NO.3, and SEQ ID NO.4 as construction sequences for the AAV9 affinity peptide ligand and affinity mediator. Among them, Figure 10 The paired RMSD matrices shown in (a) indicate that the P5-AAV9 complex forms a large area of ​​low RMSD region in the later stage of the simulation, which suggests that this representative complex has high conformational similarity and a certain convergence trend in the later stage of the simulation. Figure 10 (b) shows that the P5-AAV9 complex forms a relatively concentrated low-energy conformation region, suggesting that this representative complex can form a relatively stable low-energy conformation state.

[0098] Further interfacial interaction analysis revealed a variety of non-covalent interactions between P5 and the AAV9 target region, including electrostatic, hydrophobic, and hydrogen bonding interactions. Electrostatic interactions were primarily manifested in the charge complementarity between Arg1 and Asp532, and between Lys5 and Glu563 and Asp611. Hydrophobic interactions were mainly observed in the hydrophobic / aromatic contact between Trp3 and Ile699, and the localized hydrophobic interaction between His6 and Pro726. Hydrogen bonding interactions primarily involved sites such as P726, I560, T561, F534, E563, and E564. These results indicate that a multi-force synergistic interface involving electrostatic, hydrophobic, and hydrogen bonding can be formed between the representative candidate ligand P5 and the AAV9 target region, which can be used to explain the possible mechanisms by which candidate affinity peptides recognize the AAV9 target region.

[0099] Based on comprehensive target analysis, natural amino acid localization and docking, affinity peptide template construction, ESM3 initial peptide library generation, ADCP / MMGBSA virtual screening, Rosetta and HADDOCK fine docking, 100 ns molecular dynamics simulation, and MMPBSA binding free energy analysis, it is evident that SEQ ID NO.1, SEQ ID NO.2, SEQ ID NO.3, and SEQ ID NO.4 obtained in this embodiment can all form relatively stable complexes with the AAV9 target region, demonstrating potential application as AAV9 affinity peptide ligands. Among these, SEQ ID NO.2 is used as a representative sequence in this embodiment for further mechanism analysis, affinity medium construction, and chromatographic purification verification.

[0100] Example 2

[0101] The affinity media constructed from the screened representative candidate affinity peptide ligands and their application in the separation and purification of AAV9 were validated as follows:

[0102] (1) Preparation of affinity chromatography media

[0103] To verify the practical application capability of the candidate affinity peptide ligands screened in Example 1, this example selects SEQ ID NO.2 (P5) as a representative ligand and couples it to a Sepharose 4 Fast Flow (Sep4FF) matrix to prepare the affinity chromatography medium Sep4FF-P5 for AAV9 adsorption and purification. Following the conventional route for coupling peptide ligands to thiol-reactive agarose media, the preparation of the affinity medium mainly includes solid-phase matrix pretreatment, activated matrix preparation, peptide ligand coupling, and blocking and washing steps. The solid-phase matrix is ​​preferably a Sep4FF matrix treated with thiol pyridinization to ensure that the peptide ligand can be stably immobilized on the carrier surface and maintain good chromatographic performance. The activation, thiolization, and pyridinization steps of the matrix can be performed according to conventional methods in the art to obtain an activated medium capable of undergoing thiol-disulfide bond exchange reactions with the peptide ligand.

[0104] In the coupling step, the P5 peptide is dissolved in Tris-HCl buffer to prepare a coupling solution, preferably a 0.1 mol / L Tris-HCl buffer with pH 7.5, and 4% DMSO is added to improve the solubility of the peptide in the coupling system; the concentration of the peptide ligand solution is preferably 1-2 mg / mL. Then, 0.6-0.8 g of mercaptopyridinized medium is mixed with the above coupling solution, and the mixture is shaken at 25-30℃ for 1.5-2.5 h to ensure sufficient contact between the P5 peptide and the activated matrix and to achieve effective coupling. After the coupling reaction is complete, cysteine ​​is added to block unreacted active sites, and the mixture is then thoroughly washed with deionized water and buffer solution to remove uncoupled peptides and residual small molecule impurities, finally obtaining the affinity medium Sep4FF-P5 coupled with the P5 affinity peptide ligand. The above coupling process parameters can balance ligand immobilization efficiency, media stability, and subsequent chromatographic separation performance.

[0105] (2) Affinity chromatography separation and purification process

[0106] To determine the optimal adsorption and elution conditions for AAV9 by Sep4FF-P5, this example systematically investigated the chromatographic behavior under different pH and salt concentrations. The results are as follows: Figure 11 As shown.

[0107] Sep4FF-P5 exhibited strong adsorption capacity for AAV9 under different pH conditions, indicating that this affinity medium can effectively bind to the target viral vector over a wide pH range. Simultaneously, bound AAV9 could be effectively eluted with Gly-HCl buffer (pH 3.0), suggesting that acidic conditions promote the dissociation of the ligand-target complex. Further investigation into the effect of different salt concentrations on adsorption behavior revealed that the flow-through peak was smallest at pH 7.5 without the addition of NaCl, indicating that Sep4FF-P5 showed superior adsorption performance for AAV9 under these conditions. As the salt concentration increased, the flow-through peak gradually increased, indicating that increasing the ionic strength did not enhance adsorption but rather hindered the formation of effective binding.

[0108] Based on the molecular dynamics simulations and interaction analyses in Example 1, the binding of P5 to AAV9 involves the synergistic participation of various non-covalent interactions, including electrostatic interactions, hydrophobic interactions, hydrogen bonds, and multi-site contacts. Specifically, in the initial stage of P5 approaching and recognizing AAV9, electrostatic attraction and charge complementarity between charged residues facilitate directional contact between the ligand and the target, thereby increasing the probability of complex formation. After initial recognition, hydrophobic interactions, hydrogen bonds, and multi-site contacts further participate in interface stabilization, forming a relatively stable complex structure. In this case, although increasing the salt concentration enhances the shielding effect of the solution on charge interactions and weakens the initial electrostatic recognition and directional interaction between the ligand and AAV9, it cannot effectively disrupt the already formed hydrophobic contacts, hydrogen bond network, and multi-site synergistic interface. Therefore, it cannot improve the adsorption effect, nor is it sufficient to achieve effective desorption of the bound target.

[0109] On the other hand, low pH conditions may weaken the original charge complementarity by altering the charge state of the P5 ligand and related residues on the AAV9 capsid surface, while simultaneously perturbing the hydrogen bond and salt bridge network at the interface. This disrupts the synergistic interactions that maintain the stability of the complex, ultimately leading to the effective elution of the target virus. The combined results indicate that the adsorption process of Sep4FF-P5 on AAV9 involves a synergistic mechanism of electrostatic interactions in initial recognition, as well as hydrophobic interactions, hydrogen bonds, and multi-site contact to stabilize the binding interface, rather than a purely electrostatic-dependent ion-exchange binding mode.

[0110] Based on the above results, pH 7.5 and 0 mM NaCl were determined to be the optimal adsorption conditions for subsequent chromatographic purification experiments in this embodiment, while Gly-HCl buffer (pH 3.0) was used as an effective elution condition for subsequent separation and purification studies. Under the above optimal adsorption conditions, AAV9 was purified by chromatography using Sep4FF-P5, and the results are as follows. Figure 12 As shown. Figure 12(a) shows that AAV9 can be effectively adsorbed by Sep4FF-P5 under pH 7.5 and 0 mM NaCl conditions and can be recovered during the elution stage. Figure 12 (b) shows the SDS-PAGE analysis results of the corresponding components. The electrophoresis results show that the target-related band appears in the eluted component, while no obvious corresponding band is seen in the flow-through component, indicating that Sep4FF-P5 has good adsorption and recovery capabilities for AAV9. Furthermore, gray-scale integral analysis of the electrophoretic bands using ImageJ software shows that the relative recovery rate of AAV9 based on gray-scale integral is 64.93%. This result demonstrates that, under the conditions described in this embodiment, the affinity medium formed by the coupling of the representative affinity peptide ligand P5 can achieve effective capture and recovery of AAV9.

[0111] To further evaluate the adsorption selectivity of Sep4FF-P5 for AAV9, this example investigated its adsorption behavior on BSA solution and a mixture of Vero cell supernatant and AAV9. The results are as follows: Figure 13 As shown in the figure, when BSA was used as a non-target protein model for validation, no obvious binding or retention was observed in the chromatographic results, indicating that Sep4FF-P5 has weak adsorption of BSA, suggesting that the affinity medium has a low level of non-specific adsorption for non-target proteins. Further chromatographic experiments were performed in a mixture of Vero cell supernatant and AAV9. The results showed that most of the impurity components from the Vero cell supernatant were removed by flow-through, while the target-related components were recovered during the elution stage. The corresponding SDS-PAGE results further showed that the bands of each component were clearly separated. The flow-through component mainly consisted of impurity protein bands, while the target-related bands were significantly enriched in the eluted component, and no obvious background impurities were found. This indicates that Sep4FF-P5 can still maintain effective recognition and adsorption of AAV9 in complex sample systems. The above results show that the affinity medium not only has low non-specific protein adsorption but also can selectively capture and purify target viral vectors under complex biological raw material conditions, demonstrating good separation selectivity and practical application potential.

[0112] This embodiment demonstrates that, among the candidate affinity peptide ligands screened in Example 1, SEQ ID NO.2 was used as the representative sequence for immobilization to form the Sep4FF-P5 affinity medium. This medium exhibits superior adsorption performance for AAV9 under pH 7.5 and 0 mM NaCl conditions, and can be eluted with Gly-HCl buffer (pH 3.0). Under the conditions of this embodiment, AAV9 can be captured and recovered, with low non-specific adsorption of BSA, and AAV9 enrichment can be achieved in a complex sample system simulating Vero cell supernatant. The above results indicate that SEQ ID NO.1, SEQ ID NO.2, SEQ ID NO.3, and SEQ ID NO.4 all have the potential to be used as affinity purification ligands for AAV9; among them, SEQ ID NO.2 was used as the representative sequence for immobilization and chromatographic verification in this embodiment.

[0113] Obviously, the above embodiments are merely examples for clearly illustrating the present invention and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all embodiments here. However, obvious variations or modifications derived therefrom are still within the protection scope of the present invention.

[0114] The sequence of this invention is as follows:

[0115] SEQ ID NO.1: RVFDKHQIS

[0116] SEQ ID NO.2: RVWRKHQLS

[0117] SEQ ID NO.3: RVNVKHQRS

[0118] SEQ ID NO.4: RVDLKHQVS

[0119] SEQ ID NO.5:DGVGSSSGNWHCDSQWLGDRVITTSTRTWALPTYNNHLYKQISNSTSGGSSNDNAYFGYSTPWGYFDFNRFHCHFSPRDWQRLINNNWGFRPKRLNFKLFNIQVKEVTDNNGVKTIANNLTSTVQVFTDSDYQLPYVLGSAHEGCLPPFPADVFMIPQYGYLTLNDGSQAVGRSSFYCLEYFPSQMLRTGNNFQFSYEFENVPFHSSYAHSQSLDRLMNPLIDQYLYYLSKTINGSGQNQQTLKFSVAGPSNMAVQGRNYIPGPSYRQQRVSTTVTQNNNSEFAWPGASSWALNGRNSLMNPGPAMASHKEGEDRFFPLSGSLIFGKQGTGRDNVDADKVMITNEEEIKTTNPVATESYGQVATNHQSAQAQAQTGWVQNQGILPGMVWQDRDVYLQGPIWAKIPHTDGNFHPSPLMGGFGMKHPPPQILIKNTPVPADPPTAFNKDKLNSFITQYSTGQVSVEIEWELQKENSKRWNPEIQYTSNYYKSNNVEFAVNTEGVYSEPRPIGTRYLTRNL

Claims

1. Generating affinity peptide ligands for AAV9 antibody epitopes using protein language models, characterized by: The amino acid sequence of the affinity peptide ligand is any one of SEQ ID NO.1, SEQ ID NO.2, SEQ ID NO.3 and SEQ ID NO.

4.

2. A method for screening antibody epitope affinity peptides targeting AAV9 using protein language models, characterized by: Includes the following steps: (1) Obtain the structure of the AAV9-Fab2-4 complex and perform structural pretreatment; (2) The AAV9 capsid epitope regions identified by Fab2-4 were divided, and the binding site preferences of different amino acids in each sub-region were analyzed using natural amino acid localization analysis. (3) Extract key residues based on the localization results and construct an affinity peptide template; (4) Based on the affinity peptide template, the variable sites in the template are conditionally generated, amplified, and deduplicated using a protein language model to construct candidate affinity ligand peptides. (5) Molecular docking, binding free energy calculation, refined docking verification and molecular dynamics simulation analysis are performed on the candidate affinity peptide library to screen out candidate affinity peptide ligands.

3. The method as described in claim 2, characterized in that, In step (1), the structure file numbered 9B7N in the public database PDB is used as the target source; the chain and antibody fragments involved in binding with the AAV9 capsid are retained for interface reference; missing hydrogen atoms are added and the protonation state is adjusted, followed by local energy minimization to eliminate potential structural inconsistencies, and then converted into the PDBQT software format suitable for docking and simulation.

4. The method as described in claim 2, characterized in that, In step (2), based on the interface between Fab2-4 and AAV9, combined with spatial location and local physicochemical characteristics, the binding region is divided into sub-regions; by performing molecular docking between each residue and the epitope sub-region, comparing scores and spatial locations, the amino acid that performs best in a specific region is selected as the key residue, providing a basis for template construction.

5. The method as described in claim 2, characterized in that, in In step (3), the key residues include Arg, Val, Lys, His, Gln and Ser; the overall structure of the affinity peptide template constructed from the key residues is RV-X1-X2-KHQ-X3-S, where X1, X2 and X3 are variable sites.

6. The method as described in claim 2, characterized in that, in In step (4), the affinity peptide template composed of key residues is input into the protein language model. Under the condition of fixing the key residues R, V, K, H, Q and S, amino acid combinations are generated for the variable sites X1, X2 and X3 so that the generated sequence simultaneously meets the requirements of key residue retention, sequence context rationality and the feasibility of peptide chemical synthesis. After the generated candidate sequence is deduplicated, length consistency checked and sequence rationality screened, the initial candidate peptide sequence for virtual screening is obtained.

7. The method as described in claim 2, characterized in that, In step (5), ADCP is used to perform preliminary docking of candidate peptides with AAV9 epitopes. Due to its high precision in docking of peptide-protein complexes, reasonable binding conformations are quickly screened. Subsequently, the binding free energy was calculated using MMGBSA to evaluate the thermodynamic stability of the complex. The preliminary screening results were then refined and validated by optimizing the interface conformation and energy using Rosetta and HADDOCK to further confirm the reliability of the binding mode. Finally, the dynamic stability of the complex in the solution environment was analyzed by molecular dynamics simulation, and the contribution of key residues was determined by residue-by-residue energy decomposition using MMPBSA, thereby screening out the thermodynamically and kinetically optimal candidate affinity peptides.

8. A method for separating and purifying AAV9 using protein language models to generate antibody epitope affinity peptide ligands targeting AAV9, characterized in that... The affinity medium comprises a solid matrix and an affinity peptide ligand immobilized on the matrix, and includes the following steps: (1) Load the sample containing AAV9 onto the affinity medium to allow AAV9 to bind to the immobilized affinity peptide ligand; (2) The affinity medium is cleaned to remove unbound impurities; (3) Elute AAV9 bound to the affinity medium to achieve AAV9 capture and recovery.

9. The method according to claim 8, wherein the adsorption buffer is 20 mM pH 7.5 phosphate buffer (PB).

10. The method of claim 8, wherein the elution buffer is a pH 3.0, 0.1M Gly-HCl buffer.

Images