A method for designing cyclic peptide drugs from scratch, electronic device, and storage medium

Abstract

This invention discloses a de novo design method for cyclic peptide drugs, an electronic device, and a storage medium. The method includes: (1) screening seed short peptides; (2) peptide extension; (3) screening of candidate peptides: for candidate peptides, a protein structure prediction algorithm is used to simulate the complex structure of the peptide with the target protein; for candidate peptides whose binding ability with the target protein exceeds a preset threshold, the conformation of the peptide with the target protein in the complex structure is obtained; when the conformation of the candidate peptide forms a bicyclic peptide or a cyclic peptide with its ends connected, the peptide extension is terminated and this candidate peptide is used as a candidate structure for the cyclic peptide drug; when the conformation of the candidate peptide forms a linear peptide with a length exceeding a preset threshold, the peptide extension is terminated; otherwise, proceed to step (2). This invention can directly design cyclic peptides that can target target protein molecules without relying on prior information on protein hotspot residues, saving most of the screening process compared to designing a cyclic peptide library first and then screening.

Description

Technical Field

[0001] The present invention belongs to the field of computer-aided drug design, and more specifically, relates to a cyclic peptide drug de novo design method, electronic equipment, and storage medium. Background Art

[0002] Peptides have been used as a complementary treatment to antibodies and small molecule drugs. Compared with antibodies, peptides can bind to flat protein surfaces with high affinity and selectivity, and they can penetrate cell membranes to enter intracellular targets. Therefore, peptides provide a therapeutic space that antibodies and small molecule drugs cannot reach. Peptides are mainly divided into linear peptides and cyclic peptides. Linear peptides cannot be well drugged because they can be twisted and flipped at will and their structures are too loose. Reasonable cyclic peptide structures often have stable local secondary structures, which makes the entropy cost of cyclic peptides binding to target surfaces smaller. Cyclic peptides are divided into head-to-tail cyclic peptides, side chain and side chain cyclic peptides, side chain and terminal cyclic peptides, cyclic peptides containing disulfide bridges, and other bridged cyclic peptides according to the ring synthesis method. Natural cyclic peptides have become a major source of druggable cyclic peptides in the last century. In the past two decades, the main trend of peptide drug discovery has been to shift from naturally isolated cyclic peptides to optimized cyclic peptide analogs that improve efficacy, stability and pharmacokinetic properties. High-throughput screening of cyclopeptide libraries, such as mRNA display, DNA display, and phage display, is undoubtedly an effective way to discover cyclopeptides with better activity. However, these libraries are only a small subset of the entire cyclopeptide sequence space and cannot cover a wider sequence and conformation space. Therefore, the de novo design of cyclopeptide drugs based on target structure will also be a research direction with great potential in the future. However, there is not much work on the de novo design of cyclopeptide sequences based on target structure.

[0003] In 2017, David Baker's group at the University of Washington published more than 200 independent stable cyclic peptide structures designed by computer. They used Generaized Kinematic Closure (genKIC) to start from the poly-glycine (Poly-Gly) cyclic peptide backbone, and used L and D chiral amino acids to broaden the conformational space to obtain a series of stable peptide backbones and conduct sequence design at the same time. Based on this method, the amino acids with unsaturated NH on the main chain were replaced with main chain methylated amino acids in the Rosetta sequence design stage, and only non-polar residues were allowed in the side chain sequence design space. Using a similar method, permeable cyclic peptide molecules can be designed. To directly design cyclic peptide binders that can directly target target proteins, the sequence and conformation of cyclic peptides can be sampled on the surface of the target protein. That is, using known binding molecules as anchors, the poly-glycine backbone is unfolded and cyclized at the pocket, and after the side chains are refined, the shape complementarity coefficient or free energy is used to screen the optimal cyclic peptide, and the optimal design is evaluated using downstream experimental methods. However, the affinity of cyclic peptides designed by using ready-made cyclic peptide backbones and then mutating amino acids to target proteins is only at the micromolar level. The article published by David Baker's group in 2021 pointed out that SHA (2S-2-amino-7-sulfanylheptanoic acid) is first used to anchor into the pocket of the target protein, and then all possible classical amino acids are extended in two directions to find the solution with the highest shape complementarity or ΔΔG evaluation. The cyclic peptide backbone is extended and the amino acid design cyclic peptide is mutated to find a cyclic peptide with nanomolar affinity to the target protein. The de novo design of cyclic peptides provided by the Chinese patent document CN114333985A uses a Monte Carlo sampling algorithm framework to perform polypeptide cyclization and docking. They first docked the tripeptide with the specific epitope region of the target protein, and screened the tripeptide binding conformation with the lowest binding free energy through mutation and conformational sampling, and then extended the tripeptide to the set threshold at both ends, and then screened the candidate cyclic peptide conformations that can form a ring according to the preset backbone dihedral angle change rules, and selected the target cyclic peptide sequence after optimization.

[0004] At present, the design of cyclopeptides based on target structures can only design corresponding cyclopeptide molecules through local docking methods based on the known hot residues of various proteins, and then screen the optimal cyclopeptide molecules through energy functions. The local docking method calculates the target protein through the local domain, which limits the possibility of searching for cyclopeptide molecules that can target other epitopes. At the same time, the molecular dynamics or Rosetta docking method needs to predict the dihedral angle range, and then sample and simulate the conformation through the Monte Carlo annealing method, which often requires a lot of calculations and takes a lot of time, and the accuracy often does not meet the requirements. Summary of the invention

[0005] In view of the above defects or improvement needs of the prior art, the present invention provides a method for de novo design of cyclic peptide drugs, an electronic device, and a storage medium, the purpose of which is to screen the structure prediction and peptide structure prediction of seed short peptide binding protein complexes, and to perform sequence and structure design screening of cyclic peptide drugs without the need for prior protein hot spot residue information, so as to achieve de novo design of cyclic peptide drugs that only requires the target protein sequence, thereby solving the technical problems that the existing cyclic peptide drug design methods rely on prior protein hot spot residues or domain information, limit the target site structure of drug design, and rely on dihedral angle peptide ring optimization, which limits the structural diversity of cyclic peptide segments.

[0006] To achieve the above object, according to one aspect of the present invention, a method for de novo design of cyclic peptide drugs is provided, characterized in that it comprises the following steps:

[0007] (1) Screening seed short peptides: For short peptides of preset length, a protein structure prediction algorithm is used to simulate the complex structure between the short peptides and the target protein; the short peptides that interact with the target protein are used as seed short peptides to initialize the seed short peptides;

[0008] (2) Peptide extension: Extend a search peptide at the N-terminus and / or C-terminus of the seed short peptide according to a preset step length as a candidate peptide, and proceed to step (3);

[0009] (3) Screening of candidate peptides: For candidate peptides, a protein structure prediction algorithm is used to simulate the complex structure of the candidate peptides with the target protein; for candidate peptides whose binding ability with the target protein exceeds a preset threshold, the conformation of the candidate peptides in the complex structure with the target protein is obtained;

[0010] When the conformation of the candidate peptide segment forms a bicyclic peptide or a head-to-tail cyclic peptide, the peptide segment extension is terminated to use the candidate peptide segment as a candidate structure of a cyclic peptide drug;

[0011] When the conformation of the candidate peptide forms a linear peptide whose length exceeds a preset threshold, the peptide extension is terminated;

[0012] Otherwise, go to step (2);

[0013] Preferably, in the method for de novo design of cyclic peptide drugs, in step (3), when the conformation of the candidate peptide segment forms a bicyclic peptide, a cyclic peptide connected end to end, or when the conformation of the candidate peptide segment forms a linear peptide whose length exceeds a preset threshold, the peptide segment extension is terminated and the process proceeds to step (4); for the candidate peptide segment whose conformation forms a bicyclic peptide or a cyclic peptide connected end to end, or for the linear cyclic peptide drug candidate structure whose length exceeds a preset threshold, the cyclization optimization is performed according to the following method:

[0014] (4) Amino acid mutations are performed outside the key amino acid residues that form the ring to obtain new candidate peptides. When the conformation of the candidate peptides forms a head-to-tail cyclic peptide or lasso structure and the binding ability with the target protein exceeds a preset threshold, the candidate peptides are used as candidate structures of cyclic peptide drugs.

[0015] Preferably, in the method for de novo design of cyclic peptide drugs, the short peptide that interacts with the target protein is determined by one of the following methods:

[0016] (1-1) A short peptide whose distances between all its atoms and the target protein are less than or equal to a preset threshold is determined as a short peptide that interacts with the target protein;

[0017] (1-2) using a variety of different models to simulate the complex structure of the short peptide and the target protein, using RMSD value and / or DockQ value as screening indexes to screen the complex structure, and determining the short peptide whose screening index exceeds a preset threshold as a short peptide that interacts with the target protein;

[0018] (1-3) The short peptide obtained by method (1-1) and / or (1-2) is a short peptide that interacts with the target protein.

[0019] Preferably, the method for designing cyclic peptide drugs from scratch uses a short peptide anchored to a preset epitope as a seed short peptide for a target protein with a specific structural domain;

[0020] Preferably, the seed short peptides are screened:

[0021] A short peptide with evaluation indexes within a preset range is used as a seed short peptide, wherein the evaluation indexes include but are not limited to the number of contact residue pairs, shape complementarity coefficient, interaction area, binding free energy, or affinity.

[0022] Preferably, in the method for de novo design of cyclic peptide drugs, the amino acid residues of the search peptide in step (2) are selected from 20 natural amino acids, and the step length is preferably 1;

[0023] When the seed short peptide has a linear structure, it is preferred that the N-terminus and the C-terminus are extended separately.

[0024] When the seed short peptide has a lasso structure, a search peptide segment is extended from the free end of the lasso structure as an alternative peptide segment.

[0025] Preferably, in the method for de novo design of cyclic peptide drugs, when there is an amino acid residue in the short peptide and the distance between it and two or more amino acid residues is less than a preset threshold, and one of the two amino acid residues is the N-terminal or C-terminal residue, then the seed short peptide is judged to have a lasso structure; preferably, the preset threshold is 4.5 angstroms.

[0026] Preferably, in the method for de novo design of cyclic peptide drugs, the binding ability of the peptide segment to the protein in step (3) is quantitatively evaluated using a binding ability evaluation index, and the binding ability evaluation index includes but is not limited to: the number of contact residue pairs, shape complementarity coefficient, interaction area, binding free energy, affinity ranking, and / or phage display technology screening results.

[0027] Preferably, in the de novo design method for cyclic peptide drugs, the conformation of the candidate peptide segments forms a bicyclic peptide, specifically referring to that the distances between the N-terminal and C-terminal residues of the short peptide and two or more non-terminal residues therebetween are less than a preset threshold; the conformation of the candidate peptide segments forms a head-to-tail cyclic peptide, specifically referring to that: the distance between the N-terminal and C-terminal residues of the short peptide is less than a preset threshold; the preset threshold is 4.5 angstroms.

[0028] According to another aspect of the present invention, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for de novo design of cyclic peptide drugs provided by the present invention are implemented.

[0029] According to another aspect of the present invention, a non-transitory computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the de novo design method of cyclic peptide drugs provided by the present invention are implemented.

[0030] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0031] The present invention can directly design cyclic peptides that can target target protein molecules without relying on prior protein hotspot residue information, saving most of the screening process compared to first designing a cyclic peptide library and then screening. In addition, the present invention can achieve global docking, and can globally search for docking epitopes without protein epitope information, and complete docking at the same time, combining protein complex structure prediction and peptide segment structure prediction to quickly and comprehensively obtain alternative cyclic peptide drug molecules.

[0032] The preferred solution can also be screened according to specific epitopes, which can effectively increase the specificity of the cyclic peptide molecules. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a flow chart of the anchor module for screening seed short peptides;

[0034] Figure 2 This is the flow chart of the extension module for peptide extension and candidate peptide screening;

[0035] Figure 3 It is the optimization module flow chart of loop optimization;

[0036] Figure 4 It is the quasi-cyclic peptide molecular structure of SEQ ID NO:2;

[0037] Figure 5 It is the complex structure of the quasi-cyclic peptide molecule of SEQ ID NO:2 and SEQ ID NO:1 (SORT1);

[0038] Figure 6 It is the molecular structure of quasi-cyclic peptide of SEQ ID NO:3;

[0039] Figure 7 It is the molecular structure of the quasi-cyclic peptide of SEQ ID NO:4;

[0040] Figure 8 It is the molecular structure of SEQ ID NO:5 quasi-cyclic peptide;

[0041] Fig. 9 It is the molecular structure of SEQ ID NO:6 quasi-cyclic peptide;

[0042] Fig.10 It is the molecular structure of SEQ ID NO:7 quasi-cyclic peptide;

[0043] Fig.11 is the quasi-cyclic peptide molecular structure of SEQ ID NO: 8;

[0044] Fig.12 is the quasi-cyclic peptide molecular structure of SEQ ID NO: 10;

[0045] Fig.13 It is the complex structure of the quasi-cyclic peptide molecule of SEQ ID NO: 10 and SEQ ID NO: 9 (PD1). DETAILED DESCRIPTION

[0046] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0047] The method for de novo design of cyclic peptide drugs provided by the present invention comprises the following steps:

[0048] (1) Screening seed short peptides: For a short peptide of a preset length, a protein structure prediction algorithm is used to simulate the structure of the complex between the short peptide and the target protein; a short peptide that interacts with the target protein is used as a seed short peptide, thereby initializing the seed short peptide; the short peptide that interacts with the target protein is determined according to one of the following methods:

[0049] (1-1) A short peptide whose distances between all its atoms and the target protein are less than or equal to a preset threshold is determined as a short peptide that interacts with the target protein;

[0050] (1-2) using a variety of different models to simulate the complex structure of the short peptide and the target protein, using RMSD value and / or DockQ value as screening indexes to screen the complex structure, and determining the short peptide whose screening index exceeds a preset threshold as a short peptide that interacts with the target protein;

[0051] Compared with traditional methods, there is no need to preset the dihedral angle range and perform a time-consuming Monte Carlo annealing conformation sampling process. It can obtain a more accurate complex structure than molecular dynamics or Rosetta-based methods more quickly, and no knowledge of amino acid dihedral angles is required for operation.

[0052] (1-3) The short peptide obtained by method (1-1) and / or (1-2) is a short peptide that interacts with the target protein.

[0053] Generally speaking, if the length of the seed peptide is too short, the complex with the target protein is not representative; if the length of the seed peptide is too long, there are too many possibilities of traversing the seed peptide. The more suitable seed peptide length is 3 to 7 peptides, preferably 3 peptides.

[0054] In a preferred embodiment, for a target protein with a specific domain, a short peptide anchored to a preset epitope is used as a seed short peptide. The seed short peptide can also be further screened as needed, and a short peptide with an evaluation index within a preset range is used as a seed short peptide, wherein the evaluation index includes but is not limited to the number of contact residue pairs, shape complementarity coefficient, interaction area, binding free energy, or affinity.

[0055] (2) Peptide extension: Extend a search peptide at the N-terminus and / or C-terminus of the seed short peptide according to a preset step length as an alternative peptide, and proceed to step (3); the amino acid residues of the search peptide are selected from the 20 natural amino acids, and the step length is preferably 1.

[0056] When the seed short peptide has a linear structure, it is preferred that the N-terminus and the C-terminus are extended separately.

[0057] When the seed short peptide has a lasso structure, a search peptide segment is extended from the free end of the lasso structure as an alternative peptide segment; in a preferred embodiment, when there is an amino acid residue in the short peptide and the distance between it and two or more amino acid residues is less than a preset threshold, and one of the two amino acid residues is an N-terminal or C-terminal residue, the seed short peptide is judged to have a lasso structure; preferably, the preset threshold is 3 to 8 angstroms, preferably 4.5 angstroms;

[0058] (3) Screening of candidate peptides: For candidate peptides, a protein structure prediction algorithm is used to simulate the structure of the complex between the candidate peptides and the target protein; for candidate peptides whose binding ability to the target protein exceeds a preset threshold, the conformation of the complex between the candidate peptides and the target protein is obtained; the binding ability of the peptides to the protein is quantitatively evaluated using binding ability evaluation indicators, which include but are not limited to: the number of contact residue pairs, shape complementarity coefficient, interaction area, binding free energy, affinity ranking, and / or phage display technology screening results.

[0059] When the conformation of the candidate peptide segment forms a bicyclic peptide or a cyclic peptide connected end to end, the peptide segment extension is terminated and the candidate peptide segment is used as a candidate structure of a cyclic peptide drug or the step (4) is entered; the conformation of the candidate peptide segment forms a bicyclic peptide, specifically, the distance between the N-terminal and C-terminal residues of the short peptide and two or more non-terminal residues is less than a preset threshold; the conformation of the candidate peptide segment forms a cyclic peptide connected end to end, specifically, the distance between the N-terminal and C-terminal residues of the short peptide is less than a preset threshold; the preset threshold is 3 to 8 angstroms, preferably 4.5 angstroms.

[0060] When the conformation of the candidate peptide forms a linear peptide or a lasso peptide whose length exceeds a preset threshold, the peptide extension is terminated and the process proceeds to step (4); the preset threshold is 6 to 20 amino acid residues.

[0061] Otherwise, go to step (2).

[0062] For candidate peptides that form cyclic peptides or bicyclic peptides or linear cyclic peptides with a length exceeding a preset threshold, the following method is used to optimize the cyclization:

[0063] (4) Amino acid mutations are performed outside the key amino acid residues that form the ring to obtain new candidate peptides. When the conformation of the candidate peptides forms a head-to-tail cyclic peptide or lasso structure and the binding ability with the target protein exceeds a preset threshold, the candidate peptides are used as candidate structures of cyclic peptide drugs.

[0064] The key amino acid residues for ring formation refer to the amino acid residues at both ends of the ring structure of a polypeptide with a ring structure: for linear peptides, all residues are not key amino acids for ring formation, that is, all amino acid residues are outside the key amino acid residues for ring formation; for cyclic peptides, the amino acids at the N-terminus and C-terminus are key amino acids for ring formation; for bicyclic peptides or lasso peptides, the amino acids in contact with the middle at the N-terminus or the amino acids in contact with the middle at the C-terminus are key amino acids for ring formation.

[0065] The following are examples:

[0066] Developing cyclic peptides that can target specific targets is a huge challenge, and the de novo design of cyclic peptide drugs based on target structures is also a research direction with great potential. In recent years, the emergence of Alphafold2 and Alphafold-Multimer has greatly improved the accuracy and time of accurately predicting the structure of proteins and protein complexes. We will predict the structure of protein complexes based on artificial intelligence, without knowing the specific epitopes of the target protein in advance, and globally search for the complex structure of the target protein and the initial short peptide. Depending on the computing power, the initial short peptide can be 1 amino acid, 2 peptides, 3 peptides, 4 peptides or longer short peptides, and then extend the initial short peptide until a cyclic peptide conformation is formed or the length threshold is reached, and then screen out cyclic peptides that meet the preset conditions or directly transform linear peptides into cyclic peptides.

[0067] Example 1

[0068] The method for designing a cyclic peptide drug from scratch provided in this embodiment takes SORT1 protein as an example to design a cyclic peptide targeting SORT1, and specifically includes the following steps:

[0069] (1) Screening seed short peptides: For a short peptide of a preset length, a protein structure prediction algorithm is used to simulate the structure of the complex between the short peptide and the target protein; a short peptide that interacts with the target protein is used as a seed short peptide, thereby initializing the seed short peptide; the short peptide that interacts with the target protein is determined according to one of the following methods:

[0070] (1-1) A short peptide whose distances between all its atoms and the target protein are less than or equal to a preset threshold is determined as a short peptide that interacts with the target protein;

[0071] (1-2) using a variety of different models to simulate the complex structure of the short peptide and the target protein, using RMSD value and / or DockQ value as screening indexes to screen the complex structure, and determining the short peptide whose screening index exceeds a preset threshold as a short peptide that interacts with the target protein;

[0072] (1-3) The short peptide obtained by method (1-1) and / or (1-2) is a short peptide that interacts with the target protein.

[0073] The present embodiment is specifically as follows:

[0074] Obtain the primary sequence of SORT1, SEQ ID NO: 1; generate initial 3 peptides composed of all 20 natural amino acids, a total of 20 3 =8000.

[0075] This example uses a variety of different models to simulate the complex structure of short peptides and target proteins, and screens initial tripeptides that can interact with the target protein by judging that the atomic distance is less than 4.5 angstroms as amino acid interaction.

[0076] AlphaFold2.2.0 was used to generate the complex structures of all the initial three peptides with SORT1. Alphafold has 5 different models, with a total of 5 different complex structures of seed short peptides with target proteins.

[0077] The RMSD value or DockQ value of short peptides was used as the screening index for seed short peptides.

[0078] Align the target protein coordinates of the five complex structures in pairs: use the SVD singular value decomposition algorithm to find the best rotation and translation to minimize the RMSD value of the two structures. After aligning the structural coordinates of the target protein, only calculate the RMSD of the coordinates of the seed peptide. If the RMSD of the seed peptide is too large, it means that the docking predictions of different models are too different and the results are unreliable.

[0079] The pairwise DockQ values ​​of the five complex coordinates were used for screening, and short peptides with pairwise DockQ values ​​greater than 0.67 were selected as seed short peptides.

[0080] The DockQ value is calculated as follows:

[0081]

[0082] DockQ is mainly composed of three items, Fna (the ratio of decoy contact residues to native contact residues), LRMSD (root mean square deviation of ligand backbone atoms after alignment of receptors) and iIRMS (root mean square deviation of backbone atoms after alignment of contact interfaces), where LRMSD is the RMSD of the seed peptide mentioned above. It can be seen that DockQ is a more comprehensive evaluation parameter than the LRMSD of the seed peptide. Refer to Basu S, Wallner B. DockQ: a quality measure for protein-protein docking models [J]. PloS one, 2016, 11 (8): e0161879. Calculate the average DockQ value, mDockQ value.

[0083] According to the mDockQ value sorting, the 25 optimal (maximum mDockQ value) initial tripeptides are selected as anchoring initial tripeptides. This embodiment has no requirements for specific domains. If a specific domain is required, the first 25 initial tripeptides are preferred among the initial tripeptides targeting the specific domain.

[0084] (2) Peptide extension: Extend a search peptide at the N-terminus and / or C-terminus of the seed short peptide according to a preset step length as a candidate peptide, and proceed to step (3); the amino acid residues of the search peptide are selected from 20 natural amino acids, and the step length is 1. Specifically:

[0085] When the seed short peptide has a linear structure, the N-terminus and the C-terminus are extended respectively. In this embodiment, 1 amino acid residue is extended to the N-terminus and the C-terminus of 25 short peptides respectively, selected from 20 natural amino acids, a total of 25*20*2=1000.

[0086] When the seed short peptide has a lasso structure, a search peptide segment is extended from the free end of the lasso structure as an alternative peptide segment;

[0087] (3) Screening of candidate peptides: For candidate peptides, a protein structure prediction algorithm is used to simulate the complex structure of the candidate peptides with the target protein; for candidate peptides whose binding ability with the target protein exceeds a preset threshold, the conformation of the candidate peptides in the complex structure with the target protein is obtained; this embodiment is specifically as follows:

[0088] AlphaFold2.2.0 was used to generate the complex structures of all candidate peptides and SORT1. Based on the different complex structures generated by the five AlphaFold models, one of the models was used as a reference to calculate the DockQ values ​​of the other models and this model, and the average DockQ value, i.e., the mDockQ value, was calculated. The candidate peptides with mDockQ>0.5 were screened for structure acquisition:

[0089] When the conformation of the candidate peptide segment forms a bicyclic peptide or a head-to-tail cyclic peptide, the peptide segment extension is terminated to select the candidate peptide segment as a candidate structure of a cyclic peptide drug; the conformation of the candidate peptide segment forms a bicyclic peptide, specifically referring to the N-terminal and C-terminal residues of the short peptide being less than 4.5 angstroms apart from two or more non-terminal residues; the conformation of the candidate peptide segment forms a head-to-tail cyclic peptide, specifically referring to: the distance between the N-terminal and C-terminal residues of the short peptide is less than 4.5 angstroms.

[0090] When the conformation of the candidate peptide forms a linear peptide with a length of more than 16, the peptide extension is terminated and the process proceeds to step (4); otherwise, the process proceeds to step (2).

[0091] In steps (2) and (3), the method for determining the structure of the peptide is as follows: the distance between each amino acid and its non-adjacent residues (at least 2 residues in between) is: if the distance between the N-terminal or C-terminal amino acid and the middle amino acid is less than 4.5 angstroms, it is considered to contain a lasso peptide; if the distance between the N-terminal amino acid and the C-terminal amino acid is less than 4.5 angstroms, it is considered to be a cyclic peptide; if the distance between the N-terminal amino acid and the middle amino acid is less than 4.5 angstroms, and the distance between the C-terminal amino acid and another middle amino acid is less than 4.5 angstroms, it is considered to be a bicyclic peptide.

[0092] The binding ability of the peptide segment to the protein is quantitatively evaluated using binding ability evaluation indicators, which include but are not limited to: the number of contact residue pairs, shape complementarity coefficient, interaction area, binding free energy, affinity ranking, and / or phage display technology screening results. The binding ability evaluation method of the peptide segment to the protein in this embodiment is as follows:

[0093] Calculate the number of interaction residues N between the peptide and SORT1 protein, sort the peptides from large to small according to the number of interaction residues N, select the top 100 short peptides, and calculate the C α Atomic distance D, sorted from small to large according to D, and the first 25 short peptides were selected.

[0094] Finally, we screened out a cyclic peptide with side chain-side chain connection, SEQ ID NO: 2, which can be connected by esterification reaction between the carboxyl group of the N-terminal aspartic acid side chain and the hydroxyl group of the C-terminal serine ( Figure 4 ), and the cyclic peptide molecule binds very tightly to SORT1 ( Figure 5 ).

[0095] For linear cyclic peptide drug candidate structures with a length of more than 16 amino acid residues, the cyclization optimization was performed as follows:

[0096] (4) In addition to the key amino acid residues in the ring, amino acid mutations are performed to obtain new candidate peptides. When the conformation of the candidate peptides forms a head-to-tail cyclic peptide or lasso structure and the binding ability with the target protein exceeds a preset threshold, the candidate peptides are used as candidate structures of cyclic peptide drugs. In this embodiment, the key amino acid residues in the ring are the N-terminal and C-terminal residues. Specifically:

[0097] For SEQ ID NO:2, except for the N-terminal and C-terminal residues, other residues were mutated, inserted, or deleted for 20 natural amino acids. AlphaFold2.2.0 was used to generate the complex structure of all the short peptides and SORT1 in the previous step, and the C distance from the N-terminal to the C-terminal of the short peptide was calculated. α Atomic distance D, select short peptides with D less than 4.5 angstroms.

[0098] Finally, we screened out 5 short peptides that could form a ring.

[0099] SEQ ID NO:3, such as Figure 6 As shown in the figure, the aspartic acid at the N-terminus is very close to the serine at the C-terminus, and the distance between the Cα carbon atoms is only 3.8 angstroms. We can modify L-aspartic acid into D-aspartic acid, make the amino group of aspartic acid closer to the carboxyl group of serine, and form an amide bond between the carboxyl group and the amino group to synthesize a head-to-tail connected cyclic peptide.

[0100] SEQ ID NO:4, such as Figure 7As shown in the figure, the aspartic acid at the N-terminus is very close to the serine at the C-terminus, and the distance between the Cα carbon atoms is only 3.9 angstroms. We can modify L-aspartic acid into D-aspartic acid, make the amino group of aspartic acid closer to the carboxyl group of serine, and form an amide bond between the carboxyl group and the amino group to synthesize a head-to-tail connected cyclic peptide.

[0101] SEQ ID NO:5, such as Figure 8 As shown, the aspartic acid at the N-terminus is relatively close to the serine at the C-terminus, with the Cα carbon atom distance of 4.4 angstroms. The amino group of aspartic acid and the carboxyl group of serine are on the same side. We can directly form an amide bond between the carboxyl group and the amino group to synthesize a head-to-tail linked cyclic peptide.

[0102] SEQ ID NO:6, such as Fig. 9 As shown, the carboxyl group of the aspartic acid side chain at the N-terminus is relatively close to the hydroxyl group of the serine at the C-terminus. We can form a side chain-connected cyclic peptide through an esterification reaction between the carboxyl group and the hydroxyl group.

[0103] SEQ ID NO:7, such as Fig.10 As shown, the terminal carboxyl group of aspartic acid at the N-terminus is very close to the terminal amino group of serine at the C-terminus. We can directly form an amide bond between the carboxyl group and the amino group to synthesize a head-to-tail linked cyclic peptide.

[0104] SEQ ID NO:8, such as Fig.11 As shown, the aspartic acid amino group at the N-terminus and the carboxyl group of the serine at the C-terminus are on opposite sides. We can replace L-aspartic acid with D-aspartic acid or replace L-serine with D-serine, and directly form an amide bond between the carboxyl group and the amino group to synthesize a head-to-tail linked cyclic peptide.

[0105] The following is Example 2:

[0106] The de novo design method of cyclic peptide drugs provided in this embodiment takes PD1 protein as an example to design a cyclic peptide targeting a specific hotspot residue of PD1 to achieve the purpose of blocking the binding of PDL1 to PD1. Specifically, the following steps are included:

[0107] (1) Screening seed short peptides: For short peptides of preset length, a protein structure prediction algorithm is used to simulate the complex structure between the short peptide and the target protein; a short peptide whose distance between all atoms and the target protein is less than or equal to a preset threshold is used as a seed short peptide, thereby initializing the seed short peptide; specifically:

[0108] Obtain the main sequence of PD1, SEQ ID NO:9; obtain the hotspot residues for the binding of PD1 and PDL1; generate the initial 3 peptides composed of all 20 natural amino acids, a total of 20 3 =8000.

[0109] This example uses a variety of different models to simulate the complex structure of short peptides and target proteins, and judges that the atomic distance is less than 4.5 angstroms as amino acid interaction, and screens initial tripeptides that can interact with the hot spot residues of the target protein.

[0110] AlphaFold2.2.0 was used to generate the complex structures of all the initial three peptides with PD1. Alphafold has 5 different models, with a total of 5 different complex structures of seed short peptides with target proteins.

[0111] The RMSD value or DockQ value of short peptides was used as the screening index for seed short peptides.

[0112] The target protein coordinates of the five complex structures are aligned in pairs (SVD singular value decomposition algorithm is used to find the best rotation and translation to minimize the RMSD value of the two structures). After aligning the structural coordinates of the target protein, only the RMSD of the coordinates of the seed peptide is calculated. If the RMSD of the seed peptide is too large, it means that the docking predictions of different models are too different and the results are unreliable.

[0113] The pairwise DockQ values ​​of the five complex coordinates were used for screening, and seed short peptides with pairwise DockQ values ​​greater than 0.67 were preferred.

[0114] The DockQ value is calculated as follows.

[0115]

[0116] DockQ is mainly composed of three items: Fnat (ratio of decoy contact residues to native contact residues), LRMSD (root mean square deviation of ligand backbone atoms after alignment with receptors) and IRMS (root mean square deviation of backbone atoms after alignment with contact interfaces). RMSD is the RMSD of the seed peptide mentioned above. It can be seen that DockQ is a more comprehensive evaluation parameter than the LRMSD of the seed peptide. Basu S, Wallner B. DockQ: a quality measure for protein-protein docking models [J]. PloSone, 2016, 11 (8): e0161879. Calculate the average DockQ value, mDockQ value.

[0117] The initial tripeptides that can target the above specific hotspot residues were screened, and then sorted according to the mDockQ value, and the best 25 initial tripeptides were selected as the anchoring initial tripeptides.

[0118] (2) Peptide extension: Extend a search peptide at the N-terminus and / or C-terminus of the seed short peptide according to a preset step length as a candidate peptide, and proceed to step (3); the amino acid residues of the search peptide are selected from 20 natural amino acids, and the step length is 1. Specifically:

[0119] When the seed short peptide has a linear structure, the N-terminus and the C-terminus are extended respectively. In this embodiment, 20 natural amino acids are first extended to the N-terminus and the C-terminus of 25 short peptides respectively, for a total of 25*20*2=1000.

[0120] When the seed short peptide has a lasso structure, a search peptide segment is extended from the free end of the lasso structure as an alternative peptide segment;

[0121] (3) Screening of candidate peptides: For candidate peptides, a protein structure prediction algorithm is used to simulate the complex structure of the candidate peptides with the target protein; for candidate peptides whose binding ability with the target protein exceeds a preset threshold, the conformation of the candidate peptides in the complex structure with the target protein is obtained; this embodiment is specifically as follows:

[0122] AlphaFod2.2.0 was used to generate the complex structures of all candidate peptides and PD1. Based on the different complex structures generated by the five AlphaFold models, one of the models was used as a reference to calculate the DockQ values ​​of the other models and this model, and the average DockQ value, i.e., the mDockQ value, was calculated. The candidate peptides with mDockQ>0.5 were screened for structure acquisition:

[0123] When the conformation of the candidate peptide segment forms a bicyclic peptide or a cyclic peptide connected end to end, the peptide segment extension is terminated and the candidate peptide segment is used as a candidate structure of a cyclic peptide drug or the step (4) is entered; the conformation of the candidate peptide segment forms a bicyclic peptide, specifically, the distance between the N-terminal and C-terminal residues of the short peptide and two or more non-terminal residues is less than 4.5 angstroms; the conformation of the candidate peptide segment forms a cyclic peptide connected end to end, specifically, the distance between the N-terminal and C-terminal residues of the short peptide is less than 4.5 angstroms.

[0124] When the conformation of the candidate peptide forms a lasso peptide with a length of more than 16, the peptide extension is terminated and the process proceeds to step (4); otherwise, the process proceeds to step (2).

[0125] In steps (2) and (3), the method for determining the structure of the peptide is as follows: the distance between each amino acid and its non-adjacent residues (at least 2 residues in between) is: if the distance between the N-terminal or C-terminal amino acid and the middle amino acid is less than 4.5 angstroms, it is considered to contain a lasso peptide; if the distance between the N-terminal amino acid and the C-terminal amino acid is less than 4.5 angstroms, it is considered to be a cyclic peptide; if the distance between the N-terminal amino acid and the middle amino acid is less than 4.5 angstroms, and the distance between the C-terminal amino acid and another middle amino acid is less than 4.5 angstroms, it is considered to be a bicyclic peptide.

[0126] The binding ability of the peptide segment to the protein is quantitatively evaluated using binding ability evaluation indicators, which include but are not limited to: the number of contact residue pairs, shape complementarity coefficient, interaction area, binding free energy, affinity ranking, and / or phage display technology screening results. The binding ability evaluation method of the peptide segment to the protein in this embodiment is as follows:

[0127] Calculate the number of interaction residues N' between the peptide and the hotspot residues of PD1 protein, sort the interaction residues from large to small according to the number of interaction residues N', select the first 100 short peptides, and calculate the C from the N-terminus to the C-terminus of the short peptides. α Atomic distance D, sorted from small to large according to D, and the first 25 short peptides were selected.

[0128] Finally, we screened out a cyclic peptide with the main chain connected from head to tail, SEQ ID NO: 10, which can be connected to the free amino group of the N-terminal aspartic acid and the free carboxyl group of the C-terminal leucine through a peptide bond ( Fig.12 ), and the cyclic peptide molecule binds to the hotspot residues of PD1 to achieve the purpose of blocking the binding of PD1 and PDL1 ( Fig.13 ).

[0129] (4) Amino acid mutations are performed outside the key amino acid residues that form the ring to obtain new candidate peptides. When the conformation of the candidate peptides forms a head-to-tail cyclic peptide or lasso structure and the binding ability with the target protein exceeds a preset threshold, the candidate peptides are used as candidate structures of cyclic peptide drugs.

[0130] The key amino acid residues for ring formation refer to the amino acid residues at both ends of the ring structure of a polypeptide with a ring structure: for linear peptides, all residues are not key amino acids for ring formation, that is, all amino acid residues are outside the key amino acid residues for ring formation; for cyclic peptides, the amino acids at the N-terminus and C-terminus are key amino acids for ring formation; for bicyclic peptides or lasso peptides, the amino acids in contact with the middle at the N-terminus or the amino acids in contact with the middle at the C-terminus are key amino acids for ring formation.

[0131] In this example, no linear peptides with a conformation length exceeding 16 were found.

[0132] It will be easily understood by those skilled in the art that 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 and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for de novo design of cyclic peptide drugs, characterized in that: The following steps are involved: (1) Screening seed short peptides: For a short peptide of a preset length, a protein structure prediction algorithm is used to simulate the structure of the complex between the short peptide and the target protein; a short peptide that interacts with the target protein is used as a seed short peptide, thereby initializing the seed short peptide; the short peptide that interacts with the target protein is determined according to one of the following methods: (1-1) A short peptide whose distances between all its atoms and the target protein are less than or equal to a preset threshold is determined as a short peptide that interacts with the target protein; (1-2) using a variety of different models to simulate the complex structure of the short peptide and the target protein, using RMSD value and / or DockQ value as screening indexes to screen the complex structure, and determining the short peptide whose screening index exceeds a preset threshold as a short peptide that interacts with the target protein; (1-3) The short peptide obtained by method (1-1) and / or (1-2) is a short peptide that interacts with the target protein; (2) Peptide extension: Extend a search peptide at the N-terminus and / or C-terminus of the seed short peptide according to a preset step length as a candidate peptide, and proceed to step (3); (3) Screening of candidate peptides: For candidate peptides, a protein structure prediction algorithm is used to simulate the complex structure of the candidate peptides with the target protein; for candidate peptides whose binding ability with the target protein exceeds a preset threshold, the conformation of the candidate peptides in the complex structure with the target protein is obtained; When the conformation of the candidate peptide segment forms a bicyclic peptide or a head-to-tail cyclic peptide, the peptide segment extension is terminated to use the candidate peptide segment as a candidate structure of a cyclic peptide drug; When the conformation of the candidate peptide forms a linear peptide whose length exceeds a preset threshold, the peptide extension is terminated; Otherwise, go to step (2).

2. The method for de novo design of cyclic peptide drugs according to claim 1, characterized in that: In step (3), when the conformation of the candidate peptide segment forms a bicyclic peptide, a cyclic peptide connected end to end, or when the conformation of the candidate peptide segment forms a linear peptide whose length exceeds a preset threshold, the peptide segment extension is terminated and the process proceeds to step (4); for the candidate peptide segment whose conformation forms a bicyclic peptide or a cyclic peptide connected end to end, or for the linear cyclic peptide drug candidate structure whose length exceeds a preset threshold, the cyclization optimization is performed according to the following method: (4) Amino acid mutations are performed outside the key amino acid residues that form the ring to obtain new candidate peptides. When the conformation of the candidate peptides forms a head-to-tail cyclic peptide or lasso structure and the binding ability with the target protein exceeds a preset threshold, the candidate peptides are used as candidate structures of cyclic peptide drugs.

3. The method for de novo design of cyclic peptide drugs according to claim 1 or 2, characterized in that: For a target protein with a specific structural domain, a short peptide anchored to a preset epitope is used as a seed peptide.

4. The method for de novo design of cyclic peptide drugs according to claim 1 or 2, characterized in that: Screening of the seed short peptides: A short peptide with evaluation indexes within a preset range is used as a seed short peptide, wherein the evaluation indexes include the number of contact residue pairs, shape complementarity coefficient, interaction area, binding free energy, or affinity.

5. The method for de novo design of cyclic peptide drugs according to claim 1 or 2, characterized in that: In step (2), the amino acid residues of the searched peptide are selected from 20 natural amino acids, and the step length is preferably 1; When the seed short peptide has a linear structure, it is preferred to extend the N-terminus and the C-terminus respectively; When the seed short peptide has a lasso structure, a search peptide segment is extended from the free end of the lasso structure as an alternative peptide segment.

6. The method for de novo design of cyclic peptide drugs according to claim 5, characterized in that: When there is an amino acid residue in the short peptide and the distance between it and two or more amino acid residues is less than a preset threshold, and one of the two amino acid residues is the N-terminal or C-terminal residue, the seed short peptide is judged to have a lasso structure; preferably, the preset threshold is 4.5 angstroms.

7. The method for de novo design of cyclic peptide drugs according to claim 1 or 2, characterized in that: In step (3), the binding ability of the peptide segment to the protein is quantitatively evaluated using binding ability evaluation indicators, which include but are not limited to: the number of contact residue pairs, shape complementarity coefficient, interaction area, binding free energy, affinity ranking, and / or phage display technology screening results.

8. The method for de novo design of cyclic peptide drugs according to claim 1 or 2, characterized in that: The conformation of the candidate peptide segment forms a bicyclic peptide, specifically referring to that the distance between the N-terminal and C-terminal residues of the short peptide and two or more non-terminal residues is less than a preset threshold; the conformation of the candidate peptide segment forms a head-to-tail cyclic peptide, specifically referring to that the distance between the N-terminal and C-terminal residues of the short peptide is less than a preset threshold; the preset threshold is 4.5 angstroms.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method for de novo design of cyclic peptide drugs as claimed in any one of claims 1 to 8 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for de novo design of cyclic peptide drugs according to any one of claims 1 to 8 are implemented.

Images