A method for optimizing the structure of substituents in tetrahydrofuran-based NaV1.8 inhibitors based on reinforcement learning and inhibitors

Abstract

The present disclosure provides a method for optimizing a substituent structure of a tetrahydrofuran-based NaV1.8 inhibitor based on reinforcement learning, and the inhibitor, the method comprising: obtaining a receptor space model of the NaV1.8 inhibitor and a chemical group to be updated in an initial substituent structure; performing topological connection, group deletion or group replacement on the chemical group to obtain an updated substituent structure; obtaining a plurality of conformations by changing a spatial pose of a rotatable single bond in the updated substituent structure under the spatial constraint of the receptor space model, and obtaining a target evaluation score based on the conformation adaptability of each conformation to the receptor space model, the metabolic stability of each conformation, and the hydrophilic characteristics corresponding to the updated substituent structure, respectively; updating the policy parameters of the generation model based on the target evaluation score until the target evaluation score is greater than or equal to a preset threshold, and outputting a target substituent structure. The present disclosure can realize the synergistic optimization of the substituent adaptability, metabolic stability and hydrophilic-hydrophobic balance.

Description

Technical Field

[0001] This disclosure relates to the field of computational chemistry, and in particular to a method for optimizing the substituent structure in a tetrahydrofuran-based NaV1.8 inhibitor based on reinforcement learning, and the inhibitor thereof. Background Technology

[0002] Voltage-gated sodium channels (NaV1.8) are important sodium channel subtypes specifically expressed in peripheral sensory neurons, playing a crucial role in the generation and transmission of pain signals. Currently, when modifying the structure of substituted tetrahydrofuran-based NaV1.8 inhibitors, existing computer-aided drug design and generative artificial intelligence technologies are typically used to directly output candidate molecules or substituent structures.

[0003] However, related technologies often rely on docking scoring of receptor pockets and use static two-dimensional topological parameters as a filtering basis for physicochemical properties. Although the structure calculated based on static parameters performs well in theoretical models or single static indicators, it is difficult to balance target binding capacity and metabolic stability in the actual in vivo environment. This leads to practical problems such as short half-life or poor tissue distribution in the later stages of the obtained structure, making it difficult to directly guide the physical development of new drug compound structures. Summary of the Invention

[0004] To overcome the problems existing in related technologies, this disclosure provides a method for optimizing the substituent structure in tetrahydrofuran-based NaV1.8 inhibitors based on reinforcement learning, as well as the inhibitor itself. This method can achieve synergistic optimization of substituent adaptability, metabolic stability, and hydrophilicity-hydrophobicity balance based on the conformational adaptability, metabolic stability, and hydrophilicity-hydrophobicity balance of the substituent structure under different spatial orientations. This overcomes drug-like defects such as short half-life or poor tissue distribution in the output structure, enabling the output target substituent structure to maintain high target binding activity while possessing excellent anti-metabolic degradation performance and tissue distribution characteristics, thereby improving the success rate of new drug compound development.

[0005] According to a first aspect of the present disclosure, a method for optimizing the substituent structure in a tetrahydrofuran-based NaV1.8 inhibitor based on reinforcement learning is provided, the method comprising: Obtain the receptor spatial model of the NaV1.8 inhibitor and the chemical groups to be updated in the initial substituent structure; Based on a pre-constructed generative model, structural evolution instructions are determined, and topological connections, group deletions, or group substitutions are performed on the chemical groups according to the structural evolution instructions to obtain updated substituent structures; Under the spatial constraints of the receptor spatial model, multiple conformations are obtained by changing the spatial orientation of the rotatable single bonds in the updated substituent structure. The target evaluation score is obtained based on the conformational fit evaluation value between each conformation and the receptor spatial model, the metabolic stability evaluation value of each conformation, and the balance evaluation value between the hydrophilicity and hydrophobicity of each conformation. The strategy parameters of the generative model are updated based on the target evaluation score, and new chemical groups to be updated are determined according to the updated substituent structure. The process then returns to the step of executing the structure evolution instruction based on the pre-built generative model until the target evaluation score is greater than or equal to a preset threshold, at which point the target substituent structure is output.

[0006] In one exemplary embodiment of this disclosure, the NaV1.8 inhibitor is a tetrahydrofuran-based NaV1.8 voltage-gated channel inhibitor. Obtaining the chemical groups to be updated in the initial substituent structure of the NaV1.8 inhibitor includes: Obtain the drug skeleton structure of the NaV1.8 inhibitor, wherein the drug skeleton structure comprises a tetrahydrofuran ring, and aromatic groups and pyridine groups linked by amide bonds are attached to different sites of the tetrahydrofuran ring. The initial group at a predetermined site on the aromatic group in the drug skeleton structure is identified as the chemical group to be updated in the initial substituent structure.

[0007] In one exemplary embodiment of this disclosure, the step of performing topological linking, group deletion, or group substitution on the chemical groups according to the structural evolution instructions to obtain an updated substituent structure includes: When performing the topological connection, the target group is obtained from the pre-constructed group data set, and the three-dimensional growth vector field of the chemical group to be updated under the constraints of the acceptor space model is obtained; the spatially permissible extension region is determined based on the three-dimensional growth vector field, and the target group and the chemical group to be updated are structurally connected based on the matched chemical reaction template and the spatially permissible extension region. When performing the group deletion, the target group that contributes the least to the target evaluation score in the initial substituent structure is removed; When performing the group replacement, a three-dimensional growth vector is determined based on the three-dimensional growth vector field at the connection site of the chemical group to be updated, and the chemical group to be updated is replaced with the target group based on the matched chemical reaction template and the three-dimensional growth vector. The group data set includes hydroxyl, ether bond, methyleneoxy, methoxy and ethyl groups.

[0008] In one exemplary embodiment of this disclosure, obtaining the target evaluation score based on the conformational fit assessment value of each conformation to the receptor spatial model, the metabolic stability assessment value of each conformation, and the balance assessment value between the hydrophilicity and hydrophobicity corresponding to each conformation includes: The potential energy value of each conformation is calculated based on the three-dimensional spatial coordinates of each constituent atom in each conformation and the preset force field parameters. The lowest potential energy value in all the conformations is determined, and a weighting coefficient for each conformation is determined based on the difference between the potential energy value of each conformation and the lowest potential energy value. The smaller the difference, the larger the corresponding weighting coefficient. Using the weighting coefficients of each conformation, the conformation fit assessment value, metabolic stability assessment value, and balance assessment value of each conformation are weighted to obtain the conformation fit score, metabolic stability score, and balance score corresponding to the updated substituent structure. The target evaluation score is obtained by weighting the conformational fit score, the metabolic stability score, and the balance score.

[0009] In one exemplary embodiment of this disclosure, the method includes determining a conformational fit evaluation value between each of the conformations and the receptor spatial model: Obtain the preset spatial boundary of the updated substituent structure within the receptor spatial model; For each of the aforementioned conformations, the local free volume of the updated substituent structure within the spatial boundary is calculated, as well as the spatial repulsion volume between each constituent atom in the updated substituent structure and the acceptor atom in the acceptor spatial model. Based on the local free volume and the spatial repulsion volume, a conformational fit evaluation value is determined for each of the conformations.

[0010] In one exemplary embodiment of this disclosure, the method includes determining a metabolic stability assessment value for each of the conformations: Identify metabolically vulnerable atoms in the updated substituent structure; For each of the aforementioned conformations, the spatial solvent-accessible surface area of ​​the metabolically vulnerable atom is calculated based on the three-dimensional coordinates of the metabolically vulnerable atom in the corresponding spatial orientation. The metabolic stability assessment value for each conformation is determined based on the spatial solvent-accessible surface area; wherein the spatial solvent-accessible surface area is negatively correlated with the metabolic stability assessment value.

[0011] In one exemplary embodiment of this disclosure, the target substituent structure of the output is a linear flexible side chain containing an oxygen atom; the linear flexible side chain is one of methoxyethoxy, hydroxyethyloxy, methoxyethylethoxy, and hydroxyethylethoxy.

[0012] According to a second aspect of the present disclosure, a NaV1.8 inhibitor is provided, the NaV1.8 inhibitor comprising a drug core framework and a target substituent structure attached to the drug core framework; the target substituent structure is determined by the optimization method for the substituent structure in the NaV1.8 inhibitor described in any of the preceding claims.

[0013] In one exemplary embodiment of this disclosure, the core pharmaceutical skeleton of the NaV1.8 inhibitor is a tetrahydrofuran ring, and the aromatic group on the tetrahydrofuran ring is connected to the target substituent structure, which is one of methoxyethoxy, hydroxyethyloxy, methoxyethylethoxy, and hydroxyethylethoxy.

[0014] This disclosure provides a reinforcement learning-based method for optimizing substituent structures in tetrahydrofuran-based NaV1.8 inhibitors, along with the corresponding inhibitors. The method uses a pre-constructed generative model to determine structural evolution instructions, performing topological connections, group deletions, or group substitutions on the chemical groups to be updated. This enables the targeted evolution and generation of candidate substituent structures based on a defined receptor space model. This method can more effectively explore local chemical spaces for specific targets, improving the efficiency of discovering high-quality substituent structures.

[0015] Furthermore, under the spatial constraints of the receptor spatial model, multiple conformations are obtained by changing the spatial orientation of rotatable single bonds in the updated substituent structure, and a comprehensive evaluation is performed based on the conformational fit, metabolic stability, and hydrophilicity-hydrophobicity balance of each conformation. Thus, the dynamic conformational distribution characteristics of flexible groups within the receptor environment are introduced into the evaluation process, ensuring that the final target evaluation score not only reflects the static topological properties of the molecule but also quantifies the contribution of spatial geometric shielding effects under different rotational orientations to metabolic stability, as well as the equilibrium state of polar groups during dynamic fluctuations. This multi-dimensional dynamic evaluation mechanism enables differentiated weighting of the performance of different spatial conformations at the computational level.

[0016] Furthermore, the strategy parameters of the generative model are iteratively updated based on the target evaluation score, and the target substituent structure is finally output. Thus, the algorithm achieves synergistic optimization of adaptability, metabolic stability, and hydrophilicity-hydrophobicity balance during the optimization process. This effectively reduces prediction bias caused by neglecting three-dimensional dynamic physical characteristics, ensuring that the final output substituent structure maintains both high target binding capacity and resistance to metabolic degradation in the receptor environment. This avoids the risk of low synthesis conversion rate and loss of research value due to distortion of static model indicators, thereby improving the success rate of new drug compound development.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this disclosure, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0019] Figure 1 This disclosure is a flowchart illustrating an exemplary embodiment of a method for optimizing the substituent structure in a tetrahydrofuran-based NaV1.8 inhibitor using reinforcement learning.

[0020] Figure 2 This is a flowchart illustrating a method for determining a target evaluation score according to an exemplary embodiment of the present disclosure.

[0021] Figure 3 This disclosure illustrates a general formula for a tetrahydrofuran-based NaV1.8 inhibitor based on reinforcement learning, according to an exemplary embodiment.

[0022] Figure 4 This is a schematic diagram of the synthetic reaction of a NaV1.8 inhibitor according to an exemplary embodiment of the present disclosure.

[0023] Figure 5 This is a schematic diagram of the synthetic reaction of another NaV1.8 inhibitor according to an exemplary embodiment of the present disclosure.

[0024] Figure 6 This is a statistical chart illustrating the analgesic efficacy evaluation of a NaV1.8 inhibitor in a zebrafish pain model according to an exemplary embodiment of the present disclosure.

[0025] Figure 7 This is a statistical graph illustrating the analgesic efficacy evaluation of another NaV1.8 inhibitor in a zebrafish pain model according to an exemplary embodiment of this disclosure.

[0026] Figure 8This is a graph showing the detection results of real-time fluorescence quantification of NaV1.8 mRNA according to an exemplary embodiment of this disclosure.

[0027] Figure 9 This is a comparative image of immunofluorescence staining of NaV1.8 membrane protein expression cells according to an exemplary embodiment of this disclosure.

[0028] Figure 10 This is a diagram illustrating the concentration response of the NaV1.8 channel agonist according to an exemplary embodiment of this disclosure.

[0029] Figure 11 This is a graph illustrating the quantitative relationship between the inhibitory activity of suzetriline and its derivatives on NaV1.8 channels according to an exemplary embodiment of this disclosure.

[0030] Figure 12 This disclosure is a structural block diagram of an apparatus for optimizing the substituent structure in a tetrahydrofuran-based NaV1.8 inhibitor based on reinforcement learning, according to an exemplary embodiment.

[0031] Figure 13 This is a hardware structure diagram of a computer device shown in an embodiment of this disclosure. Detailed Implementation

[0032] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0033] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The singular forms “a,” “the,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0034] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0035] The embodiments of this disclosure will now be described in detail.

[0036] like Figure 1 As shown, Figure 1 This disclosure is a flowchart illustrating an exemplary embodiment of a method for optimizing the substituent structure in a tetrahydrofuran-based NaV1.8 inhibitor using reinforcement learning, comprising the following steps: Step 101: Obtain the receptor spatial model of the NaV1.8 inhibitor and the chemical groups to be updated in the initial substituent structure.

[0037] Step 102: Determine the structure evolution instructions based on the pre-built generative model, and perform topological connection, group deletion or group substitution on the chemical groups according to the structure evolution instructions to obtain the updated substituent structure.

[0038] Step 103: Under the spatial constraints of the receptor spatial model, multiple conformations are obtained by changing the spatial orientation of the rotatable single bonds in the updated substituent structure. The target evaluation score is obtained based on the conformational fit evaluation value between each conformation and the receptor spatial model, the metabolic stability evaluation value of each conformation, and the balance evaluation value between the hydrophilicity and hydrophobicity of each conformation.

[0039] Step 104: Update the strategy parameters of the generative model based on the target evaluation score, determine the new chemical groups to be updated according to the updated substituent structure, and return to the step of executing the structure evolution instruction based on the pre-built generative model until the target evaluation score is greater than or equal to the preset threshold, and output the target substituent structure.

[0040] The method for optimizing the substituent structure in tetrahydrofuran-based NaV1.8 inhibitors based on reinforcement learning, as provided in the exemplary embodiments of this disclosure, performs topological connections, deletions, or replacements on the chemical groups to be updated through a pre-built generation model. This enables the automatic evolution of the substituent structure based on the receptor space model, which, compared to static library screening, can broaden the chemical search space and improve the generation efficiency of high-potential candidate structures.

[0041] Furthermore, by altering the spatial orientation of the rotatable single bond under receptor spatial constraints, multiple conformations were obtained, and a comprehensive evaluation was conducted based on the conformational adaptability, metabolic stability, and hydrophilicity-hydrophobicity balance of each conformation. Thus, the dynamic conformational distribution characteristics of flexible groups in a real binding environment were introduced into the evaluation process, enabling the final target evaluation score to accurately quantify the impact of spatial geometric shielding effects under different rotational orientations on metabolic performance and to truly reflect the equilibrium state of polar groups in dynamic fluctuations.

[0042] Therefore, by iteratively updating the strategy parameters of the generative model based on the target evaluation score and outputting the target substituent structure, the generative model incorporates target binding force and physicochemical properties during the optimization process, avoiding computational prediction biases caused by neglecting three-dimensional dynamic physical characteristics. The target substituent structure output by this method can balance anti-metabolic degradation performance and tissue distribution characteristics, thus avoiding the technical defects of short half-life or poor drug-likeness in the output structure. This improves the conversion success rate of the output structure in later synthesis verification, providing a drug-value-based guiding basis for the physical development of new drug compounds.

[0043] The following will provide a detailed description of a method for optimizing the substituent structure in a tetrahydrofuran-based NaV1.8 inhibitor based on reinforcement learning in this example embodiment.

[0044] Step 101: Obtain the receptor spatial model of the NaV1.8 inhibitor and the chemical groups to be updated in the initial substituent structure.

[0045] For receptor spatial models, the three-dimensional structure of the target protein can be obtained by experimental analytical methods (such as X-ray crystallography, cryo-electron microscopy, etc.), or it can be constructed based on homology modeling or pre-trained protein structure prediction models. This model digitally defines the three-dimensional geometry, surface charge distribution, and polar / nonpolar region distribution of the NaV1.8 inhibitor binding pocket, thus providing accurate spatial boundary conditions for subsequent algorithmic modification of single bond spatial orientation and physical feature evaluation.

[0046] In some example implementations, the NaV1.8 inhibitor is a tetrahydrofuran-based NaV1.8 voltage-gated channel inhibitor, which specifically acts on peripheral sensory neurons.

[0047] Accordingly, the pharmacoskeleton structure information of the NaV1.8 inhibitor was first obtained. The pharmacoskeleton structure contains a tetrahydrofuran ring, and aromatic groups and pyridine groups linked by amide bonds are respectively attached to different sites of the tetrahydrofuran ring; For example, in this drug skeleton structure, the tetrahydrofuran ring serves as the drug skeleton structure, guiding the entire molecule to anchor within the binding pocket of the NaV1.8 channel through its specific spatial configuration; the aromatic group attached thereto can form hydrophobic interactions or π-π stacking interactions with the hydrophobic residues in the receptor pocket; while the amide bond and pyridine group attached to another point further provide hydrogen bond donors or acceptors to enhance the polar binding interaction with the receptor amino acid residues.

[0048] Subsequently, based on the spatial topological connections of the drug skeleton structure, the chemical groups to be updated are identified and determined. Specifically, the initial groups at predetermined sites on the aromatic groups in the drug skeleton structure can be identified as the chemical groups to be updated in the initial substituent structure.

[0049] For example, in a specific application scenario, the preset site can be a site on the aromatic ring that is easily metabolically cleared, and the initial group at the site can be a methoxy group.

[0050] From a medicinal chemistry perspective, in actual physiological solvent environments, initial groups (such as short-chain methoxy groups) at specific predetermined sites on aromatic groups are often highly exposed to the catalytic active sites of metabolic enzymes such as cytochrome P450, posing a high risk of metabolic transformations such as demethylation. This leads to a shortened in vivo half-life and reduced drug-like properties. Therefore, by using an algorithm to determine the structure at this predetermined site (such as the methoxy group) as the initial substituent structure and using it as the starting point for subsequent structural evolution and strategy iteration, it is possible to ensure that the generated substituent structure achieves better metabolic stability and overall drug-like potential while preserving the drug skeleton structure.

[0051] Step 102: Determine the structure evolution instructions based on the pre-built generative model, and perform topological connection, group deletion or group substitution on the chemical groups according to the structure evolution instructions to obtain the updated substituent structure.

[0052] In the exemplary embodiments of this disclosure, the pre-built generative model is a deep learning model built based on reinforcement learning. This generative model can perform feature mapping within a preset action space based on the input initial substituent structure: for topological connection actions, each action in the action space is associated with the target group identifier and the corresponding connection site in the group data set; for group replacement actions, each action in the action space is associated with the chemical group identifier to be updated in the current substituent structure, the corresponding connection site, and the target group identifier to be replaced; for group deletion actions, each action in the action space is associated with the target group identifier to be removed in the current structure and its corresponding connection site, in order to perform local structural pruning. As an example, when a specific group in the current substituent structure is identified as contributing less than a preset negative threshold to the evaluation score (i.e., the group is determined to be a performance bottleneck, such as causing severe steric hindrance or physicochemical property penalties), the probability of deleting the corresponding group will be increased to proactively stop losses and make room for subsequent evolution. When a specific site contributes to activity but does not reach a local optimum, and a group with similar physicochemical properties that can bring higher evaluation score gains is identified in the group dataset, the probability of group replacement will be increased to achieve precise optimization of local binding energy without compromising the stability of the skeleton. When the evaluation score tends to stabilize, and the three-dimensional growth vector field shows that there is still unused space allowable for extension around the connection site, the probability of topological connection will be increased to perform spatial exploration into unknown cavity regions.

[0053] For topological connections, the probability distribution of the action is recalculated based on the spatial constraints of the receptor space model and the preset chemical reaction template. This involves simulating the physical occupancy of the groups associated with each action within the receptor space, performing masking filtering on actions that exceed the spatial constraints (i.e., cause steric interference) or do not meet the synthesis conditions (i.e., do not match the chemical reaction template), normalizing the probabilities of the remaining valid actions, and sampling based on the recalculated action probability distribution to lock the target action index. This target action index includes the topological connection action, the connection site, and the target group. For group replacement, the chemical group to be updated at the current connection site is identified, the physical occupancy after the target group replaces the chemical group to be updated is simulated, and the action probability distribution is recalculated based on the receptor space constraints and the chemical reaction template. Masking filtering, normalization, and sampling are then performed to lock the associated chemical groups to be updated, the connection site, and the target group. For group deletion, when the target evaluation score in the current structure decreases, or when the contribution of a certain group is identified as negative, the connection site between the group with the lowest contribution to the target evaluation score and the backbone is obtained, thereby generating a removal instruction for that target group.

[0054] In an exemplary embodiment of this disclosure, during topological joining, a target group is obtained from a pre-constructed group data set, and the target group is structurally joined to the chemical group to be updated based on the connection sites of the chemical group to be updated. Specifically, a three-dimensional growth vector field of the chemical group to be updated under the constraints of the acceptor spatial model can be obtained; a spatially permissible extension region is determined based on the three-dimensional growth vector field, and the three-dimensional growth vector field is truncated using this region. Within the truncated vector field, the optimal three-dimensional growth vector that best matches the pocket polarity and has the lowest energy is calculated; and based on the matched chemical reaction template, preset standard bonding geometry parameters are extracted. The obtained target group is then joined to the connection sites along the three-dimensional growth vector according to the standard bonding geometry parameters, completing the structural joining. This three-dimensional growth vector field is used to characterize the geometrical capacity of the acceptor spatial model in different spatial orientations, as well as the binding stability in specific vector directions.

[0055] For example, during the initial computational optimization phase, a three-dimensional growth vector field can be constructed at the connection sites of the chemical groups to be updated. By calculating the van der Waals radii of the acceptor residues, the forbidden regions occupied by atoms are identified, thereby defining the spatially permissible extension regions that do not produce atomic overlap. The spatially permissible extension regions are mapped onto the vector field as masks. If the physical path of a growth vector enters the forbidden region, the vector weight in that direction is forcibly cleared to zero, thereby blocking the generation of invalid conformations. Furthermore, a preset chemical reaction template library is called to obtain standard bonding geometry parameters. Based on the optimal growth direction displayed by the vector field, and under the hard constraint of the spatially permissible extension regions, the target group is precisely connected to the oxygen atom site based on the bonding standard parameters, completing the structural connection. Through the synergistic constraint of the three-dimensional vector field and the reaction template, the substituent structure finally generated in this embodiment is adapted to the acceptor space model of the NaV1.8 acceptor in three-dimensional space, avoiding steric collisions with surrounding residues, thus improving the synthesis success rate compared to randomly generated groups without constraints.

[0056] When performing group deletion, the contribution of each chemical group in the substituent structure to the target evaluation score is obtained, the target group with the lowest contribution among the chemical groups is identified, and the target group is cut off and removed from its connection site with the skeleton.

[0057] When performing the group replacement, a three-dimensional growth vector is determined based on the three-dimensional growth vector field at the connection site of the chemical group to be updated. Based on the matched chemical reaction template and the three-dimensional growth vector, the chemical group to be updated is replaced with the target group. For example, the target group can be obtained from a group dataset, and bonding geometry parameters (including but not limited to standard bond lengths, bond angles, and dihedral angle constraints) can be extracted from the matched chemical reaction template as the geometric boundary of the replacement action. The three-dimensional growth vector field at the connection site of the chemical group to be updated is obtained, and the target group is matched with the optimal three-dimensional growth vector in the three-dimensional growth vector field. Under the premise of satisfying the bonding geometry parameters, the dihedral rotation orientation of the newly added target group is adjusted to align the target group with the binding stability region within the acceptor space model. Under the constraint of the spatially permissible extension region, the chemical group to be updated is replaced with the orientation-calibrated target group, thereby achieving a precise geometric fit between the replaced structure and the acceptor space model.

[0058] In some example implementations, the aforementioned group dataset includes hydroxyl, ether, methyleneoxy, methoxy, and ethyl groups. For the structural optimization of tetrahydrofuran-based NaV1.8 inhibitors, the hydroxyl group can act as a hydrogen bond donor or acceptor, replacing the original methoxy group to enhance interaction with polar residues within the target pocket, while simultaneously altering the local hydrophilic / hydrophobic distribution; the ether provides a high degree of single-bond rotational freedom, endowing the generated side chains with the physical flexibility required for multi-conformal folding; the methyleneoxy group can serve as a basic carbon-oxygen chain segment building block, participating in the physical elongation of the chain through topological linkages; the methoxy group can serve as the terminal group of flexible side chains, providing a suitable hydrophobic volume and a stable inert methyl end; the ethyl group, as a flexible unit in the carbon chain, can connect with ethers, oxygen atoms, etc., to elongate the carbon chain and regulate lipophilicity. By repeatedly calling and combining these five basic segments using a reinforcement learning-based generative model, side chains with diverse structures and flexible characteristics, such as methoxy and ethoxy groups, can be efficiently assembled starting from the linkage sites of the aromatic ring.

[0059] Step 103: Under the spatial constraints of the receptor spatial model, multiple conformations are obtained by changing the spatial orientation of the rotatable single bonds in the updated substituent structure. The target evaluation score is obtained based on the conformational fit evaluation value between each conformation and the receptor spatial model, the metabolic stability evaluation value of each conformation, and the balance evaluation value between the hydrophilicity and hydrophobicity of each conformation.

[0060] In some example implementations, the method includes determining a conformational fit assessment value for each conformation to the receptor spatial model. Specifically, this can be done according to steps 1031-1033.

[0061] Step 1031: Obtain the preset spatial boundary of the updated substituent structure within the receptor spatial model.

[0062] For example, a predefined spatial boundary is used to define the reasonable three-dimensional physical containment range reserved for the specific substituent in the binding pocket of the receptor protein. For the side chain extending outward from the aromatic ring in the NaV1.8 inhibitor, a closed or semi-closed three-dimensional envelope can be defined by scanning the surface solvent-accessible contours of amino acid residues adjacent to the binding region of the side chain in the receptor model, such as hydrophobic residues or polar residues. This envelope physically determines the maximum spatial limit and geometric contour of the reasonable extension of the updated substituent structure within the receptor.

[0063] Step 1032: For each conformation, calculate the local free volume of the updated substituent structure within the spatial boundary, and the spatial repulsion volume between each constituent atom in the updated substituent structure and the acceptor atom in the acceptor spatial model.

[0064] For example, the calculation of the local free volume is used to evaluate the filling efficiency and range of motion of the conformation in the receptor pocket cavity without overstepping the boundaries. It can be quantified by dividing the space into three-dimensional meshes within a preset spatial boundary and integrating the remaining volume not occupied by atoms. The calculation of the spatial repulsion volume is used to evaluate the steric collision. It can be quantified by calculating the volume of the geometric overlap region between each constituent atom in the conformation and the receptor protein atom. The larger the overlap region volume, the more severe the physical conflict between the conformation and the receptor in the current three-dimensional posture.

[0065] Step 1033: Determine the conformational fit evaluation value for each conformation based on the local free volume and the spatial repulsion volume.

[0066] For example, a preset monotonically decreasing mapping function (such as a reciprocal function or a negative exponential function) can be applied to the local free volume to obtain the positive fit value for each conformation. The smaller the local free volume, the more closely the conformation fits the remaining space of the acceptor cavity within the preset spatial boundary, and the larger the corresponding positive fit value. Next, a preset steric hindrance penalty coefficient is multiplied by the spatial repulsion volume to obtain the negative penalty value for the corresponding conformation. The monotonically decreasing mapping function and the steric hindrance penalty coefficient can be determined according to the actual situation, and this disclosure does not impose any special limitations. The larger the spatial repulsion volume, the more severe the physical overlap and steric conflict between the constituent atoms and acceptor atoms in the conformation, and the larger the corresponding negative penalty value. Finally, the difference between the positive fit value and the negative penalty value is calculated, that is, the negative penalty value is subtracted from the positive fit value, and the calculated result is determined as the conformation fit evaluation value for each conformation.

[0067] Through the above calculations, the conformational fit evaluation value can accurately quantify the degree of geometric complementarity between the specific three-dimensional pose and the receptor binding pocket in physical space. This allows for the effective elimination of computational structures that are reasonable in two-dimensional topological connections but cannot form effective spatial complementarity with the receptor pocket in three-dimensional physical space during model iteration.

[0068] In some example implementations, the method includes determining a metabolic stability assessment value for each conformation. Specifically, this can be illustrated as in steps 1034-1036.

[0069] Step 1034 can identify metabolically vulnerable atoms in the updated substituent structure.

[0070] For example, metabolically vulnerable atoms refer to specific atomic sites that are easily catalyzed by metabolic enzymes in the physiological environment of the body, undergoing reactions such as oxidation, dealkylation, or cleavage. Based on preset metabolic warning structural features or an empirical metabolic reaction library, the two-dimensional topology of the updated substituent structure can be scanned to identify and locate specific metabolically vulnerable atoms.

[0071] In conjunction with the aforementioned optimized scenario of tetrahydrofuran-type NaV1.8 inhibitors, if the generated substituent is a flexible side chain containing an ether bond, such as methoxyethoxy, then the metabolically vulnerable atom can be an ether bond oxygen atom that is highly susceptible to dealkylation or a free carbon atom at the end of the side chain that is highly susceptible to oxidation.

[0072] Step 1035: For each conformation, calculate the spatial solvent-accessible surface area of ​​the metabolically vulnerable atom based on the three-dimensional coordinates of the metabolically vulnerable atom in the corresponding spatial orientation.

[0073] For example, the spatial solvent-accessible surface area is used to quantify the size of the physical contact area of ​​a specific atom exposed to an external solvent environment. In actual calculations, a probe ball of a predetermined radius can be used to simulate rolling on the three-dimensional outer surface of the molecule in the current conformation. By combining the three-dimensional spatial coordinates of the metabolically vulnerable atom in the current conformation with the spatial occlusion state of surrounding neighboring atoms, the area of ​​the trajectory traced by the center of the probe ball can be calculated, thus obtaining the corresponding spatial solvent-accessible surface area. This spatial solvent-accessible surface area reflects the actual exposed area of ​​the catalytic active center of the metabolic enzyme in the current conformation, which can approach and act on the vulnerable atom in three-dimensional physical space.

[0074] Step 1036: Determine the metabolic stability assessment value for each conformation based on the spatial solvent accessible surface area, wherein the spatial solvent accessible surface area is negatively correlated with the metabolic stability assessment value.

[0075] For example, a preset monotonically decreasing mapping function can be used to numerically map the spatial solvent-accessible surface area to obtain the metabolic stability assessment value for each conformation. This monotonically decreasing mapping function can be a linear inverse proportional function or a negative exponential decay function, etc., which can be determined according to the actual situation, and this disclosure does not impose any special limitations. The above negatively correlated mapping calculation logic reflects the spatial geometric shielding effect brought about by microscopic molecular folding: specifically, for the same substituent structure, if a certain three-dimensional conformation can hide metabolically vulnerable atoms inside the nonpolar carbon skeleton through single-bond torsion and chain segment folding, its corresponding spatial solvent-accessible surface area is small, indicating that this posture effectively blocks the approach of metabolic enzymes in spatial physics. After calculation by the monotonically decreasing mapping function, a larger metabolic stability assessment value is output; conversely, if vulnerable atoms are exposed on the outside, resulting in a larger spatial solvent-accessible surface area, a smaller metabolic stability assessment value is output.

[0076] Through the specific mapping calculation method described above, the degree of physical spatial shielding brought about by the dynamic flexible folding of molecules can be transformed into a standardized quantitative value, thereby accurately evaluating the anti-degradation performance of the conformation.

[0077] In some example implementations, the method further includes determining an evaluation value for the balance between hydrophilicity and hydrophobicity corresponding to each conformation. Specifically, this can be shown in steps 1037-1038.

[0078] Step 1037: For each conformation, calculate the corresponding surface area accessible to polar solvents and the surface area accessible to non-polar solvents.

[0079] For example, based on the electronegativity of each constituent atom in the conformation or a predetermined molecular force field atom type, the atoms in the updated substituent structure are divided into a set of polar atoms (e.g., including oxygen atoms, nitrogen atoms, and hydrogen atoms bonded to them) and a set of nonpolar atoms (e.g., including carbon atoms and halogen atoms). For each specific conformation, the area integral of the external exposed shell of the aforementioned polar and nonpolar atom sets in the current conformation is performed; the resulting polar solvent-accessible surface area quantifies the size of the exposed region capable of forming polar interactions such as hydrogen bonds with external water molecules in that conformation, while the nonpolar solvent-accessible surface area quantifies the size of the exposed region of the carbon skeleton exhibiting hydrophobic properties.

[0080] Step 1038: Based on the ratio of the surface area accessible to polar solvents to the surface area accessible to non-polar solvents for each conformation, determine the balance evaluation value between hydrophilicity and hydrophobicity for each conformation.

[0081] For example, this ratio characterizes the polar distribution of the substituent structure in the folded state of this conformation. In a real physiological environment, the polar distribution of a molecule must be maintained in a specific equilibrium state to balance water solubility and tissue penetration. Specifically: if the ratio is too small (exhibiting strong hydrophobicity), although it is beneficial for the molecule to penetrate the lipid bilayer or blood-brain barrier, it can easily lead to extremely poor water solubility of the drug in blood or intestinal fluid, and accumulation and toxicity in non-specific lipid tissues; conversely, if the ratio is too large (exhibiting strong hydrophilicity), although it can maintain good water solubility, the molecule will have difficulty penetrating the hydrophobic cell membrane barrier, resulting in severely limited tissue permeability. Specifically, a corresponding reference ratio range can be set based on the specific balance requirements of drug permeability and solubility at the target site, such as the tissue where NaV1.8 is located. This reference ratio range includes an upper threshold and a lower threshold, which can be determined based on actual conditions, and this disclosure does not impose any special limitations.

[0082] When the ratio is within the reference range, it indicates that the hydrophilic-hydrophobic distribution of the conformation has reached an ideal balance, and its deviation is determined to be zero. The corresponding balance evaluation value output is the highest preset value (e.g., 1.0). When the ratio is greater than the upper threshold, it is determined that the conformation exhibits excessively high hydrophilicity, limiting its permeability. The degree of deviation can be determined based on the absolute value of the difference between the ratio and the upper threshold, and the degree of deviation can be multiplied by a preset hydrophilicity penalty coefficient to obtain a hydrophilicity penalty term. Then, a preset monotonically decreasing mapping function, such as a negative exponential function, is used to convert the hydrophilicity penalty term into a standardized balance evaluation value. When the ratio is less than the lower threshold, it is determined that the conformation exhibits excessively low hydrophobicity, easily leading to excessive enrichment of toxicity. The absolute value of the difference between the lower threshold and the ratio can be used as the degree of deviation, and the degree of deviation can be multiplied by a preset hydrophobicity penalty coefficient to obtain a hydrophobicity penalty term. Then, the monotonically decreasing mapping function is used to convert the hydrophobicity penalty term into a standardized balance evaluation value. Since the toxicity risk from excessive hydrophobic accumulation in the in vivo environment is usually higher than that from insufficient water solubility, the hydrophobic penalty coefficient is greater than the hydrophilic penalty coefficient. The specific value can be determined based on the actual situation, and this disclosure does not impose any special limitations.

[0083] By using this asymmetric transformation method that distinguishes between hydrophilic and hydrophobic properties, the generated balance assessment value not only quantifies the degree to which the molecule deviates from the balance range, but also accurately integrates the biological risk weights of different deviation directions. This guides the generative model to accurately screen out the advantageous conformations that achieve dynamic physicochemical balance, thereby avoiding the drug-like risk of excessive accumulation of molecules in non-specific lipid tissues.

[0084] Since the conformational fit assessment, metabolic stability assessment, and equilibrium assessment values ​​originate from different physical properties, their original values ​​differ in physical dimensions and magnitudes. Therefore, by employing their respective monotonically decreasing mapping functions, the calculation results can be standardized and mapped to a preset numerical range (e.g., [0,1]). The parameters of each monotonically decreasing mapping function can be set for its respective physical quantity. This method ensures the physical rationality of weight allocation in subsequent weighted calculations and provides a scale-uniform, gradient-stable reward signal for the reinforcement learning iteration of the generative model.

[0085] In some example implementations, after obtaining the three independent evaluation values ​​for each conformation, a target evaluation score can be obtained based on the conformational fit evaluation value of each conformation to the receptor spatial model, the metabolic stability evaluation value of each conformation, and the balance score between hydrophilicity and hydrophobicity corresponding to the updated substituent structure. The specific method for determining this target evaluation score is as follows: Figure 2 Steps 201-204 are shown in the diagram.

[0086] Step 201: Calculate the potential energy value of each conformation based on the three-dimensional spatial coordinates of each constituent atom and the preset force field parameters.

[0087] For example, based on the three-dimensional spatial coordinates of each constituent atom in each conformation, geometric and topological parameters such as bond lengths, bond angles, and dihedral angles corresponding to each conformation can be obtained, and the potential energy value corresponding to each conformation can be calculated by calling preset force field parameters. Specifically, this potential energy value can be obtained by accumulating the structural stress potential energy generated inside the molecule due to the deviation of geometric and topological parameters from the equilibrium position, as well as the van der Waals repulsion potential energy and electrostatic interaction potential energy generated between non-bonded atoms due to the close spatial distance. This potential energy value can be used to characterize the energy distribution state of each conformation on the potential energy surface and the stability of its physical structure. The lower the potential energy value, the closer the spatial configuration is to the natural ground state and the higher the probability of its existence in a physiological environment. The preset force field parameters may include bonding parameters, van der Waals parameters, charge parameters, and atom type parameters, etc., which can be determined according to the actual situation, and this disclosure does not impose any special limitations.

[0088] Step 202: Determine the lowest potential energy value among all conformations, and determine the weighting coefficient of each conformation based on the difference between the potential energy value of each conformation and the lowest potential energy value.

[0089] For example, the lowest potential energy value represents the most stable ground state of the substituent under the current physical environment. The relative difference between the potential energy of each other conformation and this lowest potential energy is calculated. The smaller the difference, the higher the probability of the conformation appearing in the thermal fluctuations of actual physiological temperature, and therefore it can be assigned a larger weighting coefficient. Specifically, for the first... iEach configuration, and its corresponding weight coefficient The calculation formula is as follows: In the above formula, Indicates the first substituent structure corresponding to the substituent. i The potential energy value of each conformation; Indicates all N The lowest potential energy value in each conformation; Represents the Boltzmann constant; T This indicates the preset simulated absolute temperature. N This indicates the total number of conformations corresponding to the substituent structure. j This represents the conformation index variable during traversal summation, and its value ranges from 1 to... N ; Indicates the first j The potential energy value of each conformation.

[0090] This formula allows us to calculate the weighting coefficients for each conformation. Specifically, the smaller the potential energy difference, the more stable the conformation tends to be at a specific physiological temperature, and the higher the probability that it will maintain this posture during thermal fluctuations at that specific physiological temperature. Therefore, it is assigned a larger weighting coefficient.

[0091] Step 203: Using the weighting coefficients of each conformation, the conformation fit assessment value, metabolic stability assessment value and balance assessment value of each conformation are weighted to obtain the conformation fit score, metabolic stability score and balance score corresponding to the updated substituent structure.

[0092] For example, based on the weighting coefficients determined above, the conformational fit assessment values ​​of each conformation can be weighted and summed to obtain the conformational fit score corresponding to the updated substituent structure; based on the weighting coefficients determined above, the metabolic stability assessment values ​​of each conformation can be weighted and summed to obtain the metabolic stability score corresponding to the updated substituent structure; and based on the weighting coefficients determined above, the balance assessment values ​​of each conformation can be weighted and summed to obtain the balance score corresponding to the updated substituent structure. This processing method avoids the bias caused by a few extreme values ​​in the assessment results, so that the final output scores can cover the overall characteristics of all conformations.

[0093] Step 204: The conformational fit score, metabolic stability score, and balance score are weighted and calculated to obtain the target evaluation score.

[0094] For example, the target evaluation score can be obtained by weighted summation of the scores of the three dimensions corresponding to the updated substituent structures using linear weighting or other fusion algorithms. In practice, targeted dimensional evaluation weights can be set for the three scores according to the specific preclinical needs or specific drug-likeness bottlenecks in the current target drug development. The specific weight configuration is not specifically limited in this disclosure. For example, if the main challenge faced in the current development stage is the short in vivo half-life of the drug, a relatively high weight can be dynamically assigned to the metabolic stability score to enhance the model's ability to search for metabolically stable structures. Through this flexible weighting mechanism, the originally mutually restrictive multi-dimensional dynamic physicochemical indicators can be integrated into a scalarized target evaluation score, and the strategy parameters of the generation model can be updated based on it to drive the generation model to accurately determine chemical groups in subsequent iterative calculations, achieve synergistic optimization of substituent compatibility, metabolic stability, and hydrophilicity-hydrophobicity balance, and obtain the target substituent structure.

[0095] In step 104, the strategy parameters of the generative model are updated based on the target evaluation score, and new chemical groups to be updated are determined according to the updated substituent structure. Then, the process returns to the step of executing the structure evolution instruction based on the pre-built generative model until the target evaluation score is greater than or equal to a preset threshold, at which point the target substituent structure is output.

[0096] In the exemplary embodiments of this disclosure, the pre-built generative model can be constructed using a reinforcement learning architecture based on graph neural networks, and may include a molecular feature encoding layer, a graph information transfer layer, and a policy output layer connected in sequence. The molecular feature encoding layer is used to acquire the overall structure of the current NaV1.8 inhibitor and the identified chemical groups to be updated, and convert them into molecular graph data composed of nodes and edges. Nodes correspond to atoms in the molecule, and node features include atom type, charge state, and hybrid orbital type; edges correspond to chemical bonds in the molecule, and edge features include bond type and aromaticity. This molecular feature encoding layer can map the above node features and edge features into high-dimensional initial node latent vectors and initial edge latent vectors. The graph information transfer layer is used to receive the initial node latent vectors and initial edge latent vectors, and through multi-layer graph convolution or graph attention mechanisms, performs multiple rounds of information transfer and feature aggregation along chemical bonds between adjacent atomic nodes, so that the extracted features fully integrate the drug's properties. The model considers the global topological environment of the skeletal structure and the chemical context information of local functional groups. A policy output layer receives local node latent vectors and global graph feature vectors, performs nonlinear mapping through a multilayer perceptron, and outputs an action probability distribution for the currently updated chemical functional group. This action probability distribution is mapped to a discrete action space, where options include single actions such as group deletion and combined actions with specific molecular functional groups from a pre-constructed functional group dataset, such as hydroxyl, ether, methyleneoxy, methoxy, and ethyl groups. The generative model samples this action probability distribution to determine the structural evolution instruction for the current round, which includes the specific action type and corresponding molecular functional group.

[0097] Furthermore, when performing the subsequent step of updating the policy parameters of the generative model based on the target evaluation score, the aforementioned target evaluation score can be used as a reward signal. The gradient information of the network weights in the graph information transmission layer and the policy output layer can be calculated using the policy gradient algorithm or the near-end policy optimization algorithm. These policy parameters can then be updated through the backpropagation mechanism, making the generative model more inclined to output the probability distribution of actions that can obtain high evaluation scores in subsequent iterations.

[0098] Subsequently, the updated substituent structure is treated as a new environmental state, in which the next new chemical group to be updated is identified and determined, and the process returns to the inference step of executing the generation instruction. Through this closed-loop iterative mechanism that continuously receives environmental feedback and adaptively updates the strategy, the substituent structure is driven to undergo multi-dimensional directional evolution and optimization in three-dimensional physical space until the target evaluation score calculated for the current structure is greater than or equal to a preset threshold, or when the set maximum number of iterations is reached. At this point, the iteration terminates, and the final output is a target substituent structure with comprehensive physicochemical properties that meet the expected target and has a complete topological connection relationship, which can be directly applied to subsequent molecular docking or experimental synthesis of complete side-chain structures. The preset threshold is determined according to the actual situation, and this disclosure does not impose any special limitations. If the threshold is not exceeded when the maximum number of iterations is reached, the structure corresponding to the highest historical target evaluation score recorded during the iteration process is output, thus balancing search accuracy and computational efficiency.

[0099] In the exemplary embodiments of this disclosure, after iterative optimization through the aforementioned generative model combined with multi-conformation dynamic evaluation, the final output target substituent structure is specifically manifested as a linear flexible side chain containing oxygen atoms.

[0100] For example, the linear flexible side chain can be one of methoxyethoxy, hydroxyethyloxy, methoxyethylethoxy, and hydroxyethylethoxy. From the perspective of the microscopic physical mechanism of medicinal chemistry and molecular drug design, such linear fragments rich in ether bonds or terminal hydroxyl groups not only endow the substituent structure with a high degree of conformational freedom, enabling it to adaptively twist and fold with extremely low structural stress potential energy, thereby precisely fitting and anchoring in the complex spatial pocket of the receptor target to maximize conformational adaptability; at the same time, the high density of electronegative oxygen atoms on the linear side chain can form a wide dynamic hydrogen bond network with water molecules in physiological solvents, improving the overall water solubility and hydrophilic-hydrophobic balance of the molecule; in addition, the dynamic hydration layer formed by high-frequency thermal fluctuations in the physiological environment can also generate an effective steric hindrance masking effect, thereby inhibiting the rapid recognition and catalytic degradation of the drug core skeleton by relevant metabolic enzymes, improving the overall metabolic stability, thus achieving the expected technical effect of molecular directed evolution driven by multidimensional scalar reward signals at the underlying physicochemical structural level.

[0101] First, the spatial three-dimensional coordinates of atoms in each conformation are obtained, and the internal structural stress and spatial interaction of the conformations are quantitatively analyzed in combination with preset force field parameters to obtain the potential energy value characterizing the physical stability of each conformation. Then, based on the lowest potential energy value in the conformation set, the difference between each conformation and the benchmark is calculated to assign an adaptive weight to each conformation. The conformational fit, metabolic stability and equilibrium evaluation values ​​of each individual conformation are weighted respectively, so as to transform the discrete single-point conformation features into a multi-dimensional score that can truly reflect the substituent in a dynamic environment, filtering out the evaluation noise caused by local extreme conformations. Based on preset drug-likeness constraints, the above multi-dimensional scores are weighted to output the target evaluation score, and the target evaluation score is used as a reward signal to input into the generative model, thereby driving the directional optimization of the generative model parameters, so as to accurately output the target substituent structure that meets the multi-dimensional indicators and has a complete topological connection relationship.

[0102] Corresponding to the embodiments of the aforementioned methods, this disclosure also provides a NaV1.8 inhibitor, whose general structural formula is as follows: Figure 3 As shown.

[0103] This NaV1.8 inhibitor is a voltage-gated sodium channel inhibitor of NaV1.8, and its molecular physical structure contains a drug core framework composed of... Figure 3 It can be determined that the core skeleton of the drug is based on a tetrahydrofuran ring; aromatic groups and pyridine groups linked by amide bonds are respectively attached to different sites of the tetrahydrofuran ring. For example, a pyridine group linked by an amide bond is attached to the second carbon atom of the tetrahydrofuran ring, and an aromatic group is attached to the third carbon atom of the tetrahydrofuran ring.

[0104] Based on the core framework of the drug, in order to overcome the shortcomings of existing molecules such as short oral half-life and rapid metabolic clearance, this embodiment has carried out rational structural extension of specific substitution sites on aromatic groups.

[0105] Specifically, the inhibitor comprises a target substituent structure attached to the aromatic group (i.e., Figure 3 The target substituent structure (the -OR portion on the aromatic ring) is used to replace the original methoxy group on the aromatic group, wherein R is a linear flexible side chain containing an oxygen atom. Exemplarily, the linear flexible side chain (R) is selected from any one of methoxyethoxy, hydroxyethyloxy, methoxyethylethoxy, and hydroxyethylethoxy.

[0106] While traditional monomethoxy structures are advantageous for receptor binding, they are highly susceptible to O-demethylation metabolism in vivo, leading to decreased activity and more complex metabolites. This embodiment, by introducing the aforementioned flexible ether or hydroxyl side chains onto the molecule's periphery, not only effectively regulates the molecule's hydrophobic-hydrophilic balance but also reduces dependence on a single O-demethylation metabolic pathway. Simultaneously, these oxygen-containing flexible side chains can improve the molecule's metabolic stability in vivo through steric hindrance and dynamic reconstruction of the hydrogen bond network. Therefore, this general formula maintains the high selective inhibitory activity of NaV1.8 while improving the molecule's in vivo half-life, tissue distribution, and overall tolerability, thus possessing excellent prospects for clinical translation.

[0107] To further verify the feasibility and structural accuracy of the aforementioned target substituent structures in physical chemical synthesis, this disclosure also provides the specific preparation scheme and nuclear magnetic resonance characterization results of the aforementioned NaV1.8 inhibitors.

[0108] In an exemplary embodiment of this disclosure, a method for synthesizing a hydroxyethyloxy-substituted NaV1.8 inhibitor is provided, the synthetic reaction formula of which is as follows: Figure 4 As shown, the specific steps are as follows: Take 10 mmol of the aforementioned tetrahydrofuran derivative intermediate containing a phenolic hydroxyl group as a substrate, and add 10 mmol of 2-bromoethanol as a substitution reagent and 10 mmol of potassium carbonate sequentially. The above reactants were dissolved together in 50 ml of N,N-dimethylformamide (DMF) solvent as an acid-binding agent. The reaction system was heated to 80°C and reacted at this temperature for 10 hours. After the reaction was completed, the hydroxyethyloxy derivative was obtained by separation and purification.

[0109] The obtained product was characterized by nuclear magnetic resonance to verify its structural accuracy. The characterization results are as follows: 1H NMR (600 MHz, DMSO) δ 10.68 (s, 1H), 8.49 (d, J = 5.5 Hz, 1H), 8.30 (d, J = 2.1 Hz, 1H), 8.08 (d, J = 2.8 Hz, 1H), 7.84 (dd, J = 5.5, 2.2Hz, 1H), 7.63 (d, J = 2.7 Hz, 1H), 7.24 – 7.10 (m, 2H), 5.12 (d, J = 10.7 Hz, 1H), 4.98 (t, J = 5.3 Hz, 1H), 4.41 (dd, J = 10.7, 7.3 Hz, 1H), 4.21 – 4.02(m, 2H), 3.71 (q, J = 4.8 Hz, 2H), 2.99 – 2.87 (m, 1H), 1.62 (s, 3H), 0.72 (dd, J = 7.6, 2.8 Hz, 3H).

[0110] 13C NMR (600 MHz, DMSO) δ 168.52, 165.17, 150.84, 149.95, 149.88,148.85, 148.32, 148.25, 145.55, 145.31, 145.27, 143.87, 143.78, 142.23,142.14, 128.06, 126.17, 125.43, 124.28, 122.70, 122.64, 115.16, 111.51,110.31, 110.19, 85.39, 85.21, 85.03, 84.84, 79.11, 75.34, 75.31, 59.56, 54.32, 43.99, 42.54, 39.46, 39.34, 39.20, 39.06, 38.92, 38.78, 38.65, 38.50, 36.85, 33.89, 22.08, 20.82, 11.22.

[0111] The aforementioned nuclear magnetic resonance (NMR) data confirmed the successful synthesis of the inhibitor.

[0112] In an exemplary embodiment of this disclosure, a method for synthesizing a methoxyethoxy-substituted NaV1.8 inhibitor is also provided, the synthetic reaction formula of which is as follows: Figure 5 As shown, the specific steps are as follows: Take 10 mmol of the same substrate as described above, and add 10 mmol of 2-methoxybromoethane as a substitution reagent and 10 mmol of cesium carbonate (…). The methoxyethoxy derivative was obtained by using an alkaline reagent and placing the reactants in 50 ml of acetonitrile solvent. The reaction system was heated to reflux and the reaction was continued for 10 hours. After the reaction was completed, the methoxyethoxy derivative was obtained by separation and purification.

[0113] The obtained product was characterized by nuclear magnetic resonance to verify its structural accuracy. The characterization results are as follows: 1H NMR (600 MHz, DMSO) δ 10.70 (s, 1H), 8.49 (d, J = 5.5 Hz, 1H), 8.29 (d, J = 2.1 Hz, 1H), 8.07 (d, J = 2.8 Hz, 1H), 7.84 (dd, J = 5.5, 2.2Hz, 1H), 7.63 (d, J = 2.7 Hz, 1H), 7.23 – 7.12 (m, 2H), 5.12 (d, J = 10.7 Hz,1H), 4.36 (dd, J = 10.7, 7.2 Hz, 1H), 4.28 (ddd, J = 11.2, 5.9, 2.6 Hz, 1H),4.19 (ddd, J = 11.4, 4.6, 1.8 Hz, 1H), 3.61 (qdt, J = 8.6, 5.9, 2.6 Hz, 2H), 3.27 (s, 3H), 2.84 (p, J = 7.4 Hz, 1H), 1.62 (s, 3H), 0.71 (dd, J = 7.8, 2.7Hz, 3H).

[0114] 13C NMR (600 MHz, DMSO) δ 169.59, 166.22, 151.92, 151.03, 150.96,149.94, 149.40, 149.32, 146.60, 146.09, 146.04, 144.83, 144.74, 143.19,143.10, 129.08, 127.19, 126.35, 125.30, 123.66, 116.20, 112.55, 111.46,111.34, 86.21, 86.02, 85.84, 80.09, 73.55, 73.52, 71.35, 58.47, 45.09, 43.68, 40.52, 23.10, 12.25.

[0115] The aforementioned characteristic peak data confirms the successful synthesis of the above-mentioned inhibitor.

[0116] To further verify the actual efficacy of the NaV1.8 inhibitors described in this disclosure in vivo, the analgesic efficacy of the above compounds was evaluated using a zebrafish pain model.

[0117] In some example embodiments, methoxyethoxy-substituted NaV1.8 inhibitors, namely suzetrrine methoxyethoxy derivatives (denoted as VX-548-11) and hydroxyethyloxy-substituted NaV1.8 inhibitors, namely suzetrrine hydroxyethyloxy derivatives (denoted as VX-548-21), can be selected for testing. Since the test compounds are insoluble in water or methanol, a stock solution of 50 mg / mL was prepared using dimethyl sulfoxide (DMSO), which was then further diluted to 10 μg / mL using standard zebrafish dilution water, with no visible suspended matter. After centrifugation, the actual concentration of the test compound sample solution was determined by liquid chromatography to be 1 μg / mL. Simultaneously, aspirin was used as the positive control, dissolved and diluted with standard dilution water.

[0118] Wild-type AB strain zebrafish juveniles, 4 days post-fertilization, were used in the experiment. The experiment included a normal blank control group, a model control group, a positive drug control group, and different concentrations of the test sample administered to each group, with 30 zebrafish in each group. A pain model was established using glacial acetic acid (AA) induction and incubated continuously for 72 hours. A behavioral analyzer was used to record and detect the zebrafish's behavioral trajectories within 10 minutes after model establishment. The amount of arousal activity (unit: mm) was used as a quantitative indicator to evaluate the analgesic efficacy of the test sample. Experimental data are expressed as Mean ± SE. One-way ANOVA and t-tests were used, with P < 0.05 considered statistically significant.

[0119] Figure 6 The figure shown is a statistical chart evaluating the analgesic efficacy of suzetrol methoxyethoxy derivative (VX-548-11) in a zebrafish chemically induced pain model. Figure 7 The figure shown is a statistical chart evaluating the analgesic efficacy of suzetrlin hydroxyethyloxy derivative (VX-548-21) in a zebrafish chemically induced pain model.

[0120] The control group (blank control group) was used to indicate the baseline activity level of zebrafish under normal physiological conditions. The model group (model control group, 0.025% AA) used acetic acid (AA) as a chemical analgesic, resulting in a highly significant surge in activity after stimulation, confirming the hyperactive behavior induced by acute pain stress and indicating the successful establishment of a nociceptive pain model. Aspirin-12.5 μg / mL (positive control group) served as a reference standard to demonstrate that classic analgesics can effectively reverse the abnormal increase in activity induced by chemical stimulation. The T1 dose escalation group indicated that at three concentration gradients of 0.25 μg / mL, 0.5 μg / mL, and 1 μg / mL, the activity level of zebrafish exhibited a highly significant dose-dependent decrease. Particularly at the high-dose group (1 μg / mL), the activity inhibition level was significantly better than that of the positive control group, and the statistical difference was extremely significant (P<0.0001), objectively quantifying the excellent analgesic efficacy of this target derivative.

[0121] The experimental results are as follows: 1. Verification of the effectiveness of the pain model construction: Compared with the blank control group, the arousal activity of zebrafish juveniles in the acetic acid-induced model control group increased significantly (e.g., VX-548-11 increased from 24±3 mm to 319±20 mm, or VX-548-21 increased from 15±4 mm to 307±18 mm), indicating that the zebrafish juvenile pain model constructed in this embodiment is stable and reliable, can effectively distinguish between normal physiological state and pain stress state, and is suitable for evaluating the efficacy of analgesic active substances.

[0122] 2. Verification of analgesic activity of positive control drugs: Aspirin, the positive control drug, showed excellent analgesic effects at the tested concentration (e.g., 12.5 μg / mL), significantly reducing the activity level of zebrafish juveniles. For example, the activity level of zebrafish juveniles corresponding to VX-548-11 decreased to 203±19 mm, which was highly significant compared with the model control group (P<0.001); the activity level of zebrafish juveniles corresponding to VX-548-21 decreased to 99±10 mm, which was also highly significant compared with the model control group (P<0.0001), demonstrating excellent analgesic effects.

[0123] 3. Analgesic activity test results showed that the analgesic activity of VX-548-11 exhibited a clear dose-dependent characteristic, meaning that the higher the concentration, the better the analgesic effect. Specifically, the arousal activity of zebrafish juveniles in the 0.25 μg / mL concentration group was 265±26 mm, which was lower than that of the model control group, but did not reach a statistically significant level; the arousal activity in the 0.5 μg / mL concentration group decreased to 199±21 mm, which was extremely significant compared with the model group (P<0.001); the arousal activity in the 1 μg / mL concentration group further decreased to 110±12 mm, which was extremely significant compared with the model control group (P<0.0001). The analgesic activity of (VX-548-21) exhibited a clear dose-dependent characteristic. In the 0.25 μg / mL concentration group, the arousal activity of zebrafish juveniles was 274 ± 18 mm, which was lower than the model control group, but not statistically significant. In the 0.5 μg / mL and 1 μg / mL concentration groups, the arousal activity decreased to 192 ± 21 mm and 208 ± 22 mm, respectively, showing highly significant differences compared to the model control group (P < 0.05), indicating a clear analgesic effect. These in vivo data fully demonstrate that introducing methoxyethoxy or hydroxyethyloxy side chains into the molecular backbone can significantly enhance the analgesic activity of the suzetrolene core backbone.

[0124] To verify whether compounds corresponding to the substituent structures optimized by the method of this disclosure can act on human NaV1.8 voltage-gated sodium channels, this disclosure constructs and identifies a stable human NaV1.8-expressing ND7-23 cell model.

[0125] In some example embodiments, a hybrid cell line of rat dorsal root ganglion neurons and mouse neuroblastoma (ND7-23, SCSP-5026) was used as the host cell, which did not express endogenous NaV1.8 channel protein. The cells were seeded in 6-well plates, with complete culture medium added to each well.

[0126] In some example embodiments, co-transfection was performed using liposome transfection: 5 μL of LipoBooster 3000 transfection reagent was diluted in serum-free DMEM medium; simultaneously, 1.5 μg of PiggyBac-CMV-SCN10A(human)-Puro plasmid carrying the human SCN10A gene was diluted with 0.5 μg of Super PiggyBac Transposase plasmid, and transfection enhancer was added and mixed. The two solutions were mixed to form a complex and then added dropwise to the cells. Forty-eight hours after transfection, cells were selected under pressure using 2 μg / mL puromycin for 1-2 weeks until untransfected cells died completely. Subsequently, single-clonal cell lines were obtained using limiting dilution and expanded into larger cultures.

[0127] In some example embodiments, total RNA was extracted from stable and untransfected cells and reverse transcribed into cDNA. Using β-actin as an internal control, quantitative real-time PCR was performed using SYBR qPCR Master Mix (Novizan, Q713-02). The NaV1.8 primers were as follows: front primer GTCCACTCGTGGTTCAGTTT; reverse primer TGGGAGTGGAGAGAGGTTG; β-actin primers were as follows: front primer CACCATTGGCAATGAGCGGTTC; reverse primer AGGTCTTTGCGGATGTCCACGT.

[0128] The results of real-time fluorescence quantitative PCR detection of NaV1.8 mRNA are as follows: Figure 8 As shown. In Figure 8 In the figure, the amplification curve on the left shows a typical S-shaped growth with a stable baseline, while the melting curve shows a single sharp peak, proving that the amplification product has extremely high specificity. The quantitative analysis bar chart on the right shows that the relative expression level of NaV1.8 in the stable transfected cells (transfected group) was significantly higher than that in the untransfected group (P<0.01), confirming the successful transcription of the target gene.

[0129] In some example embodiments, cells were seeded onto a fluorescent detection plate, fixed with paraformaldehyde and blocked with BSA, and then NaV1.8 specific primary antibody (Thermo, PA5-115621) and ABflo were added. ® Cells were incubated with 555-modified fluorescent secondary antibody (Abclonal, AS058) and finally stained with Hoechst 33342 at room temperature in the dark for 35 min to label the nuclei. Figure 9 As shown, untransfected ND7-23 cells only exhibited blue nuclear fluorescence labeled with Hoechst 33342; while the stably expressed ND7-23 cell membrane showed a highly significant and specific orange fluorescence signal. This result directly confirms that human NaV1.8 protein was successfully expressed in cells and correctly localized to the cell membrane. These results clearly demonstrate that the constructed stable NaV1.8-expressing ND7-23 cell line can be used for subsequent pharmacodynamic and functional studies.

[0130] To further verify the effects of the compounds corresponding to the above-mentioned substituent structures, this disclosure evaluates their inhibitory activity against the NaV1.8 channel.

[0131] In some example embodiments, the inhibitory activity of the above compounds on the NaV1.8 channel was detected by FLIPR membrane potential fluorescence method. The fluorescence intensity of the membrane potential fluorescent dye is positively correlated with the degree of cell membrane depolarization, and the channel activation and inhibition state can be intuitively reflected by the change of fluorescence signal.

[0132] To eliminate the interference of endogenous TTX-sensitive sodium channels on the experimental results, a specific inhibitor of Hainantoxin-IV was added throughout the experiment. This specific inhibitor specifically inhibits TTX-sensitive sodium channels, but has almost no inhibitory effect on NaV1.8 channels. After incubation, a mixed solution of Veratridine (final concentrations: 1.25, 2.5, 5, 10, 20, 40 μM) and Cypermethrin (final concentration: 5 μM) was added. Fluorescence intensity was measured immediately after addition, and data were continuously collected for 10 minutes. The peak response values ​​of each concentration of Veratridine were extracted, and the background fluorescence of all experimental wells was subtracted.

[0133] Figure 10 The concentration-response curve of the NaV1.8 voltage-gated sodium channel agonist Veratridine is shown. Figure 10 In the figure, the horizontal axis represents the logarithm of veratridine concentration, and the vertical axis represents the total fluorescence intensity emitted by the cells. The black curve (stable transfection group) shows a significant S-shaped upward trend, indicating that veratridine can activate NaV1.8 channels in a concentration-dependent manner; while the blue curve (untransfected group) remains at the baseline level. The experiment confirmed that under the co-stimulation of 5 μM cypermethrin, 20 μM veratridine can induce a channel activation signal close to the plateau phase maximum. Therefore, this concentration combination was selected as the standard activation condition for the inhibition experiment.

[0134] In some example embodiments, suzetril (VX-548), suzetril methoxyethoxy derivative (VX-548-11), and suzetril hydroxyethyloxy derivative (VX-548-21) were pre-incubated with cells at final concentration gradients from 0.001 nM to 1000 nM, respectively, and then an optimized activator was added for assay to obtain the inhibitory activity of the NaV1.8 channel.

[0135] like Figure 11As shown, the horizontal axis represents the logarithm of the concentration of the test compound, and the vertical axis represents the proportion of NaV1.8 channels that are still in the active state. With increasing concentration of the test compound, the normalized membrane potential fluorescence signal exhibits a typical S-shaped dose-dependent decay. Further nonlinear regression fitting analysis revealed the half-maximal inhibitory concentrations (IC50) of each compound for the NaV1.8 channel as follows: suzetril (VX-548) 0.88 nM (0.69–1.23 nM, 95% confidence interval); suzetril methoxyethoxy derivative (VX-548-11) 2.1 nM (1.64–2.79 nM, 95% confidence interval); and suzetril hydroxyethyloxy derivative (VX-548-21) 4.05 nM (3.02–5.43 nM, 95% confidence interval). The above results indicate that suzetrol derivatives VX-548-11 and VX-548-21 can specifically inhibit the NaV1.8 voltage-gated channel.

[0136] Corresponding to the embodiments of the foregoing methods, this disclosure also provides embodiments of the apparatus and the terminal to which it is applied.

[0137] like Figure 12 As shown, Figure 12 This disclosure is a block diagram of an optimization device for substituent structures in tetrahydrofuran-based NaV1.8 inhibitors based on reinforcement learning, according to an exemplary embodiment. The device includes: an acquisition module 1210, a structure evolution module 1220, a scoring module 1230, and an output module 1240. The acquisition module 1210 is used to acquire the receptor spatial model of the NaV1.8 inhibitor and the chemical groups to be updated in the initial substituent structure; The structure evolution module 1220 is used to determine the structure evolution instructions based on the pre-built generative model, and to perform topological connection, group deletion or group substitution on chemical groups according to the structure evolution instructions to obtain the updated substituent structure; The scoring module 1230 is used to obtain multiple conformations by changing the spatial orientation of the rotatable single bonds in the updated substituent structure under the spatial constraints of the receptor spatial model. The target evaluation score is obtained based on the conformational fit evaluation value between each conformation and the receptor spatial model, the metabolic stability evaluation value of each conformation, and the balance score between hydrophilicity and hydrophobicity corresponding to the updated substituent structure. Output module 1240 is used to update the strategy parameters of the generative model based on the target evaluation score, determine the new chemical groups to be updated according to the updated substituent structure, and return to execute the step of determining the structure evolution instruction based on the pre-built generative model until the target evaluation score is greater than or equal to a preset threshold, and then output the target substituent structure.

[0138] It should be noted that the device for optimizing the substituent structure in tetrahydrofuran-based NaV1.8 inhibitors based on reinforcement learning in this embodiment is used to implement the corresponding optimization method for the substituent structure in tetrahydrofuran-based NaV1.8 inhibitors based on reinforcement learning in the aforementioned method embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be elaborated here.

[0139] The embodiments of the reinforcement learning-based optimization device for the substituent structure in tetrahydrofuran-based NaV1.8 inhibitors disclosed herein can be applied to computer devices, such as servers or terminal devices. The device embodiments can be implemented in software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor reading the corresponding computer program instructions from non-volatile memory into memory and executing them. From a hardware perspective, such as... Figure 13 The diagram shown is a hardware structure diagram of a computer device containing a device for optimizing the substituent structure in a tetrahydrofuran-based NaV1.8 inhibitor based on reinforcement learning, according to an embodiment of this disclosure. Except for... Figure 13 In addition to the processor 1310, memory 1330, network interface 1320, and non-volatile memory 1340 shown, the server or electronic device where the substituent structure optimization device in the NaV1.8 inhibitor in the embodiment is located may also include other hardware depending on the actual function of the computer device, which will not be described in detail here.

[0140] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0141] The foregoing has described specific embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0142] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention applied herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not claimed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0143] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

[0144] The above description is merely a preferred embodiment of this disclosure and is not intended to limit this disclosure. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for optimizing the substituent structure in tetrahydrofuran-based NaV1.8 inhibitors based on reinforcement learning, characterized in that, include: Obtain the receptor spatial model of the NaV1.8 inhibitor and the chemical groups to be updated in the initial substituent structure; Based on a pre-constructed generative model, structural evolution instructions are determined, and topological connections, group deletions, or group substitutions are performed on the chemical groups according to the structural evolution instructions to obtain updated substituent structures; Under the spatial constraints of the receptor spatial model, multiple conformations are obtained by changing the spatial orientation of the rotatable single bonds in the updated substituent structure. The target evaluation score is obtained based on the conformational fit evaluation value between each conformation and the receptor spatial model, the metabolic stability evaluation value of each conformation, and the balance evaluation value between the hydrophilicity and hydrophobicity of each conformation. The target evaluation score is used as a reward signal to calculate the gradient information of the generative model. The strategy parameters of the generative model are updated through the backpropagation mechanism. The new chemical group to be updated is determined according to the updated substituent structure. The process returns to the step of executing the structure evolution instruction based on the pre-built generative model until the target evaluation score is greater than or equal to a preset threshold, at which point the target substituent structure is output. The pre-built generative model includes a molecular feature encoding layer, a graph information transfer layer, and a strategy output layer. The molecular feature encoding layer converts the chemical groups to be updated into molecular graph data composed of nodes and edges, and determines the initial node hidden vectors and initial edge hidden vectors. The graph information transfer layer is used to extract features based on the initial node hidden vectors and initial edge hidden vectors to obtain the chemical context information of local groups. The strategy output layer is used to output the action probability distribution and perform sampling to determine the structure evolution instructions.

2. The method according to claim 1, characterized in that, The NaV1.8 inhibitor is a tetrahydrofuran-based NaV1.8 voltage-gated channel inhibitor. The chemical groups to be updated in the initial substituent structure of the NaV1.8 inhibitor include: Obtain the drug skeleton structure of the NaV1.8 inhibitor, wherein the drug skeleton structure comprises a tetrahydrofuran ring, and aromatic groups and pyridine groups linked by amide bonds are attached to different sites of the tetrahydrofuran ring. The initial group at a predetermined site on the aromatic group in the drug skeleton structure is identified as the chemical group to be updated in the initial substituent structure.

3. The method according to claim 1, characterized in that, The step of performing topological linking, group deletion, or group substitution on the chemical groups according to the structural evolution instructions to obtain the updated substituent structure includes: When performing the topological connection, the target group is obtained from the pre-constructed group data set, and the three-dimensional growth vector field of the chemical group to be updated under the constraints of the acceptor space model is obtained; the spatially permissible extension region is determined based on the three-dimensional growth vector field, and the target group and the chemical group to be updated are structurally connected based on the matched chemical reaction template and the spatially permissible extension region. When performing the group deletion, the target group that contributes the least to the target evaluation score in the initial substituent structure is removed; When performing the group replacement, a three-dimensional growth vector is determined based on the three-dimensional growth vector field at the connection site of the chemical group to be updated, and the chemical group to be updated is replaced with the target group based on the matched chemical reaction template and the three-dimensional growth vector. The group data set includes hydroxyl, ether bond, methyleneoxy, methoxy and ethyl groups.

4. The method according to claim 1, characterized in that, The target evaluation score is obtained based on the conformational fit assessment value of each conformation to the receptor spatial model, the metabolic stability assessment value of each conformation, and the balance assessment value between the hydrophilicity and hydrophobicity of each conformation, respectively. This includes: The potential energy value of each conformation is calculated based on the three-dimensional spatial coordinates of each constituent atom in each conformation and the preset force field parameters. The lowest potential energy value in all the conformations is determined, and a weighting coefficient for each conformation is determined based on the difference between the potential energy value of each conformation and the lowest potential energy value. The smaller the difference, the larger the corresponding weighting coefficient. Using the weighting coefficients of each conformation, the conformation fit assessment value, metabolic stability assessment value, and balance assessment value of each conformation are weighted to obtain the conformation fit score, metabolic stability score, and balance score corresponding to the updated substituent structure. The target evaluation score is obtained by weighting the conformational fit score, the metabolic stability score, and the balance score.

5. The method according to claim 1, characterized in that, The method includes determining a conformational fit assessment value between each of the conformations and the receptor spatial model: Obtain the preset spatial boundary of the updated substituent structure within the receptor spatial model; For each of the aforementioned conformations, the local free volume of the updated substituent structure within the spatial boundary is calculated, as well as the spatial repulsion volume between each constituent atom in the updated substituent structure and the acceptor atom in the acceptor spatial model. Based on the local free volume and the spatial repulsion volume, a conformational fit evaluation value is determined for each of the conformations.

6. The method according to claim 1, characterized in that, The method includes determining metabolic stability assessment values ​​for each of the said conformations: Identify metabolically vulnerable atoms in the updated substituent structure; For each of the aforementioned conformations, the spatial solvent-accessible surface area of ​​the metabolically vulnerable atom is calculated based on the three-dimensional coordinates of the metabolically vulnerable atom in the corresponding spatial orientation. The metabolic stability assessment value for each conformation is determined based on the available surface area of ​​the solvent in the space. The spatial solvent-accessible surface area is negatively correlated with the metabolic stability assessment value.

7. The method according to claim 1, characterized in that, The method includes determining an evaluation value for the balance between hydrophilicity and hydrophobicity corresponding to each of the aforementioned conformations: For each of the aforementioned conformations, the corresponding surface areas accessible to polar solvents and non-polar solvents are calculated. Based on the ratio of the surface area accessible to the polar solvent to the surface area accessible to the non-polar solvent for each conformation, a balance assessment value between hydrophilicity and hydrophobicity for each conformation is determined.

8. The method according to claim 1, characterized in that, The target substituent structure is a linear flexible side chain containing an oxygen atom; the linear flexible side chain is one of methoxyethoxy, hydroxyethyloxy, methoxyethylethoxy, and hydroxyethylethoxy.

9. A NaV1.8 inhibitor, characterized in that, The NaV1.8 inhibitor comprises a drug core framework and a target substituent structure attached to the drug core framework; the target substituent structure is determined by the substituent structure optimization method for tetrahydrofuran-based NaV1.8 inhibitors based on reinforcement learning as described in any one of claims 1 to 8.

10. The NaV1.8 inhibitor according to claim 9, characterized in that, The core drug skeleton of the NaV1.8 inhibitor is a tetrahydrofuran ring, and the aromatic group on the tetrahydrofuran ring is connected to the target substituent structure, which is one of methoxyethoxy, hydroxyethyloxy, methoxyethylethoxy, and hydroxyethylethoxy.

Images