Virtual screening method and system for virtual reality-based pharmacophore

Abstract

The present application relates to the technical field of drug virtual screening, in particular to a virtual screening method and system of pharmacophore based on virtual reality. It comprises the following steps: inputting protein structure and compound library into VR scene as screening objects through a molecular resource loading module; analyzing pocket area in the protein structure by using a pocket identification module, and performing three-dimensional observation and position calibration on the pocket area in the VR scene, which is used for judging the actual position and range of the pocket area; wherein, conformation consistency analysis is performed on the pocket area based on a conformation integration mechanism, and spatial fusion is performed on the pocket area according to the conformation consistency analysis result, forming a conformation stable pocket area. The present application integrates the whole process of protein structure, pocket identification, pharmacophore construction and molecular docking in a virtual reality (VR) environment, and users can immerse themselves in three-dimensional observation of the conformation stable pocket area.

Description

Technical Field

[0001] This invention relates to the field of virtual drug screening technology, and more specifically, to a virtual screening method and system for pharmacophores based on virtual reality. Background Technology

[0002] With the continuous development of computer-aided drug design (CADD) technology, pharmacophore templates and virtual screening have become important tools for lead compound discovery. Traditional pharmacophore modeling mainly relies on two-dimensional structural analysis or static three-dimensional structural calculations, which are insufficient to fully reflect the dynamic flexibility of protein targets and the spatial complexity of ligand-target interactions. Furthermore, existing screening platforms are mostly based on desktop graphical interfaces, leaving users with a lack of intuitive understanding of the spatial relationships between molecular structures, binding pockets, and pharmacophore features. This results in highly subjective pharmacophore template construction, limited accuracy, and difficulty in efficient correction. In addition, variations in pocket shape caused by protein conformational diversity are often overlooked, making virtual screening based on a single rigid structure prone to false positives or missing truly active molecules. Therefore, this paper proposes a virtual screening method and system for pharmacophores based on virtual reality. Summary of the Invention

[0003] The purpose of this invention is to provide a virtual screening method and system for pharmacophores based on virtual reality, in order to solve the problems of low screening accuracy, high false positive rate and poor human-machine collaboration efficiency caused by the lack of flexibility in protein conformation, insufficient three-dimensional spatial interactive perception and strong subjectivity in pharmacophore modeling in the traditional virtual screening methods mentioned in the background art.

[0004] To achieve the above objectives, the present invention aims to provide a virtual screening method for pharmacophores based on virtual reality, comprising:

[0005] S1. Input protein structures and compound libraries into the VR scene as screening objects through the molecular resource loading module;

[0006] S2. Analyze the pocket region in the protein structure using the pocket recognition module, and perform three-dimensional observation and position calibration of the pocket region in the VR scene to determine the actual position and range of the pocket region;

[0007] Among them, conformational consistency analysis of pocket regions is performed based on conformational integration mechanism, and spatial fusion of pocket regions is performed based on conformational consistency analysis results to form conformationally stable pocket regions;

[0008] S3. Based on the conformationally stable pocket region, the pharmacophore analysis module generates initial functional feature points, and interactively edits and corrects the spatial location, type and tolerance parameters of the pharmacophore in the VR environment to form a pharmacophore template.

[0009] S4. Pharmacophore screening process based on pharmacophore templates, matching and scoring candidate compounds one by one, and generating a preliminary screening molecule list based on matching scores.

[0010] S5. Based on the actual position and range of the conformational stability pocket region, perform fine molecular docking on the initial screening molecule list to obtain the binding conformation and energy score of each molecule, which is used to screen the final candidate molecules.

[0011] As a further improvement to this technical solution, in step S2, the pocket recognition module is used to analyze the pocket region in the protein structure, and the pocket region is observed and its position calibrated in three dimensions in a VR scene, including the following steps:

[0012] S2.1. Rotation, movement, and magnification of proteins through three-dimensional manipulation.

[0013] S2.2. Based on the pocket recognition module, the pocket recognition algorithm is executed on the protein structure to generate potential binding sites, i.e. pocket regions, and the pocket regions are marked in a highlighted manner in the VR scene.

[0014] S2.3. Position calibration of the identified pocket regions is performed based on the geometric morphological features and spatial parameters of the protein structure.

[0015] As a further improvement to this technical solution, in step S2.2, the pocket recognition algorithm is applied to the protein structure based on the pocket recognition module to generate potential binding sites, i.e., pocket regions, including the following steps:

[0016] S2.21. Convert the atomic coordinates of the protein structure into a three-dimensional surface mesh model, and calculate the spatial coordinates and neighborhood atomic information of each mesh point;

[0017] S2.22. Perform cavity scanning on the surface mesh of the three-dimensional surface mesh model to identify the concave regions on the protein surface and the cavity volume that can accommodate ligands, and calculate the geometric center, volume and opening direction of each cavity to form a preliminary set of candidate pocket regions.

[0018] S2.23 Calculate the geometric features for each candidate pocket region to form the feature vector of the candidate pocket;

[0019] S2.24. The feature vectors of candidate pockets are evaluated by constructing a preset scoring function, and pocket regions whose evaluation results exceed the threshold n are selected and marked as potential binding sites.

[0020] S2.25. Based on the conformational integration mechanism, perform conformational consistency analysis on the pocket regions generated in step S2.24, and spatially fuse the pocket regions according to the conformational consistency analysis results to form conformationally stable pocket regions.

[0021] As a further improvement to this technical solution, in step S2.25, forming a conformationally stable pocket region includes the following steps:

[0022] Based on the pocket region generated in step S2.24, normal mode analysis is performed on the protein structure to obtain the dominant motion direction of the protein structure in the low-frequency intrinsic mode, and multiple lightweight substitution conformations are generated along the main mode within a reasonable perturbation range to form a conformation set.

[0023] For each pocket region, each conformation in the conformation set The pocket region is repositioned and its corresponding spatial occupancy area is calculated to obtain the projection results of the pocket region under different conformations, forming a conformational mapping pocket set.

[0024] Perform cross-conformity consistency analysis on the conformational mapping pocket set to generate a conformational consistency score. ;

[0025] Based on conformational consistency score With preset threshold Filtering will satisfy Candidate pocket regions are marked as conformationally stable candidate pockets;

[0026] Spatial fusion processing is performed on the conformationally stable pocket region to obtain the fused conformationally stable pocket region.

[0027] As a further improvement to this technical solution, the spatial fusion processing of the conformationally stable pocket region to obtain the fused conformationally stable pocket region involves the following steps:

[0028] Realign and scale the conformation mapping pocket set;

[0029] Based on the geometric center offset range of the conformational mapping pockets after coordinate alignment and scale normalization, a Boolean fusion strategy is applied to the boundary meshes of each conformational pocket for fusion.

[0030] The fused pocket region is sampled using voxelization and reconstructed using meshes. The stable spatial volume maintained by each conformation is calculated to form the fused conformational stable pocket region.

[0031] As a further improvement to this technical solution, in step S3, forming the pharmacophore template includes the following steps:

[0032] S3.1 The pharmacophore analysis module generates initial pharmacophore feature points based on the local structure of the protein, the distribution of key residues, and the pocket shape.

[0033] S3.2 Display the initial feature points of the pharmacophore with different identifiers near the pocket region and overlay them with the protein structure to facilitate users to observe its geometric relationship with the pocket.

[0034] S3.3, Check the distance, orientation, and geometric layout between the pharmacophore and protein residues;

[0035] S3.4 After the inspection is completed, save the pharmacophore as a pharmacophore template.

[0036] As a further improvement to this technical solution, in step S4, candidate compounds are matched and scored one by one, and a preliminary screening molecule list is generated based on the matching scores, including the following steps:

[0037] S4.1 Input the pharmacophore template as a matching rule into the filtering module;

[0038] S4.2 Set screening conditions based on the type and tolerance of pharmacophore feature points;

[0039] S4.3 Perform pre-filtering based on compound library information and output the candidate compound set after pre-filtering;

[0040] S4.4. Traverse each molecule in the candidate compound set and perform geometric alignment calculations of the pharmacophore feature points between its three-dimensional structure and the pharmacophore template.

[0041] S4.5 Compare the degree of matching between the functional group positions of the compound and the pharmacophore template, and perform multi-dimensional energy-weighted pharmacophore matching score to generate an energy-weighted matching score;

[0042] S4.6 Sort all candidate compounds from high to low according to their energy-weighted matching scores and output the initial screening list.

[0043] As a further improvement to this technical solution, step S4.5 involves performing multi-dimensional energy-weighted pharmacophore matching and scoring, including the following steps:

[0044] S4.51. Constructing basic energy contribution coefficients for different pharmacodynamic clusters. Based on the protein chemical microenvironment, the basic energy contribution coefficient is energy corrected to generate an energy-corrected contribution coefficient. ;

[0045] S4.52. Based on the spatial deviation between the matching point and the pharmacophore template point in the candidate molecule, construct a Gaussian distance decay weight for each pharmacophore feature point;

[0046] S4.53. To address the characteristics of directional pharmacophores, an angle-dependent weighting factor is introduced to evaluate the consistency between the interaction direction of candidate molecules and the pharmacophore template direction.

[0047] S4.54, Contribution coefficient The weighting factors of spatial weight and angle dependence are jointly integrated to calculate a multidimensional weighted score for each pharmacophore feature point, and the sum of the weighted scores of all pharmacophore feature points is used as the overall energy weighted matching score of the candidate molecule.

[0048] As a further improvement to this technical solution, step S5, obtaining the binding conformation and energy score of each molecule, includes the following steps:

[0049] S5.1 Based on the initial screening molecule list, each candidate molecule in the initial screening molecule list is imported into the docking module, and the three-dimensional mesh of the conformational stability pocket region is used as the constraint space.

[0050] S5.2 Perform conformational sampling and rotation-translation optimization on each candidate molecule to generate multiple initial binding postures that can enter the pocket region, and perform collision checks and geometric reachability verification on each posture to generate the verified conformation.

[0051] S5.3. Perform local energy minimization again on the verified conformation, and generate the final comprehensive binding energy score by combining the interaction strength, conformation stability and pocket space complementarity.

[0052] S5.4. Sort all candidate molecules according to the comprehensive binding energy score, generate a candidate molecule set, and present the binding conformation of protein and ligand, key interactions and docking search space in three dimensions through the docking visualization module.

[0053] On the other hand, the present invention provides a virtual screening system for pharmacophores based on virtual reality, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the virtual screening method for pharmacophores based on virtual reality as described above.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] 1. This invention relates to a virtual screening method and system for pharmacophores based on virtual reality. By integrating the entire process of protein structure, pocket identification, pharmacophore construction, and molecular docking into a virtual reality (VR) environment, users can immerse themselves in 3D observation, spatial calibration, and interactive editing of conformationally stable pocket regions. Especially in the pharmacophore template construction stage, users can intuitively drag and modify feature point types and tolerance parameters, and verify their spatial matching relationship with key residues in real time, effectively overcoming modeling biases caused by insufficient spatial perception in traditional two-dimensional interfaces. Simultaneously, the "conformationally stable pocket region" generated by the conformation integration mechanism significantly improves the adaptability of the pharmacophore template to protein flexibility changes, making subsequent virtual screening closer to the real biological environment, thereby greatly improving the initial screening hit rate and lead compound discovery efficiency.

[0056] 2. This invention relates to a virtual screening method and system for pharmacophores based on virtual reality. It integrates pre-screening using physicochemical rules, multi-dimensional pharmacophore matching scoring with energy correction, and fine molecular docking based on conformationally stable pocket regions, forming a three-tiered progressive screening system of "coarse screening—fine screening—verification." Specifically, the pharmacophore matching scoring incorporates local chemical microenvironment correction (such as pH and charge density), Gaussian spatial decay weights, and directional angle-dependent factors, ensuring that the score reflects not only geometric matching but also the realism of physicochemical interactions. Finally, through VR visualization of the docking results, users can intuitively evaluate the ligand binding conformation, key interactions, and the rationality of the docking box settings, supporting rapid decision-making and structural optimization. This process significantly shortens the drug discovery cycle and reduces experimental trial-and-error costs while ensuring screening accuracy. Attached Figure Description

[0057] Figure 1 This is a flowchart of the overall method of the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0059] Example 1: Please refer to Figure 1 As shown, this embodiment provides a virtual screening method for pharmacophores based on virtual reality, including the following steps:

[0060] S1. Input protein structures and compound libraries into the VR scene as screening objects through the molecular resource loading module;

[0061] In this embodiment, the molecular resource loading module includes a local molecular resource module, a website resource retrieval module, a PDB molecular library retrieval module, and a QR code loading module.

[0062] The local molecular resource module (Load From File) is used to load molecular resources from the user's local storage. The user places the .pdb format molecular file in the preset path, and the system reads the file through this module and loads the three-dimensional molecular structure into the visualization window, realizing the rapid import and visualization of local molecular resources, and providing input data for subsequent pocket recognition and pharmacophore analysis.

[0063] The website resource retrieval module (Load From Internet) is used to automatically retrieve the corresponding molecular files from the online database based on the PDB ID (protein data identifier) ​​provided by the user and load them into the visualization window; the system provides virtual keyboard input and control functions, enabling users to easily enter the PDB ID and complete the acquisition of remote molecular resources, providing technical support for users to access structural data without local storage;

[0064] The PDB molecular library retrieval module (Load From Resources) is used to select target molecules from the software's built-in or preset molecular libraries, including drug molecules, target proteins, ligands, and other experimental or test compounds, and load the selected molecules into a visualization window for 3D display. This module supports the management of multiple categories of molecular resources, ensuring that users can quickly access commonly used molecular libraries that have been organized, facilitating molecular screening and docking operations.

[0065] The QR code loading module (Load By QRCode) parses the target molecular file from the QR code containing PDB file information and loads it into the visualization window, enabling rapid acquisition of molecular resources. This module provides users with a convenient QR code-based import solution, simplifying the molecular data acquisition process, and is especially suitable for application scenarios where molecular information can be quickly collected on mobile devices or in the laboratory.

[0066] Users can select different display modes through the Style module, such as Hide, Wireframe, Stick, Ball, BallStick, Cartoon, Surface, AEO (Atomic Ellipsoid Model, usually referring to anisotropic displacement ellipsoid), and AEO+ (Enhanced Atomic Ellipsoid Model), to make protein structures more suitable for observation and identification of potential binding pockets.

[0067] S2. Analyze the pocket region in the protein structure using the pocket recognition module, and perform three-dimensional observation and position calibration of the pocket region in the VR scene to determine the actual position and range of the pocket region;

[0068] Among them, conformational consistency analysis of pocket regions is performed based on conformational integration mechanism, and spatial fusion of pocket regions is performed based on conformational consistency analysis results to form conformationally stable pocket regions;

[0069] In this embodiment, a pocket recognition module is used to analyze the pocket region in the protein structure, and the pocket region is observed and its position calibrated in three dimensions in a VR scene, including the following steps:

[0070] S2.1. The protein is rotated, moved and magnified through three-dimensional manipulation so that the surface grooves and internal cavity areas of the protein can be viewed from different angles, providing a spatial reference for pocket identification;

[0071] The three-dimensional operations include Rotate, Move, Scale, and Auto operations. Rotate is a rotation operation used to adjust the orientation of protein molecules in three-dimensional space to observe their surface grooves and internal cavity regions from different angles. Move is a translation operation used to change the position of molecules in three-dimensional visualization space to adjust the center of the observation viewpoint. Scale is a scaling operation used to change the display scale of molecules in three-dimensional space to focus on detailed regions of the protein structure, such as active sites, thereby providing a spatial reference for pocket identification and position calibration. Auto is an automatic control operation used to control molecules to remain stationary or rotate.

[0072] S2.2. Based on the pocket recognition module, the pocket recognition algorithm is executed on the protein structure (i.e., the system automatically executes the pocket recognition algorithm on the protein structure using the pocket recognition module (XPocket module) in the "Function" menu module) to generate potential binding sites, i.e. pocket regions, and the pocket regions are marked in the VR scene by highlighting or marking the boundaries.

[0073] In this process, users can observe the pocket area from a 360° free-view perspective and immerse themselves in the VR environment to view the pocket's depth, volume, opening direction, and internal structure to confirm whether it conforms to the biologically significant binding site.

[0074] Furthermore, the "Function" menu modules include the Sequence module (sequence viewing function, used to view the sequence information of molecules), the Water module (water molecule management function, which can perform related management operations on water molecules in the molecular structure), the XPocket module (pocket recognition function, which can identify pocket regions in the molecular structure), the XPharma module (pharmacophore features, drug molecule labeling function, which can label drug molecules for differentiation and analysis), the Atom module (atomic-level manipulation function, which supports fine-grained manipulation of atoms in molecules), and the Dock Result module (dock docking result visualization function, which presents the molecular docking simulation results in a visual way).

[0075] The process of generating potential binding sites, i.e., pocket regions, by performing a pocket recognition algorithm on the protein structure based on the pocket recognition module includes the following steps:

[0076] S2.21. Convert the atomic coordinates of the protein structure into a three-dimensional surface mesh model, and calculate the spatial coordinates and neighboring atomic information of each mesh point for subsequent groove and cavity detection. Specifically, construct a molecular surface potential energy field based on the input protein structure atomic coordinates, and generate a continuous three-dimensional surface mesh through spatial sampling or surface extraction algorithms such as Marching Cubes. Then, for each generated mesh point, perform a neighborhood query in the atomic KD-Tree or octree spatial index structure using its three-dimensional coordinates, retrieve all atoms closest to the mesh point within a preset radius, and record their atom type, van der Waals radius, and charge properties. Finally, output the spatial coordinates, normal vector, and associated set of neighboring atoms for each mesh point.

[0077] S2.22. Perform cavity scanning on the surface mesh of the 3D surface mesh model to identify the concave regions on the protein surface and the cavity volume that can accommodate ligands. Calculate the geometric center, volume, and opening direction of each cavity to form a preliminary set of candidate pocket regions. Specifically: First, based on the normal information of the mesh points and the relative distance between the mesh, identify the concave regions on the surface, and use spatial bounding boxes or voxelization methods to search for continuous voids inside the protein that are not occupied by atoms. Then, perform connectivity analysis on the mesh cells of these concave regions and connected voids to extract the voxel set of each independent cavity. Calculate its geometric center (e.g., centroid), overall cavity volume, and opening direction along the mesh normal or the direction of the maximum opening cross section based on the voxel distribution. Finally, uniformly mark the cavity regions that can accommodate small molecule ligands as candidate pockets, forming a preliminary set of pocket regions.

[0078] S2.23. Calculate the geometric features of each candidate pocket region, including depth, curvature, and pocket opening size, to form a feature vector for the candidate pocket. Specifically, when calculating the geometric features of each candidate pocket region, the system first extracts basic data describing the shape from the inner and outer surface meshes of the cavity, such as the normal of the mesh vertices, the boundary contour lines, and the shortest distance from the mesh to the surrounding atoms. Then, the pocket depth is calculated by the shortest distance along the path from the deepest point inside the pocket to the opening on the outer surface. The principal curvature and average curvature of the concave region are obtained by local surface fitting (such as quadratic surface fitting), and the pocket opening size is obtained by measuring the maximum cross-sectional area at the pocket boundary or the radius of the circumscribed circle of the boundary. Finally, the above values ​​of depth, curvature, and diameter are combined into a feature vector in a preset order as the structured input for subsequent pocket scoring.

[0079] S2.24. The feature vectors of candidate pockets are evaluated by constructing a preset scoring function, and pocket regions whose evaluation results exceed the threshold n are selected and marked as potential binding sites.

[0080] The preset scoring function is as follows:

[0081] ;

[0082] In the formula, The evaluation result, i.e., the evaluation score, For depth, For curvature, For the size of the pocket opening, The weighting coefficient is the depth factor. The curvature is a weighting coefficient. The weighting coefficients for the pocket opening size are dimensionless and determined through expert experience. For reference length (e.g., 10 angstroms or protein feature size);

[0083] S2.25. Perform conformational consistency analysis on the pocket regions generated in step S2.24 based on the conformational integration mechanism, and spatially fuse the pocket regions according to the conformational consistency analysis results to form conformationally stable pocket regions (the volume, shape and opening direction of the pocket will change significantly when the protein structure is in different conformational states (such as open-close conformation or induced fit state).

[0084] The conformational integration mechanism, which analyzes the conformational consistency of pocket regions and forms conformationally stable pocket regions, primarily addresses the problem of unstable binding pockets or inconsistent spatial positions caused by conformational changes in protein structures (such as opening, closing, twisting, or respiratory movements) during virtual screening. Traditional methods typically rely on a single static protein structure for pocket identification, failing to accurately reflect the dynamic flexibility of proteins under physiological conditions. This may identify false pockets that exist only in specific conformations or ignore key binding sites, thus reducing the accuracy and reliability of virtual screening. The conformational integration mechanism, by generating multiple alternative protein conformations and analyzing the spatial consistency and stability of pockets under different conformations, can capture persistent binding pockets in the dynamic behavior of proteins, reducing false positives caused by conformational fluctuations and improving the robustness of pocket identification. Its role is to provide a more realistic and stable binding environment for subsequent pharmacophore construction and molecular docking, ensuring that screened candidate molecules not only match static structures but also adapt to conformational changes in proteins, thereby improving the success rate of virtual screening and its practical application value.

[0085] Based on the conformational integration mechanism, conformational consistency analysis is performed on the pocket region to form a conformationally stable pocket region, including the following steps:

[0086] Based on the pocket region generated in step S2.24, normal mode analysis is performed on the protein structure to obtain the dominant motion direction of the protein structure in the low-frequency intrinsic mode. Multiple lightweight substitution conformations are generated along the dominant mode within a reasonable perturbation range to form a conformation set. ( This set of conformations (total number of conformations) is used to simulate conformational fluctuations that proteins may exhibit under natural flexibility, providing input for subsequent evaluation of the stability of candidate pockets. Specifically, an elastic network model (ENM) is first constructed based on the input protein's three-dimensional atomic coordinates. Using amino acid residues as nodes and distance constraints as springs, a simplified force field describing the overall flexibility of the protein is established. When constructing the elastic network model (ENM) based on the protein's three-dimensional atomic coordinates, the following repeatable parameters are used: Node settings: The α-carbon atom (Ca) of each amino acid residue is used as the node, and the node coordinates follow the original Ca atom coordinates in the PDB file; for residues without Ca such as glycine, the main chain carbon atom (C) coordinates are used as nodes; only the spatial distance of Ca atoms is considered. (Angle) node pairs construct a spring, with the initial force constant of the spring being uniformly set to Subsequently, the Hessian matrix of the elastic network is calculated, and its eigenvalues ​​and eigenvectors are solved. Several of the lowest-frequency intrinsic modes are selected as the dominant motion directions reflecting the overall opening, closing, torsion, and respiratory changes of the protein. Next, using the original structure as a reference, the atomic coordinates are continuously translated and deformed along the positive and negative directions of each low-frequency mode within a preset perturbation amplitude (e.g., ±1~3 normalized modal units), generating multiple slightly shifted and energy-acceptable alternative conformations. Finally, all alternative conformations generated along different modes are compiled into a conformation set. This is used for subsequent conformational consistency analysis and pocket stability assessment;

[0087] For each pocket region, each conformation in the conformation set The pocket region is repositioned and its corresponding spatial occupancy area is calculated, including the geometric center, pocket boundary mesh, and local surface shape. The projection results of the pocket region under different conformations are obtained, forming a conformational mapping pocket set. This is used to describe the spatial fluctuation characteristics of pockets as their conformation changes. Specifically, for each pocket region, during mapping within the conformation set, the system first maps the pocket geometric center, boundary mesh, and surface description parameters from the original structure to each conformation using coordinate alignment (e.g., rigid body superposition based on amino acid residues). In the coordinate system; subsequently, in the configuration Local mesh sampling and cavity reconstruction are re-performed on the 3D surface. The spatial occupancy area of ​​the pocket is updated according to the new positions of the surrounding atoms, and the geometric properties of the pocket in this conformation, such as geometric center, boundary profile, deepest point, local concavity shape, and opening direction, are recalculated. Through the above steps, the pocket in the conformation can be obtained. The complete spatial projection below forms the corresponding conformational mapping pocket. Finally, the above mapping process is repeated for all conformations in the conformation set to obtain the conformation mapping pocket set. This is used for subsequent conformational consistency assessment and stable pocket extraction;

[0088] Perform cross-conformity consistency analysis on the conformational mapping pocket set to generate a conformational consistency score. This includes overlap and center offset;

[0089] Among them, overlap For (dimensionless):

[0090] ;

[0091] Center offset is (Unit: angstroms)

[0092] ;

[0093] Then normalize the scores to make them more consistent.

[0094] ;

[0095] Consistency score for:

[0096] ;

[0097] In the formula, For the first in the set of constructs The space occupied by each pocket For the first in the set of constructs The space occupied by each pocket To calculate the volume of a spatial region, For the first and the Euclidean distance of the geometric centers of the pockets The adjustable parameter controls the intensity of the offset's influence on the score (positive number, set according to pocket size). The weighting coefficient for overlap. The weighting coefficients for the center offset are dimensionless and determined through expert experience. The maximum allowable offset distance is empirically set at 5 angstroms;

[0098] Based on conformational consistency score With preset threshold Filtering will satisfy Candidate pocket regions are marked as conformationally stable candidate pockets; otherwise, they are considered unstable pockets that are susceptible to conformational fluctuations and are removed.

[0099] Spatial fusion processing is performed on the conformationally stable pocket region to obtain the fused conformationally stable pocket region;

[0100] Furthermore, obtaining the fused conformationally stable pocket region involves the following steps:

[0101] The conformational mapping pocket set is re-aligned (so that all pockets can be compared under the same reference frame) and scaled (to eliminate the bias caused by rotation, translation or small scale differences between different conformations and to ensure the comparability of subsequent fusion analysis) so that it is in the same reference coordinate system used in the consistency analysis in step S2.25.

[0102] Based on the geometric center offset range of the conformational mapping pockets after coordinate alignment and scale normalization, a Boolean fusion strategy is performed on the boundary meshes of each conformational pocket to fuse them (using the intersection strategy of regions where most conformations coexist, or the soft union method based on weight allocation, to integrate the spatial information of pockets under different conformations and obtain a preliminary fused pocket region). Specifically, when performing Boolean fusion on the conformational mapping pockets, the system first completes the coordinate alignment and scale normalization of each conformational pocket in the same reference coordinate system, so that its geometric center and overall size can be directly compared; then, the offset range of each pocket is calculated based on the aligned geometric center position to determine which mesh points coexist in most conformations; next, a Boolean intersection operation (AND) is performed on the boundary meshes of each conformational pocket to retain the spatial regions occupied by all conformations simultaneously, thereby removing pocket parts that are occasional or unique to a single conformation. At the same time, a soft union processing can be performed on some regions in combination with a weight strategy to integrate the spatial information under different conformations; finally, the fused conformational stable pocket region is generated through mesh reconstruction and voxel sampling to ensure that the extracted pocket region exists in multiple conformations and is geometrically stable.

[0103] The fused pocket region undergoes voxel sampling and mesh reconstruction to calculate the common stable spatial volume across all conformations, forming a fused conformationally stable pocket region. This step removes pocket portions specific to occasional or single conformations. In this way, stable spatial features present in multiple conformations can be extracted, generating the final fused conformationally stable pocket region. Specifically, during voxel sampling and mesh reconstruction of the fused pocket region, the system first divides the fused boundary mesh into regular 3D voxel meshes, assesses the occupancy of each voxel in the mapped pockets of each conformation, and counts the frequency of each voxel across all conformations. Subsequently, voxels occupied in most or all conformations are selected based on a preset threshold, retaining these commonly existing stable voxels as the core pocket region. Then, surface reconstruction (e.g., Marching) is performed on the selected voxels. The Cubes algorithm generates a continuous three-dimensional boundary mesh, while simultaneously calculating the overall volume, geometric center, and surface shape of the fusion pocket. Finally, the resulting continuous mesh is the conformationally stable pocket region after fusion, ensuring that the region exists in multiple conformations and has stable geometric features, which can be used for subsequent pharmacophore construction and molecular docking.

[0104] S2.3 Users can further adjust the perspective and perform position calibration on the identified pocket region based on the geometric morphological features of the protein structure and spatial parameters such as pocket boundaries, center point, and opening direction to ensure that the identification results are consistent with the actual geometric morphology of the protein structure. This is used for subsequent pharmacophore construction and docking box setting, and to confirm the key parameters of the pocket region, including the center point position, boundary range, and opening direction. These parameters will be used to set the PocketCenter and Pocket Size in subsequent docking steps.

[0105] S3. Based on the conformationally stable pocket region, the pharmacophore analysis module generates initial functional feature points, and interactively edits and corrects the spatial location, type and tolerance parameters of the pharmacophore in the VR environment to form a pharmacophore template.

[0106] In this embodiment, based on the conformationally stable pocket region determined in step S2, the user enters the visual perspective near the pocket, providing a spatial positioning basis for pharmacophore identification.

[0107] The process of forming a pharmacophore template includes the following steps:

[0108] S3.1. Based on the pharmacophore analysis module (XPharma module), initial pharmacophore feature points are generated by analyzing the local protein structure, key residue distribution, and pocket shape. These feature points include aromatic rings, hydrogen bond donors, hydrogen bond acceptors, hydrophobic regions, positively charged centers (or cations), and negatively charged centers (or anions). Specifically, the local protein structure within the conformationally stable pocket region is analyzed, including amino acid residue types, spatial arrangement of the main chain and side chains, and their accessible surfaces. Then, key functional groups (such as hydroxyl, amino, carboxyl, and aromatic rings) and their distribution locations are identified. Combined with pocket geometry and depth information, spatial regions that interact with ligands through hydrogen bonding, hydrophobic interactions, ionic interactions, or aromatic stacking are determined. Finally, initial pharmacophore feature points are automatically generated within these regions and identified according to type (such as hydrogen bond donors, hydrogen bond acceptors, aromatic rings, hydrophobic points, and positive / negative charge centers). Their three-dimensional coordinates, direction vectors, and spatial tolerances are recorded for subsequent interactive adjustment and optimization in the VR environment, forming a complete set of initial pharmacophore points.

[0109] S3.2 Display the initial feature points of the pharmacophore near the pocket region with different identifiers (such as spheres, icons, color codes) and overlay them with the protein structure to facilitate users to observe their geometric relationship with the pocket.

[0110] Users can utilize VR 3D manipulation tools such as "Rotate, Move, Scale" to observe from multiple angles whether pharmacophores are in reasonable spatial positions, such as whether they are close to key residues or key functional groups. Users can select any pharmacophore and modify its type to aromatic ring, hydrogen bond donor / acceptor, hydrophobic, anion / cation, etc., through the VR controller to correct points that may be automatically identified inaccurately by the system. Users can directly grab pharmacophores through the VR controller and drag their spatial positions to make them more accurately fit the spatial features inside the protein pocket, such as being close to hydrogen bond residues or hydrophobic patches. Users can set matching tolerances for each pharmacophore, such as geometric parameters like recognition radius and direction vector restrictions, for spatial scoring standards during subsequent compound matching. If the automatically identified feature points do not meet the actual pocket requirements, users can delete redundant points and add new pharmacophores based on the VR fine view to improve matching accuracy.

[0111] S3.3. Users check the distance, orientation and geometric layout between the pharmacophore and protein residues in the VR scene to ensure that the generated template can truly reflect the optimal binding mode of the ligand in the pocket.

[0112] S3.4 After the inspection is completed, the user saves the pharmacophore as a pharmacophore template for use in subsequent virtual screening and molecular docking processes.

[0113] S4. Pharmacophore screening process based on pharmacophore templates, matching and scoring candidate compounds one by one, and generating a preliminary screening molecule list based on matching scores.

[0114] In this embodiment, candidate compounds are matched and scored one by one, and a preliminary screening molecule list is generated based on the matching scores. This includes the following steps:

[0115] In this embodiment, the user loads all candidate compounds using methods such as "Load From File / Load From Internet / Load From Resources", so that they can be called one by one by the screening algorithm in the VR environment;

[0116] S4.1. The pharmacophore template is used as the matching rule input to the screening module. The screening module is used to automatically evaluate and sort the candidate compound set based on the preset pharmacophore template and related screening conditions, thereby identifying molecules with a high degree of matching with the target protein pocket. The module compares and calculates the three-dimensional structure of each candidate compound with the pharmacophore template according to parameters such as pharmacophore feature point type, spatial tolerance, geometric angle and missing match tolerance, generates a matching score, and outputs a preliminary screening molecule list based on the score results, providing a candidate molecule set for subsequent fine molecular docking.

[0117] S4.2. Set screening conditions according to the type and tolerance of pharmacophore feature points, including at least: matching tolerance radius (the range of spatial deviation allowed for each feature point), minimum requirement for the number of matching points (the minimum number of pharmacophore points to match), allowable proportion of missing matching points (e.g., allowing 1 missing point), and geometric angle deviation limit (e.g., directional H-bond feature). These screening conditions affect the final matching score of each compound.

[0118] S4.3. Pre-filtering is performed based on compound library information (such as molecular weight, polarity, structural complexity, Lipinski rules, etc.). This involves extracting the basic physicochemical parameters (such as molecular weight, cLogP, number of hydrogen bond donors / acceptors, polar surface area, structural complexity, number of rotatable bonds, etc.) of each compound in the library and comparing these parameters with preset drug compatibility rules, such as Lipinski rules restricting the molecular weight, lipophilicity, and hydrogen bond characteristics of oral drugs, and Veber rules restricting the polar surface area and number of rotatable bonds. Simultaneously, the system combines a structural rule library to exclude molecules containing highly reactive groups, toxicity warning fragments, and non-synthetic structures. Only compounds that simultaneously meet all filtering conditions are retained for the next stage, thereby rapidly narrowing the candidate space and improving the efficiency and reliability of subsequent pharmacophore matching. The pre-filtered set of candidate compounds eligible for pharmacophore matching is then output to eliminate molecules that do not meet the pharmacophore structural requirements.

[0119] S4.4. Traverse each molecule in the candidate compound set and perform geometric alignment calculations of its 3D structure with the pharmacophore template for pharmacophore feature points. Specifically, the system first generates or loads a 3D conformation for each compound and extracts functional groups (such as hydrogen donor atoms, hydrogen acceptor atoms, aromatic ring centers, hydrophobic carbon chains, positive and negative charge centers, etc.) that can serve as pharmacophore matching points. Then, these functional points are paired with corresponding feature points in the pharmacophore template according to type, and a geometric alignment algorithm (such as least squares rigid body alignment) is used to calculate the optimal spatial transformation matrix, so that the feature point cloud of the compound and the 3D feature point cloud of the pharmacophore template overlap to the greatest extent in space and direction. During the alignment process, the distance deviation, direction vector matching degree and tolerance range are evaluated simultaneously, and matching methods that cannot meet the geometric constraints are eliminated. Finally, the optimal alignment conformation of the compound and the pharmacophore template is obtained, providing a basis for subsequent matching score calculations.

[0120] S4.5 Compare the degree of matching between the functional group positions of the compound and the pharmacophore template, and perform multi-dimensional energy-weighted pharmacophore matching score to generate an energy-weighted matching score;

[0121] The multi-dimensional energy-weighted pharmacophore matching scoring method primarily addresses the oversimplification of scoring functions in traditional pharmacophore screening methods. Traditional methods typically score based solely on geometric matching of feature points (e.g., distance, direction), neglecting the local chemical microenvironment of the protein binding pocket (e.g., charge distribution, acidity / basicity) and the differences in energy contributions between different interaction types. This results in scoring results that fail to accurately reflect the true physicochemical trends of molecular binding potential, thus reducing the hit rate and lead compound quality in virtual screening. The multi-dimensional energy-weighted matching scoring method integrates basic energy contribution, local chemical microenvironment correction, spatial geometric deviation attenuation, and directional interaction weights to construct a scoring system that more closely approximates real biophysical processes. Its advantage lies in elevating simple geometric matching to an energy-weighted multi-dimensional evaluation, enabling the scoring to not only distinguish whether a molecule "similar" to the pharmacophore template but also predict the "quality" and "strength" of its binding interaction. Its effect is to significantly improve the enrichment effect and reliability of virtual screening, prioritizing the selection of high-quality candidate molecules that are highly complementary to the target in both geometry and interaction energy.

[0122] Perform multi-dimensional energy-weighted pharmacophore matching scoring, including the following steps:

[0123] S4.51. Construct fundamental energy contribution coefficients for different pharmacophore sites (such as hydrogen bond acceptors, hydrogen bond donors, aromatic rings, hydrophobic sites, positive and negative charge sites, etc.). The energy contribution coefficient of the basic energy is corrected based on the protein chemical microenvironment (such as local acidity / alkalinity and charge density) to generate an energy-corrected contribution coefficient. This allows the pharmacophore score to reflect the influence of the protein's real chemical environment on the strength of the interaction, such as local acidity / basicity and charge density, thereby improving the ability of the matching score to express the trend of the actual binding energy and improving the accuracy of virtual screening.

[0124] After energy correction:

[0125] ;

[0126] ;

[0127] In the formula, This refers to the numbering or indexing of pharmacophore feature points. For the first Quantitative factors of chemical microenvironment characteristics, such as acidity / basicity and charge density. For the corresponding number Weighting coefficients for microenvironment-like characteristics are used to adjust the intensity of the influence of different environmental factors on energy correction. This refers to the local acid-base parameters near the pharmacophore point. The local charge density near the pharmacophore. The weighting of the effect of local acidity / alkalinity on energy correction. The weighting of the effect of local charge density on energy correction;

[0128] S4.52. Based on the spatial deviation between the matching point in the candidate molecule and the pharmacophore template point (the three-dimensional coordinates are finally determined after VR editing), a Gaussian distance decay weight is constructed for each pharmacophore feature point (the pharmacophore feature point here is the final set of pharmacophore feature points after VR editing of the initial functional feature points). Based on the spatial deviation between the pharmacophore template point and the corresponding point of the candidate molecule, a Gaussian spatial decay weight is constructed to measure the "degree of geometric position matching". By penalizing the spatial deviation, only candidate molecules that are geometrically close to the template can obtain a high score, making the scoring system more in line with the spatial specificity requirements of molecular recognition and improving the structural rationality and target selectivity of the screening results.

[0129] The Gaussian distance decay weight is:

[0130] ;

[0131] In the formula, The spatial Gaussian weights for the pharmacophore points (i.e., Gaussian distance decay weights) are used to describe the decay caused by spatial bias. The distance between the candidate molecule matching point and the pharmacophore template point is expressed in Å (3D spatial distance deviation). This is a spatial tolerance parameter, representing the permissible spatial offset for the corresponding pharmacophore type (the larger the value, the higher the tolerance), and is measured in Å.

[0132] S4.53. For directional pharmacophore features (hydrogen bonds, ionic interactions), an angle-dependent weighting factor is introduced. This factor is used to evaluate the consistency between the interaction direction of candidate molecules and the template direction, ensuring that interactions with directional requirements (e.g., hydrogen bonds must be close to linear) can be accurately modeled. Molecules that do not meet the interaction geometry requirements will automatically receive a lower score, thereby improving the physicochemical accuracy of the pharmacophore template and the reliability of screening.

[0133] Among them, an angle-dependent weighting factor is introduced. (used to describe the impact of directional deviation on matching quality):

[0134] ;

[0135] In the formula, The angle between the direction of the candidate molecule interaction and the direction of the template pharmacophore. For optimal interaction angles, such as the directionality of hydrogen bonds, the common range is 120° to 180°. This is the angle tolerance parameter, representing the allowable range of angle offset.

[0136] S4.54, Contribution coefficient The weighting factors of spatial weight and angle dependence are jointly integrated to calculate a multidimensional weighted score for each pharmacophore feature point, and the sum of the weighted scores of all pharmacophore feature points is used as the overall energy weighted matching score of the candidate molecule.

[0137] Among them, the calculation of multidimensional weighted scores :

[0138] ;

[0139] Overall energy-weighted matching score for:

[0140] ;

[0141] S4.6 Sort all candidate compounds from high to low according to their energy-weighted matching scores and output the initial screening list.

[0142] S5. Based on the actual position and range of the conformational stability pocket region, perform fine molecular docking on the initial screening molecule list to obtain the binding conformation and energy score of each molecule, which is used to screen the final candidate molecules.

[0143] In this embodiment, the binding conformation and energy score of each molecule are obtained, including the following steps:

[0144] S5.1 Based on the initial screening molecule list, each candidate molecule in the initial screening molecule list is imported into the docking module, and the three-dimensional mesh of the conformational stability pocket region is used as the constraint space.

[0145] S5.2. Perform conformational sampling and rotation-translation optimization on each candidate molecule to generate multiple initial binding postures that can enter the pocket region. Perform collision checks and geometric reachability verification on each posture to generate the verified conformation. Specifically, based on the internal conformational degrees of freedom of the candidate molecule (such as rotation bonds, ring conformations, and side chain flips), generate several different molecular conformations. Within the conformationally stable pocket region, perform systematic translation and rotation scans on each conformation in three-dimensional space to allow the molecule to attempt to enter the pocket at different positions and orientations. Simultaneously, perform collision detection and geometric reachability verification (when verifying the potential binding conformations of each candidate molecule, first perform conformational sampling and rotation-translation optimization on each conformation). Collision detection is performed on each atom in the pocket grid and the residues within the pocket to determine whether the distance between atoms or between atoms and the pocket boundary is lower than a preset minimum tolerance threshold. If there is overlap or severe proximity, the conformation is marked as inaccessible. Then, geometric reachability verification is performed, that is, the connectivity analysis of the path from the pocket entrance to each functional group of the molecule is performed to ensure that each key group can successfully reach its target binding site. If the path is blocked or there is a spatially closed region, the conformation is eliminated. Conformations that have severe spatial conflicts with protein atoms are eliminated, thereby obtaining multiple sets of initial binding postures that can enter the pocket region, ensuring sufficient coverage of the docking search space.

[0146] S5.3. The local energy minimization process is repeated on the verified conformations. A final comprehensive binding energy score is generated by combining interaction strength, conformational stability, and pocket space complementarity. Specifically: First, energy minimization is performed on the molecular atoms within a local area, including adjusting bond lengths, bond angles, dihedral angles, and non-bonded interactions to eliminate local stress and obtain the optimal conformational state. Then, the interaction energies between the molecule and pocket residues are calculated, including van der Waals forces, electrostatic interactions, hydrogen bonds, and hydrophobic interactions, while also considering the molecule's own conformational stability, such as internal strain energy and the adaptability of flexible groups. Based on this, the complementarity of the molecule in the pocket space is analyzed, i.e., the degree of matching between the molecular volume and the pocket groove, and the proximity of the key group positions to the key pocket residues. Finally, the above factors are weighted according to preset weights to generate a final comprehensive binding energy score for each candidate molecule, used for ranking and screening the optimal binding conformation.

[0147] S5.4. Sort all candidate molecules according to the comprehensive binding energy score to generate a candidate molecule set. Then, use the docking visualization module to present the protein-ligand binding conformation, key interactions, and docking search space (including the size and position of the Dock Box) in three dimensions. (This step is implemented through the Dock Result module. The docking visualization module provides a three-dimensional visualization of the protein-ligand binding conformation and docking search space. The center position, size, and orientation of the docking search space can be adjusted by the user. Candidate conformations are sorted from low to high binding energy and displayed in an overlay manner. Key interaction sites (including hydrogen bonds, salt bridges, and hydrophobic interactions) are marked with color mapping. The module also provides functions for selecting, saving, exporting, and further optimizing conformations to assist users in interactive analysis and structure optimization.)

[0148] Example 2: This example provides a virtual screening system for pharmacophores based on virtual reality, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the virtual screening method for pharmacophores based on virtual reality described in any one of the above examples.

[0149] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A virtual screening method for pharmacophores based on virtual reality, characterized in that, Includes the following steps: S1. Input protein structures and compound libraries into the VR scene as screening objects through the molecular resource loading module; S2. Analyze the pocket region in the protein structure using the pocket recognition module, and perform three-dimensional observation and position calibration of the pocket region in the VR scene to determine the actual position and range of the pocket region; Among them, conformational consistency analysis of pocket regions is performed based on conformational integration mechanism, and spatial fusion of pocket regions is performed based on conformational consistency analysis results to form conformationally stable pocket regions; S3. Based on the conformationally stable pocket region, the pharmacophore analysis module generates initial functional feature points, and interactively edits and corrects the spatial location, type and tolerance parameters of the pharmacophore in the VR environment to form a pharmacophore template. S4. Pharmacophore screening process based on pharmacophore templates, matching and scoring candidate compounds one by one, and generating a preliminary screening molecule list based on matching scores. S5. Based on the actual position and range of the conformational stability pocket region, perform fine molecular docking on the initial screening molecule list to obtain the binding conformation and energy score of each molecule, which is used to screen the final candidate molecules. In step S2, the pocket recognition module is used to analyze the pocket region in the protein structure, and the pocket region is observed and its position is calibrated in three dimensions in a VR scene, including the following steps: S2.

1. Rotate, move, and magnify proteins through three-dimensional manipulation; S2.

2. Based on the pocket recognition module, the pocket recognition algorithm is executed on the protein structure to generate potential binding sites, i.e. pocket regions, and the pocket regions are marked in a highlighted manner in the VR scene. S2.

3. Position calibration of the identified pocket regions based on the geometric morphological features and spatial parameters of protein structure; In step S2.2, the pocket recognition algorithm is applied to the protein structure based on the pocket recognition module to generate potential binding sites, i.e., pocket regions, including the following steps: S2.

21. Convert the atomic coordinates of the protein structure into a three-dimensional surface mesh model, and calculate the spatial coordinates and neighborhood atomic information of each mesh point; S2.

22. Perform cavity scanning on the surface mesh of the three-dimensional surface mesh model to identify the concave regions on the protein surface and the cavity volume that can accommodate ligands, and calculate the geometric center, volume and opening direction of each cavity to form a preliminary set of candidate pocket regions. S2.23 Calculate the geometric features for each candidate pocket region to form the feature vector of the candidate pocket; S2.

24. The feature vectors of candidate pockets are evaluated by constructing a preset scoring function, and pocket regions whose evaluation results exceed the threshold n are selected and marked as potential binding sites. S2.

25. Perform conformational consistency analysis on the pocket regions generated in step S2.24 based on the conformational integration mechanism, and spatially fuse the pocket regions according to the conformational consistency analysis results to form conformationally stable pocket regions. In step S2.25, forming a conformationally stable pocket region includes the following steps: Based on the pocket region generated in step S2.24, normal mode analysis is performed on the protein structure to obtain the dominant motion direction of the protein structure in the low-frequency intrinsic mode, and multiple lightweight substitution conformations are generated along the main mode within a reasonable perturbation range to form a conformation set. For each pocket region, each conformation in the conformation set The pocket region is repositioned and its corresponding spatial occupancy area is calculated to obtain the projection results of the pocket region under different conformations, forming a conformational mapping pocket set. Perform cross-conformity consistency analysis on the conformational mapping pocket set to generate a conformational consistency score. ; Based on conformational consistency score With preset threshold Filtering will satisfy Candidate pocket regions are marked as conformationally stable candidate pockets; Spatial fusion processing is performed on the conformationally stable pocket region to obtain the fused conformationally stable pocket region; The spatial fusion process for the conformationally stable pocket region to obtain the fused conformationally stable pocket region involves the following steps: Realign and scale the conformation mapping pocket set; Based on the geometric center offset range of the conformational mapping pockets after coordinate alignment and scale normalization, a Boolean fusion strategy is applied to the boundary meshes of each conformational pocket for fusion. Voxelized sampling and mesh reconstruction are performed on the fused pocket region to calculate the stable spatial volume maintained under each conformation, thus forming the fused conformational stable pocket region; In step S4, candidate compounds are matched and scored one by one, and a preliminary screening molecule list is generated based on the matching scores. This includes the following steps: S4.1 Input the pharmacophore template as a matching rule into the filtering module; S4.2 Set screening conditions based on the type and tolerance of pharmacophore feature points; S4.3 Perform pre-filtering based on compound library information and output the candidate compound set after pre-filtering; S4.

4. Traverse each molecule in the candidate compound set and perform geometric alignment calculations of the pharmacophore feature points between its three-dimensional structure and the pharmacophore template. S4.5 Compare the degree of matching between the functional group positions of the compound and the pharmacophore template, and perform multi-dimensional energy-weighted pharmacophore matching score to generate an energy-weighted matching score; S4.6 Sort all candidate compounds from high to low according to their energy-weighted matching scores and output the initial screening list; In step S4.5, a multi-dimensional energy-weighted pharmacophore matching score is performed, including the following steps: S4.

51. Constructing basic energy contribution coefficients for different pharmacodynamic clusters. Based on the protein chemical microenvironment, the basic energy contribution coefficient is energy corrected to generate an energy-corrected contribution coefficient. ; S4.

52. Based on the spatial deviation between the matching point and the pharmacophore template point in the candidate molecule, construct a Gaussian distance decay weight for each pharmacophore feature point; S4.

53. To address the characteristics of directional pharmacophores, an angle-dependent weighting factor is introduced to evaluate the consistency between the interaction direction of candidate molecules and the pharmacophore template direction. S4.54, Contribution coefficient The weighting factors of spatial weight and angle dependence are jointly integrated to calculate a multidimensional weighted score for each pharmacophore feature point, and the sum of the weighted scores of all pharmacophore feature points is used as the overall energy-weighted matching score of the candidate molecule.

2. The virtual screening method for pharmacophores based on virtual reality according to claim 1, characterized in that: In step S3, forming a pharmacophore template includes the following steps: S3.1 The pharmacophore analysis module generates initial pharmacophore feature points based on the local structure of the protein, the distribution of key residues, and the pocket shape. S3.2 Display the initial feature points of the pharmacophore with different identifiers near the pocket region and overlay them with the protein structure to facilitate users to observe its geometric relationship with the pocket. S3.3, Check the distance, orientation, and geometric layout between the pharmacophore and protein residues; S3.4 After the inspection is completed, save the pharmacophore as a pharmacophore template.

3. The virtual screening method for pharmacophores based on virtual reality according to claim 1, characterized in that: In step S5, the binding conformation and energy score of each molecule are obtained, including the following steps: S5.1 Based on the initial screening molecule list, each candidate molecule in the initial screening molecule list is imported into the docking module, and the three-dimensional mesh of the conformational stability pocket region is used as the constraint space. S5.2 Perform conformational sampling and rotation-translation optimization on each candidate molecule to generate multiple initial binding postures that can enter the pocket region, and perform collision checks and geometric reachability verification on each posture to generate the verified conformation. S5.

3. Perform local energy minimization again on the verified conformation, and generate the final comprehensive binding energy score by combining the interaction strength, conformation stability and pocket space complementarity. S5.

4. Sort all candidate molecules according to the comprehensive binding energy score, generate a candidate molecule set, and present the binding conformation of protein and ligand, key interactions and docking search space in three dimensions through the docking visualization module.

4. A virtual screening system for pharmacophores based on virtual reality, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes a computer program to implement the virtual screening method for pharmacophores based on virtual reality as described in any one of claims 1-3.

Images