A virtual screening method based on fast retrieval of ligand docking poses

Abstract

The application discloses a ligand docking pose virtual screening method based on fast retrieval, comprising the following steps: preprocessing ligand conformation information data, and establishing an index table with a retrieval tree space index structure; inputting a known active ligand conformation, and utilizing the retrieval tree space index structure to quickly retrieve and screen potential candidate conformations similar to the query conformation; taking the most similar top-k query result as a top-k conformation result; evaluating the top-k conformation result obtained through retrieval, comparing the actual RMSD value of the output top-k candidate conformation and the natural conformation, verifying the accuracy of the screening result, and further optimizing the screening strategy. The application is based on the spatial data of a ligand molecule, utilizes a three-dimensional space retrieval tree to organize a data structure and create an index, so that the search range is reduced, and therefore the best docking pose structure can be quickly retrieved from massive ligand structure data, and the prediction performance is effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of computer-aided drug design and content retrieval, specifically relating to a virtual screening method for ligand docking postures based on rapid retrieval. Background Technology

[0002] Predicting the docking posture between proteins and ligands plays a crucial role in computer-aided biopharmaceutical research, and improving prediction and screening efficiency is a key aspect. With the emergence of protein design technology, more potential proteins are being discovered, exhibiting increasingly diverse properties and functions. Consequently, the demand for rapidly screening optimal ligand docking postures is continuously rising, and new proteins are constantly emerging. However, the difficulty in quickly updating relevant protein property data and ligand docking posture data poses a significant challenge to computer-aided drug prediction.

[0003] Traditional methods typically generate several combinations of docking postures, from which the most suitable set is selected. The logic of posture prediction and selection requires utilizing existing knowledge and comprehensively considering intramolecular forces, intermolecular forces, and local and global protein information. However, using a "blind search" approach is time-consuming, costly, and lacks accuracy. Furthermore, neural network-based docking posture prediction methods usually require massive amounts of well-labeled posture docking data, while existing biomedical information is limited and data labeling is incomplete. Therefore, existing methods inevitably miss or overlook some important intermolecular information.

[0004] Drug discovery methods also include high-throughput screening-based drug discovery (HTSBDD) and structure-based drug discovery (SBDD). HTSBDD methods utilize phenotypic screening, but are based on biochemical experiments and are therefore less efficient. Structure-based drug discovery methods, which use molecular docking calculations, require high computational power, and accuracy remains a bottleneck.

[0005] Content-based fast retrieval methods can leverage existing information databases and perform rapid filtering through feature value indexing, improving search efficiency and ensuring the validity of results. Currently, spatial data retrieval methods mainly include spatial indexing and dimensionality reduction. The former is a data structure used to organize and manage spatial data, effectively reducing the search scope and improving query efficiency. It has numerous successful applications in fields such as Geographic Information Systems (GIS), computer graphics, and robot navigation. Especially in GIS, spatial indexing methods play a crucial role in traffic flow analysis and land use planning. In the field of drug discovery, indexing methods also exist, specifically fragment-based drug discovery (FBDD). These methods use small molecule fragments as starting points and employ indexing or other methods to find lead compounds that bind to the target. However, they do not address the spatial data contained in the ligand during pose docking, nor do they utilize corresponding indexes to optimize the search strategy. Summary of the Invention

[0006] Purpose of the invention: To overcome the shortcomings of existing technologies, this invention provides a virtual screening method for ligand docking postures based on rapid retrieval. It employs a spatial search tree, indexing the spatial positional relationships of ligand conformation molecules, and optimizes the splitting and screening strategies. This avoids the complexity of stepwise trial and error in traditional prediction methods and the loss of important intermolecular information in machine learning prediction methods. It fully utilizes existing ligand conformation molecule information and greatly improves the accuracy and efficiency of predicting ligand molecule docking postures.

[0007] Technical Solution: To achieve the above objectives, this invention provides a virtual screening method for ligand docking postures based on rapid retrieval, comprising the following steps:

[0008] S1: Preprocess the ligand conformation information data to establish an index table with a retrieval tree spatial index structure;

[0009] S2: Input a known active ligand conformation, use the retrieval tree spatial index structure to quickly retrieve and filter potential candidate conformations similar to the query conformation, and take the most similar top-k query results as the top-k conformation results;

[0010] S3: Evaluate the top-k conformation results obtained from the search, compare the actual RMSD values ​​of the output top-k candidate conformations with those of the native conformations, verify the accuracy of the screening results, and further optimize the screening strategy.

[0011] Furthermore, in step S1, the bioinformatics data of the ligand conformations are preprocessed based on the spatial positional relationship of the ligand conformations. The specific processing steps are as follows:

[0012] A1: Process the CASF-2016 ligand docking candidate dataset in PDBbind to obtain the biological information of different conformations of a single ligand, and at the same time obtain the RMSD values ​​between different candidate conformations and the natural ligand conformations.

[0013] A2: Extract the spatial position information of atoms in each conformation, convert their corresponding three-dimensional structure into a set of feature points, each feature point containing coordinate and type information; cover all feature points of the conformation in space with a minimum bounding box;

[0014] A3: Construct a retrieval tree from the set of feature points in a hierarchical structure, where each leaf node stores a feature point and each non-leaf node stores the minimum bounding box of its child nodes.

[0015] Furthermore, the biological information of different conformations of a single ligand in step A1 includes molecular system information, constituent atom information, bond value information, and substructure information.

[0016] Furthermore, step S2 specifically includes the following steps:

[0017] B1: Extract the spatial position information of atoms in the known active ligand conformations, and convert the three-dimensional structure of the known active ligand conformations into a set of feature points, each of which contains coordinate and type information;

[0018] B2: The feature points of the known active ligand conformations are searched in the established search tree, and similarity matching is performed with the candidate conformations to be screened in a hierarchical manner. The RMSD value, i.e., the similarity score, is calculated based on the number and distance of the matched feature points.

[0019] B3: Sort the candidate conformations to be screened according to the similarity score, select the candidate conformations with the highest scores as candidate ligand conformations, and obtain the top-k docking posture results.

[0020] Furthermore, in step B2, the feature points of the known active ligand conformations are searched in the established search tree, and the search includes top-down search and bottom-up search.

[0021] The top-down retrieval process is as follows: First, the region where the instance is located is found from the root node using the retrieval method. Then, the region is further segmented based on the target's attributes to find the next layer region. This process is iterated until the final conformation output result is obtained.

[0022] Bottom-up retrieval specifically involves: first, determining the atomic positional relationships based on preprocessed data, and then distinguishing different instances through clustering and metric learning.

[0023] Furthermore, step B2 employs a similarity score calculation method based on spatial location relationships, as detailed below:

[0024] The formula for calculating the similarity score of candidate conformations based on spatial location relationships is as follows:

[0025]

[0026] Where x ij x is the j-th conformation in the top-k results j All points in the neighborhood, y i y represents all points in the y-neighborhood of the natural conformation, and n is the sum of the number of atoms contained in a single ligand. The distance between each point in formula (1) is defined as the deviation difference or deviation error of the candidate conformation. The smaller the RMSD value, the closer the candidate conformation molecule is to the known active conformation.

[0027] In order to minimize the value of formula (1), that is, to obtain the docking posture closest to the natural conformation, the minimum value is taken in the calculation set, and the resulting conformation is the conformation molecule with the smallest deviation in the candidate conformation library. After obtaining the conformation with the smallest deviation in the search tree, backtracking is performed to obtain the conformation with the smallest deviation error outside the obtained conformation. This process is repeated until the top-k conformation query results are output.

[0028] Further, step S3 specifically includes:

[0029] C1: Evaluate the screening results; calculate the RMSD value of all candidate conformations, sort the results to obtain the k smallest conformations, compare them with the top-k results obtained from the retrieval, calculate the accuracy and compare the time taken;

[0030] C2: Optimize the filtering strategy of the search tree; optimize the splitting strategy, rebuild the search tree index items, and re-evaluate. The strategy with the highest accuracy was obtained through multiple experiments.

[0031] Furthermore, the optimization splitting strategy in step C2 includes linear splitting, binary splitting, quadtree splitting, etc.

[0032] The filtering in step C2 of this invention is mainly based on the idea of ​​ThresholdAlgorithm (TA). When merging regions, it is necessary to merge the two regions that are most "cognitively" related together. Therefore, the weight of different attribute parameters will affect the sorting, thereby affecting the merging and splitting strategy.

[0033] This invention utilizes spatial data of ligand molecules and a three-dimensional spatial retrieval tree to organize data structures and create an index, thereby reducing the search range and enabling the rapid retrieval of the optimal docking posture structure from massive ligand structure data, effectively improving prediction performance.

[0034] This invention utilizes the spatial relationship of different ligand conformations to index the ligand database, enabling rapid screening of the top-k docking posture structures most similar to known conformations from a large-scale ligand candidate library, while ensuring the accuracy of the selected conformations.

[0035] Beneficial effects: Compared with existing technologies, this invention uses a spatial retrieval tree to index the spatial positional relationships of ligand conformation molecules, optimizes the splitting and screening strategies, avoids the complexity of stepwise trial and error in traditional prediction methods and the defects of losing important intermolecular information in machine learning prediction methods, realizes the full utilization of existing ligand conformation molecule information, and greatly improves the accuracy and efficiency of predicting ligand molecule docking posture. It can quickly retrieve the optimal docking posture structure from massive ligand structure data, and can play an important role in computer-aided drug design in drug discovery and design. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the overall process of the method of the present invention;

[0037] Figure 2 This is a schematic diagram of the conformational preprocessing flow for the ligand docking posture of the present invention;

[0038] Figure 3 This is a schematic diagram of the retrieval and query process for ligand docking postures in this invention. Detailed Implementation

[0039] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0040] This invention provides a virtual screening method for ligand docking postures based on rapid retrieval, such as... Figure 1 As shown, it includes the following steps:

[0041] S1: Preprocess the ligand conformation information data to establish an index table with a retrieval tree spatial index structure;

[0042] S2: Input a known active ligand conformation, use the retrieval tree spatial index structure to quickly retrieve and filter potential candidate conformations similar to the query conformation, and take the most similar top-k query results as the top-k conformation results;

[0043] S3: Evaluate the top-k conformation results obtained from the search, compare the actual RMSD values ​​of the output top-k candidate conformations with those of the native conformations, verify the accuracy of the screening results, and further optimize the screening strategy.

[0044] Reference Figure 2 In step S1 of this embodiment, the bioinformatics data of the ligand conformations are preprocessed based on the spatial positional relationship of the ligand conformations. The specific processing steps are as follows:

[0045] A1: Process the CASF-2016 ligand docking candidate dataset in PDBbind to obtain biological information of different conformations of a single ligand, including molecular system information, constituent atom information, bond value information, substructure information, etc., and at the same time obtain the RMSD value between different candidate conformations and the natural ligand conformation.

[0046] A2: Extract the spatial position information of atoms in each conformation, convert their corresponding three-dimensional structure into a set of feature points, each feature point containing coordinate and type information; cover all feature points of the conformation in space with a minimum bounding box;

[0047] A3: Construct a retrieval tree from the set of feature points in a hierarchical structure, where each leaf node stores a feature point and each non-leaf node stores the minimum bounding box of its child nodes.

[0048] Reference Figure 3 Step S2 in this embodiment specifically includes the following steps:

[0049] B1: Extract the spatial position information of atoms in the known active ligand conformations, and convert the three-dimensional structure of the known active ligand conformations into a set of feature points, each of which contains coordinate and type information;

[0050] B2: The feature points of the known active ligand conformations are searched in the established search tree, and similarity matching is performed with the candidate conformations to be screened in a hierarchical manner. The RMSD value, i.e., the similarity score, is calculated based on the number and distance of the matched feature points.

[0051] In this step, the feature points of the known active ligand conformations are searched in the established search tree. The search includes top-down search and bottom-up search.

[0052] The top-down retrieval process is as follows: First, the region where the instance is located is found from the root node using the retrieval method. Then, the region is further segmented based on the target's attributes to find the next layer region. This process is iterated until the final conformation output result is obtained.

[0053] Bottom-up retrieval specifically involves: first, determining the atomic positional relationships based on preprocessed data, and then distinguishing different instances through clustering and metric learning.

[0054] This embodiment uses a similarity score calculation method based on spatial location relationships, as detailed below:

[0055] The formula for calculating the similarity score of candidate conformations based on spatial location relationships is as follows:

[0056]

[0057] Where x ij x is the j-th conformation in the top-k results j All points in the neighborhood, y i y represents all points in the y-neighborhood of the natural conformation, and n is the sum of the number of atoms contained in a single ligand. The distance between each point in formula (1) is defined as the deviation difference or deviation error of the candidate conformation. The smaller the RMSD value, the closer the candidate conformation molecule is to the known active conformation.

[0058] In order to minimize the value of formula (1), that is, to obtain the docking posture closest to the natural conformation, the minimum value is taken in the calculation set, and the resulting conformation is the conformation molecule with the smallest deviation in the candidate conformation library. After obtaining the conformation with the smallest deviation in the search tree, backtracking is performed to obtain the conformation with the smallest deviation error outside the obtained conformation. This process is repeated until the top-k conformation query results are output.

[0059] B3: Sort the candidate conformations to be screened according to the similarity score, select the candidate conformations with the highest scores as candidate ligand conformations, and obtain the top-k docking posture results.

[0060] The specific process of step S3 in this embodiment is as follows:

[0061] C1: Evaluate the screening results; calculate the RMSD value of all candidate conformations, sort the results to obtain the k smallest conformations, compare them with the top-k results obtained from the retrieval, calculate the accuracy and compare the time taken;

[0062] C2: Optimize the filtering strategy of the search tree; optimize the splitting strategy, including linear splitting, binary splitting, quadtree splitting, etc., reconstruct the search tree index items, and re-evaluate them. The strategy with the highest accuracy was obtained through multiple experiments.

[0063] This embodiment also provides a virtual screening system for ligand docking posture based on fast retrieval. The system includes a network interface, a memory, and a processor. The network interface is used to receive and send signals during the process of sending and receiving information with other external network elements. The memory is used to store computer program instructions that can run on the processor. The processor is used to execute the steps of the consensus method described above when running the computer program instructions.

[0064] This embodiment also provides a computer storage medium storing a computer program that, when executed by a processor, can implement the methods described above. The computer-readable medium can be considered tangible and non-transitory. Non-limiting examples of non-transitory tangible computer-readable media include non-volatile memory circuitry (e.g., flash memory circuitry, erasable programmable read-only memory circuitry, or masked read-only memory circuitry), volatile memory circuitry (e.g., static random access memory circuitry or dynamic random access memory circuitry), magnetic storage media (e.g., analog or digital magnetic tape or hard disk drive), and optical storage media (e.g., CD, DVD, or Blu-ray disc). The computer program includes processor-executable instructions stored on at least one non-transitory tangible computer-readable medium. The computer program may also include or depend on stored data. The computer program may include a basic input / output system (BIOS) for interacting with the hardware of a dedicated computer, device drivers for interacting with specific devices of the dedicated computer, one or more operating systems, user applications, background services, background applications, etc.

[0065] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0066] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0067] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0068] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

Claims

1. A virtual screening method for ligand docking postures based on fast retrieval, characterized in that, Includes the following steps: S1: Preprocess the ligand conformation information data to establish an index table with a retrieval tree spatial index structure; S2: Input a known active ligand conformation, use the retrieval tree spatial index structure to quickly retrieve and filter potential candidate conformations similar to the query conformation, and take the most similar top-k query results as the top-k conformation results; S3: Evaluate the top-k conformation results obtained from the search, compare the actual RMSD values ​​of the output top-k candidate conformations with those of the native conformations, verify the accuracy of the screening results, and further optimize the screening strategy; In step S1, the bioinformatics data of the ligand conformations are preprocessed based on their spatial positional relationships. The specific processing steps are as follows: A1: Process the candidate dataset for ligand docking to obtain the biological information of different conformations of a single ligand, and at the same time obtain the RMSD values ​​between different candidate conformations and the natural ligand conformations. A2: Extract the spatial position information of atoms in each conformation, convert their corresponding three-dimensional structure into a set of feature points, each feature point containing coordinate and type information; cover all feature points of the conformation in space with a minimum bounding box; A3: Construct a retrieval tree from the set of feature points according to a hierarchical structure, where each leaf node stores a feature point and each non-leaf node stores the minimum bounding box of its child nodes; Step S2 specifically includes the following steps: B1: Extract the spatial position information of atoms in the known active ligand conformations, and convert the three-dimensional structure of the known active ligand conformations into a set of feature points, each of which contains coordinate and type information; B2: The feature points of the known active ligand conformations are searched in the established search tree, and similarity matching is performed with the candidate conformations to be screened in a hierarchical manner. The RMSD value, i.e., the similarity score, is calculated based on the number and distance of the matched feature points. B3: Sort the candidate conformations to be screened according to the similarity score, select the candidate conformations with the highest scores as candidate ligand conformations, and obtain the top-k docking posture results.

2. The virtual screening method for ligand docking posture based on fast retrieval as described in claim 1, characterized in that, The biological information of different conformations of a single ligand in step A1 includes molecular system information, constituent atom information, bond value information, and substructure information.

3. The virtual screening method for ligand docking posture based on fast retrieval according to claim 1, characterized in that, In step B2, the feature points of the known active ligand conformations are searched in the established search tree. The search includes top-down search and bottom-up search. The top-down retrieval process is as follows: First, the region where the instance is located is found from the root node using the retrieval method. Then, the region is further segmented based on the target's attributes to find the next layer region. This process is iterated until the final conformation output result is obtained. Bottom-up retrieval specifically involves: first, determining the atomic positional relationships based on preprocessed data, and then distinguishing different instances through clustering and metric learning.

4. The virtual screening method for ligand docking posture based on fast retrieval according to claim 1, characterized in that, Step B2 employs a similarity score calculation method based on spatial location relationships, as detailed below: The formula for calculating the similarity score of candidate conformations based on spatial location relationships is as follows: Where x ij x is the j-th conformation in the top-k results j All points in the neighborhood, y i Let y be all points in the y-neighborhood of the natural conformation, and n be the sum of the number of atoms contained in a single ligand. The distance between each point in formula (1) is defined as the deviation difference or deviation error of the candidate conformation. In order to minimize the value of formula (1), that is, to obtain the docking posture closest to the natural conformation, the minimum value is taken in the calculation set, and the resulting conformation is the conformation molecule with the smallest deviation in the candidate conformation library. After obtaining the conformation with the smallest deviation difference in the search tree, backtrack upwards to obtain the conformation with the smallest deviation error outside the obtained conformation, and continue iterating until the top-k conformation query results are output.

5. The virtual screening method for ligand docking posture based on fast retrieval according to claim 1, characterized in that, Step S3 specifically involves: C1: Evaluate the screening results; calculate the RMSD value of all candidate conformations, sort the results to obtain the k smallest conformations, compare them with the top-k results obtained from the retrieval, calculate the accuracy and compare the time taken; C2: Optimize the filtering strategy of the search tree; optimize the splitting strategy, rebuild the search tree index items, and re-evaluate. The strategy with the highest accuracy was obtained through multiple experiments.

6. The virtual screening method for ligand docking posture based on fast retrieval according to claim 5, characterized in that, The optimization splitting strategy in step C2 includes linear splitting, binary splitting, and quadtree splitting.

Images