A 3D molecule sampling optimization method and device, electronic equipment and storage medium

Abstract

The application relates to the technical field of 3D molecule generation, and discloses a 3D molecule sampling optimization method and device, electronic equipment and a storage medium, the method comprising the following steps: step 101, determining an exploration sampling budget and an optimization sampling budget of a current round based on a total sampling budget of the current round; step 102, performing initial sampling based on the exploration sampling budget to obtain an initial 3D molecule set; step 103, determining the sampling probability of each fragment appearing based on the initial 3D molecule set; step 104, continuing sampling based on the optimization sampling budget and the sampling probability of each fragment to obtain an intermediate 3D molecule set; and step 105, determining the exploration sampling budget and the optimization sampling budget of the next round based on the quality of the 3D molecules in the intermediate 3D molecule set, and returning to step 102 until a preset round is completed. The application effectively realizes the synchronous improvement of sampling efficiency and molecule quality.

Description

Technical Field

[0001] This invention relates to the field of 3D molecular generation technology, specifically to a 3D molecular sampling optimization method, apparatus, electronic device, and storage medium. Background Technology

[0002] 3D molecular generation is a core technology for targeted drug development in the field of AI-driven pharmaceutical research. This technology generates small molecule conformations that match the target pocket through a model, thereby outputting high-quality active molecules and directly impacting the drug development cycle. Lingo3DMol, as a representative 3D molecular generation model, combines fragmented SMILES encoding and local and global coordinate prediction, and is at the forefront in terms of indicators such as drug similarity, synthetic accessibility, and binding affinity.

[0003] However, current 3D molecular generation models are mostly based on multinomial sampling schemes in the sampling stage. Although widely used, this scheme completes all sampling inferences at once based on the initial probability distribution of the model, lacking dynamic adjustment and feedback mechanisms, which leads to a waste of computational resources and unstable quality of generated molecules. Summary of the Invention

[0004] This invention provides a 3D molecular sampling optimization method, device, electronic device, and storage medium to solve the problem in the prior art where the initial probability distribution completes all sampling inference at once and lacks dynamic adjustment and feedback mechanisms.

[0005] In a first aspect, the present invention provides a 3D molecular sampling optimization method, the method comprising: Step 101: Based on the total sampling budget for the current round, determine the exploration sampling budget and the optimized sampling budget for the current round; Step 102: Perform initial sampling based on the exploratory sampling budget to obtain an initial 3D molecular set; Step 103: Based on the initial 3D molecular set, determine the sampling probability of each fragment that appears; Step 104: Based on the optimized sampling budget and the sampling probability of each segment, continue sampling to obtain the intermediate 3D molecule set; Step 105: Based on the quality of the 3D molecules in the intermediate 3D molecule set, determine the exploration sampling budget and optimize the sampling budget for the next round, and return to step 102 until the preset round is completed.

[0006] This invention transforms the traditional one-time sampling process into a multi-round iterative optimization process. First, it explores the sampling budget to discover new high-quality fragments. Then, it optimizes the sampling budget to fully sample and optimize based on the discovered high-quality fragments. Furthermore, the dynamic budget adjustment mechanism adaptively balances exploration and utilization based on sampling results, avoiding resource waste or insufficient exploration that might result from a fixed allocation ratio. This effectively achieves a simultaneous improvement in sampling efficiency and molecular weight.

[0007] In one alternative implementation, based on an initial 3D molecular set, the sampling probability of each appearing fragment is determined, including: The 3D molecules in the initial 3D molecule set are segmented into segments based on a preset segmentation algorithm; Identify the atoms involved in NCI and assign them to the corresponding segments; Count the number of NCI actions for each segment; Determine the historical frequency of each segment; The NCI score for each segment is determined based on the number of NCI actions and the historical frequency of occurrence. Based on the NCI score, the sampling probability of each segment is determined.

[0008] In one optional implementation, the NCI score for each segment is determined based on the number of NCI actions and historical frequency of occurrence, including: In determining the number of NCI actions for each segment, the number of the first hit by expert marking and the number of the second hit by tool judgment; The NCI score for each segment is determined based on the first occurrence, the second occurrence, and the historical frequency of occurrence.

[0009] In this embodiment, the high-quality fragments already available in the exploration phase are efficiently utilized, and the high-scoring fragments are given higher sampling weights. This allows the generative model to prioritize the construction of new molecules based on these validated high-quality fragments, which can significantly improve molecular quality and increase resource utilization.

[0010] In one optional implementation, before segmenting the 3D molecules in the initial 3D molecule set into segments based on a preset segmentation algorithm, the method further includes: Verify the geometric structure of 3D molecules in the initial 3D molecular assembly; Molecular mechanics optimization was performed on 3D molecules that passed geometric structure verification to obtain a preliminary set of 3D molecules; Determine the NCI value of 3D molecules in the initial screening 3D molecule set; 3D molecules with NCI values ​​greater than a preset screening threshold are screened to obtain an initial set of finely screened 3D molecules; the initial set of finely screened 3D molecules is used for fragment segmentation.

[0011] This embodiment establishes a multi-level quality filtration system, which can effectively ensure that the screened molecules have good 3D structure and bioactivity potential, thereby improving the accuracy of subsequent 3D molecule sampling optimization.

[0012] In one optional implementation, the preset segmentation algorithm is: Choose non-cyclic single bonds; do not split at CH bonds; and ensure that the split fragments contain cyclic structures or important functional groups.

[0013] In one alternative implementation, sampling continues based on an optimized sampling budget and the sampling probability of each fragment to obtain an intermediate 3D molecular set, including: The sampled data is stored in a fragment tree structure; the sampled data includes: the 3D conformation of each fragment, the identifier of each fragment, the NCI score of each fragment, the sampling probability of each fragment, the number of times it has been sampled, and the optimized sampling budget; Traverse all nodes in the fragment tree structure and perform full node sampling to obtain the intermediate 3D molecular set.

[0014] In this embodiment, a fragment tree structure is used to achieve hierarchical organization, full attribute recording, full node sampling, and count control, which transforms fragmented molecular construction into a manageable, iterative, and optimizable structure. This significantly reduces search complexity and improves generation efficiency, which is conducive to the efficient generation of high-quality molecules under multiple cycles.

[0015] In one optional implementation, after the preset number of rounds is completed, the method further includes: Filter 3D molecules in the intermediate 3D molecule set whose NCI value is greater than the preset target threshold; A preset number of 3D molecules with the highest NCI values ​​are selected as target 3D molecules.

[0016] In this embodiment, without changing the model parameters, the probability distribution of subsequent sampling is continuously optimized by recording the paths that have already been sampled. This solves the problem of balancing high quality and diversity caused by relying solely on model experience, and enables the sampling of high-quality molecules while maintaining sufficient diversity.

[0017] In a second aspect, the present invention provides a 3D molecular sampling optimization device, the device comprising: The budget determination module is used to determine the exploration sampling budget and the optimized sampling budget for the current round based on the total sampling budget for the current round. The initial sampling module is used to perform initial sampling based on the exploratory sampling budget to obtain an initial 3D molecular set; The sampling frequency determination module is used to determine the sampling probability of each fragment based on the initial 3D molecular set; An optimized sampling module is used to continue sampling based on an optimized sampling budget and the sampling probability of each fragment to obtain an intermediate 3D molecular set; The loop module is used to determine the exploration sampling budget for the next round and optimize the sampling budget based on the mass of 3D molecules in the intermediate 3D molecule set, until the preset round is completed.

[0018] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the 3D molecular sampling optimization method of the first aspect or any corresponding embodiment described above.

[0019] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the 3D molecular sampling optimization method of the first aspect or any corresponding embodiment described above.

[0020] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the 3D molecular sampling optimization method of the first aspect or any corresponding embodiment described above.

[0021] It should be noted that the 3D molecular sampling optimization device, electronic equipment, and computer-readable storage medium provided by this invention correspond to the 3D molecular sampling optimization method described above. Therefore, regarding the beneficial effects of the 3D molecular sampling optimization device, electronic equipment, and computer-readable storage medium, please refer to the description of the corresponding beneficial effects of the 3D molecular sampling optimization method above, and will not be repeated here. Attached Figure Description

[0022] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating the 3D molecular sampling optimization method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of segmentation according to an embodiment of the present invention; Figure 3 This is a structural block diagram of a 3D molecular sampling optimization device according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0026] Among related technologies, the three core drawbacks of multinomial sampling are mainly reflected in budget allocation, structure-activity correlation, and exploration and utilization. First, because sampling consumes the entire budget at once and cannot be dynamically adjusted based on the generated quality, 60%-70% of the sampling results are filtered out due to NCI (non-hydrogen bonded interactions, used to describe various weak intermolecular or intramolecular interactions other than hydrogen bonds) coverage below 0.6, resulting in wasted computational power. Second, sampling relies solely on the overall probability distribution of the molecule, failing to effectively correlate key indicators such as NCI with molecular fragments; therefore, it cannot generate highly active molecular fragments. Finally, due to the lack of a dynamic sampling mechanism, multinomial sampling cannot prioritize the exploration of new fragments or deeply utilize highly active fragments, resulting in insufficient molecular diversity and a low activity probability.

[0027] This invention aims to overcome the above-mentioned problems by leveraging the advantages of dynamic retrieval enhancement, thereby improving the efficiency and activity of molecule generation and advancing the development of targeted drugs.

[0028] In view of this, according to an embodiment of the present invention, a 3D molecular sampling optimization method embodiment is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0029] This embodiment provides a 3D molecular sampling optimization method, which can be used in servers, terminals, mobile terminals, etc. Figure 1 This is a flowchart of a 3D molecular sampling optimization method according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps: Step 101: Based on the total sampling budget for the current round, determine the exploration sampling budget and the optimized sampling budget for the current round.

[0030] In this embodiment, the total sampling budget for the current round can be understood as the total number of times molecules are generated in the current round, and this total sampling budget is divided into an exploratory sampling budget and an optimization sampling budget. The exploratory sampling budget is used to explore molecular diversity, while the optimization sampling budget is used to improve molecular quality. For example, if the total sampling budget for the current round is set to 100,000 times, the exploratory sampling budget is 10% of the total sampling budget, i.e., 10,000 times, specifically used to "find new fragments"; the optimization sampling budget is set to 90% of the total sampling budget, i.e., 90,000 times, to generate more high-quality, strongly binding molecules based on the new fragments found by the exploratory sampling budget.

[0031] Step 102: Perform initial sampling based on the exploratory sampling budget to obtain an initial 3D molecular set.

[0032] Lingo3DMol is used as the 3D molecule generation model in this embodiment. The 3D molecule generation model performs initial sampling according to the set exploration sampling budget to generate multiple 3D molecules, thereby obtaining an initial 3D molecule set.

[0033] Step 103: Based on the initial 3D molecular set, determine the sampling probability of each fragment that appears.

[0034] In this embodiment, the sampling probability of a fragment can be determined based on its frequency of occurrence in the initial 3D molecular set. For example, the higher the frequency of a fragment's occurrence, the higher its sampling probability; conversely, the lower the frequency of a fragment's occurrence, the lower its sampling probability. Fragments with higher sampling probabilities are more likely to participate in molecule generation during the sampling phase based on the optimized sampling budget. In other words, fragments with higher sampling probabilities participate in more sampling attempts.

[0035] Step 104: Based on the optimized sampling budget and the sampling probability of each segment, continue sampling to obtain the intermediate 3D molecular set.

[0036] Step 105: Based on the quality of the 3D molecules in the intermediate 3D molecule set, determine the exploration sampling budget and optimize the sampling budget for the next round, and return to step 102 until the preset round is completed.

[0037] In this embodiment, the quality of 3D molecules can be determined based on their NCI value. For example, an NCI value greater than or equal to a preset threshold, such as 0.6, indicates that the 3D molecule has high quality. This embodiment can also determine the next round of exploration sampling budget and optimization sampling budget based on the proportion of high-quality molecules in the intermediate 3D molecule set. For example, if the proportion of high-quality molecules α < 20%, it indicates that the current sampling quality is low, requiring an expanded exploration scope, and the exploration sampling budget is increased to 15%; when 20% ≤ α < 40%, the current exploration sampling budget of 10% is maintained; when α ≥ 40%, it indicates that many high-quality fragments have been discovered, and the exploration sampling budget should be reduced to 5%, focusing on the in-depth utilization of highly active fragments. A moderate exploration sampling budget can discover new high-quality fragments without excessively consuming computational resources; while a sufficient optimization sampling budget ensures that the discovered high-quality fragments can be fully sampled and optimized.

[0038] In some optional implementations, the preset number of rounds in this embodiment can be 3 to 8, which can be set according to actual needs.

[0039] In this embodiment, the traditional one-time sampling is transformed into a multi-round iterative optimization process. First, new high-quality fragments are discovered by exploring the sampling budget. Then, the sampling budget is optimized to fully sample and optimize based on the discovered high-quality fragments. Furthermore, the dynamic budget adjustment mechanism can adaptively balance exploration and utilization according to the sampling effect, avoiding resource waste or insufficient exploration that may be caused by a fixed ratio allocation, and effectively achieving a simultaneous improvement in sampling efficiency and molecular mass.

[0040] This embodiment provides a 3D molecular sampling optimization method, which can be used on servers, terminals, mobile terminals, etc. The process includes the following steps: Step 201: Based on the total sampling budget for the current round, determine the exploration sampling budget and the optimized sampling budget for the current round. See details below. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0041] Step 202: Perform initial sampling based on the exploratory sampling budget to obtain an initial 3D molecular set. See details below. Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0042] Step 203: Based on the initial 3D molecular set, determine the sampling probability of each fragment that appears.

[0043] Specifically, step S202 includes: Step S2021: The 3D molecules in the initial 3D molecule set are segmented into segments based on a preset segmentation algorithm.

[0044] This embodiment employs an improved fragmentation algorithm to ensure that the segmented fragments possess chemical significance and structural integrity. In some optional implementations, the preset fragmentation algorithm is as follows: selects non-cyclic single bonds; does not segment at CH bonds; and ensures that the segmented fragments contain cyclic structures or important functional groups.

[0045] In other words, the cleavage fragments must be selected from non-cyclic single bonds, i.e., preferentially choosing single bonds connecting rings or rings to the chain; they must not be connected to hydrogen atoms, i.e., avoid cleavage at CH bonds to maintain the integrity of functional groups; and at least one end must be cyclic to ensure that the cleaved fragment contains a cyclic structure or important functional groups. The molecular fragment cleavage effect is referenced. Figure 2 As shown, Figure 2 The image on the left shows the original molecule, while the different colors on the right represent different segments.

[0046] Step S2022: Identify the atoms involved in NCI and assign them to the corresponding fragments.

[0047] The atoms involved in NCI can be located and assigned to the corresponding fragments using prolif (a Python library specifically designed for analyzing protein-ligand interactions, capable of extracting interaction fingerprints from molecular dynamics simulations, experimental structures, and molecular docking).

[0048] Step S2023: Count the number of NCI actions for each segment.

[0049] Step S2024: Determine the historical frequency of each segment. That is, combine the historical frequency of each segment in previous rounds.

[0050] Step S2025: Determine the NCI score for each segment based on the number of NCI actions.

[0051] Step S2025 above includes: determining the first hit count of expert markers and the second hit count of tool judgments in the NCI action counts for each segment; determining the NCI score for each segment based on the first hit count, the second hit count, and historical frequency. For example, the number of NCI action counts S for each segment is counted, and S is decomposed to obtain the number of expert NCI hits S1 and the number of Prolif additional judgment NCI hits S2 for each segment.

[0052] In this embodiment, the NCI score of the fragment can be calculated based on the following formula: NCI Score = ((S1×1.0+S2×0.5) / S_max)×0.7+(1-historical frequency)×0.3.

[0053] Where S_max is the total number of times the fragment appears.

[0054] In this embodiment, a depth-first search (DFS) method can be used to traverse the molecular fragment graph, starting with a fragment where S1>0, to explore large fragments consisting of multiple NCI-hit fragments. If an intermediate connecting fragment does not hit an NCI, the NCI score of the large fragment is penalized.

[0055] Step S2026: Determine the sampling probability of each segment based on the NCI score.

[0056] Based on the fragment NCI score, all fragments and large fragments are prioritized and sorted, and the sampling probability of fragments is assigned according to the score as weight when sampling based on the optimized sampling budget.

[0057] In this embodiment, the high-quality fragments already available in the exploration phase are efficiently utilized, and the high-scoring fragments are given higher sampling weights. This allows the generative model to prioritize the construction of new molecules based on these validated high-quality fragments, which can significantly improve molecular quality and increase resource utilization.

[0058] In some optional implementations, prior to step S2021, the following steps are also included: Step a1 involves verifying the geometric structure of the 3D molecules in the initial 3D molecular assembly. This verification checks for obvious structural errors such as atomic overlap and abnormal bond lengths.

[0059] Step a2 involves performing molecular mechanics optimization on the 3D molecules that have passed geometric structure verification to obtain an initial set of 3D molecules. Molecular mechanics optimization can eliminate unreasonable 3D conformations. Further filtering can be performed based on hydrogen bond utilization, synthetic accessibility, and drug-likeness.

[0060] Step a3: Determine the NCI value of the 3D molecules in the initial 3D molecule set.

[0061] Step a4: Screen 3D molecules with NCI values ​​greater than the preset screening threshold to obtain an initial set of finely screened 3D molecules; the initial set of finely screened 3D molecules is used for fragment segmentation.

[0062] If expert-annotated NCI data is available, an NCI coverage rate ≥ 0.6 can be used as the screening threshold. If no expert NCI data is available, the Prolif tool is used to perform NCI confidence statistics as the estimated NCI, and the screening threshold is adjusted based on the confidence level. Currently, Prolif supports the identification of hydrophobic interactions, hydrogen bonds, salt bridges, π-π stacking, and halogen bonds.

[0063] In this embodiment, a multi-level quality filtration system was established, which can effectively ensure that the screened molecules have good 3D structure and bioactivity potential, thereby improving the accuracy of subsequent 3D molecule sampling optimization.

[0064] Step 204: Based on the optimized sampling budget and the sampling probability of each segment, continue sampling to obtain the intermediate 3D molecular set.

[0065] The selected segment can be used as a prompt, that is, a hint, to input the generative model (such as Lingo3DMol) to continue sampling.

[0066] Step 205: Based on the quality of the 3D molecules in the intermediate 3D molecule set, determine the exploration sampling budget and optimize the sampling budget for the next round, and return to step 202 until the preset number of rounds is completed. For details, please refer to [link to relevant documentation]. Figure 1 Step S105 of the illustrated embodiment will not be described again here.

[0067] Step 206: Filter the 3D molecules in the intermediate 3D molecule set whose NCI value is greater than the preset target threshold.

[0068] Step 207: Select a preset number of 3D molecules with the highest NCI values ​​as target 3D molecules.

[0069] It can screen for NCI values ​​≥ 0.7 and other screening channels, and output a Top-K high-quality molecule set.

[0070] In this embodiment, without changing the model parameters, the probability distribution of subsequent sampling is continuously optimized by recording the paths that have already been sampled. This solves the problem of balancing high quality and diversity caused by relying solely on model experience, and enables the sampling of high-quality molecules while maintaining sufficient diversity.

[0071] In some alternative implementations, sampling continues based on an optimized sampling budget and the sampling probability of each fragment to obtain an intermediate 3D molecular set, including: The sampled data is stored in a fragment tree structure. The sampled data includes: the 3D conformation of each fragment, the identifier of each fragment, the NCI score of each fragment, the sampling probability of each fragment, the number of times it has been sampled, and the optimized sampling budget. All nodes on the fragment tree structure are traversed to perform full node sampling to obtain the intermediate 3D molecule set.

[0072] In this embodiment, the fragment 3D configuration (including the fragment's parent node), fragment identifier, NCI score, number of samples C, and optimized sampling budget C_max can be used as node attributes of the fragment tree structure. The root node can be a user-given fragment or a high-frequency fragment that hits the expert NCI, and child nodes can be extended new fragments. Fragments at the same level are mutually exclusive.

[0073] In this embodiment, full node sampling can be performed based on fragment priority. Priority can be calculated as Priority = NCI Score + 0.5 × I (I = 1 for new fragments, 0 for old fragments, encouraging exploration). Full node sampling involves traversing all nodes in the tree (including intermediate nodes) and selecting nodes based on Priority as the weight, rather than sampling only leaf nodes. When the number of samples C reaches C_max, Priority is set to 0, and sampling stops.

[0074] In this embodiment, a fragment tree structure is used to achieve hierarchical organization, full attribute recording, full node sampling, and count control, which transforms fragmented molecular construction into a manageable, iterative, and optimizable structure. This significantly reduces search complexity and improves generation efficiency, which is conducive to the efficient generation of high-quality molecules under multiple cycles.

[0075] The overall workflow of this embodiment is as follows: First, the total sampling budget M for the current round is set. Using an exploration and optimization mode, the first round allocates the exploration sampling budget for initial sampling. After quality filtering, the generated molecules are evaluated by expert NCI or predicted with confidence using the Prolif tool. The NCI effect is then attributed to the fragment level, and fragment scores are calculated. This fragment information is stored in a tree structure. Subsequent sampling uses the fragment level as a prompt, and the fragment sampling probability in the optimization mode is dynamically adjusted based on the fragment scores. The entire process involves N iterations, with each iteration optimizing the sampling strategy based on the results of the previous round, ultimately outputting a high-quality molecule set.

[0076] The dynamic budget allocation mechanism based on quality feedback proposed in this embodiment overcomes the limitations of traditional one-time sampling and achieves a balance between exploration and utilization. Furthermore, this embodiment constructs a cyclical closed-loop optimization framework of "exploration-attribution-optimization." A complete cyclical optimization process is designed, continuously improving the quality of molecule generation through N iterations; a learning mechanism based on historical data is established, which can continuously accumulate knowledge of structure-NCI relationships, effectively improving the quality of molecule generation.

[0077] In this embodiment, the generation effect was evaluated at three targets: 7vu6, 8gcy, and mcl1. A sampling budget of 20,000 runs was used to evaluate different sampling schemes, focusing on differences in NCI coverage, the number of non-repeating molecules with an NCI greater than or equal to 0.7, and the proportion of invalid samples (molecules with an NCI < 0.7 were defined as invalid samples). As shown in Table 1, the scheme of this invention significantly improves the NCI level compared to polynomial sampling, thereby optimizing sampling efficiency and overall generation performance.

[0078] Table 1. Comparison of Results

[0079] This embodiment also provides a 3D molecular sampling optimization device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0080] This embodiment provides a 3D molecular sampling optimization device, such as... Figure 3 As shown, the device includes: Budget determination module 301 is used to determine the exploration sampling budget and the optimized sampling budget for the current round based on the total sampling budget for the current round; The initial sampling module 302 is used to perform initial sampling based on the exploratory sampling budget to obtain an initial 3D molecular set; The sampling frequency determination module 303 is used to determine the sampling probability of each fragment based on the initial 3D molecular set; The optimized sampling module 304 is used to continue sampling based on the optimized sampling budget and the sampling probability of each segment to obtain an intermediate 3D molecular set; The loop module 305 is used to determine the exploration sampling budget for the next round and optimize the sampling budget based on the mass of the 3D molecules in the intermediate 3D molecule set, until the preset round is completed.

[0081] In some alternative embodiments, the device further includes: The generation module is used to filter 3D molecules in the intermediate 3D molecule set whose NCI value is greater than a preset target threshold; and selects a preset number of 3D molecules with the highest NCI values ​​as target 3D molecules.

[0082] In some optional implementations, the initial sampling module 302 is specifically used for: segmenting the 3D molecules in the initial 3D molecule set into segments based on a preset segmentation algorithm; identifying the atoms participating in NCI and assigning them to the corresponding segments; counting the number of NCI actions for each segment; determining the historical occurrence frequency of each segment; determining the NCI score for each segment based on the number of NCI actions and the historical occurrence frequency; and determining the sampling probability of each segment based on the NCI score. It is also specifically used for: determining the first hit count and the second hit count as determined by the tool in the number of NCI actions for each segment; and determining the NCI score for each segment based on the first hit count, the second hit count, and the historical occurrence frequency.

[0083] In some alternative embodiments, the device includes: The screening module is used to verify the geometric structure of 3D molecules in the initial 3D molecular set; to perform molecular mechanics optimization on the 3D molecules that pass the geometric structure verification to obtain a preliminary 3D molecular set; to determine the NCI value of the 3D molecules in the preliminary 3D molecular set; to screen 3D molecules with NCI values ​​greater than a preset screening threshold to obtain a finely screened initial 3D molecular set; and to use the finely screened initial 3D molecular set for fragment segmentation.

[0084] In one optional implementation, the preset fragment segmentation algorithm is as follows: select non-cyclic single bonds; and do not segment at CH bonds; and ensure that the segmented fragments contain cyclic structures or important functional groups.

[0085] In some optional implementations, the optimized sampling module 304 is specifically used to: store the sampling data in a fragment tree structure; the sampling data includes: the 3D conformation of each fragment, the identifier of each fragment, the NCI score of each fragment, the sampling probability of each fragment, the number of times it has been sampled, and the optimized sampling budget; and traverse all nodes on the fragment tree structure to perform full node sampling to obtain an intermediate 3D molecule set.

[0086] The 3D molecular sampling optimization device provided in this embodiment of the invention can execute the 3D molecular sampling optimization method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0087] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0088] The following is a detailed reference. Figure 4 This diagram illustrates a structural schematic suitable for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 401, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 402 or a program loaded from memory 408 into random access memory (RAM) 403. The RAM 403 also stores various programs and data required for the operation of the electronic device. The processor 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0089] Typically, the following devices can be connected to I / O interface 405: input devices 406 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 407 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 408 including, for example, magnetic tapes, hard disks, etc.; and communication devices 409. Communication device 409 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 4 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0090] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 409, or installed from a memory 408, or installed from a ROM 402. When the computer program is executed by the processor 401, it performs the functions defined in the 3D molecular sampling optimization method of the embodiments of the present invention.

[0091] Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0092] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the 3D molecular sampling optimization method shown in the above embodiments is implemented.

[0093] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0094] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A 3D molecular sampling optimization method, characterized in that, The method includes: Step 101: Based on the total sampling budget for the current round, determine the exploration sampling budget and the optimized sampling budget for the current round; Step 102: Perform initial sampling based on the exploratory sampling budget to obtain an initial 3D molecular set; Step 103: Based on the initial 3D molecule set, determine the sampling probability of each fragment; wherein, the steps include: segmenting the 3D molecules in the initial 3D molecule set into fragments using a preset fragment segmentation algorithm; identifying the atoms participating in NCI and assigning them to the corresponding fragments; counting the number of NCI actions for each fragment; determining the historical occurrence frequency of each fragment; determining the first hit count by expert marking and the second hit count by tool judgment in the number of NCI actions for each fragment; determining the NCI score for each fragment based on the first hit count, the second hit count, and the historical occurrence frequency; and determining the sampling probability of each fragment based on the NCI score. Step 104: Based on the optimized sampling budget and the sampling probability of each segment, continue sampling to obtain an intermediate 3D molecule set; Step 105: Based on the quality of the 3D molecules in the intermediate 3D molecule set, determine the exploration sampling budget and the optimized sampling budget for the next round, and return to step 102 until the preset round is completed.

2. The method according to claim 1, characterized in that, Before segmenting the 3D molecules in the initial 3D molecule set into segments based on a preset segmentation algorithm, the following steps are also included: The geometric structure of the 3D molecules in the initial 3D molecular set is verified. Molecular mechanics optimization was performed on 3D molecules that passed geometric structure verification to obtain a preliminary set of 3D molecules; Determine the NCI value of the 3D molecules in the initial screening 3D molecule set; 3D molecules with NCI values ​​greater than a preset screening threshold are screened to obtain the initial set of finely screened 3D molecules; the initial set of finely screened 3D molecules is used for fragment segmentation.

3. The method according to claim 1, characterized in that, The preset segmentation algorithm is as follows: Choose non-cyclic single bonds; do not split at CH bonds; and ensure that the split fragments contain cyclic structures or important functional groups.

4. The method according to claim 1, characterized in that, The sampling process continues based on the optimized sampling budget and the sampling probabilities of each segment to obtain an intermediate 3D molecular set, including: The sampled data is stored in a fragment tree structure; the sampled data includes: the 3D conformation of each fragment, the identifier of each fragment, the NCI score of each fragment, the sampling probability of each fragment, the number of times it has been sampled, and the optimized sampling budget; By traversing all nodes in the fragment tree structure and performing full node sampling, the intermediate 3D molecule set is obtained.

5. The method according to claim 1, characterized in that, After the preset number of rounds is completed, the following is also included: Filter the intermediate 3D molecule set to include 3D molecules whose NCI value is greater than a preset target threshold. A predetermined number of 3D molecules with the highest NCI values ​​are selected as target 3D molecules.

6. A 3D molecular sampling optimization device, characterized in that, The device includes: The budget determination module is used to determine the exploration sampling budget and the optimized sampling budget for the current round based on the total sampling budget for the current round. The initial sampling module is used to perform initial sampling based on the exploratory sampling budget to obtain an initial 3D molecular set; A sampling frequency determination module is used to determine the sampling probability of each segment based on the initial 3D molecule set; wherein the module includes: segmenting the 3D molecules in the initial 3D molecule set into segments using a preset segmentation algorithm; identifying the atoms participating in NCI and assigning them to the corresponding segments; counting the number of NCI actions for each segment; determining the historical occurrence frequency of each segment; determining the first hit count and the second hit count of the tool in the number of NCI actions for each segment; determining the NCI score of each segment based on the first hit count, the second hit count, and the historical occurrence frequency; and determining the sampling probability of each segment based on the NCI score. An optimized sampling module is used to continue sampling based on the optimized sampling budget and the sampling probability of each segment to obtain an intermediate 3D molecular set; The loop module is used to determine the exploration sampling budget and the optimized sampling budget for the next round based on the mass of the 3D molecules in the intermediate 3D molecule set, until the preset round is completed.

7. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory storing computer instructions, and the processor executing the computer instructions to perform the 3D molecular sampling optimization method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the 3D molecular sampling optimization method according to any one of claims 1 to 5.

Images