Data Processing Method and Device for Drug Optimization

Abstract

The present application discloses a data processing method and apparatus for drug optimization. The method includes: obtaining drug data to be optimized, where the drug data to be optimized is relevant data of the drug to be optimized; performing multiple preprocessings on the drug data to be optimized based on a preset reinforcement learning model to obtain multiple process-optimized drug data, where the multiple preprocessings are preset molecular processes based on the molecular structure, and the multiple process-optimized drug data are drug data with multiple different molecular structures obtained after the drug to be optimized undergoes the preprocessing; and performing attribute screening processing on the multiple process-optimized drug data to obtain a target optimized drug data set, where the target optimized drug data set includes multiple process-optimized drug data that meet the preset attributes. The technical problem of low data optimization efficiency in the process of drug molecule optimization in the prior art is solved, and the technical effect of improving the data optimization efficiency of drug molecules is achieved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and more particularly, to a data processing method and device for drug optimization. Background Art

[0002] In the process of new drug research and development, small molecule drugs are an important part. For the research and development of small molecule drugs, it usually goes through stages such as identifying targets, finding hit compounds, optimizing hit compounds, finding lead compounds, and optimizing lead compounds. For different stages, there are different requirements for molecule generation and optimization. With the rapid development of computer technology, various molecule generation models have begun to be applied to all stages of small molecule drug research and development.

[0003] In the prior art, algorithms for generating small molecule drugs by using machine learning technology include molecular generation methods such as recurrent neural network algorithms and generative adversarial networks. The inventor found that the generative adversarial network model is difficult to train, the setting of hyperparameters needs to be adjusted repeatedly, the training process is usually slow, and the process is unpredictable. The model is not easy to converge and sometimes even causes the problem of model collapse, resulting in low efficiency of drug optimization processing.

[0004] Therefore, there is a technical problem of low data optimization efficiency in the process of drug molecule optimization in the prior art. Summary of the Invention

[0005] The main object of the present application is to provide a data processing method and device for drug optimization, so as to solve the technical problem of low data optimization efficiency in the process of drug molecule optimization in the prior art and improve the data optimization efficiency in the process of drug molecule optimization.

[0006] To achieve the above object, in the first aspect of the present application, a data processing method for drug optimization is proposed, including:

[0007] Obtain drug data to be optimized, where the drug data to be optimized is related data of the drug to be optimized;

[0008] Perform multiple preprocessings on the drug data to be optimized based on a preset reinforcement learning model to obtain multiple process-optimized drug data, where the multiple preprocessings are preset molecule processings based on the molecular structure, and the multiple process drug-optimized data are drug data with multiple different molecular structures obtained after the preprocessing of the drug to be optimized; and

[0009] Perform attribute screening processing on the multiple process-optimized drug data to obtain a target optimized drug data set, where the target optimized drug data set includes multiple process-optimized drug data that meet preset attributes.

[0010] Optionally, performing a plurality of preprocessings on the drug data to be optimized based on a preset reinforcement learning model, and obtaining a plurality of process-optimized drug data, including:

[0011] Performing feature recognition on the drug data to be optimized to obtain molecular structure feature data, where the molecular structure feature data is data used to represent the molecular structure features of the drug to be optimized;

[0012] Performing molecular cutting processing on the drug data to be optimized according to the molecular structure feature data to obtain a plurality of molecular fragment data, where the plurality of molecular fragment data is data corresponding to a plurality of molecular fragments in the drug to be optimized;

[0013] Performing fragment update processing on the first molecular fragment data to obtain a plurality of updated molecular fragment data, where the first molecular fragment data is any one of the plurality of molecular fragment data, and the plurality of updated molecular fragment data corresponds to the first molecular fragment data; and

[0014] Performing update processing on the drug data to be optimized according to the plurality of updated molecular fragment data to obtain the plurality of process-optimized drug data.

[0015] Optionally, performing fragment update processing on the first molecular fragment data to obtain a plurality of updated molecular fragment data, including:

[0016] Performing first fragment modification processing on the first molecular fragment data to obtain first modified molecular fragment data, where the first modified molecular fragment data is molecular fragment data obtained by modifying the first molecular fragment at a first position;

[0017] Performing second fragment modification processing on the first modified molecular fragment data to obtain second modified molecular fragment data, where the second modified molecular fragment data is molecular fragment data obtained by modifying the first modified molecular fragment at a second position;

[0018] Obtaining the plurality of updated molecular fragment data, where the plurality of updated molecular fragment data includes the first modified molecular fragment data and the second modified molecular fragment data.

[0019] Optionally, performing update processing on the drug data to be optimized according to the plurality of updated molecular fragment data to obtain the plurality of process-optimized drug data, including:

[0020] Performing recognition on the plurality of updated molecular fragment data to obtain first position feature data, where the first position feature data is data used to represent the molecular fragment position of the first molecular fragment in the drug to be optimized, and the first molecular fragment corresponds to the first molecular fragment data; and

[0021] Replace the first molecular fragment at the first position in the drug to be optimized respectively according to the plurality of updated molecular fragment data, to obtain the plurality of process-optimized drug data, wherein the first position corresponds to the first position feature data.

[0022] Optionally, perform attribute screening processing on the plurality of process-optimized drug data, and the obtained target optimized drug data set includes:

[0023] Perform attribute calculation processing on the plurality of process-optimized drug data to obtain a plurality of attribute data to be processed, wherein the plurality of attribute data to be processed respectively correspond to the plurality of process-optimized drug data;

[0024] Screen the plurality of attribute data to be processed according to the preset predicted attribute data, to judge whether the plurality of attributes to be processed corresponding to the plurality of attribute data to be processed are within the attribute range corresponding to the preset predicted attribute data; and

[0025] If the attribute to be processed corresponding to the attribute data to be processed is within the attribute range corresponding to the preset predicted attribute data, obtain the target optimized drug data set, wherein the target optimized drug data set includes the plurality of process-optimized drug data corresponding to the plurality of attribute data to be processed.

[0026] Optionally, after performing attribute screening processing on the plurality of process-optimized drug data to obtain the target optimized drug data set, the method further includes:

[0027] Identify the target optimized drug data set to obtain a plurality of target optimized drug data;

[0028] Perform model update training processing on the preset reinforcement learning model according to the plurality of target optimized drug data to obtain an updated reinforcement learning model;

[0029] Perform molecular structure-based molecular update processing on the drug data to be optimized based on the updated reinforcement learning model to obtain a plurality of updated process-optimized drug data; and

[0030] Perform attribute screening processing on the plurality of updated process-optimized drug data to obtain an updated target optimized drug data set.

[0031] According to a second aspect of the present application, a data processing device for drug optimization is provided, including:

[0032] A data acquisition module, configured to acquire drug data to be optimized, wherein the drug data to be optimized is related data of the drug to be optimized;

[0033] A preprocessing module for performing multiple preprocessings on the drug data to be optimized based on a preset reinforcement learning model to obtain multiple process-optimized drug data, where the multiple preprocessings are preset molecular processes based on the molecular structure, and the multiple process-optimized drug data are drug data with multiple different molecular structures obtained after the preprocessing of the drug to be optimized; and

[0034] A screening module for performing attribute screening on the multiple process-optimized drug data to obtain a target optimized drug data set, where the target optimized drug data set includes multiple process-optimized drug data that meet the preset attributes.

[0035] Optionally, the preprocessing module includes:

[0036] An identification module for identifying the characteristics of the drug data to be optimized to obtain molecular structure characteristic data, where the molecular structure characteristic data is data used to represent the molecular structure characteristics of the drug to be optimized;

[0037] A cutting module for performing molecular cutting on the drug data to be optimized according to the molecular structure characteristic data to obtain multiple molecular fragment data, where the multiple molecular fragment data are data corresponding to multiple molecular fragments in the drug to be optimized;

[0038] A fragment update module for performing fragment update on the first molecular fragment data to obtain multiple updated molecular fragment data, where the first molecular fragment data is any one of the multiple molecular fragment data, and the multiple updated molecular fragment data correspond to the first molecular fragment data; and

[0039] A drug update module for updating the drug data to be optimized according to the multiple updated molecular fragment data to obtain the multiple process-optimized drug data.

[0040] According to the third aspect of the present application, a computer-readable storage medium is provided, where the computer-readable storage medium stores computer instructions, and the computer instructions are used to cause the computer to execute the above-mentioned data processing method for drug optimization.

[0041] According to the fourth aspect of the present application, an electronic device is provided, including: at least one processor; and a memory communicatively connected to the at least one processor; where the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to cause the at least one processor to execute the above-mentioned data processing method for drug optimization.

[0042] The technical solutions provided by the embodiments of the present application may include the following beneficial effects:

[0043] In this application, by obtaining the drug data to be optimized, performing multiple preprocessings on the drug data to be optimized based on a preset reinforcement learning model, where the multiple preprocessings are preset molecular processes based on the molecular structure, multiple process-optimized drug data are obtained, and an attribute screening process is performed on the multiple process-optimized drug data to obtain a target optimized drug data set, which contains multiple process-optimized drug data that meet the preset attributes. By performing preprocessing based on the molecular structure on the drug to be optimized through the preset reinforcement learning model, drug data with multiple different molecular structures are obtained, solving the technical problem of low data optimization efficiency in the existing drug molecule optimization process and achieving the technical effect of improving the data optimization efficiency of drug molecules. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The drawings constituting a part of this application are used to provide a further understanding of this application, making other features, objectives, and advantages of this application more obvious. The schematic embodiments and descriptions thereof of this application are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0045] Figure 1 is a flowchart of a data processing method for drug optimization provided by this application;

[0046] Figure 2 is a flowchart of a data processing method for drug optimization provided by this application;

[0047] Figure 3 is a flowchart of a data processing method for drug optimization provided by this application;

[0048] Figure 4 is a flowchart of a data processing method for drug optimization provided by this application;

[0049] Figure 5 is a schematic structural diagram of a data processing device for drug optimization provided by this application;

[0050] Figure 6 is a schematic structural diagram of another data processing device for drug optimization provided by this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] In order to enable those skilled in the art to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of this application.

[0052] It should be noted that the terms "first", "second", etc. in the description, claims and above-mentioned drawings of this application are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that such data used can be interchanged under appropriate circumstances for the embodiments of this application described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0053] In this application, the orientation or positional relationship indicated by the terms "upper", "lower", "left", "right", "front", "rear", "top", "bottom", "inner", "outer", "middle", "vertical", "horizontal", "lateral", "longitudinal", etc. is based on the orientation or positional relationship shown in the drawings. These terms are mainly used to better describe this application and its embodiments, and are not used to limit that the indicated devices, elements or components must have a specific orientation or be constructed and operated in a specific orientation.

[0054] Moreover, in addition to being able to represent an orientation or positional relationship, some of the above terms may also be used to represent other meanings. For example, the term "upper" may also be used to represent a certain attachment relationship or connection relationship in some cases. For those of ordinary skill in the art, the specific meanings of these terms in this application can be understood according to specific circumstances.

[0055] In addition, the terms "install", "set", "be provided with", "connect", "be connected", "be sleeved" should be understood in a broad sense. For example, "connect" can be a fixed connection, a detachable connection, or an integral structure; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or there can also be internal communication between two devices, elements or components. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0056] The drug data set in the embodiments of this application is a data set of drug compounds existing in the drug R & D process, the reference drug data set is a data set of multiple active drug compounds obtained from an open-source database, and the target optimized drug data set is a data of drug compounds that meet the preset attribute requirements obtained after optimization processing.

[0057] Figure 1 is a flowchart of a data processing method for drug optimization provided by this application, as Figure 1 shown, the method includes the following steps:

[0058] S101: Obtain the drug data to be optimized;

[0059] The drug data to be optimized is the relevant data of the drug to be optimized. During the process of drug research and development, first confirm the target target, and determine the drug compound based on the target target. During the process of drug research and development, there are multiple stages of drug development, and the requirements for optimizing the drug compound are different in different stages. Different optimization processes are performed on the drug compound in different stages. The first research and development stage corresponds to the first drug data to be optimized, and the first optimization process is performed on the first drug data to be optimized. The second research and development stage corresponds to the second drug data to be optimized, and the second optimization process is performed on the second drug data to be optimized. If the current drug research and development stage is the first research and development stage, the first data to be optimized is obtained, and the first data to be optimized is the relevant data of the first drug to be optimized existing in the first research and development stage.

[0060] S102: Perform multiple preprocessings on the drug data to be optimized based on a preset reinforcement learning model to obtain multiple process-optimized drug data;

[0061] The multiple preprocessings are multiple preset molecule processings based on the molecular structure, and the multiple process drug optimization data are the drug data with multiple different molecular structures obtained after the drug to be optimized is preprocessed.

[0062] Preprocess according to the molecular structure of the drug to be optimized. The drug data to be optimized is updated based on the molecular structure to obtain multiple process drug optimization data. The multiple process drug optimization data respectively correspond to multiple process drugs. The multiple process drugs are the multiple process drugs generated during the optimization of the drug to be optimized. The molecular structures corresponding to the multiple process drugs are different from the molecular structure of the drug to be optimized.

[0063] By performing preprocessing on the drug to be optimized with molecular structure updates, multiple process drugs with different molecular structures are obtained. By processing the multiple process drugs, the optimization process of the drug to be optimized is realized. By updating the molecular structure of the drug to be optimized based on the dimension of the molecular structure, the optimization efficiency of the drug molecule is improved.

[0064] Figure 2 This is the flowchart of a data processing method for drug optimization provided by the present application. As Figure 2 shown, the method includes the following steps:

[0065] S201: Perform feature recognition on the drug data to be optimized to obtain molecular structure feature data;

[0066] The molecular structure feature data is data used to represent the molecular structure features of the drug to be optimized. Among them, the molecular structure feature data corresponds to the optimization rules at different stages of drug research and development. The molecular structure feature data is determined according to the optimization rules at the current stage of drug research and development. The molecular structure feature data can be the number of atoms in a molecular fragment, molecular weight, extended single bonds on the cleavage ring, etc., which is data used to represent the molecular structure features in the molecular fragments of the drug to be optimized.

[0067] S202: Perform molecular cleavage processing on the drug data to be optimized according to the molecular structure feature data, and obtain multiple molecular fragment data;

[0068] The multiple molecular fragment data are the data corresponding to multiple molecular fragments in the drug to be optimized. The molecular fragments in the drug to be optimized are cleaved according to the molecular structure feature data to obtain multiple molecular fragments. The multiple molecular fragments are obtained by molecular cleavage processing of the drug to be optimized. For example, when performing molecular cleavage, cleavage rules are set based on the molecular structure feature data, and cleavage rules are set according to the number of atoms in the molecular fragment, molecular weight, extended single bonds on the cleavage ring, cleavage points, etc. For example, rules such as the number of atoms not exceeding 14 and the number of cleavage points not exceeding 5 are used to perform molecular cleavage processing on the drug to be optimized to obtain multiple molecular fragment data. Cleavage rules can be set according to one molecular structure feature in the molecular structure feature data, or cleavage rules can be set according to multiple molecular structure features in the molecular structure feature data, and the drug to be optimized is subjected to molecular fragment cleavage according to the set cleavage rules.

[0069] In the embodiments of the present application, the drug data to be optimized is subjected to molecular cleavage processing through the molecular structure feature data of the drug to be optimized, and multiple different molecular fragments of the drug to be optimized are obtained, reducing the complexity of the molecular fragments after the molecular cleavage processing of the drug to be processed, improving the diversity of the molecular fragments obtained after cleavage, improving the diversity of the optimization directions of the drug to be optimized, and increasing the probability of obtaining the target optimized drug, thereby realizing the improvement of the efficiency of optimizing drug molecules.

[0070] S203: Perform fragment update processing on the first molecular fragment data to obtain multiple updated molecular fragment data;

[0071] The first molecular fragment data is any one of the multiple molecular fragment data. The multiple updated molecular fragment data corresponds to the first molecular fragment data. The first molecular fragment data is the data of the first molecular fragment. After the first molecular fragment undergoes fragment update processing, multiple updated molecular fragments are obtained. The multiple updated molecular fragments are obtained by performing multiple updates at different positions in the first molecular fragment. The multiple updated molecular fragment data are the data of the multiple updated molecular fragments.

[0072] Perform a first fragment modification process on the first molecular fragment data to obtain first modified molecular fragment data, where the first modified molecular fragment data is the molecular fragment data obtained after modifying the first molecular fragment at the first position; perform a second fragment modification process on the first modified molecular fragment data to obtain second modified molecular fragment data, where the second modified molecular fragment data is the molecular fragment data obtained after modifying the first modified molecular fragment at the second position; obtain a plurality of updated molecular fragment data, and the plurality of updated molecular fragment data includes the first modified molecular fragment data and the second modified molecular fragment data.

[0073] There are multiple updatable positions in the first molecular fragment. Update the first molecular fragment sequentially according to the multiple updatable positions. For example, perform a first update process on the first molecular fragment at the first update position to obtain a first updated molecular fragment; perform a second update process on the first updated molecular fragment at the second update position to obtain a second updated molecular fragment; perform a third update process on the first updated molecular fragment and the second updated molecular fragment respectively at the third update position to obtain third and fourth updated molecular fragments corresponding to the first updated molecular fragment and the second updated molecular fragment respectively. Each update process only changes one position, and the updated molecular fragment data obtained from the previous update process is used as the molecular fragment to be updated in the next update process. Based on a preset number of iterations, obtain a plurality of updated molecular fragments and obtain data corresponding to the plurality of updated molecular fragments.

[0074] In another alternative embodiment of the present application, a method for updating molecular fragments is provided. Set the update positions according to the similarity between the updated molecular fragment and the first molecular fragment. The first update position is located at the farthest end of the molecular structure of the first molecular fragment, the second update position is located at the second farthest end of the molecular structure of the first molecular fragment, and the third update position is located at the proximal end of the first molecular fragment. When updating the first molecular fragment, perform a first update on the first molecular fragment at the first update position to obtain a first updated molecular fragment; perform a second update on the first updated molecular fragment at the second update position to obtain a second updated molecular fragment; perform a third update on the second updated molecular fragment at the third update position to obtain a third updated molecular fragment. By performing sequential updates at different positions according to the structural similarity between the updated molecular fragment and the first molecular fragment, screen the properties of the process-optimized drug obtained by separately updating the molecular fragments at different update positions, and obtain a process-optimized drug that meets the preset properties.

[0075] In the process of updating the molecular fragment, the first molecular fragment is coded based on similarity to obtain the first molecular fragment code. The first molecular fragment code can be a computer coding language such as binary code. The end of the first molecular fragment code corresponds to the end of the first molecular fragment structure. By updating the end of the first molecular fragment code, such as replacing the binary value of the end of the first molecular fragment code with 0-1, the updated molecular fragment obtained after the update is screened for preset attributes, thereby achieving the update of the molecular fragment to meet the preset attributes when the structural similarity between the updated molecular fragment and the original molecular fragment is from high to low.

[0076] In an embodiment of the present application, by updating the molecular fragments according to the set update position order, the process optimization drugs are obtained according to the molecular structure similarity from high to low, thereby achieving the goal of obtaining the target optimized drug with smaller molecular structure difference from the drug to be optimized while meeting the requirements of the preset properties, and achieving the goal of obtaining the target optimized drug that meets the preset properties while maintaining the structural similarity as much as possible.

[0077] In another optional embodiment of the present application, a method for reinforcing a learning model is proposed to obtain a reference drug data set, where the reference drug data set includes multiple reference drug data, which are data used to characterize active compounds. Open source data on multiple active compounds for a certain target can be obtained from the ChEMBL or ZINC database.

[0078] By performing molecular structure-based cutting processing on multiple reference drug data, multiple reference molecular fragment data are obtained;

[0079] Similarity calculation is performed on multiple reference molecular fragment data to obtain multiple reference similarity data, wherein the reference similarity is the similarity between any two reference molecular fragments in the multiple reference molecular fragments, and the reference similarity data is data used to represent the reference similarity. A balanced binary tree based on the similarity of drug molecular fragments is constructed based on the multiple reference similarity data. The combination of molecular fragment structure and the order of connection points is always unique, so different drug molecular fragment structures can only exist in one leaf of the tree. Ultimately, the fragments of each drug molecule are stored in the balanced binary tree.

[0080] In an embodiment of the present application, a balanced binary tree based on molecular fragment similarity is constructed by using the acquired reference drug data set. When performing molecular cutting and molecular fragment updating on the drug to be optimized, the molecular fragments to be updated are encoded by the balanced binary tree. The molecular fragment encoding can monitor and track the degree of change of the molecular fragments, thereby realizing a controllable drug optimization process.

[0081] In another optional embodiment of the present application, a method is provided for updating molecular fragments according to a reinforcement learning model to obtain a process-optimized drug. The first molecular fragment data is encoding data for representing the structure of the molecular fragment. The fragment update of the first molecular fragment is tracked by the first molecular fragment encoding data, so as to facilitate monitoring the degree of molecular change of the first molecular fragment. For example, in the process of optimizing a certain drug, a first molecular fragment A is obtained after segmentation processing. The first molecular fragment obtained by processing the first molecular fragment A by the reinforcement learning model is encoded as 100110. The first molecular fragment is updated according to the preset update position. The first update is performed at the first position to obtain 100111. The second update is performed at the second position to obtain 100100 and 100101. The third update is performed at the third position to obtain 100010, 100011, 100000 and 100001. When the preset number of iterative updates is reached, multiple updated molecular fragment data are obtained. By encoding the molecular fragments and tracking the degree of change of the molecular fragments, the monitorability of the molecular fragment update is improved, which facilitates the improvement of the controllability of the drug optimization process during the drug optimization process. In the process of updating the molecular fragments, by updating the molecular fragments at different positions according to the preset molecular fragment structural similarity, the target optimized drug with smaller molecular structure difference from the drug to be optimized is obtained, and the target optimized drug that meets the preset properties is obtained through property screening under the condition of maintaining structural similarity as much as possible.

[0082] S204: updating the drug data to be optimized according to the multiple updated molecular fragment data to obtain multiple process optimized drug data.

[0083] Figure 3 A flow chart of a data processing method for drug optimization provided in this application, such as Figure 3 As shown, the method includes the following steps:

[0084] S301: Identify multiple updated molecular fragment data to obtain first position feature data;

[0085] The first position characteristic data is used to represent the data of the molecular fragment position of the first molecular fragment in the drug to be optimized. The first molecular fragment corresponds to the first molecular fragment data. By identifying the molecular fragments to be updated corresponding to multiple updated molecular fragments, for example, multiple updated molecular fragment data are obtained based on the update of the first molecular fragment data. Therefore, the positions corresponding to the multiple updated molecular fragments in the drug molecule to be optimized are the positions corresponding to the first molecular fragment, and the first position characteristic data are identified.

[0086] S302: Replace the first molecular fragment at the first position in the drug to be optimized according to multiple updated molecular fragment data, obtaining multiple process-optimized drug data.

[0087] The first position feature data is data used to represent the first position feature. Replace the first molecular fragment in the drug to be optimized according to the obtained multiple updated molecular fragment data. Replace the first molecular fragment in sequence according to the preset order of the multiple updated molecular fragments, obtaining multiple process-optimized drug data. The multiple process-optimized drug data is optimized based on the molecular structure of the drug. By performing optimization processing on the drug to be optimized based on the molecular structure, multiple process-optimized drug data with different branched structures is obtained. By tracking the updates of different updated molecular fragments, the tracking of the drug optimization process is realized, facilitating the monitoring of the drug molecule optimization process and improving the data optimization efficiency of the optimization processing of the drug to be optimized.

[0088] S103: Perform attribute screening processing on multiple process-optimized drug data to obtain a target optimized drug data set.

[0089] The target optimized drug data set includes multiple process-optimized drug data that meet the preset attributes.

[0090] Perform attribute calculation processing on multiple process-optimized drug data to obtain multiple to-be-processed attribute data. Among them, the multiple to-be-processed attribute data respectively correspond to the multiple process-optimized drug data. Perform preset attribute calculations on the multiple target optimized drug data. The preset attributes are determined according to the drug optimization rules in the current drug R & D stage, and may include a first preset attribute, a second preset attribute, and a third preset attribute. For example, the polar surface area (PSA), quantitative evaluation of drug-likeness (QED), octanol-water partition coefficient (LOGP), etc. Calculate the first preset attribute, the second preset attribute, and the third preset attribute of the multiple process-optimized drug data respectively, obtaining multiple to-be-processed attribute data. The multiple to-be-processed attribute data includes multiple to-be-processed first preset attribute data, multiple to-be-processed second preset attribute data, and multiple to-be-processed third preset attribute data. The multiple to-be-processed first preset attribute data respectively correspond to the multiple process-optimized drug data, the multiple to-be-processed second preset attribute data respectively correspond to the multiple process-optimized drug data, and the multiple to-be-processed third preset attribute data respectively correspond to the multiple process-optimized drug data.

[0091] Filter multiple to-be-processed attribute data according to preset predicted attribute data to determine whether multiple to-be-processed attributes corresponding to the multiple to-be-processed attribute data are within the attribute range corresponding to the preset predicted attribute data; if the to-be-processed attribute corresponding to the to-be-processed attribute data is within the attribute range corresponding to the preset predicted attribute data, obtain a target optimized drug dataset, where the target optimized drug dataset includes multiple process-optimized drug data corresponding to the multiple to-be-processed attribute data.

[0092] Filter multiple to-be-processed attribute data according to preset predicted attribute data. For example, filter and process multiple to-be-processed first preset attribute data corresponding to multiple process-optimized drug data through the first preset predicted attribute data respectively; filter and process multiple to-be-processed second preset attribute data corresponding to multiple process-optimized drug data through the second preset predicted attribute data respectively; filter and process multiple to-be-processed third preset attribute data corresponding to multiple process-optimized drug data through the third preset predicted attribute data respectively. Only when the to-be-processed attribute data corresponding to the process-optimized drug data meets the preset conditions, and the preset conditions are that the to-be-processed attribute data are all within the attribute range corresponding to the corresponding predicted preset attribute data. For example, the to-be-processed first preset attribute data is within the attribute range corresponding to the first preset predicted attribute data, the to-be-processed second preset attribute data is within the attribute range corresponding to the second preset predicted attribute data, and the to-be-processed third preset attribute data is within the attribute range corresponding to the third preset predicted attribute data, then the process-optimized drug data meets the preset conditions. When obtaining multiple process-optimized drug data that meet the preset conditions, a target optimized drug dataset is obtained.

[0093] In another alternative embodiment of the present application, a method for obtaining ideal attribute data is proposed. By obtaining a reference drug dataset, the reference drug dataset includes multiple reference drug data. Through attribute analysis of multiple reference drug data, where the multiple attributes of the reference drug can be calculated and processed by open-source attribute calculation software to obtain ideal attribute data. The ideal attribute data is data used to represent the ideal attribute range. Through the ideal attribute range, attribute screening is performed on multiple process-optimized drug data corresponding to the drug to be optimized, and a target optimized drug dataset is obtained.

[0094] In another alternative embodiment of the present application, a data processing method for drug optimization is provided, such as Figure 4 shown Figure 4 is a flowchart of a data processing method for drug optimization provided by the present application, such as Figure 4 shown, and the method includes the following steps:

[0095] S401: Identify the target optimized drug dataset to obtain multiple target optimized drug data;

[0096] Obtain multiple target optimized drug data in the target optimized drug dataset, and obtain multiple process optimized drug data, where the process optimized drug data is the data of the process optimized drug that does not meet the preset attribute screening rules. The target optimized drug data includes the attribute data corresponding to the target optimized drug, such as the first target optimized drug attribute data, the second target optimized drug attribute data, and the third target optimized drug attribute data. The process optimized drug data includes the attribute data corresponding to the process optimized drug, such as the first process optimized drug attribute data, the second process optimized drug attribute data, and the third process optimized drug attribute data. Update and train the preset reinforcement learning model based on the first target optimized drug attribute data and the first process optimized drug attribute data to obtain the first attribute updated reinforcement learning model. The first attribute updated reinforcement learning model is more convenient for obtaining the process optimized drug within the ideal range of the first attribute compared to the preset reinforcement learning model, improving the efficiency of obtaining the target optimized drug.

[0097] S402: Perform model update training processing on the preset reinforcement learning model according to multiple target optimized drug data to obtain an updated reinforcement learning model;

[0098] In an alternative embodiment of the present application, a method for updating and training a reinforcement learning model is provided. The reinforcement learning model in the embodiments of the present application can be obtained based on an actor-critic model. The drug to be optimized is processed for update based on molecular fragments through the reinforcement learning model, the position information of the updated molecular fragments in the drug molecule is calculated, preset attribute calculations are performed on the process optimized drugs obtained by changing different molecular fragments, and the updates at different positions in the drug molecule are adjusted according to the attribute calculation results, improving the efficiency of drug optimization.

[0099] The reinforcement learning model in the embodiments of the present application gives a higher reward to the molecules that produce attributes within the ideal range. The reward for meeting a given attribute range is negatively correlated with the difficulty of finding the compound of this target optimized drug. The calculation method of the reward function in the i-th generation is as follows:

[0100] N

[0099] , , , i , i , ,

[0100] =∑check(p i ,s), Where M is the total number of molecules in the i-th generation, Pi is the number of molecules in the i-th generation that meet the multi-parameter optimization goal, Ni is the number of molecules in the i-th generation that meet the multi-attribute optimization goal and are similar, s is the similarity, δ is the threshold constant, β is the learning rate, and check(pi, s) is an indicator function that has a value of 1 when the similarity of molecule pi is greater than the threshold δ, and 0 otherwise. This reward function ensures that the model achieves the goal of multi-attribute optimization while ensuring the molecular structure similarity. By reducing the degree of difference in the molecular structure between the updated drug compound and the pre-updated drug compound during the process of updating the drug compound, the molecular structure similarity between the updated and pre-updated drug compounds is improved, and multi-attribute optimization of the drug compound is achieved while ensuring the structural similarity between the updated and pre-updated drug compounds.

[0101] For example, during the optimization process of a drug B to be optimized, after the reinforcement learning model performs molecular fragment update processing on the drug B to be optimized, process-optimized drugs B1, B2, and B3 are obtained. The process-optimized drug data B1, B2, and B3 are process-optimized drugs obtained by updating the molecular fragments of the drug molecule to be optimized at the first, second, and third positions respectively. Calculate the data of the first attribute h of the process-optimized drugs B1, B2, and B3, which are h1, h2, and h3 respectively. Calculate the data of the second attribute j of the process-optimized drugs B1, B2, and B3, which are j1, j2, and j3 respectively. Calculate the data of the third attribute k of the process-optimized drugs B1, B2, and B3, which are k1, k2, and k3 respectively. Update and train the reinforcement learning model according to the data of the first attribute h of the process-optimized drugs B1, B2, and B3, which are h1, h2, and h3, to obtain an updated reinforcement learning model. The updated reinforcement learning model realizes the adaptation of the drug to be optimized to the first attribute, improving the efficiency of obtaining process-optimized drugs that meet the first attribute. According to the results of the dynamic update, h1 and h2 are within the ideal range of the first attribute h, and h3 is outside the ideal range of the first attribute h. Update and train the reinforcement learning model through h1, h2, and h3 to improve the processing of further optimizing the molecular structure of the drug to be optimized based on the process-optimized drugs B1 and B2. Perform molecular update processing on the process-optimized drugs B1 and B2 through the updated reinforcement learning model to obtain B11, B12, B21, and B22. Update the reinforcement learning model again through the ideal range of the first attribute h until the preset number of iterations is reached. The process of optimizing the drug to be optimized based on the second attribute j and the third attribute k is the same as the optimization process of the first attribute h. By setting the above dynamic update drug optimization process, multi-attribute optimization processing of the drug to be optimized can be achieved, improving the data processing efficiency of drug optimization processing.

[0102] S403: Perform molecular update processing based on the molecular structure on the drug data to be optimized using the updated reinforcement learning model, obtaining multiple optimized drug data during the update process;

[0103] S404: Perform attribute screening processing on the multiple optimized drug data during the update process to obtain an updated target optimized drug data set.

[0104] In the embodiments of the present application, by updating the reinforcement learning model, iterative optimization processing of the drug data to be optimized is achieved. Through the dynamically updated drug optimization process, multi-attribute optimization processing of the drug to be optimized is realized, improving the data processing efficiency of drug optimization processing.

[0105] Figure 5 The structural schematic diagram of a data processing device for drug optimization provided by the present application is as follows Figure 5 shown, the device includes:

[0106] A data acquisition module 51, configured to acquire drug data to be optimized, where the drug data to be optimized is the relevant data of the drug to be optimized;

[0107] A preprocessing module 52, configured to perform multiple preprocessings on the drug data to be optimized based on a preset reinforcement learning model, obtaining multiple optimized drug data during the process, where the multiple preprocessings are preset molecular processing based on the molecular structure, and the multiple optimized drug data during the process are drug data with multiple different molecular structures obtained after the preprocessing of the drug to be optimized; and

[0108] A screening module 53, configured to perform attribute screening processing on the multiple optimized drug data during the process to obtain a target optimized drug data set, where the target optimized drug data set includes multiple optimized drug data during the process that meet the preset attributes.

[0109] Figure 6 The structural diagram of another data processing device for drug optimization provided by the present application is as follows Figure 6 shown, the device includes:

[0110] An identification module 61, configured to perform feature identification on the drug data to be optimized, obtaining molecular structure feature data, where the molecular structure feature data is data used to represent the molecular structure features of the drug to be optimized;

[0111] A cutting module 62, configured to perform molecular cutting processing on the drug data to be optimized according to the molecular structure feature data, obtaining multiple molecular fragment data, where the multiple molecular fragment data are data corresponding to multiple molecular fragments in the drug to be optimized;

[0112] A fragment update module 63 is configured to perform fragment update processing on the first molecular fragment data to obtain a plurality of updated molecular fragment data, where the first molecular fragment data is any one of the plurality of molecular fragment data, and the plurality of updated molecular fragment data corresponds to the first molecular fragment data; and

[0113] A drug update module 64 is configured to perform update processing on the drug data to be optimized according to the plurality of updated molecular fragment data to obtain the plurality of process-optimized drug data.

[0114] The specific manners of the execution operations of the above-mentioned units in the embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0115] In summary, in the present application, by obtaining the drug data to be optimized, performing a plurality of preprocessings on the drug data to be optimized based on a preset reinforcement learning model, where the plurality of preprocessings are preset molecular processings based on the molecular structure, to obtain a plurality of process-optimized drug data, performing attribute screening processing on the plurality of process-optimized drug data to obtain a target optimized drug data set, and the target optimized drug data set includes a plurality of process-optimized drug data that meet the preset attributes. By performing preprocessing based on the molecular structure on the drug to be optimized through the preset reinforcement learning model, drug data with a plurality of different molecular structures is obtained, solving the technical problem of low data optimization efficiency in the prior art during the drug molecule optimization process, and achieving the technical effect of improving the data optimization efficiency of drug molecules.

[0116] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0117] Obviously, those skilled in the art should understand that the above-mentioned units or steps of the present application can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device, so that they can be stored in a storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module to implement. In this way, the present application is not limited to any specific combination of hardware and software.

[0118] The above are only the preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A data processing method for drug optimization, characterized in that, Including: Obtain the drug data to be optimized, where the drug data to be optimized is the relevant data of the drug to be optimized; Perform multiple preprocessings on the drug data to be optimized based on a preset reinforcement learning model to obtain multiple process-optimized drug data, where the multiple preprocessings are preset molecular processes based on the molecular structure, and the multiple process-optimized drug data are drug data with multiple different molecular structures obtained after the drug to be optimized undergoes the preprocessings; and Perform attribute screening processing on the multiple process-optimized drug data to obtain a target optimized drug data set, where the target optimized drug data set includes multiple process-optimized drug data that meet the preset attributes; Among them, performing multiple preprocessings on the drug data to be optimized based on a preset reinforcement learning model includes: performing molecular cutting and molecular fragment updating on the drug to be optimized through a balanced binary tree, and performing encoding processing on the molecular fragment to be updated through a balanced binary tree to obtain an updated molecular fragment encoding; Among them, obtain a reference drug data set, where the reference drug data set includes multiple reference drug data, and the reference drug data are data used to characterize active compounds; perform cutting processing based on the molecular structure on the multiple reference drug data to obtain multiple reference molecular fragment data; perform similarity calculation processing on the multiple reference molecular fragment data to obtain multiple reference similarity data, and construct a balanced binary tree based on the similarity of drug molecular fragments based on the multiple reference similarity data; Perform molecular fragment-based update processing on the drug to be optimized through a reinforcement learning model, calculate the position information of the updated molecular fragment in the drug molecule, perform preset attribute calculation on the process-optimized drug obtained according to the changes of different molecular fragments, adjust the updates at different positions in the drug molecule according to the attribute calculation results, and the reinforcement learning model gives a higher reward to the molecule that generates attributes within the ideal range, where the reward for meeting a given attribute range is negatively correlated with the difficulty of finding the compound of the target optimized drug, and the calculation method of the reward function in the i-th generation is as follows: N i = ∑ check(p i , s), where M is the total number of molecules in the i-th generation, Pi is the number of molecules in the i-th generation that meet the multi-parameter optimization goal, Ni is the number of molecules in the i-th generation that meet the multi-attribute optimization goal and are similar, s is the similarity, δ is the threshold constant, β is the learning rate, and check(pi, s) is an indicator function that has a value of 1 when the similarity of molecule pi is greater than the threshold δ and 0 otherwise.

2. The data processing method according to claim 1, wherein Performing multiple preprocessings on the drug data to be optimized based on a preset reinforcement learning model to obtain multiple process-optimized drug data includes: Perform feature recognition on the drug data to be optimized to obtain molecular structure feature data, where the molecular structure feature data is data used to represent the molecular structure features of the drug to be optimized; Perform molecular cutting processing on the drug data to be optimized according to the molecular structure feature data to obtain multiple molecular fragment data, where the multiple molecular fragment data are data corresponding to multiple molecular fragments in the drug to be optimized; Perform fragment update processing on the first molecular fragment data to obtain multiple updated molecular fragment data, where the first molecular fragment data is any one of the multiple molecular fragment data, and the multiple updated molecular fragment data correspond to the first molecular fragment data; and Perform update processing on the drug data to be optimized according to the multiple updated molecular fragment data to obtain the multiple process-optimized drug data.

3. The data processing method according to claim 2, wherein Perform fragment update processing on the first molecular fragment data to obtain multiple updated molecular fragment data, including: Perform first fragment modification processing on the first molecular fragment data to obtain first modified molecular fragment data, where the first modified molecular fragment data is the molecular fragment data obtained after modifying the first molecular fragment at the first position; Perform second fragment modification processing on the first modified molecular fragment data to obtain second modified molecular fragment data, where the second modified molecular fragment data is the molecular fragment data obtained after modifying the first modified molecular fragment at the second position; Obtain the multiple updated molecular fragment data, where the multiple updated molecular fragment data includes the first modified molecular fragment data and the second modified molecular fragment data.

4. The data processing method according to claim 2, wherein Perform update processing on the drug data to be optimized according to the multiple updated molecular fragment data to obtain the multiple process-optimized drug data, including: Identify the multiple updated molecular fragment data to obtain first position feature data, where the first position feature data is the data used to represent the molecular fragment position of the first molecular fragment in the drug to be optimized, and the first molecular fragment corresponds to the first molecular fragment data; and Replace the first molecular fragment at the first position in the drug to be optimized according to the multiple updated molecular fragment data respectively to obtain the multiple process-optimized drug data, where the first position corresponds to the first position feature data.

5. The data processing method according to claim 1, wherein Perform attribute screening processing on the multiple process-optimized drug data to obtain a target optimized drug data set, including: Perform attribute calculation processing on the multiple process-optimized drug data to obtain multiple attribute data to be processed, where the multiple attribute data to be processed correspond to the multiple process-optimized drug data respectively; Screen the multiple attribute data to be processed according to the preset predicted attribute data to determine whether the multiple attributes corresponding to the multiple attribute data to be processed are within the attribute range corresponding to the preset predicted attribute data; and If the attribute corresponding to the attribute data to be processed is within the attribute range corresponding to the preset predicted attribute data, obtain the target optimized drug data set, where the target optimized drug data set includes the multiple process-optimized drug data corresponding to the multiple attribute data to be processed.

6. The data processing method according to claim 1, characterized in that After performing attribute screening processing on the multiple process-optimized drug data to obtain a target optimized drug data set, the method further includes: Identify the target optimized drug data set to obtain multiple target optimized drug data; Perform model update training processing on the preset reinforcement learning model according to the multiple target optimized drug data to obtain an updated reinforcement learning model; Perform molecular update processing based on the molecular structure on the drug data to be optimized based on the updated reinforcement learning model to obtain multiple updated process-optimized drug data; and Perform attribute screening processing on the multiple updated process-optimized drug data to obtain an updated target optimized drug data set.

7. A data processing device for drug optimization, characterized in that Including: A data acquisition module for acquiring data of a drug to be optimized, where the data of the drug to be optimized is related data of the drug to be optimized; A preprocessing module for performing multiple preprocessings on the data of the drug to be optimized based on a preset reinforcement learning model to obtain multiple process-optimized drug data, where the multiple preprocessings are preset molecular processes based on the molecular structure, and the multiple process-optimized drug data are drug data with multiple different molecular structures obtained after the preprocessing of the drug to be optimized; and A screening module for performing attribute screening on the multiple process-optimized drug data to obtain a target optimized drug data set, where the target optimized drug data set includes multiple process-optimized drug data that meet the preset attributes; Among them, performing multiple preprocessings on the data of the drug to be optimized based on a preset reinforcement learning model includes: performing molecular cutting and molecular fragment updating on the drug to be optimized through a balanced binary tree, and encoding the molecular fragments to be updated through a balanced binary tree to obtain updated molecular fragment encodings; Among them, obtaining a reference drug data set, where the reference drug data set includes multiple reference drug data, and the reference drug data are data for characterizing active compounds; performing cutting processing based on the molecular structure on the multiple reference drug data to obtain multiple reference molecular fragment data; performing similarity calculation processing on the multiple reference molecular fragment data to obtain multiple reference similarity data, and constructing a balanced binary tree based on the similarity of drug molecular fragments based on the multiple reference similarity data; Performing molecular fragment-based updating processing on the drug to be optimized through a reinforcement learning model, calculating the position information of the updated molecular fragments in the drug molecule, performing preset attribute calculation on the process-optimized drugs obtained according to the changes of different molecular fragments, adjusting the updates at different positions in the drug molecule according to the attribute calculation results, and the reinforcement learning model gives higher rewards to molecules that generate attributes within an ideal range, where the reward for meeting a given attribute range is negatively correlated with the difficulty of finding the compound of this target optimized drug, and the calculation method of the reward function in the i-th generation is as follows: N i = ∑check(p i , s), where Where M is the total number of molecules in the i-th generation, Pi is the number of molecules that meet the multi-parameter optimization target in the i-th generation, Ni is the number of molecules that meet the multi-attribute optimization target and are similar in the i-th generation, s is the similarity, δ is the threshold constant, β is the learning rate, and check(pi, s) is an indicator function, whose value is 1 when the similarity of molecule pi is greater than the threshold δ, otherwise it is 0.

8. The data processing device according to claim 7, wherein The preprocessing module includes: An identification module for performing feature identification on the data of the drug to be optimized to obtain molecular structure feature data, where the molecular structure feature data are data for representing the molecular structure features of the drug to be optimized; A cutting module for performing molecular cutting on the data of the drug to be optimized according to the molecular structure feature data to obtain multiple molecular fragment data, where the multiple molecular fragment data are data corresponding to multiple molecular fragments in the drug to be optimized; A fragment update module, configured to perform fragment update processing on first molecular fragment data to obtain a plurality of updated molecular fragment data, wherein the first molecular fragment data is any one of the plurality of molecular fragment data, and the plurality of updated molecular fragment data corresponds to the first molecular fragment data; and A drug update module, configured to update the drug data to be optimized according to the plurality of updated molecular fragment data to obtain the plurality of process-optimized drug data.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the data processing method for drug optimization according to any one of claims 1-6.

10. An electronic device, characterized in that, Comprising: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to cause the at least one processor to execute the data processing method for drug optimization according to any one of claims 1-6.

Images