Structure-property relationship analysis method and electronic equipment based on multidimensional interpretability

Abstract

The present disclosure relates to a structure-activity relationship analysis method and electronic device based on multidimensional interpretability, the method comprising: obtaining a series of ligand molecules; calculating target feature data of the series of ligand molecules, the target feature data including molecular descriptors, parent core structures, substituent groups and protein-ligand interactions; obtaining the real molecular activity data of each ligand molecule obtained based on experiments; using the target feature data of the ligand molecule as sample data, using the real molecular activity data as label data of the corresponding ligand molecule, and training the target machine learning model based on the sample data and label data; analyzing the target machine learning model based on the Shapley addition and interpretation algorithm to obtain the Shapley value of each target feature data; determining the influence of each target feature data on the molecular activity based on the Shapley value. The method aims to promote the intelligence and precision of structure-activity relationship analysis based on the interpretability of artificial intelligence models based on multidimensional feature representation and the Shapley addition and interpretation algorithm.

Description

Technical Field

[0001] The present disclosure relates to the field of drug-assisted design, and in particular, to a structure-activity relationship analysis method and electronic equipment based on multi-dimensional interpretability. Background Art

[0002] Drug development is a large and competitive market, with a growing demand for rapid discovery and development of new drugs. The traditional drug development process requires a lot of time and resources, and there are high costs and uncertainties. Therefore, the market demand for efficient, intelligent and precise drug development methods is gradually increasing. In recent years, with the continuous improvement of computer performance, the application of artificial intelligence in drug design has developed rapidly.

[0003] Structure-Activity Relationship (SAR) analysis is a key step in drug design. It guides the design and optimization of new drugs by studying the relationship between the chemical structure of a compound and its biological activity. Traditional SAR analysis methods usually require a large amount of experimental data and complex computational fitting, which is cumbersome and time-consuming. With the development of machine learning and deep learning algorithms, the application of AI in SAR analysis has gradually become popular. AI technology can quickly process and analyze a large amount of compound data to obtain the structure-activity relationship of a series of compounds, and use it to predict the activity and toxicity of potential candidate drugs and optimize the molecular structure. Compared with traditional methods, AI-driven SAR analysis can significantly shorten the drug development cycle, reduce costs, and increase the success rate.

[0004] However, in related technologies, artificial intelligence still has certain problems in SAR analysis, such as the single dimension of feature description, which affects the accuracy of SAR analysis, and the opaque internal operation of the computational model, and the decision-making process and prediction output results are difficult to explain, leading to ethical issues such as trust. Summary of the invention

[0005] The purpose of the present disclosure is to provide a structure-activity relationship analysis method and electronic equipment based on multidimensional interpretability, aiming to expand the feature description dimension, enhance the interpretability of the model, and provide more comprehensive and in-depth information support for structure-activity relationship analysis.

[0006] In order to achieve the above-mentioned object, the present disclosure provides, in a first aspect, a structure-activity relationship analysis method based on multidimensional interpretability, the method comprising:

[0007] Obtain a series of ligand molecules;

[0008] Calculating target feature data of the series of ligand molecules, wherein the target feature data includes a molecular descriptor, a core structure, a substitution group relative to the core structure, and a protein-ligand interaction between the ligand molecules and a target protein;

[0009] Obtaining experimentally obtained real molecular activity data of each of the ligand molecules;

[0010] Using the target feature data of the ligand molecule as sample data, using the real molecular activity data as label data of the corresponding ligand molecule, and training a target machine learning model based on the sample data and the label data;

[0011] Analyzing the target machine learning model based on the Shapley addition and interpretation algorithm to obtain the Shapley value of each target feature data, wherein the Shapley value is used to characterize the correlation between the effect of the target feature data on the molecular activity;

[0012] The influence of each target characteristic data on the molecule activity is determined based on the Shapley value.

[0013] Optionally, determining the influence of each target characteristic data on the molecular activity based on the Shapley value comprises:

[0014] When the Shapley value is positive, it is determined that the larger the Shapley value is, the greater the positive effect of the target characteristic data corresponding to the Shapley value on the molecular activity is;

[0015] When the Shapley value is negative, it is determined that the larger the absolute value of the Shapley value is, the greater the negative impact of the target characteristic data corresponding to the Shapley value on the molecular activity is.

[0016] When the Shapley value is zero, it is determined that the target characteristic data corresponding to the Shapley value has no effect on the molecular activity.

[0017] Optionally, the method further comprises:

[0018] Obtain the molecule to be optimized;

[0019] Based on the influence of each target characteristic data determined by the Shapley value on the molecular activity, the structure of the molecule to be optimized is optimized to increase the molecular activity of the molecule to be optimized.

[0020] Optionally, the structural optimization of the molecule to be optimized based on the influence of each target characteristic data determined by the Shapley value on the molecule activity comprises:

[0021] Obtaining candidate molecular descriptors, selecting a molecular descriptor with a positive Shapley value from the candidate molecular descriptors, and adding a molecular substructure described by the molecular descriptor with a positive Shapley value to the molecule to be optimized; and / or

[0022] Reducing the molecular substructures described by the molecular descriptors corresponding to negative Shapley values ​​in the molecule to be optimized; and / or

[0023] The molecular substructure in the molecule to be optimized is replaced by the molecular substructure described by the molecular descriptor corresponding to the larger Shapley value.

[0024] Optionally, the structural optimization of the molecule to be optimized based on the influence of each target characteristic data determined by the Shapley value on the molecule activity comprises:

[0025] The mother core in the molecule to be optimized is replaced with a mother core corresponding to a larger Shapley value.

[0026] Optionally, the structural optimization of the molecule to be optimized based on the influence of each target characteristic data determined by the Shapley value on the molecule activity comprises:

[0027] Obtaining candidate substituent groups, selecting a substituent group with a positive Shapley value from the candidate substituent groups, and adding the substituent group with a positive Shapley value to the molecule to be optimized; and / or

[0028] Reducing the substituent groups with negative corresponding Shapley values ​​in the molecule to be optimized; and / or

[0029] The substituent groups in the molecule to be optimized are replaced with substituent groups corresponding to a larger Shapley value.

[0030] Optionally, the structural optimization of the molecule to be optimized based on the influence of each target characteristic data determined by the Shapley value on the molecule activity comprises:

[0031] Obtaining candidate protein-ligand interactions, selecting a protein-ligand interaction with a positive Shapley value from the candidate protein-ligand interactions, and adjusting the structure of the molecule to be optimized so that the molecule to be optimized and the target protein generate the protein-ligand interaction with a positive Shapley value; and / or

[0032] Adjusting the structure of the molecule to be optimized to reduce the protein-ligand interaction between the molecule to be optimized and the target protein with a negative Shapley value; and / or

[0033] The structure of the molecule to be optimized is adjusted to replace the protein-ligand interaction between the molecule to be optimized and the target protein with a protein-ligand interaction with a larger Shapley value.

[0034] Optionally, the calculating target characteristic data of the series of ligand molecules comprises:

[0035] Calculating a molecular descriptor for each of the ligand molecules, wherein the molecular descriptor includes molecular low-dimensional structure information and molecular 3D structure information;

[0036] Calculating the core structure of the series of ligand molecules and the substituent groups of each ligand molecule relative to the core structure;

[0037] A target protein of each of the ligand molecules is selected and a 3D conformation of the target protein is obtained, the 3D conformation of the ligand molecule is generated, the 3D conformation of the ligand molecule is molecularly docked with the 3D conformation of the target protein, and after the molecular docking is completed, the protein-ligand interaction between the ligand molecule and the target protein is calculated, wherein the protein-ligand interaction includes amino acid residues and protein-ligand interaction types.

[0038] Optionally, the molecular low-dimensional structure information includes molecular 0D structure information, molecular 1D structure information, molecular 2D structure information and molecular property descriptors, and / or the molecular 3D structure information includes molecular shape, three-dimensional atomic space coordinates, shape complexity, and best fit plane.

[0039] According to a second aspect of the present disclosure, there is provided an electronic device, including:

[0040] a memory having a computer program stored thereon;

[0041] A processor is used to execute the computer program in the memory to implement the steps of any one of the methods described in the first aspect above.

[0042] Through the above technical solution, after obtaining a series of ligand molecules, their target feature data are calculated. The target feature data includes molecular descriptors, parent core structures, substitution groups, and protein-ligand interactions, which greatly broadens the dimension of the calculation input data and provides more data basis for the calculation, thereby improving the accuracy of the calculation. After that, the target feature data is used together with the real molecular activity data to train the target machine learning model, and the target machine learning model is analyzed using the Shapley addition and interpretation algorithm to calculate the Shapley value of each target feature data, so as to use the Shapley value analysis to determine the impact of each target feature data on the molecular activity, and then realize the molecular structure-activity relationship analysis. The Shapley addition and interpretation algorithm analyzes and infers the decision-making process and prediction results of the machine learning model, which can clearly understand the influence of each feature on the model prediction, understand the model decision-making process, and on this basis improve the efficiency of molecular structure-activity relationship analysis.

[0043] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings:

[0045] Figure 1 It is a flowchart of a structure-activity relationship analysis method based on multidimensional interpretability according to an exemplary embodiment.

[0046] Figure 2 The figure is a flowchart of multi-dimensional feature data calculation according to an exemplary embodiment.

[0047] Figure 3 is another flow chart of a structure-activity relationship analysis method based on multidimensional interpretability according to an exemplary embodiment.

[0048] Figure 4 is another flow chart of a structure-activity relationship analysis method based on multidimensional interpretability according to an exemplary embodiment.

[0049] Figure 5 is a schematic diagram of molecular structure optimization according to an exemplary embodiment.

[0050] Figure 6 is another schematic diagram showing molecular structure optimization according to an exemplary embodiment.

[0051] Figure 7 It is a block diagram of a structure-activity relationship analysis device based on multi-dimensional interpretability according to an exemplary embodiment.

[0052] Figure 8 It is a block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0053] The specific implementation of the present disclosure is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the present disclosure, and is not used to limit the present disclosure.

[0054] Figure 1 is a flowchart of a structure-activity relationship analysis method based on multidimensional interpretability according to an exemplary embodiment, see Figure 1 , the method comprising:

[0055] S101, obtaining a series of ligand molecules.

[0056] S102, calculating target feature data of the series of ligand molecules, wherein the target feature data includes a molecular descriptor, a core structure, a substituent group relative to the core structure, and a protein-ligand interaction between the ligand molecule and the target protein.

[0057] S103, obtaining the real molecular activity data of each ligand molecule obtained based on experiments.

[0058] S104, using the target feature data of the ligand molecule as sample data, using the real molecular activity data as label data of the corresponding ligand molecule, and training a target machine learning model based on the sample data and the label data.

[0059] S105, analyzing the target machine learning model based on the Shapley addition and interpretation algorithm to obtain the Shapley value of each target feature data, wherein the Shapley value is used to characterize the correlation between the target feature data and the molecular activity.

[0060] S106, determining the influence of each target characteristic data on the molecular activity based on the Shapley value.

[0061] In step S101, a group or multiple groups of molecules with similar structures containing multiple ligand molecules can be obtained from an internal compound library, a commercial compound library, etc. as a series of ligand molecules. Specifically, the acquisition of ligand molecules can be completed by obtaining their SMILES. The acquisition process can be achieved through wired or wireless communication. SMILES is a specification that describes molecular structures using ASCII strings, mainly used to represent the structures of organic molecules and other chemical substances. SMILES uses a series of characters to represent atoms, bonds and molecular structures.

[0062] After step S101 is completed, step S102 may be entered. In step S102, each target feature data may be calculated by obtaining relevant data of a series of ligand molecules, such as SMILES, and the target feature data may be described using at least four types of feature information, including molecular descriptors, parent core structures, substitution groups relative to the parent core structures, and protein-ligand interactions between the ligand molecules and the target protein.

[0063] Optionally, in step S102, calculating target characteristic data of the series of ligand molecules includes:

[0064] S1021, calculating a molecular descriptor of each of the ligand molecules, wherein the molecular descriptor includes molecular low-dimensional structure information and molecular 3D structure information.

[0065] S1022, calculating the parent core structure of the series of ligand molecules and the substitution group of each ligand molecule relative to the parent core structure.

[0066] S1023, selecting a target protein of each of the ligand molecules and obtaining a 3D conformation of the target protein, generating a 3D conformation of the ligand molecule, performing molecular docking between the 3D conformation of the ligand molecule and the 3D conformation of the target protein, and calculating the protein-ligand interaction between the ligand molecule and the target protein after the molecular docking is completed, wherein the protein-ligand interaction includes amino acid residues and protein-ligand interaction types.

[0067] In step S1021, the molecular descriptor may include molecular low-dimensional structure information and molecular 3D structure information, wherein the molecular low-dimensional structure information may include molecular 0D structure information, molecular 1D structure information, molecular 2D structure information and molecular property descriptors, and the molecular 0D structure information may be described by 0D descriptors, which are mainly composed of structural descriptors and quantitative descriptors, such as molecular weight, number of atoms, number of various rings and number of double bonds. The molecular 1D structure information may be described by 1D descriptors, which are composed of structural fragments or molecular fingerprints, such as some specific groups such as keto, azido, amino, etc. The molecular 2D structure information may be described by 2D descriptors, such as topological descriptors. Molecular property descriptors are used to describe the basic properties of molecules such as some physical and chemical properties.

[0068] In view of the structure-activity relationship analysis requirements of the present invention, in order to further expand the dimension of input data, optionally, the molecular 0D structural information includes molecular weight, chemical bond type, the molecular 1D structural information includes the number of hydrogen bond donors, the number of hydrogen bond acceptors, and the polar surface area, the molecular 2D structural information includes molecular shape, hybridization type, and molecular property descriptors include fat solubility, water solubility, and drug-like properties. In a specific embodiment, the molecular low-dimensional structural information includes a total of 210 kinds of characteristic data, which can be input into the calculation module for calculation. The molecular 3D structural information is used to describe the three-dimensional structure of the molecule. Optionally, the molecular 3D structural information includes molecular shape, three-dimensional atomic space coordinates, shape complexity, and best fit plane. By calling the relevant algorithm in the prior art, the SMILES of each ligand molecule in the series of ligand molecules can be directly used to calculate the corresponding above-mentioned molecular descriptors.

[0069] After step S1021 is completed, step S1022 may be executed. In step S1022, optionally, the core structure of the series of ligand molecules and the substituent groups of each ligand molecule relative to the core structure are calculated, including:

[0070] S10221, based on the FMCS algorithm, cluster analysis is performed on the series of ligand molecules to obtain one or more classes of ligand molecule sets, and the maximum common substructure of each class of ligand molecule sets is calculated as the parent core structure of the class of ligand molecule sets.

[0071] S10222, for each ligand molecule in each type of ligand molecule set, breaking the chemical bond between the ligand molecule and the parent core structure of the type of ligand molecule set to obtain a substitution group of the ligand molecule relative to the parent core structure.

[0072] In step S102221, the frequent maximum common substructure FMCS (Frequent Maximum Common Substructure) algorithm is a method for identifying the maximum common substructure between a group of molecules in chemical informatics, and the maximum common substructure (MCS) problem is modeled as a graph isomorphism problem, and subgraph enumeration and subgraph isomorphism testing are used to process multiple MCS problems. The maximum common substructure is the maximum structure that each ligand molecule has in each class of ligand molecule set. The SMILES of the series of ligand molecules can be used as the input of the FMCS algorithm, so that the series of ligand molecules are first clustered and analyzed to obtain one or more classes of ligand molecule combinations, and each class of ligand molecule set has a similar structure, and then the maximum common substructure of each class of ligand molecule set is calculated as its parent core structure. Of course, in possible implementations, other calculation methods can also be used to calculate the parent core structure.

[0073] After step S10221 is completed, step S10222 may be executed. In step S10222, for each ligand molecule in each type of ligand molecule set, the chemical bonds between these ligand molecules and the parent core structure of the series of ligand molecules to which they belong are broken to obtain the substitution group of the ligand molecule relative to the parent core structure.

[0074] After step S1022 is completed, step S1023 is executed. Molecular docking is a process in which two or more molecules recognize each other through geometric matching and energy matching. Figure 2 In step S1023, the target protein of each ligand molecule can be selected by using relevant computing software, and its 3D conformation can be obtained. The 3D conformation of the target protein can be pre-built into the protein database. Generate the 3D conformation of the ligand molecule, and perform molecular docking between the 3D conformation of the ligand molecule and the 3D conformation of the target protein. After the molecular docking calculation is completed, the protein-ligand interaction between the ligand molecule and the target protein can be calculated. Among them, the protein-ligand interaction can include amino acid residues and protein-ligand interaction types, for example, (94|CYS-HygB) represents the hydrogen bond formed between the ligand and the 94th cysteine, cysteine ​​is an amino acid residue, and the hydrogen bond is a protein-ligand interaction type.

[0075] In step S103, the real molecular activity data is the real molecular activity data obtained through experiments, which is used to characterize the molecular activity. This data is difficult to obtain through computer simulation and needs to be obtained through experiments. Step S103 can be performed after step S102 or before step S101, as long as step S103 is performed before step S104, and the present invention does not make any specific restrictions on this.

[0076] After step S103 is completed, step S104 may be entered. In step S104, see Figure 3 , the molecular descriptors of the ligand molecule calculated in the above steps, the core structure, the substitution groups relative to the core structure, and the protein-ligand interaction between the ligand molecule and the target protein can be used as sample data, and the acquired real molecular activity data can be used as label data to form a training set for the target machine learning model. The target machine learning model is trained using the sample data and label data in the training set, so that the trained target machine learning model has the ability to predict the corresponding molecular activity based on the target feature data.

[0077] Optionally, the target machine learning model can be selected according to actual conditions, such as a random forest model, a neural network model and an XGBoost model. In a possible embodiment, the XGBoost model is preferably used. As an integrated learning algorithm based on a gradient-boosted decision tree, the XGBoost model shows excellent performance when processing tabular data, and its model effect is comparable to the top level of machine learning, and sometimes even exceeds the performance of deep learning models. Although large language models (LLMs) have made significant progress in the field of text data processing in recent years, when processing small-scale but high-quality tabular data, the accuracy of XGBoost is still better than many deep learning models and LLMs, and it is also significantly ahead of LLMs in interpretability, with scalability, high efficiency, strong generalization ability and the advantages of avoiding overfitting. The model algorithm is a strong learner built with a classification regression tree as the basic component unit in an iterative manner, which can greatly improve the accuracy and robustness of SAR analysis. In view of this, when performing molecular virtual screening based on structure-activity relationship analysis of drug molecules, XGBoost is undoubtedly a more ideal choice.

[0078] After step S104 is completed, step S105 may be executed. Figure 3 In step S105, the training result of the target machine learning model, such as the XGBoost model, is used as a new round of input and incorporated into the Shaplegi plus interpretation SHAP algorithm, so that the model can be deeply analyzed. The intrinsic information of the model is comprehensively analyzed and mined from multiple angles, and the activity information is mined based on SAR analysis. Specifically, it can include: identifying positive factors that promote activity improvement, analyzing negative factors that lead to reduced activity, determining the key factors that have the most significant impact on activity, and identifying secondary factors that have relatively small impact on activity.

[0079] In response to the "black box" problem of machine learning, some machine learning explainable algorithms in related technologies can solve it. Explainable algorithms are subdivided into intrinsic explainability and ex post explainability based on different stages of artificial intelligence algorithms. There are usually two methods for intrinsic explainability: choosing a simple model and reducing the number of input features. Explainability is achieved by limiting the complexity of the model and building a model with a simple structure, such as linear regression and decision trees, but this method often sacrifices model accuracy; ex post explainability refers to the use of some explanatory methods after the AI ​​model is trained and built to analyze and infer the decision process / prediction results. This method is very flexible and can adapt to different problems. Among them, Shapley additive exPlanation (SHAP) is one of the most popular ex post explanation methods in recent years.

[0080] The Shapley addition and interpretation algorithm is based on game theory and is a model-independent machine learning interpretation method that provides a solution for calculating the feature contribution of a single prediction of any machine learning model. Based on the Shapley addition and interpretation algorithm, the contribution of each input data of the machine learning model to the output data can be calculated, which is represented by the Shapley value. In step S105, after the XGBoost model is incorporated into the Shapley addition and interpretation algorithm for computational analysis, the Shapley values ​​of molecular descriptors, parent core structures, substitution groups, and protein-ligand interactions can be calculated. The Shapley value can be used to characterize the correlation between the effect of target feature data on molecular activity, or the contribution of target feature data to the output data of the target machine learning model, i.e., molecular activity. On the basis of the prediction model, by introducing the Shapley addition and interpretation framework, the prediction results of the model can be explained and visualized in a fine-grained manner. Researchers can clearly understand the contribution of each feature to the model prediction, so as to better understand the intrinsic relationship between the structural information and activity of drug molecules and protein targets.

[0081] After step S105 is completed, step S106 may be executed. In step S106, structure-activity relationship analysis may be performed based on the Shapley value to determine the influence of each target feature data such as molecular descriptors, parent core structure, substitution groups, and protein-ligand interaction on the molecular activity. Optionally, in step S106, the influence of each target feature data on the molecular activity is determined based on the Shapley value, including:

[0082] S1061, when the Shapley value is positive, determining that the larger the Shapley value is, the greater the positive impact of the target characteristic data corresponding to the Shapley value on the molecular activity is.

[0083] S1062, when the Shapley value is negative, determining that the greater the absolute value of the Shapley value, the greater the negative impact of the target characteristic data corresponding to the Shapley value on the molecular activity.

[0084] S1063, when the Shapley value is zero, determining that the target characteristic data corresponding to the Shapley value has no effect on the molecular activity.

[0085] In step S1061, if the Shapley value is a positive number, it can be indicated that the target feature data corresponding to the Shapley value has a positive effect on the molecular activity, that is, the feature corresponding to the target feature data will increase the molecular activity, and the larger the Shapley value, the greater the positive effect or positive contribution to the molecular activity, and the greater the improvement in the molecular activity caused by it.

[0086] In step S1062, if the Shapley value is a negative number, it may indicate that the target feature data corresponding to the Shapley value has a negative impact on the molecular activity, that is, the feature corresponding to the target feature data will reduce the molecular activity, and the larger the absolute value of the Shapley value, the greater the negative impact or negative contribution on the molecular activity, and the greater the reduction in molecular activity caused by it.

[0087] In step S1063, if the Shapley value is zero, it may indicate that the target characteristic data corresponding to the Shapley value has no effect on the molecular activity.

[0088] According to the different situations of the Shapley values ​​from step S1061 to step S1063, the influence of each target feature data on the molecular activity can be determined according to the positive and negative and size of the Shapley value. For example, for the molecular descriptors PEOE_VSA6, SPS, 3D_OverLap and 3D_USRCAT, if the calculated Shapley values ​​are greater than 0, they can be identified as positive factors, and then the molecular substructures corresponding to these molecular descriptors can be determined according to their Shapley values ​​to have a positive influence on the molecular activity, which can increase the activity of the molecule. For the molecular descriptors VSA_EState4, SMR_VSA10 and FpDensityMorgan, if their corresponding Shapley values ​​are all less than 0, they can be identified as negative factors, and then the molecular substructures corresponding to these molecular descriptors can be determined according to the Shapley values ​​to have a negative influence on the molecular activity, which can reduce the activity of the molecule.

[0089] The meanings of the molecular descriptors are as follows:

[0090] PEOE_VSA6: (Partial Equalization of Orbital Electronegativity-Volume Surface Area 6) is a descriptor based on charge distribution and volume surface area. It uses partial charge (PEOE) to calculate the volume surface area (VSA) of a specific region in a molecule. PEOE_VSA6 specifically refers to the 6th volume surface area segment, which is usually used to describe the surface area within a certain charge range in a molecule.

[0091] SPS: (Surface Polar Surface Area) SPS describes the polar surface area of ​​the molecular surface. It calculates the sum of the surface areas of polar atoms (such as oxygen, nitrogen, etc.) on the molecular surface.

[0092] 3D_OverLap: describes the degree of overlap between two molecules in three-dimensional space. It is usually used to compare the three-dimensional structural similarity of two molecules.

[0093] 3D_USRCAT: (Ultrafast Shape Recognition with Categorical descriptors) 3D_USRCAT is a fast 3D shape recognition method that combines shape and chemical features. It describes molecules by calculating their shape features (such as volume, surface, etc.) and chemical features (such as hydrogen bond donors / acceptors, hydrophobicity, etc.).

[0094] VSA_EState4: (Volume Surface Area-Estate 4) VSA_EState4 is a descriptor based on the EState metric and the volume surface area (VSA). The EState metric (Electrotopological State) is a parameter that describes the electronic topological state of atoms in a molecule, taking into account the charge distribution of the atoms and the surrounding atomic environment. VSA_EState4 specifically refers to the 4th volume surface area segment, which is used to describe the surface area within a specific EState range in a molecule.

[0095] SMR_VSA10: (Surface Molar Refractivity-Volume SurfaceArea 10) SMR_VSA10 is a descriptor based on molar refractivity (MR) and volume surface area (VSA). Molar refractivity is a physical property of molecules, which is related to the polarity and electron cloud density of the molecule. SMR_VSA10 specifically refers to the 10th volume surface area segment, which is used to describe the surface area within a specific molar refractive index range in a molecule.

[0096] FpDensityMorgan: is a descriptor based on Morgan fingerprints (also known as ECFC fingerprints, Extended Connectivity Fingerprints). Morgan fingerprints are bit vectors that describe the structure of a molecule and are generated by recursively extending the neighborhood of an atom. FpDensityMorgan describes the density of Morgan fingerprints, i.e. the number of fingerprint bits per unit length.

[0097] In the analysis of the core and substituents, if the Shapley values ​​of the core skeleton_a and skeleton_b are both positive and the Shapley value of skeleton_a is larger, it can be determined that skeleton_a is more helpful in improving the activity of the molecule. If the Shapley values ​​of cycloalkane and sulfonyl substituents are negative, they may have an adverse effect on the activity of the molecule. In the analysis of protein-ligand interactions, if the Shapley values ​​of the hydrogen bond formed by the ligand with the 94th cysteine ​​(94|CYS-HygB) and the salt bridge formed with the 101st aspartic acid (101|ASP-SaB) are both positive and large, it can be determined that they help to improve the activity of the molecule and the improvement effect is significant. If the Shapley value of the hydrophobic interaction between the ligand and the 23rd leucine (23|LEU-HypI) is small, it can be determined that its contribution to the activity of the molecule is negligible.

[0098] Through the above technical solution, after obtaining a series of ligand molecules, their target feature data are calculated. The target feature data includes molecular descriptors, parent core structures, substitution groups, and protein-ligand interactions, which greatly broadens the dimension of the calculation input data and provides more data basis for the calculation, thereby improving the accuracy of the calculation. After that, the target feature data is used together with the real molecular activity data to train the target machine learning model, and the target machine learning model is analyzed using the Shapley addition and interpretation algorithm to calculate the Shapley value of each target feature data, so as to use the Shapley value analysis to determine the impact of each target feature data on the molecular activity, and then realize the molecular structure-activity relationship analysis. The Shapley addition and interpretation algorithm analyzes and infers the decision-making process and prediction results of the machine learning model, which can clearly understand the influence of each feature on the model prediction, understand the model decision-making process, and on this basis improve the efficiency of molecular structure-activity relationship analysis.

[0099] Figure 4 is another flow chart of a structure-activity relationship analysis method based on multidimensional interpretability according to an exemplary embodiment, see Figure 4 , the method comprising:

[0100] S401, obtaining a series of ligand molecules.

[0101] S402, calculating target feature data of the series of ligand molecules, wherein the target feature data includes a molecular descriptor, a core structure, a substituent group relative to the core structure, and a protein-ligand interaction between the ligand molecule and the target protein.

[0102] S403, obtaining the real molecular activity data of each ligand molecule obtained based on experiments.

[0103] S404, using the target feature data of the ligand molecule as sample data, using the real molecular activity data as label data of the corresponding ligand molecule, and training a target machine learning model based on the sample data and the label data.

[0104] S405, analyzing the target machine learning model based on the Shapley addition and interpretation algorithm to obtain the Shapley value of each target feature data, wherein the Shapley value is used to characterize the correlation between the target feature data and the molecular activity.

[0105] S406, determining the influence of each target characteristic data on the molecular activity based on the Shapley value.

[0106] S407, obtaining the molecule to be optimized.

[0107] S408, performing structural optimization on the molecule to be optimized based on the influence of each target characteristic data determined by the Shapley value on the molecular activity, so as to increase the molecular activity of the molecule to be optimized.

[0108] Steps S401 to S406 may refer to steps S101 to S106 respectively.

[0109] In step S407, one or more molecules to be optimized may be obtained from an internal compound library, a commercial compound library, etc. Specifically, the SMILES of the molecules to be optimized may be obtained.

[0110] After step S407 is completed, step S408 may be entered. In step S408, the structure of the molecule to be optimized may be optimized according to the structure-activity relationship analysis result of the molecule in step S406, i.e., the influence of the target characteristic data on the molecule activity, so as to increase the molecular activity of the molecule to be optimized and improve its potential possibility of having a drug effect.

[0111] Optionally, in step S408, the structure of the molecule to be optimized is optimized based on the influence of each target characteristic data determined by the Shapley value on the molecule activity, including:

[0112] Obtaining candidate molecular descriptors, selecting a molecular descriptor with a positive Shapley value from the candidate molecular descriptors, and adding a molecular substructure described by the molecular descriptor with a positive Shapley value to the molecule to be optimized; and / or

[0113] Reducing the molecular substructures described by the molecular descriptors corresponding to negative Shapley values ​​in the molecule to be optimized; and / or

[0114] The molecular substructure in the molecule to be optimized is replaced by the molecular substructure described by the molecular descriptor corresponding to the larger Shapley value.

[0115] In step S408, the structure of the molecule to be optimized can be optimized according to the influence of the molecular descriptor determined by the Shapley value on the molecular activity. At least one optimization method can be adopted among the three methods of adding, deleting and replacing the molecular substructure corresponding to the molecular descriptor of the molecule to be optimized.

[0116] Specifically, the Shapley values ​​of the candidate molecular descriptors can be calculated with reference to the steps of step S401 to step S405, and the molecular descriptors with positive Shapley values ​​with good optimization effects can be selected. For example, the molecular descriptors with Shapley values ​​greater than or equal to the threshold can be selected, and the molecular substructures corresponding to the molecular descriptors are added to the corresponding parts of the molecules to be optimized. If the Shapley values ​​of the molecular descriptors corresponding to the molecular substructures in the molecules to be optimized are negative, it means that the molecular activity will be reduced. According to the actual situation, at least some of the molecular substructures with negative Shapley values ​​can be deleted to reduce the number of their structures. In addition, the molecular substructures of the molecules to be optimized can be replaced with molecular substructures with larger Shapley values ​​to increase the molecular activity.

[0117] Using the above example, see Figure 5 In the optimization process in box S501, molecular descriptors such as PEOE_VSA6, SPS, 3D_OverLap and 3D_USRCAT in area S501a are identified as positive factors, which help improve the activity of the molecule; while molecular descriptors such as VSA_EState4, SMR_VSA10 and FpDensityMorgan in area S501b show negative factors, which may reduce the activity of the molecule. Based on the indications of these molecular descriptors, when we optimize the molecules to be optimized, we should consider introducing substituents with strong electronegativity to improve the activity of the molecules; at the same time, increase the three-dimensional spatial overlap between the candidate molecules and the highly active reference molecules, and reduce the ring structure in the molecule or use a simpler ring system.

[0118] Optionally, in step S408, the structure of the molecule to be optimized is optimized based on the influence of each target characteristic data determined by the Shapley value on the molecule activity, including:

[0119] The mother core in the molecule to be optimized is replaced with a mother core corresponding to a larger Shapley value.

[0120] Using the previous example, see Figure 5 In the optimization process in box S502, if the Shapley value of the mother nucleus skeleton_a in region S502a is larger than that of the mother nucleus skeleton_b in region S502b, the mother nucleus skeleton_a is more helpful in improving the activity of the molecule; in the subsequent molecular optimization strategy, skeleton_a or its bioisosteric analogue should be preferentially selected as the mother nucleus. If the mother nucleus of the molecule to be optimized is skeleton_b, it can be replaced by skeleton_a.

[0121] Optionally, in step S408, the structure of the molecule to be optimized is optimized based on the influence of each target characteristic data determined by the Shapley value on the molecule activity, including:

[0122] Obtaining candidate substituent groups, selecting a substituent group with a positive Shapley value from the candidate substituent groups, and adding the substituent group with a positive Shapley value to the molecule to be optimized; and / or

[0123] Reducing the substituent groups with negative corresponding Shapley values ​​in the molecule to be optimized; and / or

[0124] The substituent groups in the molecule to be optimized are replaced with substituent groups corresponding to a larger Shapley value.

[0125] Specifically, the structure of the molecule to be optimized can be optimized according to the influence of the substituent groups on the molecular activity determined by the Shapley value. At least one of the three optimization methods of adding, deleting and replacing the substituent groups in the molecule to be optimized can be adopted.

[0126] The Shapley values ​​of the candidate substituent groups can be calculated with reference to the steps of step S401 to step S405, and the substituent groups with positive Shapley values ​​with good optimization effects can be selected therefrom. For example, a substituent group with a Shapley value greater than or equal to a threshold value can be selected, and the substituent group can be added to the corresponding part of the molecule to be optimized. If there are substituent groups with negative Shapley values ​​in the molecule to be optimized, it means that they will reduce the activity of the molecule, and at least some of the substituent groups with negative Shapley values ​​can be deleted according to the actual situation to reduce their number. In addition, the substituent groups of the molecule to be optimized can also be replaced with substituent groups with larger Shapley values ​​to increase the activity of the molecule.

[0127] Using the previous example, see Figure 5In the optimization process in block S502, cycloalkane or sulfonyl substituents may have an adverse effect on the activity of the molecule. Therefore, in the subsequent molecular optimization strategy, strong electronegative groups, such as halogenated hydrocarbons such as trifluoromethyl, are preferred in the selection of substituents, while (quasi) sulfonyl groups are avoided as much as possible. Figure 5 In the figure, the factors shown in area S502a are positive factors that are conducive to improving the activity of the molecule, and the factors shown in area S502b are negative factors that lead to a decrease in the activity of the molecule.

[0128] Optionally, in step S408, the structure of the molecule to be optimized is optimized based on the influence of each target characteristic data determined by the Shapley value on the molecule activity, including:

[0129] Obtaining candidate protein-ligand interactions, selecting a protein-ligand interaction with a positive Shapley value from the candidate protein-ligand interactions, and adjusting the structure of the molecule to be optimized so that the molecule to be optimized and the target protein generate the protein-ligand interaction with a positive Shapley value; and / or

[0130] Adjusting the structure of the molecule to be optimized to reduce the protein-ligand interaction between the molecule to be optimized and the target protein with a negative Shapley value; and / or

[0131] The structure of the molecule to be optimized is adjusted to replace the protein-ligand interaction between the molecule to be optimized and the target protein with a protein-ligand interaction with a larger Shapley value.

[0132] Specifically, the structure of the molecule to be optimized can be optimized according to the effect of the candidate protein-ligand interaction determined by the Shapley value on the molecular activity. At least one of the three optimization methods of adding, deleting and replacing the candidate protein-ligand interaction can be used for the molecule to be optimized.

[0133] The Shapley value of the candidate protein-ligand interaction can be calculated with reference to the steps of step S401 to step S405, and the protein-ligand interaction with a positive Shapley value having a good optimization effect can be selected therefrom. For example, the protein-ligand interaction with a Shapley value greater than or equal to a threshold value can be selected, and the molecular structure to be optimized is adjusted according to the effect, for example, a group can be added so that the molecule to be optimized can produce the protein-ligand interaction with the target protein. If there is a protein-ligand interaction with a negative Shapley value between the molecule to be optimized and the target protein, it means that it will reduce the molecular activity, and the molecular structure to be optimized can be adjusted according to actual conditions, for example, its corresponding group is deleted to reduce the protein-ligand interaction between the molecule to be optimized and the target protein. In addition, the group of the molecule to be optimized can also be replaced, so that the Shapley value of the protein-ligand interaction between the molecule to be optimized and the target protein after the replacement group is larger than that before the replacement.

[0134] Using the previous example, see Figure 5 In the optimization process in box S503, the factors shown in area S503a are positive factors that are conducive to improving the activity of the molecule, and the factors shown in area S503b are negative factors that lead to reduced activity of the molecule. For example, the hydrogen bond formed by the ligand and the 94th cysteine ​​(94|CYS-HygB) and the salt bridge formed with the 101st aspartic acid (101|ASP-SaB) both showed significant activity enhancement effects. In contrast, the hydrophobic interaction between the ligand and the 23rd leucine (23|LEU-HypI) contributed little to the activity enhancement. Based on these findings, when making molecular improvements, we should consider retaining or enhancing the interaction between (94|CYS-HygB) and (101|ASP-SaB), while reducing or replacing the hydrophobic interaction of (23|LEU-HypI) to explore other interaction modes that are more conducive to activity enhancement.

[0135] In addition, see Figure 6 For example, for the molecule to be optimized whose parent core is pyrimidine amide and whose molecular structure is as follows:

[0136]

[0137] See also Figure 6 In region S601, in a possible embodiment, the R1 group of the molecule to be optimized is the substituent group in region 601a. ​​Through structure-activity analysis, it is found that the Shapley value of the substituent group in region 601a is smaller than the Shapley value of the substituent group in region 601b. Further analysis shows that the substituent group in region 601b can form hydrogen bonds with Asp-100 to fix it, which has a greater improvement in the activity of the molecule. Therefore, when optimizing the molecule to be optimized whose parent core is pyrimidine amide, the substituent group in region 601a can be replaced by a substituent group in region 601b.

[0138] In a possible embodiment, the R1 group is -H, which has a Shapley value smaller than that of the hydroxyl group -OH. Analysis shows that the outer hydrophilic group such as -OH can form a water bridge with the conservative water molecule, so -H can be replaced by -OH.

[0139] See also Figure 6 In region S602, in a possible embodiment, the R2 substituent group is -H in region 602b, and its Shapley value is smaller than that of the substituent group in region 602b. Through analysis, it is found that the substituent group in region 602b can form a hydrophobic effect, which has a greater improvement on the molecular activity. Therefore, -H can be replaced by a hydrophobic group in region 602a.

[0140] See also Figure 6In region S603, in one possible embodiment, the R2 substituent group is an amidopyrimidine shown in region 603a, and its Shapley value is smaller than that of the substituent group in region 603b. Through analysis, it is found that the substituent group in region 603b has a three-dimensional spatial effect and has a greater improvement in molecular activity. Therefore, the amidopyrimidine shown in region 603a can be replaced by a substituent group in region 603b.

[0141] Figure 7 is a block diagram of a structure-activity relationship analysis device based on multi-dimensional interpretability according to an exemplary embodiment, see Figure 7 , the device comprises:

[0142] A ligand molecule acquisition module is used to acquire a series of ligand molecules;

[0143] A calculation module, for calculating target characteristic data of the series of ligand molecules, wherein the target characteristic data includes a molecular descriptor, a parent core structure, a substitution group relative to the parent core structure, and a protein-ligand interaction between the ligand molecules and the target protein;

[0144] A data acquisition module, used to acquire the real molecular activity data of each ligand molecule obtained based on experiments;

[0145] A training module, used to use the target feature data of the ligand molecule as sample data, use the real molecular activity data as label data of the corresponding ligand molecule, and train a target machine learning model based on the sample data and the label data;

[0146] A model interpretation module, used to analyze the target machine learning model based on the Shapley addition and interpretation algorithm to obtain the Shapley value of each target feature data, wherein the Shapley value is used to characterize the correlation between the target feature data and the molecular activity;

[0147] The structure-activity relationship analysis module is used to determine the influence of each target characteristic data on the molecular activity based on the Shapley value.

[0148] Optionally, the structure-activity relationship analysis module is also used for:

[0149] When the Shapley value is positive, it is determined that the larger the Shapley value is, the greater the positive effect of the target characteristic data corresponding to the Shapley value on the molecular activity is;

[0150] When the Shapley value is negative, it is determined that the larger the absolute value of the Shapley value is, the greater the negative impact of the target characteristic data corresponding to the Shapley value on the molecular activity is.

[0151] When the Shapley value is zero, it is determined that the target characteristic data corresponding to the Shapley value has no effect on the molecular activity.

[0152] Optionally, the device further comprises:

[0153] A molecule acquisition module to be optimized, used to acquire the molecule to be optimized;

[0154] The molecular optimization module is used to perform structural optimization on the molecule to be optimized based on the influence of each target characteristic data determined by the Shapley value on the molecular activity, so as to increase the molecular activity of the molecule to be optimized.

[0155] Optionally, the molecular optimization module is further used for:

[0156] Obtaining candidate molecular descriptors, selecting a molecular descriptor with a positive Shapley value from the candidate molecular descriptors, and adding a molecular substructure described by the molecular descriptor with a positive Shapley value to the molecule to be optimized; and / or

[0157] Reducing the molecular substructures described by the molecular descriptors corresponding to negative Shapley values ​​in the molecule to be optimized; and / or

[0158] The molecular substructure in the molecule to be optimized is replaced by the molecular substructure described by the molecular descriptor corresponding to the larger Shapley value.

[0159] Optionally, the molecular optimization module is further used for:

[0160] The mother core in the molecule to be optimized is replaced with a mother core corresponding to a larger Shapley value.

[0161] Optionally, the molecular optimization module is further used for:

[0162] Obtaining candidate substituent groups, selecting a substituent group with a positive Shapley value from the candidate substituent groups, and adding the substituent group with a positive Shapley value to the molecule to be optimized; and / or

[0163] Reducing the substituent groups with negative corresponding Shapley values ​​in the molecule to be optimized; and / or

[0164] The substituent groups in the molecule to be optimized are replaced with substituent groups corresponding to a larger Shapley value.

[0165] Optionally, the molecular optimization module is further used for:

[0166] Obtaining candidate protein-ligand interactions, selecting a protein-ligand interaction with a positive Shapley value from the candidate protein-ligand interactions, and adjusting the structure of the molecule to be optimized so that the molecule to be optimized and the target protein generate the protein-ligand interaction with a positive Shapley value; and / or

[0167] Adjusting the structure of the molecule to be optimized to reduce the protein-ligand interaction between the molecule to be optimized and the target protein with a negative Shapley value; and / or

[0168] The structure of the molecule to be optimized is adjusted to replace the protein-ligand interaction between the molecule to be optimized and the target protein with a protein-ligand interaction with a larger Shapley value.

[0169] Optionally, the calculation module is further used for:

[0170] Calculating a molecular descriptor for each of the ligand molecules, wherein the molecular descriptor includes molecular low-dimensional structure information and molecular 3D structure information;

[0171] Calculating the core structure of the series of ligand molecules and the substituent groups of each ligand molecule relative to the core structure;

[0172] A target protein of each of the ligand molecules is selected and a 3D conformation of the target protein is obtained, the 3D conformation of the ligand molecule is generated, the 3D conformation of the ligand molecule is molecularly docked with the 3D conformation of the target protein, and after the molecular docking is completed, the protein-ligand interaction between the ligand molecule and the target protein is calculated, wherein the protein-ligand interaction includes amino acid residues and protein-ligand interaction types.

[0173] Optionally, the molecular low-dimensional structure information includes molecular 0D structure information, molecular 1D structure information, molecular 2D structure information and molecular property descriptors, and / or the molecular 3D structure information includes molecular shape, three-dimensional atomic space coordinates, shape complexity, and best fit plane.

[0174] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0175] Figure 8 FIG. 1 is a block diagram of an electronic device according to an exemplary embodiment. Figure 8 As shown, the electronic device 700 may include: a processor 701 and a memory 702. The electronic device 700 may also include one or more of a multimedia component 703, an input / output (I / O) interface 704, and a communication component 705.

[0176] The processor 701 is used to control the overall operation of the electronic device 700 to complete all or part of the steps in the structure-activity relationship analysis method based on multi-dimensional interpretability. The memory 702 is used to store various types of data to support the operation of the electronic device 700. For example, these data may include instructions for any application or method used to operate on the electronic device 700, and application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 702 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, referred to as EPROM), programmable read-only memory (Programmable Read-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, magnetic disk or optical disk. The multimedia component 703 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 702 or sent through the communication component 705. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 704 provides an interface between the processor 701 and other interface modules, and the other interface modules may be keyboards, mice, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 705 is used for wired or wireless communication between the electronic device 700 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IOT, eMTC, or other 5G, etc., or a combination of one or more of them, is not limited here. Therefore, the corresponding communication component 705 may include: Wi-Fi module, Bluetooth module, NFC module, etc.

[0177] In an exemplary embodiment, the electronic device 700 can be implemented by one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned structure-activity relationship analysis method based on multi-dimensional interpretability.

[0178] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the structure-activity relationship analysis method based on multidimensional interpretability are implemented. For example, the computer-readable storage medium can be the memory 702 including the program instructions, and the program instructions can be executed by the processor 701 of the electronic device 700 to complete the structure-activity relationship analysis method based on multidimensional interpretability.

[0179] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings; however, the present disclosure is not limited to the specific details in the above embodiments. Within the technical concept of the present disclosure, a variety of simple modifications can be made to the technical solution of the present disclosure, and these simple modifications all fall within the protection scope of the present disclosure.

[0180] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.

[0181] In addition, various embodiments of the present disclosure may be arbitrarily combined, and as long as they do not violate the concept of the present disclosure, they should also be regarded as the contents disclosed by the present disclosure.

Claims

1. A structure-activity relationship analysis method based on multidimensional interpretability, characterized in that: The method comprises: Obtain a series of ligand molecules; Calculating target feature data of the series of ligand molecules, wherein the target feature data includes a molecular descriptor, a core structure, a substituent group relative to the core structure, and a protein-ligand interaction between the ligand molecules and a target protein; Obtaining experimentally obtained real molecular activity data of each of the ligand molecules; Using the target feature data of the ligand molecule as sample data, using the real molecular activity data as label data of the corresponding ligand molecule, and training a target machine learning model based on the sample data and the label data; Analyzing the target machine learning model based on the Shapley addition and interpretation algorithm to obtain the Shapley value of each target feature data, wherein the Shapley value is used to characterize the correlation between the effect of the target feature data on the molecular activity; The influence of each target characteristic data on the molecule activity is determined based on the Shapley value.

2. The method according to claim 1, wherein determining the influence of each target characteristic data on the molecular activity based on the Shapley value comprises: When the Shapley value is positive, it is determined that the larger the Shapley value is, the greater the positive effect of the target characteristic data corresponding to the Shapley value on the molecular activity is; When the Shapley value is negative, it is determined that the larger the absolute value of the Shapley value is, the greater the negative impact of the target characteristic data corresponding to the Shapley value on the molecular activity is. When the Shapley value is zero, it is determined that the target characteristic data corresponding to the Shapley value has no effect on the molecular activity.

3. The method according to claim 2, characterized in that The method further comprises: Obtain the molecule to be optimized; Based on the influence of each target characteristic data determined by the Shapley value on the molecular activity, the structure of the molecule to be optimized is optimized to increase the molecular activity of the molecule to be optimized.

4. The method according to claim 3, characterized in that The structural optimization of the molecule to be optimized based on the influence of each target characteristic data determined by the Shapley value on the molecular activity comprises: Obtaining candidate molecular descriptors, selecting a molecular descriptor with a positive Shapley value from the candidate molecular descriptors, and adding a molecular substructure described by the molecular descriptor with a positive Shapley value to the molecule to be optimized; and / or Reducing the molecular substructures described by the molecular descriptors corresponding to negative Shapley values ​​in the molecule to be optimized; and / or The molecular substructure in the molecule to be optimized is replaced by the molecular substructure described by the molecular descriptor corresponding to the larger Shapley value.

5. The method according to claim 3, characterized in that: The structural optimization of the molecule to be optimized based on the influence of each target characteristic data determined by the Shapley value on the molecular activity comprises: The mother core in the molecule to be optimized is replaced with a mother core corresponding to a larger Shapley value.

6. The method according to claim 3, characterized in that The structural optimization of the molecule to be optimized based on the influence of each target characteristic data determined by the Shapley value on the molecular activity comprises: Obtaining candidate substituent groups, selecting a substituent group with a positive Shapley value from the candidate substituent groups, and adding the substituent group with a positive Shapley value to the molecule to be optimized; and / or Reducing the substituent groups with negative corresponding Shapley values ​​in the molecule to be optimized; and / or The substituent groups in the molecule to be optimized are replaced with substituent groups corresponding to a larger Shapley value.

7. The method according to claim 3, characterized in that The structural optimization of the molecule to be optimized based on the influence of each target characteristic data determined by the Shapley value on the molecular activity comprises: Obtaining candidate protein-ligand interactions, selecting a protein-ligand interaction with a positive Shapley value from the candidate protein-ligand interactions, and adjusting the structure of the molecule to be optimized so that the molecule to be optimized and the target protein generate the protein-ligand interaction with a positive Shapley value; and / or Adjusting the structure of the molecule to be optimized to reduce the protein-ligand interaction between the molecule to be optimized and the target protein with a negative Shapley value; and / or The structure of the molecule to be optimized is adjusted to replace the protein-ligand interaction between the molecule to be optimized and the target protein with a protein-ligand interaction with a larger Shapley value.

8. The method according to claim 1, characterized in that The step of calculating target characteristic data of the series of ligand molecules comprises: Calculating a molecular descriptor for each of the ligand molecules, wherein the molecular descriptor includes molecular low-dimensional structure information and molecular 3D structure information; Calculating the core structure of the series of ligand molecules and the substituent groups of each ligand molecule relative to the core structure; A target protein of each of the ligand molecules is selected and a 3D conformation of the target protein is obtained, the 3D conformation of the ligand molecule is generated, the 3D conformation of the ligand molecule is molecularly docked with the 3D conformation of the target protein, and after the molecular docking is completed, the protein-ligand interaction between the ligand molecule and the target protein is calculated, wherein the protein-ligand interaction includes amino acid residues and protein-ligand interaction types.

9. The method according to claim 8, characterized in that The molecular low-dimensional structure information includes molecular 0D structure information, molecular 1D structure information, molecular 2D structure information and molecular property descriptors, and / or the molecular 3D structure information includes molecular shape, three-dimensional atomic space coordinates, shape complexity, and best fit plane.

10. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the method according to any one of claims 1 to 9.

Images