Multi-target drug screening and optimizing method based on machine learning

Abstract

The invention belongs to the technical field of drug screening, and provides a multi-target drug screening and optimizing method based on machine learning, which comprises the following steps: S1, carrying out feature screening on compound molecular descriptors by utilizing a machine learning algorithm, and determining the molecular descriptor which has the most significant influence on biological activity; s2, constructing a quantitative structure-activity relation model, and predicting the biological activity of the compound; s3, based on a multi-objective optimization algorithm, combining absorption, distribution, metabolism, excretion and toxicity properties of the compound; according to the method, feature screening is carried out on compound molecule descriptors by utilizing a machine learning algorithm, a QSAR model is constructed to predict the biological activity, ADMET property classification prediction is combined, the biological activity, ADMET properties and pharmacokinetic characteristics of drugs are comprehensively considered, and the accuracy and accuracy of drug activity prediction are improved. The method can efficiently and accurately screen and optimize drug candidate compounds with potential drug effects and good pharmacokinetic characteristics, so that the success rate of drug research and development is increased, and the research and development cost and time are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of drug screening, and specifically is a multi-objective drug screening and optimization method based on machine learning. Background Art

[0002] Drug development is a complex and time-consuming process. Traditional methods usually rely on a large number of experimental screening and verification, which is not only costly but also inefficient. With the rapid development of machine learning technology, its application in the field of drug screening and optimization has gradually emerged.

[0003] However, existing machine learning-based drug screening methods often only focus on a single target, such as biological activity or pharmacokinetic properties, while ignoring the need for multi-target optimization in drug development; at the same time, the prediction of the ADMET properties (absorption, distribution, metabolism, excretion, toxicity) of compounds lacks accuracy and efficiency, which limits the success rate and efficiency of drug development.

[0004] To this end, those skilled in the art have proposed a multi-objective drug screening and optimization method based on machine learning to solve the problems raised by the background technology. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a multi-objective drug screening and optimization method based on machine learning, so as to solve the problem that the drug screening method based on machine learning in the prior art often only focuses on a single target, such as biological activity or pharmacokinetic properties, but ignores the multi-objective optimization requirements in drug research and development; at the same time, the prediction of the ADMET properties (absorption, distribution, metabolism, excretion, toxicity) of the compound also lacks accuracy and efficiency, which limits the success rate and efficiency of drug research and development.

[0006] A multi-objective drug screening and optimization method based on machine learning, comprising: S1. Use machine learning algorithms to perform feature screening on compound molecular descriptors and determine the molecular descriptors that have the most significant impact on biological activity; S2. Construct quantitative structure-activity relationship (QSAR) models to predict the biological activity of compounds; S3. Based on the multi-objective optimization algorithm, combined with the absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of the compounds, classify and predict the drug candidate compounds; S4. Based on the QSAR model and ADMET property classification prediction results, screen and optimize drug candidate compounds with potential efficacy and good pharmacokinetic properties.

[0007] Preferably, in step S1, the machine learning algorithm includes one of grey relational analysis and BP neural network optimized by genetic algorithm.

[0008] Preferably, in step S2, the QSAR model is constructed by an SVR model to predict the biological activity of the compound. When constructing the model, the objective function and constraint formula of the SVR model are introduced. The objective function of the SVR model usually includes a regularization term and an error term, which is used to find a balance between fitting accuracy and model complexity.

[0009] Preferably, the ADMET property classification prediction uses a logistic regression model to predict CYP3A4 and micronucleus (MN) characteristics, a BP neural network model to predict human oral bioavailability (HOB) characteristics, and an SVR model to predict Caco-2 and hERG characteristics.

[0010] Preferably, the multi-objective optimization algorithm comprehensively considers the biological activity, ADMET properties and pharmacokinetic characteristics of the drug, and screens out the optimal drug candidate compounds through weight allocation and iterative optimization. The multi-objective optimization algorithm converts the multi-objective optimization problem into a single-objective optimization problem by using a weighted sum method.

[0011] Preferably, in step S4, the screening and optimization of drug candidate compounds with potential efficacy and good pharmacokinetic properties adopts a simulated annealing algorithm to optimize the screening process of drug candidate compounds.

[0012] Preferably, the method further includes verifying the rationality of the screened molecular descriptors using information entropy theory, and evaluating the stability and accuracy of the model through sensitivity analysis.

[0013] Preferably, a multi-objective optimization drug screening and optimization system based on machine learning, using the above-mentioned multi-objective drug screening and optimization method based on machine learning, comprises: The machine learning algorithm module performs feature screening on the molecular descriptors of compounds to determine the molecular descriptors that have the most significant impact on biological activity; Quantitative structure-activity relationship model building module for predicting the biological activity of compounds; The multi-objective optimization algorithm module classifies and predicts drug candidate compounds based on the absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of the compounds; The drug candidate compound screening and optimization module screens and optimizes drug candidate compounds with potential efficacy and good pharmacokinetic properties based on the QSAR model and ADMET property classification prediction results.

[0014] A processor is configured to execute the above-mentioned multi-objective drug screening and optimization method based on machine learning.

[0015] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-mentioned multi-objective drug screening and optimization method based on machine learning.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention uses a machine learning algorithm to perform feature screening on the molecular descriptors of the compound and determine the molecular descriptors that have the most significant impact on biological activity, thereby narrowing the screening range and improving the screening efficiency.

[0017] 2. The present invention improves the accuracy and reliability of drug screening and optimization by using the QSAR model to predict the biological activity of compounds and combining it with the ADMET property classification prediction and comprehensive consideration of multiple objectives for optimization.

[0018] 3. The present invention uses a multi-objective optimization algorithm to comprehensively consider the biological activity, ADMET properties and pharmacokinetic characteristics of the drug, screen out the optimal drug candidate compounds, and improve the success rate of drug research and development.

[0019] 4. This method reduces the number of experimental screening and verification, and reduces the cost and time of drug development.

[0020] 5. The present invention verifies the rationality of the screened molecular descriptors through information entropy theory, and uses sensitivity analysis to evaluate the stability and accuracy of the model, further improving the reliability and practicality of the method. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flow chart of the multi-objective drug screening and optimization method based on machine learning of the present invention; Figure 2 This is a framework diagram of the multi-objective optimization drug screening and optimization system based on machine learning of the present invention. DETAILED DESCRIPTION

[0022] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0023] Embodiment: The present invention provides a multi-objective drug screening and optimization method based on machine learning, such as Figure 1 As shown, including: S1. Use machine learning algorithms to perform feature screening on compound molecular descriptors and determine the molecular descriptors that have the most significant impact on biological activity; S2. Construct quantitative structure-activity relationship (QSAR) models to predict the biological activity of compounds; S3. Based on the multi-objective optimization algorithm, combined with the absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of the compounds, classify and predict the drug candidate compounds; S4. Based on the QSAR model and ADMET property classification prediction results, screen and optimize drug candidate compounds with potential efficacy and good pharmacokinetic properties.

[0024] From the above, we can see that by using machine learning algorithms to perform feature screening on compound molecular descriptors, constructing QSAR models to predict biological activity, and combining ADMET property classification prediction, and comprehensively considering the biological activity, ADMET properties and pharmacokinetic properties of drugs, it is possible to efficiently and accurately screen and optimize drug candidate compounds with potential efficacy and good pharmacokinetic properties, thereby improving the success rate of drug research and development and reducing research and development costs and time.

[0025] Further, in step S1, the machine learning algorithm includes one of grey relational analysis and BP neural network optimized by genetic algorithm; The grey relational analysis includes the grey relational degree formula and the relational coefficient formula: The grey relational formula is as follows: ; in, It represents the grey relational degree between the i-th reference sequence and the j-th comparison sequence. is the correlation coefficient at the kth moment (or data point), and n is the total number of data points; The correlation coefficient formula is as follows: ; in, and are the values ​​of the i-th reference sequence and the j-th comparison sequence at the k-th moment, ρ is the resolution coefficient, which is taken as 0.5; The BP neural network formula of the BP neural network optimized by the genetic algorithm includes a forward propagation formula and a back propagation formula: The forward propagation formula is as follows: ; ; Where W represents the weight matrix, and They are The weighted inputs and activation outputs of the layer, and They are The weights and biases of the layers, is the activation function; The back propagation formula is as follows: ; ; ; ; in, represents the displacement change rate, represents the coordinates of the Lth layer, It is The error term of the layer, W represents the weights connecting neurons in a neural network, is the loss function, Represents element-wise multiplication.

[0026] From the above, it can be seen that by using grey correlation analysis and BP neural network optimized by genetic algorithm as machine learning algorithms, it is possible to accurately perform feature screening on the molecular descriptors of compounds and determine the molecular descriptors that have the most significant impact on biological activity; grey correlation analysis can quantify the degree of correlation between different molecular descriptors and biological activity by calculating the grey correlation degree and correlation coefficient, thereby improving the accuracy of screening; and the BP neural network optimized by genetic algorithm utilizes the global search capability of genetic algorithm and the nonlinear mapping capability of BP neural network to further optimize the feature screening process, improve screening efficiency and accuracy, and lay a solid foundation for subsequent drug screening and optimization.

[0027] Further, in step S2, the QSAR model is constructed by the SVR model to predict the biological activity of the compound. When constructing the model, the objective function and constraint condition formula of the SVR model are introduced. The objective function of the SVR model usually includes a regularization term and an error term, which is used to find a balance between fitting accuracy and model complexity. The formula is as follows: ; ; ; ; Among them, w is the weight vector, b is the bias term, represents the target value, The input vector Function mapped to high-dimensional space, C is the regularization parameter, is the allowable error range, and is a slack variable.

[0028] From the above, we can see that constructing a QSAR model through the SVR model to predict the biological activity of compounds can effectively balance the fitting accuracy and complexity of the model; by introducing the objective function and constraint formula of the SVR model, the model can find an optimal solution during the training process, so that while minimizing the error, the complexity of the model is controlled through the regularization term to prevent overfitting; this method can not only improve the prediction accuracy of the QSAR model, but also enhance the generalization ability of the model, providing reliable biological activity prediction results for subsequent drug screening and optimization.

[0029] Furthermore, the ADMET property classification prediction uses a logistic regression model to predict CYP3A4 and micronucleus (MN) characteristics, a BP neural network model to predict human oral bioavailability (HOB) characteristics, and an SVR model to predict Caco-2 and hERG characteristics; the prediction formula of the logistic regression model is as follows: ; in, is the Sigmoid function, w is the weight vector, and b is the bias term; The BP neural network model prediction involves weighted summation and activation function operations between multiple layers of neurons, and its formula is as follows: ; ; in, It is The weighted sum of the inputs to the jth neuron in the layer, It is from -1 layer from the i-th neuron to the The weight of the jth neuron in the layer, It is The bias term of the jth neuron in the layer, It is The output (activation value) of the jth neuron in the layer, f is the activation function.

[0030] From the above, we can see that by adopting multiple machine learning models to construct ADMET property classification prediction, the key pharmacokinetic and toxicological properties of the compound can be predicted comprehensively and accurately; the logistic regression model is suitable for dealing with binary classification problems, such as the prediction of CYP3A4 and micronucleus (MN) properties, and its simplicity and efficiency make the prediction results stable and interpretable; the BP neural network model can effectively capture nonlinear relationships through complex interactions between multiple layers of neurons, and is suitable for predicting complex human oral bioavailability (HOB) properties; and the SVR model performs well in processing continuous variable predictions, such as the prediction of Caco-2 and hERG properties, and its strong generalization ability ensures the accuracy of the prediction; this method of comprehensive use of multiple machine learning models improves the diversity and accuracy of ADMET property predictions, and provides a more comprehensive and reliable basis for drug screening and optimization.

[0031] Furthermore, the multi-objective optimization algorithm comprehensively considers the biological activity, ADMET properties and pharmacokinetic characteristics of the drug, and screens out the optimal drug candidate compound through weight allocation and iterative optimization. The multi-objective optimization algorithm converts the multi-objective optimization problem into a single-objective optimization problem by using a weighted sum method. The formula of the multi-objective optimization algorithm is as follows: ; Where M is the number of objective functions, is the weight of the ith objective function, is the i-th objective function in the candidate compound The value on .

[0032] From the above, it can be seen that by comprehensively considering the biological activity, ADMET properties and pharmacokinetic characteristics of the drug and converting the multi-objective optimization problem into a single-objective optimization problem, the multi-dimensional decision-making problem in drug screening is effectively solved; this method can balance the importance of each optimization goal according to actual needs through weight distribution, ensuring that the screened drug candidate compounds perform well in multiple key performance indicators; at the same time, through the iterative optimization process, the algorithm can continuously explore better solutions and ultimately screen out drug candidate compounds with the highest comprehensive performance; this comprehensive optimization strategy significantly improves the efficiency and accuracy of drug screening, providing a strong guarantee for the success of drug research and development.

[0033] Furthermore, in step S3, a particle swarm optimization (PSO) algorithm is introduced to assist the multi-objective optimization algorithm in parameter optimization and candidate drug screening. The formula of the particle swarm optimization (PSO) algorithm is as follows: Position update formula: ; in, represents the position of the i-th particle at the t-th iteration, represents the velocity of the i-th particle at the t+1th iteration; Speed ​​update formula: ; Among them, w is the inertia weight, which is used to control the influence of the current velocity of the particle on the next velocity; and are learning factors, which represent the particle's movement to its historical optimal position ( ) and the global optimal position ( gbest ) The weights of learning; and is a random number between [0,1], used to increase the randomness of the search; is the optimal historical position of the ith particle, gbest is the current global optimal position.

[0034] From the above, we can see that by introducing the particle swarm optimization (PSO) algorithm in the multi-objective optimization stage, we can further accelerate the convergence speed of the optimization process and improve the efficiency and effect of multi-objective optimization. The PSO algorithm simulates the intelligent behavior of the group and efficiently searches for the global optimal solution in the solution space, providing strong support for drug screening and optimization.

[0035] Furthermore, in step S4, the screening and optimization of drug candidate compounds with potential efficacy and good pharmacokinetic properties uses a simulated annealing algorithm to optimize the screening process of drug candidate compounds; the temperature update formula of the simulated annealing algorithm is as follows: ; in, α is the temperature drop coefficient, is the temperature at the kth step.

[0036] From the above, it can be seen that by using the simulated annealing algorithm to optimize the screening process of drug candidate compounds, this method can significantly improve the efficiency and effect of screening; the simulated annealing algorithm simulates the physical annealing process, randomly explores and accepts poor solutions with a certain probability in the search space, thereby avoiding falling into the local optimal solution and ensuring the search for the global optimal solution; the "cooling" rate of the search process is controlled by the temperature update formula. As the temperature gradually decreases, the algorithm gradually reduces the probability of accepting poor solutions and guides the search process to converge to the global optimal solution; this method can not only improve the diversity of screening, but also ensure that the drug candidate compounds finally screened are optimal in both potential efficacy and pharmacokinetic properties, providing strong support for the success of drug research and development.

[0037] Furthermore, it also includes using information entropy theory to verify the rationality of the screened molecular descriptors, and evaluating the stability and accuracy of the model through sensitivity analysis. The calculation formula of the information entropy is as follows: ; Where X is the value set of the molecular descriptor, p ( x ) is the probability of taking the value x; The gradient formula for the sensitivity analysis is as follows: ; Where f is the model output, is an input parameter.

[0038] From the above, we can see that in the process of drug screening and optimization, using information entropy theory to verify the rationality of the screened molecular descriptors and evaluating the stability and accuracy of the model through sensitivity analysis can significantly improve the scientificity and reliability of the method; the calculation of information entropy can quantify the amount of information carried by the molecular descriptors, thereby verifying the rationality of their screening and ensuring that the selected descriptors have sufficient information contribution to the prediction of drug properties; and sensitivity analysis reveals the stability and robustness of the model by evaluating the response of the model output to small changes in the input parameters, ensuring that the accuracy of the model in the prediction process is not affected by parameter fluctuations; the combination of these two methods not only enhances the scientificity of the drug screening process, but also improves the reliability of the screening results, laying a solid foundation for the success of drug research and development.

[0039] Working principle: Molecular descriptors with significant effects on biological activity are screened through machine learning algorithms, and QSAR models are constructed to predict the biological activity of compounds. Combined with ADMET property classification prediction, a multi-objective optimization algorithm is used to comprehensively consider biological activity, ADMET properties and pharmacokinetic characteristics. The screening process is optimized through simulated annealing algorithm, and information entropy theory and sensitivity analysis are used to verify the rationality of the screening results and the stability of the model, so as to efficiently and accurately screen and optimize drug candidate compounds.

[0040] Furthermore, the effects of a multi-objective drug screening and optimization method based on machine learning in this embodiment and an existing drug screening method based on machine learning (comparative example) were compared to obtain the following table:

[0041] As can be seen from the above table, the multi-objective drug screening and optimization method based on machine learning in this embodiment shows significant advantages over the existing drug screening method based on machine learning in terms of the number of target optimization, screening efficiency, prediction accuracy, model stability, cost and time, optimization of screening results, and rationality of screening results.

[0042] A multi-objective optimization drug screening and optimization system based on machine learning, such as Figure 2 As shown, the above-mentioned multi-objective drug screening and optimization method based on machine learning includes: The machine learning algorithm module performs feature screening on the molecular descriptors of compounds to determine the molecular descriptors that have the most significant impact on biological activity; Quantitative structure-activity relationship model building module for predicting the biological activity of compounds; The multi-objective optimization algorithm module classifies and predicts drug candidate compounds based on the absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of the compounds; The drug candidate compound screening and optimization module screens and optimizes drug candidate compounds with potential efficacy and good pharmacokinetic properties based on the QSAR model and ADMET property classification prediction results.

[0043] The embodiment of the present application provides an electronic device, which is applicable to the above-mentioned multi-objective drug screening and optimization method based on machine learning, including: Memory, used to protect computer programs and data; Processor, used to run system programs.

[0044] An embodiment of the present application provides a computer storage medium, which is suitable for the above-mentioned multi-objective drug screening and optimization method based on machine learning, and performs hierarchical confidentiality management on the above-mentioned system and data in accordance with confidentiality management requirements.

[0045] Those skilled in the art will appreciate that the embodiments of the present application may be provided as a system or a computer program product. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0046] The present application is described with reference to the flowcharts and / or block diagrams of the devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0047] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0048] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0049] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0050] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0051] Computer readable media include permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0052] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, commodity or device including the elements.

[0053] The embodiments of the present invention are provided for the purpose of illustration and description. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present invention. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A multi-objective drug screening and optimization method based on machine learning, characterized in that: include: S1. Use machine learning algorithms to perform feature screening on compound molecular descriptors and determine the molecular descriptors that have the most significant impact on biological activity; S2. Constructing a QSAR model to predict the biological activity of the compound; the QSAR model is constructed by an SVR model to predict the biological activity of the compound. When constructing the model, the objective function and constraint condition formula of the SVR model are introduced. The objective function of the SVR model usually includes a regularization term and an error term, which is used to find a balance between fitting accuracy and model complexity; S3. Based on the multi-objective optimization algorithm and combined with ADMET properties, classify and predict drug candidate compounds; S4. Based on the QSAR model and ADMET property classification prediction results, screen and optimize drug candidate compounds with potential efficacy and good pharmacokinetic properties.

2. A multi-objective drug screening and optimization method based on machine learning as claimed in claim 1, characterized in that: In step S1, the machine learning algorithm includes one of grey relational analysis and BP neural network optimized by genetic algorithm.

3. A multi-objective drug screening and optimization method based on machine learning as claimed in claim 1, characterized in that: The ADMET property classification prediction uses a logistic regression model to predict CYP3A4 and micronucleus characteristics, a BP neural network model to predict human oral bioavailability characteristics, and an SVR model to predict Caco-2 and hERG characteristics.

4. The multi-objective drug screening and optimization method based on machine learning as claimed in claim 1, characterized in that: The multi-objective optimization algorithm comprehensively considers the biological activity, ADMET properties and pharmacokinetic characteristics of the drug, and screens out the optimal drug candidate compound through weight allocation and iterative optimization. The multi-objective optimization algorithm converts the multi-objective optimization problem into a single-objective optimization problem by using a weighted sum method.

5. The multi-objective drug screening and optimization method based on machine learning as claimed in claim 1, characterized in that: In step S4, the screening and optimization of drug candidate compounds with potential efficacy and good pharmacokinetic properties adopts a simulated annealing algorithm to optimize the screening process of drug candidate compounds.

6. A multi-objective drug screening and optimization method based on machine learning as claimed in claim 5, characterized in that: It also includes the use of information entropy theory to verify the rationality of the screened molecular descriptors, and the evaluation of the stability and accuracy of the model through sensitivity analysis.

7. A multi-objective optimization drug screening and optimization system based on machine learning, characterized in that: A multi-objective drug screening and optimization method based on machine learning according to any one of claims 1 to 6, comprising: The machine learning algorithm module performs feature screening on the molecular descriptors of compounds to determine the molecular descriptors that have the most significant impact on biological activity; Quantitative structure-activity relationship model building module for predicting the biological activity of compounds; The multi-objective optimization algorithm module classifies and predicts drug candidate compounds based on their absorption, distribution, metabolism, excretion, and toxicity properties; The drug candidate compound screening and optimization module screens and optimizes drug candidate compounds with potential efficacy and good pharmacokinetic properties based on the QSAR model and ADMET property classification prediction results.

8. A processor, characterized in that: The method is configured to execute the multi-objective drug screening and optimization method based on machine learning according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the multi-objective drug screening and optimization method based on machine learning described in any one of claims 1 to 6 is implemented.

Images