Intelligent medicine screening method and device and medium

By dynamically monitoring drug demand and building an optimized drug screening model, using knowledge distillation optimization channels to improve drug screening efficiency and accuracy, solving the problems of low efficiency and poor accuracy of drug screening in the existing technology, and achieving faster drug screening results .

CN118983111AInactive Publication Date: 2024-11-19QIDONG FANGJING BIOTECHNOLOGY CO LTD
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411462773.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-19
Publication Date
2024-11-19
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art has low screening efficiency and poor accuracy during drug screening, making it difficult to meet the needs of rapid treatment and new drug development.

Method used

By obtaining drug feature sets based on drug public database, dynamically monitor drug demand, constructing and optimizing drug screening models, using knowledge distillation optimization channels to improve model performance, and generating a second drug screening model for intelligent screening.

Benefits of technology

It improves the efficiency and accuracy of drug screening, and can find effective drugs more quickly to meet the needs of rapid treatment and new drug development.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118983111A_ABST
    Figure CN118983111A_ABST
Patent Text Reader

Abstract

The invention discloses an intelligent drug screening method and device and a medium, and relates to the technical field of drug screening, and the method comprises the steps: collecting and obtaining a drug feature set based on a drug disclosure database; dynamically monitoring drug requirements to obtain a drug screening range, performing comprehensive analysis according to the drug screening range in combination with the drug feature set, and constructing a first drug screening model; a knowledge distillation optimization channel is constructed, and the knowledge distillation optimization channel comprises a plurality of distillation optimization branches; synchronizing the drug feature set to a plurality of distillation optimization branches, and outputting a plurality of distillation results; performing distillation optimization on the first drug screening model based on the plurality of distillation results to generate a second drug screening model; and performing intelligent drug screening based on the second drug screening model. The technical problems of low screening efficiency and poor accuracy in the drug screening process in the prior art are solved, and the technical effect of improving the drug screening efficiency and accuracy is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of drug screening, and in particular to an intelligent drug screening method, device and medium. Background Art

[0002] In the modern medical field, drug screening is a crucial part of the drug development process. When the human body is infected with pathogens such as viruses or bacteria, it is usually necessary to quickly and accurately find effective drugs for treatment. For example, in the case of viral infections such as colds or flu, it is necessary to screen out antiviral drugs for specific viruses, and in the case of inflammation caused by bacterial infections, it is necessary to find symptomatic antibacterial and anti-inflammatory drugs. Traditional drug screening methods mainly rely on laboratory experiments and clinical experience, which is time-consuming and labor-intensive, with low screening efficiency, and it is difficult to meet the needs of rapid treatment and new drug development. Summary of the invention

[0003] The embodiments of the present application provide an intelligent drug screening method, device and medium, which solve the technical problems of low screening efficiency and poor accuracy in the drug screening process in the prior art.

[0004] In view of the above problems, the embodiments of the present application provide an intelligent drug screening method, device and medium.

[0005] A first aspect of an embodiment of the present application provides an intelligent drug screening method, the method comprising: Based on the collection of public drug databases, a drug feature set is acquired; drug demand is dynamically monitored to obtain a drug screening range, and a first drug screening model is constructed by performing a comprehensive analysis based on the drug screening range and the drug feature set; a knowledge distillation optimization channel is constructed, wherein the knowledge distillation optimization channel includes multiple distillation optimization branches; the drug feature set is synchronized to the multiple distillation optimization branches, and multiple distillation results are output; the first drug screening model is distilled and optimized based on the multiple distillation results to generate a second drug screening model; and intelligent drug screening is performed based on the second drug screening model.

[0006] A second aspect of an embodiment of the present application provides an electronic device, comprising: a memory for storing executable instructions; and a processor for executing the executable instructions stored in the memory to implement an intelligent drug screening method provided by the present application.

[0007] A third aspect of an embodiment of the present application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements an intelligent drug screening method provided by the present application.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application collects and obtains drug feature sets based on public drug databases; dynamically monitors drug demand to obtain drug screening ranges, conducts comprehensive analysis based on drug screening ranges combined with drug feature sets, and constructs a first drug screening model; constructs a knowledge distillation optimization channel, wherein the knowledge distillation optimization channel includes multiple distillation optimization branches; synchronizes the drug feature set to multiple distillation optimization branches, and outputs multiple distillation results; performs distillation optimization on the first drug screening model based on multiple distillation results to generate a second drug screening model; and performs intelligent drug screening based on the second drug screening model. The present invention solves the technical problems of low screening efficiency and poor accuracy in the process of drug screening in the prior art, dynamically monitors drug demand and conducts comprehensive analysis based on drug feature sets, constructs and optimizes drug screening models, uses knowledge distillation optimization channels to improve model performance, and finally optimizes the screening model through multiple distillation results, thereby achieving the technical effect of improving drug screening efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic diagram of a smart drug screening method provided in an embodiment of the present application; Figure 2 A schematic diagram of a process for constructing a knowledge distillation optimization channel in an intelligent drug screening method provided in an embodiment of the present application; Figure 3 This is a schematic diagram of the structure of an exemplary electronic device of the present application.

[0011] Description of the reference numerals: bus 300 , receiver 301 , processor 302 , transmitter 303 , memory 304 , bus interface 305 . DETAILED DESCRIPTION

[0012] The embodiments of the present application solve the technical problems of low screening efficiency and poor accuracy in the drug screening process in the prior art by providing an intelligent drug screening method, equipment and medium. The drug screening model is constructed and optimized by dynamically monitoring drug demand and conducting comprehensive analysis in combination with drug feature sets. The knowledge distillation optimization channel is used to improve model performance. Finally, the screening model is optimized through multiple distillation results, thereby achieving the technical effect of improving drug screening efficiency and accuracy.

[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0014] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0015] Embodiment 1, as Figure 1 As shown, the present application provides an intelligent drug screening method, the method comprising: Step S100: Acquire a drug feature set based on a public drug database.

[0016] In the embodiments of the present application, a drug feature set is collected from a public drug database, including chemical structure, physical properties and biological activity. Among them, chemical structure refers to the composition and arrangement of drug molecules, describing how the atoms in the drug molecules are connected and how these atoms are bonded. Physical properties refer to the physical characteristics of the drug, such as solubility, melting point, boiling point, density, etc. Biological activity involves the interaction between the drug and the organism, including the effect of the drug on specific biomarkers, the toxicity of the drug, and the effect of the drug at different doses.

[0017] Step S200: dynamically monitoring drug demand to obtain a drug screening range, performing a comprehensive analysis based on the drug screening range and the drug feature set, and constructing a first drug screening model.

[0018] In an embodiment of the present application, first, the demand for drugs is dynamically monitored, for example, the patient's condition and medication are tracked in real time through an electronic medical record system. In addition, the current drug demand is obtained through the drug use records and inventory data of the hospital pharmacy. Through these data sources, it is determined which drugs are frequently used in treatment, which drugs are in increasing demand, or which new drugs may need to be introduced. Next, the drug screening range is determined based on the dynamic monitoring results, that is, the type and quantity of drugs that need to be screened. This step is achieved by classifying and analyzing the collected drug use data through data mining and basic statistical analysis to identify high-demand drug categories. For example, cluster analysis techniques, such as hierarchical clustering, are used to determine the demand levels of different categories of drugs.

[0019] The drug screening range is then combined with the drug feature set for comprehensive analysis, including chemical structure analysis, using chemical informatics software to calculate the chemical descriptors of drug molecules, including molecular weight, molecular formula, bonding pattern, etc. Molecular fingerprints and similarity algorithms are used to evaluate the structural similarity between drug molecules. Then physical property analysis is performed, integrating the physical property data of the drug such as solubility, melting point, boiling point, density, etc., using principal component analysis to reduce the dimension and cluster the physical property data, and identify drug groups with similar physical properties. Then, biological activity analysis is performed. The biological activity data of the drug is normalized to eliminate data differences under different experimental conditions. K-means clustering is used to cluster the normalized activity data to identify drugs with similar biological activity patterns.

[0020] Then, data fusion and feature selection are performed. The chemical structure, physical properties and biological activity data are fused together to form a comprehensive data matrix. Finally, a comprehensive scoring model is constructed. Machine learning models such as support vector machines (SVMs) are used to train the fused data and construct a comprehensive scoring model. The performance of the model is evaluated using cross-validation techniques to ensure the robustness and generalization ability of the model on different data sets. The candidate drugs are input into the comprehensive scoring model, the comprehensive score of each drug is calculated, and the screening criteria are set according to the comprehensive score, and the drugs with the highest scores are selected as candidate drugs. Through the above dynamic monitoring and comprehensive analysis steps, the first drug screening model is finally constructed.

[0021] Step S300: constructing a knowledge distillation optimization channel, wherein the knowledge distillation optimization channel includes multiple distillation optimization branches.

[0022] In an embodiment of the present application, a knowledge distillation optimization channel is constructed. The knowledge distillation optimization channel is a multi-level optimization architecture, which includes multiple distillation optimization branches, each branch optimizing different types of drug characteristics.

[0023] Further, such as Figure 2 As shown, in the method provided in the embodiment of the application, constructing a knowledge distillation optimization channel also includes: Based on the drug feature set, multi-dimensional analysis is performed to obtain multiple drug chemical structure features, multiple drug physical property features, and multiple drug biological activity features; based on the multiple drug chemical structure features, the multiple drug physical property features, and the multiple drug biological activity features, normalization is performed to set multi-dimensional screening criteria; according to the multi-dimensional screening criteria, the knowledge distillation optimization channel is divided to generate a first distillation optimization branch, a second distillation optimization branch, and a third distillation optimization branch.

[0024] In an embodiment of the present application, a drug feature set is first subjected to multidimensional analysis to obtain multiple drug chemical structure features, multiple drug physical property features, and multiple drug biological activity features. After obtaining these features, normalization is performed, and feature data of different scales and units are converted to the same scale using standardization or minimum-maximum scaling. Next, based on the normalized multiple drug chemical structure features, physical property features, and biological activity features, a multidimensional screening criterion is set. The multidimensional screening criterion is a rule and criterion for screening and evaluating drug features, combining the chemical, physical, and biological properties of the drug. Using multivariate statistical analysis techniques, such as principal component analysis and factor analysis, the normalized data is subjected to dimensionality reduction processing to determine the screening dimensions and criteria. According to the set multidimensional screening criteria, the knowledge distillation optimization channel is divided to generate the first distillation optimization branch, the second distillation optimization branch, and the third distillation optimization branch.

[0025] Through the above steps, the knowledge distillation optimization channel is divided and three distillation optimization branches are generated, each branch is optimized for chemical structure, physical properties and biological activity characteristics respectively.

[0026] Step S400: synchronizing the drug feature set to the multiple distillation optimization branches, and outputting multiple distillation results.

[0027] In an embodiment of the present application, the extracted drug feature sets are synchronized to multiple distillation optimization branches. During the synchronization process, these feature data are respectively passed to the corresponding distillation optimization branches. Each distillation optimization branch focuses on a specific type of drug feature. In each distillation optimization branch, the key knowledge in the feature data is extracted and optimized through the training and verification of the deep learning model. This process is called knowledge distillation, which improves the performance and efficiency of the model by transferring the knowledge in the complex model to the simpler model. The optimized feature data are called distillation results, which contain high-quality feature information processed by the deep learning model.

[0028] Finally, multiple distillation results are output, each of which contains optimized chemical structure characteristics, physical property characteristics, and biological activity characteristics.

[0029] Furthermore, in the method provided in the embodiment of the application, the drug feature set is synchronized to the multiple distillation optimization branches, and multiple distillation results are output, and the method further includes: The drug feature set is trained according to the multidimensional screening standard, and a first distillation logic, a second distillation logic, and a third distillation logic are determined according to the training results; the first distillation optimization branch performs distillation processing on the multiple drug chemical structure features based on the first distillation logic to obtain chemical structure distillation results; the second distillation optimization branch performs distillation processing on the multiple drug physical property features based on the second distillation logic to obtain physical property distillation results; the third distillation optimization branch performs distillation processing on the multiple drug biological activity features based on the third distillation logic to obtain biological activity distillation results; the chemical structure distillation results, the physical property distillation results, and the biological activity distillation results are integrated to obtain the multiple distillation results.

[0030] In the embodiment of the present application, the drug feature set is first trained according to the multidimensional screening criteria to ensure that each drug feature is fully analyzed and optimized. In this process, the chemical structure characteristics, physical property characteristics and biological activity characteristics of the drug are trained separately. The data preprocessing link standardizes or normalizes the drug feature data, such as using Min-Max standardization or Z-Score standardization. In terms of feature extraction, chemical informatics software such as RDKit is used to extract chemical structure features and calculate molecular descriptors; physical property characteristics are obtained through experimental data; biological activity characteristics collect data on the effect of drugs on biomarkers, toxicity data, etc.

[0031] In the model training stage, the chemical structure features are trained using support vector machines; the physical property features are reduced in dimension using PCA and then classified by clustering analysis; the biological activity features are clustered and classified by K-means clustering method. According to the training results, three main distillation logics are determined, namely, the first distillation logic is based on knowledge distillation of relationships, using GNNs to model the relationship of drug chemical structure features, calculate the similarity between molecules, evaluate molecular fingerprints and use Tanimoto similarity algorithm for comparison; the second distillation logic is based on knowledge distillation of features, using PCA to reduce the dimension of physical property features, identify the main components, and then classify them by K-means clustering; the third distillation logic is based on knowledge distillation of responses, using neural network models (such as CNNs) to analyze biological activity features, and performing regression analysis based on the response curve of experimental data to evaluate drug efficacy and toxicity.

[0032] Next, the drug feature set is synchronized to the corresponding distillation optimization branch. The first distillation optimization branch distills the chemical structure features of multiple drugs based on the first distillation logic, uses multidimensional scaling (MDS) to reduce the dimension, and generates chemical structure distillation results; the second distillation optimization branch uses PCA and K-means clustering to process the physical property features based on the second distillation logic, optimizes the main components according to the feature importance ranking, and generates physical property distillation results; the third distillation optimization branch is based on the third distillation logic, analyzes the biological activity features through neural networks, performs weighted allocation and descending arrangement based on drug activity efficacy information and interaction information, prunes unimportant information, and generates biological activity distillation results.

[0033] Finally, the chemical structure distillation results, physical property distillation results and biological activity distillation results are integrated to form multiple optimized distillation results.

[0034] Furthermore, in the method provided in the embodiment of the application, the first distillation optimization branch performs distillation processing on the multiple drug chemical structure features based on the first distillation logic to obtain a chemical structure distillation result, and further includes: Construct similarity probability distribution function: ; in, It is used to characterize the similarity divergence of the relationship between the chemical structure feature and the chemical structure distillation result. The chemical structure characteristics are subjected to dimensionality reduction processing by constraining the chemical structure characteristics to be close to 0. Characterize the divergence of the similarity of the relationship between the chemical structure feature and the chemical structure distillation result, Characterizing the structural feature points in the chemical structure feature With preset chemical structure feature points The probability of relationship similarity is Characterizing the structural feature points in the distillation results of the chemical structure Chemical structure characteristics of target drugs ; using the similarity probability distribution function to calculate the similarity probability distribution of the chemical structure features in the first distillation logic to generate a similarity probability distribution result; using the similarity probability distribution result to perform distillation processing on the multiple drug chemical structure features to obtain the chemical structure distillation result.

[0035] In the embodiment of the present application, the similarity probability distribution of the chemical structure features in the first distillation logic is calculated using the similarity probability distribution function to generate a similarity probability distribution result. By minimizing the KL divergence, the similarity between the extracted chemical structure features and the distillation result is maximized, thereby obtaining an optimized chemical structure distillation result.

[0036] Furthermore, in the method provided in the embodiment of the application, the third distillation optimization branch performs distillation processing on the multiple drug biological activity characteristics based on the second distillation logic to obtain a biological activity distillation result, and further includes: Based on the analysis of the multiple drug biological activity characteristics, multiple drug activity efficacy information and multiple drug interaction information are determined; the multiple drug activity efficacy information and the multiple drug interaction information are weighted using the second distillation logic to obtain a weight coefficient allocation result; the multiple drug activity efficacy information and the multiple drug interaction information are arranged in descending order according to the weight coefficient allocation result to generate an information list; based on the second distillation logic, the information list is pruned according to a preset last digit ratio threshold to obtain the bioactivity distillation result.

[0037] In the examples of the present application, the biological activity data of the drug are first obtained through the biological database, which includes the effect of the drug on specific biomarkers, toxicity, and effects at different doses. In order to ensure the consistency and comparability of the data, the collected biological activity data are standardized, such as by Z-score standardization, to eliminate data differences under different experimental conditions.

[0038] Next, feature engineering methods are used to extract drug activity information, such as IC50 and EC50 values, which reflect the potency of the drug. In addition, information on drug interactions, such as the synergy index, is extracted to evaluate the effect of drug combinations.

[0039] After feature extraction, the second distillation logic is used to assign weights to this information. The specific method is to use a machine learning model, such as a random forest or gradient boosting decision tree, to train the model with bioactivity data, evaluate the importance of each feature, and obtain the weight assignment result. The feature importance score output by the model is used to calculate the weight coefficient of each feature.

[0040] Then, the activity efficacy information and interaction information are sorted in descending order according to the weight coefficient to generate an information list. The purpose of descending order is to prioritize and retain feature information that is of higher importance in the screening process, thereby improving the accuracy and efficiency of the screening. Based on the second distillation logic, the information list is pruned according to the preset last-place ratio threshold. Pruning refers to deleting information with lower weight and less importance to reduce computational complexity and improve model performance. The set last-place ratio threshold, for example 10%, means deleting the 10% information with the lowest weight. Through the above steps, the bioactivity distillation results are finally obtained.

[0041] Step S500: performing distillation optimization on the first drug screening model based on the multiple distillation results to generate a second drug screening model.

[0042] In an embodiment of the present application, multiple distillation results are input into the first drug screening model for distillation optimization. Distillation optimization is a method of knowledge extraction and transfer, which improves the performance of the student model by transferring the knowledge of a complex model (teacher model) to a simpler model (student model). In this process, deep learning techniques, such as convolutional neural networks or recursive neural networks, are used to optimize the training of the first drug screening model. Specifically, knowledge extraction is first performed to extract key features from the chemical structure distillation results, physical property distillation results, and biological activity distillation results. Next, model training is performed, and these key features are used to train the first drug screening model to optimize model parameters. Cross-validation techniques are used during training to evaluate the robustness and generalization ability of the model on different data sets. After that, loss function optimization is performed, and a distillation loss function is introduced to measure the difference between the teacher model and the student model. Commonly used distillation loss functions include mean square error and cross entropy loss. By minimizing the distillation loss function, the student model can effectively learn the knowledge of the teacher model. Finally, the model is adjusted, and the structure and parameters of the model are adjusted according to the distillation loss results to gradually improve the performance and prediction accuracy of the model. The gradient descent algorithm is used for parameter adjustment to accelerate the convergence of the model.

[0043] Through the above method, high-quality feature information of multiple distillation results is passed to the first drug screening model, comprehensive distillation optimization is performed, and finally the second drug screening model is generated.

[0044] Step S600: performing intelligent drug screening based on the second drug screening model.

[0045] Furthermore, the method provided in the application embodiment also includes: Extract drug activity information, drug structure information, and drug property information based on the multiple distillation results; extract target points, and traverse the drug structure information and the drug property information according to the drug activity information to match them with the target points to generate a drug matching coefficient; perform weighted calculation on the drug matching coefficient based on a first drug screening model to obtain the first drug candidate result, which includes multiple real-time screening indexes.

[0046] In the embodiment of the present application, drug activity information, drug structure information and drug property information are first extracted from multiple distillation results. Drug activity information involves the interaction between drugs and organisms, including the effects of drugs on specific biomarkers, toxicity and effects at different doses. Drug structure information describes the composition and arrangement of drug molecules, such as molecular weight, molecular formula and bonding mode. Drug property information includes physical properties of drugs, such as solubility, melting point, boiling point and density.

[0047] Next, extract the target. The target refers to the specific biological molecule or marker that the drug acts on. In the process of drug design and screening, the selection of the target is crucial. According to the extracted drug activity information, the structural information and property information of the drug are traversed to match the target. The matching process is completed by calculating the interaction strength and similarity between the drug and the target to generate the drug matching coefficient.

[0048] Then, based on the first drug screening model, the drug matching coefficient is weighted. Weighted calculation means weighting the matching coefficient according to the importance of each feature to obtain a comprehensive score, and finally the first drug candidate result is obtained. The first drug candidate result contains multiple real-time screening indexes, which are used to evaluate the performance of the drug under different screening conditions.

[0049] Furthermore, in the method provided in the embodiment of the application, intelligent drug screening is performed based on the second drug screening model, and further includes: A distillation loss function is introduced to generate a distillation loss result by calculating the first drug candidate result and the second drug candidate result in combination with the multiple real-time screening indexes; The distillation loss function is: ; in, Characterizing the distillation loss results, Characterizing the first drug candidate results obtained by the first drug screening model The second drug candidate result obtained by the second drug screening model The divergence between Characterizes the distillation temperature coefficient, Characterize any real-time screening index of the first drug candidate results obtained by the first drug screening model; based on the distillation loss result, synchronize the multiple distillation results to the second drug screening model for intelligent drug screening.

[0050] In the embodiment of the present application, the first candidate drug result and the second candidate drug result are first obtained through the previous steps. The two results are obtained through similar processes, that is, based on drug activity information, drug structure information and drug property information, target targets are extracted, drug matching coefficients are calculated, and weighted calculations are performed to obtain the first candidate drug result and the second candidate drug result. Next, the distillation loss function is used for calculation to generate a distillation loss result.

[0051] During the calculation process, the output of the first and second drug screening models is softened, that is, the output probability is adjusted to make it smoother, so as to better capture the subtle differences between the models. This step is achieved by introducing the distillation temperature coefficient t. When the temperature is higher, the output probability distribution becomes smoother and the distillation loss is larger, so that the student model can better learn the knowledge of the teacher model.

[0052] Then the output of the first drug screening model and the output of the second drug screening model are calculated through the distillation loss function to obtain a loss value. This loss value is used to measure the similarity between the two models. Finally, based on the distillation loss result, multiple distillation results are combined and synchronized to the second drug screening model for intelligent drug screening.

[0053] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects: This application collects and obtains drug feature sets based on public drug databases; dynamically monitors drug demand to obtain drug screening ranges, conducts comprehensive analysis based on drug screening ranges combined with drug feature sets, and constructs a first drug screening model; constructs a knowledge distillation optimization channel, wherein the knowledge distillation optimization channel includes multiple distillation optimization branches; synchronizes the drug feature set to multiple distillation optimization branches, and outputs multiple distillation results; performs distillation optimization on the first drug screening model based on multiple distillation results to generate a second drug screening model; and performs intelligent drug screening based on the second drug screening model. The present invention solves the technical problems of low screening efficiency and poor accuracy in the process of drug screening in the prior art, dynamically monitors drug demand and conducts comprehensive analysis based on drug feature sets, constructs and optimizes drug screening models, uses knowledge distillation optimization channels to improve model performance, and finally optimizes the screening model through multiple distillation results, thereby achieving the technical effect of improving drug screening efficiency and accuracy.

[0054] Embodiment 2

[0055] Based on the inventive concept of an intelligent drug screening method in the aforementioned embodiment, the present application also provides an electronic device, comprising: at least one processor; a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the steps of any one of the methods described in the aforementioned embodiment one.

[0056] Figure 3 This is a schematic diagram of the structure of an exemplary electronic device of this application. Figure 3 In the embodiment, the bus architecture is represented by bus 300, which may include any number of interconnected buses and bridges, and bus 300 connects various circuits including one or more processors represented by processor 302 and memory represented by memory 304. Bus 300 may also connect various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. Bus interface 305 provides an interface between bus 300 and receiver 301 and transmitter 303. Receiver 301 and transmitter 303 may be the same element, namely a transceiver, providing a unit for communicating with various other devices over a transmission medium. Processor 302 is responsible for managing bus 300 and general processing, while memory 304 may be used to store data used by processor 302 when performing operations.

[0057] Embodiment 3

[0058] Based on the same inventive concept as the intelligent drug screening method in the aforementioned embodiment, the present application also provides a computer-readable storage medium, on which a computer program is stored, and the computer program implements the steps of any one of the methods described in the above embodiment one when executed.

[0059] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0060] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0061] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. An intelligent drug screening method, characterized in that: The method comprises: Obtain drug feature sets based on public drug databases; Dynamically monitor drug demand to obtain a drug screening range, conduct a comprehensive analysis based on the drug screening range and the drug feature set, and construct a first drug screening model; Constructing a knowledge distillation optimization channel, wherein the knowledge distillation optimization channel includes multiple distillation optimization branches; Synchronizing the drug feature set to the multiple distillation optimization branches, and outputting multiple distillation results; Performing distillation optimization on the first drug screening model based on the multiple distillation results to generate a second drug screening model; Intelligent drug screening is performed based on the second drug screening model.

2. The intelligent drug screening method according to claim 1, characterized in that: Construct a knowledge distillation optimization channel, including: Perform multidimensional analysis based on the drug feature set to obtain multiple drug chemical structure features, multiple drug physical property features, and multiple drug biological activity features; Performing normalization based on the multiple drug chemical structure characteristics, the multiple drug physical property characteristics, and the multiple drug biological activity characteristics, and setting a multidimensional screening standard; The knowledge distillation optimization channel is divided according to the multidimensional screening standard to generate a first distillation optimization branch, a second distillation optimization branch, and a third distillation optimization branch.

3. The intelligent drug screening method according to claim 2, characterized in that: Synchronizing the drug feature set to the multiple distillation optimization branches and outputting multiple distillation results, the method comprising: Training the drug feature set according to the multidimensional screening criteria, and determining a first distillation logic, a second distillation logic, and a third distillation logic according to the training results; The first distillation optimization branch performs distillation processing on the plurality of drug chemical structure features based on the first distillation logic to obtain a chemical structure distillation result; The second distillation optimization branch performs distillation processing on the plurality of drug physical property characteristics based on the second distillation logic to obtain a physical property distillation result; The third distillation optimization branch performs distillation processing on the plurality of drug biological activity characteristics based on the third distillation logic to obtain a biological activity distillation result; The chemical structure distillation results, the physical property distillation results, and the biological activity distillation results are integrated to obtain the multiple distillation results.

4. The intelligent drug screening method according to claim 3, characterized in that: The first distillation optimization branch performs distillation processing on the plurality of drug chemical structure features based on the first distillation logic to obtain a chemical structure distillation result, and the method includes: Construct similarity probability distribution function: ; in, It is used to characterize the similarity divergence of the relationship between the chemical structure feature and the chemical structure distillation result. The chemical structure characteristics are subjected to dimensionality reduction processing by constraining the chemical structure characteristics to be close to 0. Characterize the divergence of the similarity of the relationship between the chemical structure feature and the chemical structure distillation result, Characterizing the structural feature points in the chemical structure feature With preset chemical structure feature points The probability of relationship similarity is Characterizing the structural feature points in the distillation results of the chemical structure Chemical structure characteristics of target drugs The probability of relationship similarity; Calculating the similarity probability distribution of chemical structure features in the first distillation logic using the similarity probability distribution function to generate a similarity probability distribution result; The plurality of drug chemical structure features are distilled using the similarity probability distribution result to obtain the chemical structure distillation result.

5. The intelligent drug screening method according to claim 3, characterized in that: The third distillation optimization branch performs distillation processing on the plurality of drug biological activity characteristics based on the second distillation logic to obtain a biological activity distillation result, and the method includes: Analyze the multiple drug biological activity characteristics to determine multiple drug activity efficacy information and multiple drug interaction information; Using the second distillation logic, weighting the plurality of drug activity efficacy information and the plurality of drug interaction information is performed to obtain a weight coefficient allocation result; Arrange the plurality of drug activity efficacy information and the plurality of drug interaction information in descending order according to the weight coefficient allocation result to generate an information list; Based on the second distillation logic, the information list is pruned according to a preset last-place ratio threshold to obtain the bioactivity distillation result.

6. The intelligent drug screening method according to claim 1, characterized in that: Methods include: extracting drug activity information, drug structure information, and drug property information based on the multiple distillation results; Extracting the target point, and matching the drug structure information and the drug property information with the target point according to the drug activity information to generate a drug matching coefficient; The drug matching coefficient is weightedly calculated based on the first drug screening model to obtain a first drug candidate result, wherein the first drug candidate result includes multiple real-time screening indexes.

7. The intelligent drug screening method according to claim 6, characterized in that: Based on the second drug screening model, intelligent drug screening is performed, and the method includes: A distillation loss function is introduced to generate a distillation loss result by calculating the first drug candidate result and the second drug candidate result in combination with the multiple real-time screening indexes; The distillation loss function is: ; in, Characterizing the distillation loss results, Characterizing the first drug candidate results obtained by the first drug screening model The second drug candidate result obtained by the second drug screening model The divergence between Characterizes the distillation temperature coefficient, Characterize any real-time screening index of the first drug candidate results obtained by the first drug screening model; Based on the distillation loss result, the plurality of distillation results are synchronized to the second drug screening model for intelligent drug screening.

8. An electronic device, characterized in that: The electronic device comprises: A memory for storing executable instructions; The processor is used to implement an intelligent drug screening method according to any one of claims 1 to 7 when executing the executable instructions stored in the memory.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, an intelligent drug screening method as described in any one of claims 1 to 7 is implemented.

Citation Information

Patent Citations

  • Drug molecule generation and optimization method based on recurrent neural network and multi-task reinforcement learning

    CN114842922A

  • Model training method and device based on knowledge distillation, equipment and medium

    CN117313830A

  • Drug target binding affinity prediction method based on drug bimodal characteristics

    CN118298908A

  • Drug-target interaction prediction method based on joint representation learning and meta-path

    CN118298909A

  • Drug activity screening method and device

    CN118506911A