WO2023134061A1 - 基于人工智能的药物特征信息确定方法及装置 - Google Patents

基于人工智能的药物特征信息确定方法及装置 Download PDF

Info

Publication number
WO2023134061A1
WO2023134061A1 PCT/CN2022/089689 CN2022089689W WO2023134061A1 WO 2023134061 A1 WO2023134061 A1 WO 2023134061A1 CN 2022089689 W CN2022089689 W CN 2022089689W WO 2023134061 A1 WO2023134061 A1 WO 2023134061A1
Authority
WO
WIPO (PCT)
Prior art keywords
drug
molecular structure
feature
structure image
classification result
Prior art date
Application number
PCT/CN2022/089689
Other languages
English (en)
French (fr)
Inventor
王俊
Original Assignee
平安科技(深圳)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2023134061A1 publication Critical patent/WO2023134061A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/24Classification techniques
    • G06F18/241Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/22Matching criteria, e.g. proximity measures
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16CCOMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/50Molecular design, e.g. of drugs

Definitions

  • the present application relates to the field of intelligent medical technology, in particular to a method and device for determining drug characteristic information based on artificial intelligence.
  • the present application provides an artificial intelligence-based method and device for determining drug characteristic information, the main purpose of which is to solve the problem of poor determination accuracy of existing drug characteristic information.
  • a method for determining drug characteristic information based on artificial intelligence including:
  • the image data of the molecular structure of the drug is classified and processed to obtain the classification result of the molecular structure image.
  • the training is completed;
  • a feature matching operation is performed on the molecular structure image classification result based on the drug feature processing flow to obtain drug feature information of the target drug.
  • a device for determining drug characteristic information based on artificial intelligence including:
  • the acquisition module is used to acquire the drug molecular structure image data of the target drug
  • the processing module is used to classify the drug molecular structure image data based on the image classification model that has been trained to obtain the molecular structure image classification result.
  • the image classification model is based on the self-attention mechanism to mine molecular structure node features and pass
  • the auxiliary information of the graph topology structure is obtained after the layered pooling of the graph convolutional network and the completion of training;
  • a retrieval module configured to retrieve a drug feature processing flow that matches the classification result of the molecular structure image
  • a matching module configured to perform a feature matching operation on the molecular structure image classification result based on the drug feature processing flow, to obtain drug feature information of the target drug.
  • a computer-readable storage medium on which computer-readable instructions are stored, wherein, when the computer-readable instructions are executed by a processor, an artificial intelligence-based method for determining drug characteristic information is implemented ,include:
  • the image data of the molecular structure of the drug is classified and processed to obtain the classification result of the molecular structure image.
  • the training is completed;
  • a feature matching operation is performed on the molecular structure image classification result based on the drug feature processing flow to obtain drug feature information of the target drug.
  • a computer device including a memory, a processor, and computer-readable instructions stored in the memory and operable on the processor, wherein the computer-readable instructions are executed by the processor Realize the drug feature information determination method based on artificial intelligence, including:
  • the image data of the molecular structure of the drug is classified and processed to obtain the classification result of the molecular structure image.
  • the training is completed;
  • a feature matching operation is performed on the molecular structure image classification result based on the drug feature processing flow to obtain drug feature information of the target drug.
  • the technical solution provided by the embodiment of the present application has at least the following advantages:
  • the present application provides a method and device for determining drug feature information based on artificial intelligence, which realizes the determination of drug features based on artificial intelligence, greatly speeds up the recognition accuracy of drug features, reduces the complexity and time-consuming of artificial identification, and makes it possible to pass
  • the identification of drug molecular structure to determine drug characteristics is greatly accelerated, thereby improving the effectiveness of drug feature matching disease, and realizing an intelligent drug feature determination.
  • FIG. 1 shows a flow chart of a method for determining drug characteristic information based on artificial intelligence provided by an embodiment of the present application
  • FIG. 2 shows a flow chart of another method for determining drug characteristic information based on artificial intelligence provided by the embodiment of the present application
  • FIG. 3 shows a schematic diagram of a training structure of a graph convolutional neural network model provided by an embodiment of the present application
  • Fig. 4 shows a flow chart of another method for determining drug characteristic information based on artificial intelligence provided by the embodiment of the present application
  • Fig. 5 shows a composition block diagram of an artificial intelligence-based drug characteristic information determination device provided by an embodiment of the present application
  • FIG. 6 shows a schematic structural diagram of a computer device provided by an embodiment of the present application.
  • AI artificial intelligence
  • the embodiments of the present application may acquire and process relevant data based on artificial intelligence technology.
  • artificial intelligence is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. .
  • Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation/interaction systems, and mechatronics.
  • Artificial intelligence software technology mainly includes computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning/deep learning.
  • a method for determining drug characteristic information based on artificial intelligence is provided, and the application of the method to computer equipment such as a server is used as an example for illustration, wherein the server can be an independent Servers can also provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery network (Content Delivery Network, CDN), and large Cloud servers for basic cloud computing services such as data and artificial intelligence platforms, such as intelligent medical systems, digital medical platforms, etc.
  • the above method comprises the following steps:
  • the executor may be an intelligent management system with an information push function, for example, an intelligent medical system, a data medical platform, and the like.
  • the current execution subject is an intelligent medical system
  • the target drug is a drug that is suitable for determining the characteristics of the drug.
  • the drug molecular structure image data of the target drug is a molecule representing the target drug using a graph structure, where the drug
  • the image content in the molecular structure image data is the atom-chemical bond structure of the target drug molecule. From the image content, the characteristic content of the molecular structure in the form of node-edge, atomic number, charge number, etc.
  • the classification of image data can be abstracted, so that based on the The classification of image data obtains a classification implementation method for the molecular structure of drugs, that is, through the graph neural network, it can capture the local relationship of the graph and automatically learn the graph attributes by passing the information of nodes and edges, so as to efficiently perform graph classification tasks. .
  • the drug molecular structure image data in the embodiment of the present application is obtained by loading the drug molecular structure image data of the target drug generated by the intelligent medical system as the current execution subject based on the computer software for making molecular structure diagrams.
  • the operator can obtain the drug molecular structure image data matching the target drug based on the drug database already stored in the current intelligent medical system, or make it through the molecular structure creation application program, and make it in the specified file format in the intelligent medical system acquisition, which is not specifically limited in the embodiment of this application.
  • the drug molecular structure image data contains graph nodes and edges
  • the graph nodes contain entity information, such as atoms in compounds
  • the edges contain relationship information between entities, such as atoms in compound image data
  • the model training is performed in advance to obtain the image classification model to classify the drug molecular structure diagram data, and to obtain the molecular structure image classification results.
  • the corresponding classification result of the molecular structure image is the classification result representing different atoms-chemical bonds, so as to determine the molecular characteristics of the drug based on the classification result of the molecular structure image.
  • the image classification model is based on the self-attention mechanism to mine the characteristics of molecular structure nodes and perform hierarchical pooling on the graph convolutional network through the auxiliary information of the graph topology.
  • a Transformer-inspired module is used to calculate a self-attention score as a selection criterion , to mine molecular structure node features.
  • the first layer in the graph convolutional neural network selects high-scoring nodes with a learnable scoring function L2Pool to delete unnecessary nodes.
  • the function L2Pool relies on self-attention and the enhanced The auxiliary information of the graph topology, in this way, can retain as much information as possible while compressing the node size of the original graph sample data, thereby improving the accuracy of the image classification model for graph classification.
  • the drug feature processing process is used to characterize the process of determining the characteristics of applicable drugs.
  • the drug feature processing process includes pre-configured drug molecular composition features, drug molecular attribute features, drug molecular structure
  • the intelligent medical system in the embodiment of the present application pre-stores the corresponding relationship between different drug feature processing procedures and different molecular structure image classification results, so as to further determine the characteristics of the atom-chemical bond classification results, For example, if the classification result of atom a-chemical bond 1 matches the drug feature processing flow of drug molecular composition features and drug molecular attribute features, the drug feature processing flow of drug molecular composition features and drug molecular attribute features is called, so as to The processing node of feature and drug molecular attribute feature performs feature matching processing on atom a-chemical bond 1, and obtains the drug feature information of the target drug for the drug molecular composition feature and drug molecular attribute feature.
  • the feature matching operation is to perform one-to-one correspondence matching on the classification results of atoms-chemical bonds contained in the molecular structure image classification results according to the determined drug feature processing flow, that is, based on all existing data stored in the intelligent medical system.
  • the similarity calculation is carried out on the drug molecular composition characteristics, drug molecular attribute characteristics, drug molecular structure and atom-chemical bond classification results, so as to determine the drug characteristic information of the target drug.
  • step 102 classifies the drug molecular structure image data based on the trained image classification model, and obtains sub-structure image classification results.
  • the method further included:
  • a graph convolutional network is constructed according to the characteristics of the image data.
  • a molecular structure image sample data to be classified is image data in the form of graph nodes and edges.
  • a self-attention score is calculated as a selection criterion through a Transformer-inspired module.
  • V Through the graph convolutional network (Graph Convolutional Network, GCN) model, and in order to avoid excessive smoothing, it is limited to the graph convolutional network GCN shallow architecture, which limits the performance of its model, through the initial residual and identity Mapped GCNII (Graph Convolutional Network via Initial residual and Identity mapping) to construct V to display the global structure and capture the information interaction between graph nodes according to the structural dependencies of graph nodes, that is, to complete the mining of molecular structure node features .
  • GCN graph Convolutional Network
  • a rough image data is obtained by sampling and reducing the size of graph nodes.
  • the importance score is calculated for graph nodes based on sampling to retain the most important prior
  • the k graph nodes and the connections between them generate coarse image data, and realize the pooling of each network layer of the graph convolutional network.
  • the pooling model that performs pooling on each layer obtains a more effective graph representation through the global structure information and local structure information of the drug molecular structure image data, and the auxiliary information includes the drug molecular structure image data
  • the global structure information and the local structure information that is, the pooling of each network layer of the graph convolutional network that introduces the self-attention mechanism is performed through auxiliary information.
  • the image classification model is obtained by performing model training on the pooled graph convolutional network through the molecular structure image sample data as the model training sample.
  • step 202 determines the graph topology enhanced self-attention mechanism includes: constructing skip connections for the input layer of the graph convolutional network based on elementary residuals, and using identity mapping The weight matrix is unitized, and the unitized graph convolutional network is combined with a self-attention mechanism to obtain a graph topology enhanced self-attention mechanism.
  • the initial residual and the GCNII of the identity map are used to construct the V in the four-head attention.
  • the initial residual constructs a skip connection in the input layer, and the weight matrix is processed through the identity mapping Unitization processing, that is, identity mapping adds the unit matrix to the weight matrix to combine the unitized graph convolutional network with the self-attention mechanism to increase the network depth of GCNII, prevent excessive smoothing and continuously improve GCNII. performance.
  • the input of the graph convolutional neural network is an image data structure with graph nodes or edges, which includes the adjacency matrix A of the image data and the corresponding feature attribute information X.
  • the graph convolutional neural network trains the implicit vector representation of each graph node in the image data according to the graph structure and the attributes of the input nodes. The goal is to make the vector representation contain enough powerful expression information to help each graph node perform information Extraction, and finally the information vector representation of the entire graph can be obtained.
  • the molecular level information representation of the entire molecular compound can be extracted through the characteristics of the atomic nodes and the edge chemical bond information between atoms. .
  • the main process of graph convolutional neural network model learning is to iteratively aggregate and update the neighbor information of graph nodes in graph data.
  • each graph node updates its own information by aggregating the features of its neighbor nodes and its own features in the upper layer, and usually performs nonlinear transformation on the aggregated information.
  • each graph node can obtain the neighbor node information within the corresponding hop number, and use a new graph coarsening method based on Transformer self-attention mechanism and network topology information of image data, such as through coarse pooling
  • the method is realized by two features of node feature and graph topology, which are not specifically limited in this embodiment of the application.
  • step 103 before step 103 calls the drug feature processing flow matching the molecular structure image classification result, the method further includes: receiving a trigger for the drug molecular structure image data molecular classification tasks.
  • the drug feature processing flow contains at least one processing node in the drug molecular composition feature, drug molecular attribute feature, and drug molecular structure that are pre-configured to match the classification results of different molecular structure images, in order to meet the requirements of different drug features
  • the operational requirements of information matching enable different operators to automatically and flexibly execute the drug feature processing process based on the intelligent medical system, and configure molecular classification tasks for the drug feature processing process in advance, so that operators can trigger molecular classification tasks to execute different drug molecular structures The trigger of the drug characteristic processing flow.
  • the molecular classification task is used to represent the corresponding processing nodes in the drug feature processing flow for different drug molecular structure image data, that is, different processing nodes can be triggered by different molecular classification tasks, so as to have specific Consistently complete the execution of the drug feature processing flow for the classification results of different molecular structure images.
  • step 103 calls the drug feature processing flow that matches the classification result of the molecular structure image, including: analyzing the processing node of the molecular structure image data of the drug in the molecular classification task, and processing the obtained molecular structure image based on the classification The classification result is matched with the processing node; and at least one processing node matched with the molecular structure image classification result is called.
  • the processing node is at least one of the drug molecular composition feature, drug molecular attribute feature, and drug molecular structure matching in the drug feature processing flow, that is, the processing node includes a drug molecular composition feature processing node, a drug Molecular attribute feature processing node, drug molecular structure processing node, at this time, because the molecular classification task has the expected processing node for the drug molecular structure image data before image classification processing, but not all drug molecular structure image data After the image classification, it can still be executed according to the processing node.
  • the processing nodes of drug molecular structure image data in the molecular classification task of target drug a are drug molecular composition feature processing nodes, drug molecular attribute feature processing nodes, and drug molecular structure processing nodes
  • the processing node corresponding to the atom a-chemical bond 1 in the molecular structure image classification result matches the drug molecular composition feature processing node and the drug molecular structure processing node, and the drug molecular composition feature processing node and the drug molecular structure processing node are called , for processing node execution.
  • a matching relationship list can be entered in advance, so as to accurately match the classification results of different molecular structure images with the processing nodes.
  • the matching relationship list Not specifically limited.
  • step 104 performs a feature matching operation on the classification result of the molecular structure image based on the drug feature processing flow to obtain the drug feature of the target drug.
  • Information includes:
  • the corresponding processing nodes are called, such as the drug molecular composition feature processing node, the drug molecular attribute feature processing node, and the drug molecular structure processing node After that, the corresponding matching operation is performed.
  • the matching operation for each processing node is to calculate the similarity of the molecular structure image classification results according to each processing node, so as to determine the drug characteristic information of the target drug based on the obtained similarity.
  • the molecular structure image classification result is calculated by similarity with the drug molecular composition features stored in the intelligent medical system, because the molecular structure image classification result contains the image of the atom-chemical bond Specifically, the similarity calculation is performed on the image corresponding to the molecular composition characteristics of the drug, so as to determine whether the calculated similarity value is greater than the first similarity threshold, and if so, determine that the target drug contains the drug characteristic information of the molecular composition characteristics of the drug . In the drug molecular attribute feature processing node, the similarity calculation is performed by comparing the molecular structure image classification results with the drug molecular attribute features stored in the intelligent medical system.
  • Similarity calculation is performed on the image corresponding to the drug molecular attribute feature, so as to determine whether the calculated similarity value is greater than the second similarity threshold, and if so, determine that the target drug contains drug feature information of the drug molecular attribute feature.
  • the drug molecular structure processing node calculates the similarity between the molecular structure image classification result and the drug molecular structure stored in the intelligent medical system. Similarity calculation is performed on the corresponding images, so as to determine whether the calculated similarity value is greater than the third similarity threshold, and if so, determine that the target drug contains drug characteristic information of the drug molecular structure.
  • the molecular composition characteristics of the drug are used to characterize the characteristic content of different molecules in the drug, for example, molecular composition features such as aromatic hydrocarbons and methyl groups, and the molecular property characteristics of the drug are used to characterize the physical or molecular properties of different molecules in the drug.
  • Chemical properties such as toluene-easily oxidized and other chemical properties
  • drug molecular structure is used to characterize the chemical bond structure between molecules, for example, aromatic hydrocarbons have the basic molecular structure of benzene rings, therefore, the composition characteristics of drug molecules, drug molecular properties
  • the processing nodes corresponding to the features and molecular structure of the drug match the classification results of the molecular structure image.
  • the molecular structure image classification results are numerically processed, so that the drug molecular composition characteristics, drug molecular attribute characteristics, drug molecular structure progress number and molecular structure image classification results to be calculated similarity are put into one data unit Perform similarity calculation to complete the matching process.
  • the step is to perform a feature matching operation on the classification result of the molecular structure image based on the drug feature processing flow, and after obtaining the drug feature information of the target drug, the method further The method includes: obtaining a disease characteristic database, judging whether the drug characteristic information and each disease characteristic information in the disease characteristic database have an antagonistic attribute; if they have the antagonistic attribute, outputting the disease characteristic information.
  • the disease characteristic information related to the drug characteristic information, so as to perform the treatment operation of the relevant disease.
  • the disease characteristic information corresponding to various known diseases is recorded in the disease characteristic database stored in the intelligent medical system.
  • the disease characteristic information is used to represent the content of the disease caused by the disease to the human body, for example, a certain disease It will cause blood pressure to be higher than 180mmHg, and a certain disease will cause a great decrease in adrenaline, etc.
  • the antagonistic attribute is whether there is an attribute that interacts with each other between the drug characteristic information and the disease characteristic information, for example, the drug characteristic information Whether it has a boosting effect on the low blood pressure in the disease characteristic information, if there is an antagonistic attribute, it means that the target drug can be used as the treatment or symptom relief of the disease corresponding to the disease characteristic information, and has been output to the user for viewing .
  • the step of judging whether the drug feature and each disease feature information in the disease feature database has an antagonistic attribute includes: acquiring an antagonistic attribute list; Information and the markers corresponding to the biological feature information and the chemical feature information determine whether they have an antagonistic attribute.
  • the biological characteristic information, chemical Feature information marks whether there is a medical connection between different drug specific effect information.
  • This mark is an antagonistic attribute with medical relevance determined through medical experiments.
  • the drug molecular attribute feature is recorded in the list of antagonistic attributes a
  • the biological feature information and chemical feature information of different disease feature information are respectively used to be pre-entered into the intelligent medical system.
  • the symptom characteristics of the disease such as pain, fever, blood item value, cell mass, etc.
  • the chemical characteristic information refers to the symptom characteristics of the chemical composition of the disease, hormone values, etc., which are not specifically limited in the embodiment of the present application.
  • the embodiment of the present application provides a method for determining drug characteristic information based on artificial intelligence. Compared with the prior art, the embodiment of the present application obtains the drug molecular structure image data of the target drug; The image data of the molecular structure of the drug is classified and processed to obtain the classification result of the molecular structure image.
  • the image classification model is to mine the molecular structure node features based on the self-attention mechanism and perform hierarchical pooling on the graph convolutional network through the auxiliary information of the graph topology.
  • the obtained training is completed; the drug feature processing flow matched with the molecular structure image classification result is called; based on the drug feature processing flow, a feature matching operation is performed on the molecular structure image classification result to obtain the target drug Drug feature information, realize the determination of drug features based on artificial intelligence, greatly speed up the recognition accuracy of drug features, reduce the complexity and time-consuming of human identification, and greatly accelerate the speed of determining drug features through the identification of drug molecular structures , so as to improve the effectiveness of drug feature matching disease, and realize an intelligent drug feature determination.
  • an embodiment of the present application provides an artificial intelligence-based device for determining drug characteristic information, as shown in Figure 5, the device includes:
  • An acquisition module 31, configured to acquire drug molecular structure image data of the target drug
  • the processing module 32 is used to classify the drug molecular structure image data based on the image classification model that has been trained to obtain the molecular structure image classification result.
  • the image classification model is based on the self-attention mechanism to mine molecular structure node features and After layered pooling of the graph convolutional network through the auxiliary information of the graph topology, the training is completed;
  • a call module 33 configured to call a drug feature processing flow that matches the classification result of the molecular structure image
  • the matching module 34 is configured to perform a feature matching operation on the molecular structure image classification result based on the drug feature processing flow to obtain drug feature information of the target drug.
  • the device also includes:
  • a determination module configured to determine a self-attention mechanism for graph topology enhancement, and introduce the self-attention mechanism into the input layer of the graph convolutional network to mine molecular structure node features;
  • the parsing module is used to determine auxiliary information based on the graph topology obtained by parsing the drug molecular structure image data, and based on the auxiliary information, analyze each network layer of the graph convolution network that introduces the self-attention mechanism Performing pooling, the auxiliary information includes global structure information and local structure information of the drug molecular structure image data;
  • the training module is used to perform model training on the pooled graph convolutional network through molecular structure image sample data to obtain an image classification model.
  • the determination module is specifically configured to construct a skip connection to the input layer of the graph convolutional network based on the elementary residual, and perform unitization processing on the weight matrix through the identity mapping, and unitize the unitized
  • the graph convolutional network is combined with the self-attention mechanism to obtain a graph topology-enhanced self-attention mechanism.
  • the device also includes:
  • the receiving module receives the molecular classification task triggered by the molecular structure image data of the drug, and the molecular classification task is used to represent the corresponding processing node in the drug feature processing flow for different molecular structure image data of the drug;
  • the calling module includes:
  • An analysis unit configured to analyze the processing node of the molecular structure image data of the drug in the molecular classification task, and match the classification result of the molecular structure image based on the classification processing with the processing node, and the processing node is the drug Executing at least one of drug molecule composition features, drug molecule attribute features, and drug molecule structure matching in the feature processing flow;
  • a calling unit configured to call at least one processing node matching the molecular structure image classification result.
  • the matching module includes:
  • a calculation unit used to calculate the similarity between the classification result of the molecular structure image and the composition characteristics of the drug molecule, the attribute characteristics of the drug molecule, and the molecular structure of the drug;
  • the first determining unit is configured to determine the drug feature information including the drug molecular composition feature if the first similarity between the molecular structure image classification result and the drug molecular composition feature is greater than a preset first similarity threshold; and / or,
  • the second determining unit is configured to determine the drug feature information including the drug molecule attribute feature if the second similarity between the molecular structure image classification result and the drug molecule attribute feature is greater than a preset second similarity threshold; and / or,
  • the third determining unit is configured to determine the drug feature information including the drug molecular structure if the third similarity between the molecular structure image classification result and the drug molecular structure is greater than a preset third similarity threshold.
  • the device also includes:
  • a judging module configured to acquire a disease characteristic database, and judge whether the drug characteristic information and each disease characteristic information in the disease characteristic database have an antagonistic attribute
  • An output module configured to output the disease characteristic information if it has the antagonistic attribute.
  • the judging module includes:
  • the acquiring unit is used to acquire an antagonism attribute list, the antagonism attribute list records the biological characteristic information and chemical characteristic information of different disease characteristic information, and marks whether there is a medical association between the special effect information of different drugs;
  • a judging unit configured to judge whether it has an antagonistic attribute based on the marks corresponding to the drug feature information, the biological feature information, and the chemical feature information.
  • the embodiment of the present application provides a device for determining drug characteristic information based on artificial intelligence. Compared with the prior art, the embodiment of the present application obtains the drug molecular structure image data of the target drug; The image data of the molecular structure of the drug is classified and processed to obtain the classification result of the molecular structure image.
  • the image classification model is to mine the molecular structure node features based on the self-attention mechanism and perform hierarchical pooling on the graph convolutional network through the auxiliary information of the graph topology.
  • the obtained training is completed; the drug feature processing flow matched with the molecular structure image classification result is called; based on the drug feature processing flow, a feature matching operation is performed on the molecular structure image classification result to obtain the target drug Drug characteristic information, realize the determination of drug characteristics based on artificial intelligence, greatly speed up the recognition accuracy of drug characteristics, reduce the complexity and time-consuming of human identification, and greatly speed up the determination of drug characteristics through the identification of drug molecular structure , so as to improve the effectiveness of drug feature matching disease, and realize an intelligent drug feature determination.
  • a computer-readable storage medium stores at least one executable instruction, and the computer-executable instruction can perform the determination of drug characteristic information based on artificial intelligence in any of the above method embodiments method.
  • the computer-readable storage medium may be non-volatile or volatile.
  • FIG. 6 shows a schematic structural diagram of a computer device provided according to an embodiment of the present application.
  • the specific embodiment of the present application does not limit the specific implementation of the computer device.
  • the computer device may include: a processor (processor) 402, a communication interface (Communications Interface) 404, a memory (memory) 406, and a communication bus 408.
  • processor processor
  • Communication interface Communication Interface
  • memory memory
  • the processor 402 , the communication interface 404 , and the memory 406 communicate with each other through the communication bus 408 .
  • the communication interface 404 is used to communicate with network elements of other devices such as clients or other servers.
  • the processor 402 is configured to execute the program 410, specifically, may execute the relevant steps in the above embodiment of the method for determining drug characteristic information based on artificial intelligence.
  • the program 410 may include program codes including computer operation instructions.
  • the processor 402 may be a central processing unit CPU, or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.
  • the one or more processors included in the computer device may be of the same type, such as one or more CPUs, or may be of different types, such as one or more CPUs and one or more ASICs.
  • the memory 406 is used to store the program 410 .
  • the memory 406 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.
  • the program 410 can specifically be used to make the processor 402 perform the following operations:
  • the image data of the molecular structure of the drug is classified and processed to obtain the classification result of the molecular structure image.
  • the training is completed;
  • a feature matching operation is performed on the molecular structure image classification result based on the drug feature processing flow to obtain drug feature information of the target drug.
  • each module or each step of the above-mentioned application can be realized by a general-purpose computing device, and they can be concentrated on a single computing device, or distributed in a network composed of multiple computing devices Alternatively, they may be implemented in program code executable by a computing device so that they may be stored in a storage device to be executed by a computing device, and in some cases, in an order different from that shown here
  • the steps shown or described are carried out, or they are separately fabricated into individual integrated circuit modules, or multiple modules or steps among them are fabricated into a single integrated circuit module for implementation.
  • the present application is not limited to any specific combination of hardware and software.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Biology (AREA)
  • Molecular Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Mathematical Physics (AREA)
  • Biophysics (AREA)
  • Chemical & Material Sciences (AREA)
  • Medicinal Chemistry (AREA)
  • Pharmacology & Pharmacy (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Crystallography & Structural Chemistry (AREA)
  • Image Analysis (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

本申请公开了一种基于人工智能的药物特征信息确定方法及装置,涉及智能医疗技术领域,主要目的在于解决现有药物特征信息确定准确性差的问题。主要包括:获取目标药物的药物分子结构图像数据;基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;调取与所述分子结构图像分类结果匹配的药物特征处理流程;基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息。主要用于基于人工智能的药物特征信息的确定。

Description

基于人工智能的药物特征信息确定方法及装置
本申请要求与2022年01月11日提交中国专利局、申请号为202210026435.8申请名称为“基于人工智能的药物特征信息确定方法及装置”的中国专利申请的优先权,其全部内容通过引用结合在申请中。
技术领域
本申请涉及一种智能医疗技术领域,特别是涉及一种基于人工智能的药物特征信息确定方法及装置。
背景技术
近年来,智能医疗技术的应用领域已经从临床治疗逐步向药物研发方向发展,越来越多的人工智能技术涉足于药物对不同病症的适用情况的分析,从而准确找到适用于临床治疗的药物。尤其是针对药物的分子结构进行研究,从而基于药物特征来确定适合患者的治疗方案或者病症的治疗。发明人意识到目前基于药物分子结构的研究均是采用物理实验方式来确定药物特征,然而,这样的药物分子结构识别过程较慢,无法有效的使用到临床治疗中,从而使得基于药物特征匹配病症在智能医疗中的使用效率较低,因此,亟需一种基于人工智能的药物特征信息确定方法来解决上述问题。
发明内容
有鉴于此,本申请提供一种基于人工智能的药物特征信息确定方法及装置,主要目的在于解决现有药物特征信息确定准确性差的问题。
依据本申请一个方面,提供了一种基于人工智能的药物特征信息确定方法,包括:
获取目标药物的药物分子结构图像数据;
基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;
调取与所述分子结构图像分类结果匹配的药物特征处理流程;
基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息。
依据本申请另一个方面,提供了一种基于人工智能的药物特征信息确定装置,包括:
获取模块,用于获取目标药物的药物分子结构图像数据;
处理模块,用于基于已完成训练的图像分类模型对所述药物分子结构图像数据进行 分类处理,得到分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;
调取模块,用于调取与所述分子结构图像分类结果匹配的药物特征处理流程;
匹配模块,用于基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息。
根据本申请的又一方面,提供了一种计算机可读存储介质,其上存储有计算机可读指令,其中,所述计算机可读指令被处理器执行时实现基于人工智能的药物特征信息确定方法,包括:
获取目标药物的药物分子结构图像数据;
基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;
调取与所述分子结构图像分类结果匹配的药物特征处理流程;
基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息。
根据本申请的再一方面,提供了一种计算机设备,包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机可读指令,其中,所述计算机可读指令被处理器执行时实现基于人工智能的药物特征信息确定方法,包括:
获取目标药物的药物分子结构图像数据;
基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;
调取与所述分子结构图像分类结果匹配的药物特征处理流程;
基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息。
借由上述技术方案,本申请实施例提供的技术方案至少具有下列优点:
本申请提供了一种基于人工智能的药物特征信息确定方法及装置,实现基于人工智能的药物特征的确定,大大加快了药物特征的识别准确性,减少人为识别的复杂度以及耗时,使得通过药物分子结构的识别来确定药物特征的速度大大加快,从而提高了药物特征匹配病症的有效性,并实现了一种智能化的药物特征确定。
上述说明仅是本申请技术方案的概述,为了能够更清楚了解本申请的技术手段,而可依照说明书的内容予以实施,并且为了让本申请的上述和其它目的、特征和优点能够更明显易懂,以下特举本申请的具体实施方式。
附图说明
通过阅读下文优选实施方式的详细描述,各种其他的优点和益处对于本领域普通技术人员将变得清楚明了。附图仅用于示出优选实施方式的目的,而并不认为是对本申请的限制。而且在整个附图中,用相同的参考符号表示相同的部件。在附图中:
图1示出了本申请实施例提供的一种基于人工智能的药物特征信息确定方法流程图;
图2示出了本申请实施例提供的另一种基于人工智能的药物特征信息确定方法流程图;
图3示出了本申请实施例提供的一种图卷积神经网络模型训练结构示意图;
图4示出了本申请实施例提供的又一种基于人工智能的药物特征信息确定方法流程图;
图5示出了本申请实施例提供的一种基于人工智能的药物特征信息确定装置组成框图;
图6示出了本申请实施例提供的一种计算机设备的结构示意图。
具体实施方式
下面将参照附图更详细地描述本公开的示例性实施例。虽然附图中显示了本公开的示例性实施例,然而应当理解,可以以各种形式实现本公开而不应被这里阐述的实施例所限制。相反,提供这些实施例是为了能够更透彻地理解本公开,并且能够将本公开的范围完整的传达给本领域的技术人员。
本申请实施例可以基于人工智能技术对相关的数据进行获取和处理。其中,人工智能(Artificial Intelligence,AI)是利用数字计算机或者数字计算机控制的机器模拟、延伸和扩展人的智能,感知环境、获取知识并使用知识获得最佳结果的理论、方法、技术及应用系统。
人工智能基础技术一般包括如传感器、专用人工智能芯片、云计算、分布式存储、大数据处理技术、操作/交互系统、机电一体化等技术。人工智能软件技术主要包括计算机视觉技术、机器人技术、生物识别技术、语音处理技术、自然语言处理技术以及机器学习/深度学习等几大方向。
基于此,在一个实施例中,如图1所示,提供了一种基于人工智能的药物特征信息确定方法,以该方法应用于服务器等计算机设备为例进行说明,其中,服务器可以是独立的服务器,也可以是提供云服务、云数据库、云计算、云函数、云存储、网络服务、云通信、中间件服务、域名服务、安全服务、内容分发网络(Content Delivery Network,CDN)、以及大数据和人工智能平台等基础云计算服务的云服务器,如智能医疗系统、数字医疗平台等。上述方法包括以下步骤:
101、获取目标药物的药物分子结构图像数据。
本申请实施例中,执行主体可以是带有信息推送功能的智能管理系统,例如,智能医疗系统、数据医疗平台等。示例性的,当前执行主体为智能医疗系统,目标药物为适用于待进行与药物特征确定的药物,对应的,目标药物的药物分子结构图像数据为使用图结构表示目标药物的分子,其中,药物分子结构图像数据中的图像内容为目标药物分子的原子-化学键结构,从图像内容中可以抽象得到以节点-边形式的空间特征、原子序数、电荷数等分子结构的特征内容,从而可以基于对图像数据的分类,得到对药物分子结构的一种分类实现方法,即通过图神经网络可以通过传递节点和边的信息等特定,捕捉图的局部关系自动学习图属性,从而高效的进行图分类任务。
需要说明的是,本申请实施例中的药物分子结构图像数据为作为当前执行主体的智能医疗系统基于制作分子结构图的计算机软件生成目标药物的药物分子结构图像数据后进行加载得到的,此时,操作人员可以基于已经存储于当前智能医疗系统中的药物数据库获取与目标药物匹配的药物分子结构图像数据,也可以通过分子结构制作应用程序进行制作,并以智能医疗系统中的指定文件格式进行获取,本申请实施例不做具体限定。
102、基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果。
本申请实施例中,由于药物分子结构图像数据中带有包含图节点和边,其中,图节点包含了实体信息,如化合物中的原子,边包含实体间的关系信息,如化合物图像数据中原子间的化学键,为了针对药物分子结构图像数据进行分类,以得到药物分子分类结果进行药物特征信息的确定,预先进行模型训练得到图像分类模型以对药物分子结构图数据进行分类处理,得到分子结构图像分类结果。其中,由于对药物分子结构图像数据是进行图神经网络的分类,对应得到的分子结构图像分类结果即为表示不同原子-化学键的分类结果,以便基于分子结构图像分类结果确定药物分子特征。
需要说明的是,为了解决图池化不足的限制,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的,即构建图卷积神经网络后,在训练过程中,基于在图样本数据的邻接矩阵A和特征矩阵X的基础上,使用Transformer启发的模块计算一个自注意力分数作为选择标准,以挖掘分子结构节点特征。同时,在图卷积神经网络中的第一层选择具有可学习得分函数L2Pool的高得分节点,以删除不需要的节点,此时,函数L2Pool依赖于自注意力和图卷积网络中增强的图拓扑结构的辅助信息,通过这种方式,可以在压缩原始图样本数据的节点规模的同时,保留尽可能多的信息,从而提高图像分类模型进行图分类的准确性。
103、调取与所述分子结构图像分类结果匹配的药物特征处理流程。
本申请实施例中,由于分子结构图像分类结果中包含有不同原子-化学键的分类结果,为了增加与药物特征信息确定的准确性,不同原子-化学键的分类结果对应不同的 药物特征处理流程。其中,药物特征处理流程用于表征适用药物进行特征确定的流程,药物特征处理流程中包含有预先配置的对不同分子结构图像分类结果进行匹配的药物分子组成特征、药物分子属性特征、药物分子结构的处理节点,对不同处理节点进行任意组合而得到的处理流程,从而启动在各个处理节点处进行针对性的药物特征确定,从而确定是否具有疾病对抗性,以确定为针对某些疾病的治疗药物。
需要说明的是,本申请实施例中的智能医疗系统中预先存储有不同药物特征处理流程与不同分子结构图像分类结果之间的对应关系,以便对原子-化学键的分类结果进行进一步地特征确定,例如,原子a-化学键1的分类结果与药物分子组成特征、药物分子属性特征的药物特征处理流程匹配,则调取药物分子组成特征、药物分子属性特征的药物特征处理流程,以便基于药物分子组成特征、药物分子属性特征的处理节点对原子a-化学键1进行特征匹配处理,得到目标药物针对药物分子组成特征、药物分子属性特征的药物特征信息。
104、基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息。
本申请实施例中,特征匹配操作即为按照已经确定的药物特征处理流程对分子结构图像分类结果中包含的原子-化学键的分类结果进行一一对应匹配,即基于智能医疗系统中存储全部现存的药物分子组成特征、药物分子属性特征、药物分子结构与原子-化学键的分类结果进行相似度计算,从而确定目标药物的药物特征信息。
在一个本申请实施例中,为了进一步限定及说明,如图2所示,步骤102基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分67子结构图像分类结果之前,所述方法还包括:
201、基于药物分子结构的节点数、邻接矩阵以及特征矩阵构建图卷积网络;
202、确定图拓扑增强的自注意力机制,并将所述自注意力机制引入所述图卷积网络的输入层;
203、基于所述药物分子结构图像数据中解析得到的图拓扑结构,确定辅助信息,并基于所述辅助信息对所述引入所述自注意力机制的图卷积网络的各网络层进行池化;
204、通过分子结构图像样本数据对完成池化后的图卷积网络进行模型训练,得到图像分类模型。
本申请实施例中,为了实现对图像数据进行分类,因此,针对图像数据的特征,构建图卷积网络。如图3所示,一个待进行分类的分子结构图像样本数据为包含图节点和边形式的图像数据,因此,在构建图卷积网络时,基于药物分子结构的节点数、邻接矩阵、特征矩阵进行构建,即图像数据表示为G=(A,X),其中,A和X分别表示邻接矩阵和特征矩阵,图像数据中的图节点的个数为n,节点特征维度为d,则图卷积网络GNN通过图的拓扑结构和特征矩阵,为每个图节点生成用于学习的一个f维的图卷积网络表示H=[h 1,h 2,...,h n] T=GNN(A,X),H∈R n×f。同时,在邻接矩阵A和特征矩阵X的基础上,通过Transformer启发的模块计算一个自注意力分数作为选择标准。为了确保在计算复杂度较强的情况下,本申请实施例中,
确定自注意力机制为四头注意力,即表示为MH(Q,K,V)=[O 1,...,O h]W o;O i=Att(QW i Q,KW i K,VW i V)=Att(QW i Q,KW i K,GCNII i V(H,A)),其中,V、K、Q是固定的单个值,Q、K和V对应的学习参数矩阵为WQ,WK和WV,此外,定义dmodel为多头注意力函数的输出维度。为了通过图卷积网络(Graph Convolutional Network,GCN)模型来计算V,并为了避免过度平滑而局限于图卷积网络GCN浅层体系结构,限制其模型性能的问题,通过初始残差和恒等映射的GCNII(Graph Convolutional Network via Initial residual and Identity mapping)来构建V,以显示化利用全局结构并根据图节点的结构依赖性捕获图节点之间的信息交互,即完成对分子结构节点特征的挖掘。
需要说明的是,在将自注意力机制引入图卷积网络的输入层时,即第l层选择具有可学习得分函数L2Pool的高得分图节点i(l+1)∈R nl+1i(l+1)∈Rnl+1,以删除不需要的图节点,表示为:y (l)=L2Pool(Att,H (l),A (l));i (l+1)=top k(y (l)),其中,函数L2Pool依赖于多头注意力和GCNII增强的拓扑信息,top k()函数通过丢弃得分较低的士节点来对前k个图节点进行采样,以在压缩图像数据的图节点规模的同时,保留尽可能多的信息,以便引入所述自注意力机制的图卷积网络的各网络层进行池化。其中,为了构建多尺度的图像数据,在每一个池化层,通过采样缩减图节点的规模得到一个粗化图像数据,此时,基于采样为图节点计算重要性得分,以保留最重要的前k个图节点以及它们间的连接关系生成粗化图像数据,实现图卷积网络的各网络层的池化。在此过程中,对各层进行池化的池化模型通过药物分子结构图像数据的全局结构信息以及局部结构信息,得到更有效的图表示,,所述辅助信息包括所述药物分子结构图像数据的全局结构信息以及局部结构信息,即通过辅助信息对所述引入自注意力机制的图卷积网络的各网络层进行池化。最后,通过作为模型训练样本的分子结构图像样本数据对完成池化后的图卷积网络进行模型训练,得到图像分类模型。
在一个本申请实施例中,为了进一步限定及说明,步骤202确定图拓扑增强的自注意力机制包括:基于初等残差对所述图卷积网络的输入层构建跳跃连接,并通过恒等映射对权重矩阵进行单位化处理,将单位化处理后的所述图卷积网络与自注意力机制结合,得到图拓扑增强的自注意力机制。
为了避免因度平滑而局限于图卷积网络的浅层体系结构,而限制其模型性能的问题,本申请实施例中,初始残差和恒等映射的GCNII来构建四头注意力中的V。具体的,由于GCNII为带有初始剩余连接和恒等映射的GCN,在图卷积网络中的每一层中,初始残差在输入层构造一个跳跃连接,并通过恒等映射对权重矩阵进行单位化处理,即恒等映射将单位矩阵添加到权重矩阵,以将单位化处理后的图卷积网络与自注意力机制结合,以增加GCNII的网络深度时,防止过度平滑并持续改善GCNII的性能。
其中,图拓扑增强的自注意力机制定义为:GCHII(H,A)=σ(((1-α)AH+αH 0)((1-β)I n)+βW));
Figure PCTCN2022089689-appb-000001
其中,α、β为超参数,I n为GCNII中的单位矩阵,给定图节点嵌入H∈R n×d及其邻接矩阵A,我们使用4层GCNII模型来构造值V,从而实现引入图拓扑信息优化重要的图节点的得分。
需要说明的是,图卷积神经网络的输入为带图节点或连边的图像数据结构,即包括该图像数据的邻接矩阵A和对应的特征属性信息X。图卷积神经网络根据图结构和输入 节点属性,训练图像数据中每个图节点的隐式向量表示,目标是让该向量表示包含足够强大的表达信息,使其能够帮助每个图节点进行信息抽取,最后可以获得整个图的信息向量表示,如对于一个原子和化学键组成的分子图,通过原子节点的特征,和原子之间的连边化学键信息,抽取出整个分子化合物的分子级别的信息表示。图卷积神经网络模型学习的主要过程是通过迭代对图数据中图节点的邻居信息进行聚合和更新。在每一次迭代中,每一个图节点通过聚合邻居节点的特征及自己在上一层的特征来更新自己的信息,通常也会对聚合后的信息进行非线性变换。通过堆叠多层网络,每个图节点可以获取到相应跳数内的邻居节点信息,以基于Transformer自注意力机制和图像数据的网络拓扑信息的新的图粗化方法,如通过粗化池化方法包括节点特征、图拓扑两个特征来实现,本申请实施例不做具体限定。
在一个本申请实施例中,为了进一步限定及说明,步骤103调取与所述分子结构图像分类结果匹配的药物特征处理流程之前,所述方法还包括:接收对所述药物分子结构图像数据触发的分子分类任务。
本申请实施例中,由于药物特征处理流程中包含预先配置的对不同分子结构图像分类结果进行匹配的药物分子组成特征、药物分子属性特征、药物分子结构中至少一个处理节点,为了满足不同药物特征信息匹配的操作需求,使得不同操作者基于智能医疗系统进行自动、灵活地执行药物特征处理流程,预先为药物特征处理流程配置分子分类任务,以便操作者进行触发分子分类任务来执行不同药物分子结构的药物特征处理流程的触发。其中,所述分子分类任务用于表征对不同药物分子结构图像数据执行所述药物特征处理流程中对应的处理节点,即针对不同的处理节点,可以通过不同的分子分类任务进行触发,以便具有针对性的完成对不同分子结构图像分类结果的药物特征处理流程的执行。
对应的,步骤103调取与所述分子结构图像分类结果匹配的药物特征处理流程包括:解析所述分子分类任务中所述药物分子结构图像数据的处理节点,并基于分类处理得到的分子结构图像分类结果与所述处理节点进行匹配;调取与所述分子结构图像分类结果所匹配的至少一个处理节点。
本申请实施例中,所述处理节点为所述药物特征处理流程中执行药物分子组成特征、药物分子属性特征以及药物分子结构匹配中的至少一个,即处理节点包括药物分子组成特征处理节点、药物分子属性特征处理节点、药物分子结构处理节点,此时,由于分子分类任务中带有未进行图像分类处理之前药物分子结构图像数据所预期进行的处理节点,但并不是所有的药物分子结构图像数据在进行图像分类后仍可以按照处理节点进行执行,因此,首先解析分子分类任务中药物分子结构图像数据的处理节点,然后基于分类处理得到的分子结构图像分类结果与处理节点进行匹配,以便调取最终与分子结构图像分类结果所匹配的至少一个处理节点。例如,若目标药物a的分子分类任务中药物分子结构图像数据的处理节点为药物分子组成特征处理节点、药物分子属性特征处理 节点、药物分子结构处理节点,当对药物分子结构图像数据进行分类处理后,得到的分子结构图像分类结果中的原子a-化学键1所对应的处理节点匹配药物分子组成特征处理节点、药物分子结构处理节点,则调取药物分子组成特征处理节点、药物分子结构处理节点,以进行处理节点的执行。其中,对于各分子结构图像分类结果与处理节点之间的匹配,可以通过预先录入一个匹配关系列表,以便准确对不同的分子结构图像分类结果与处理节点进行匹配,本申请实施例对匹配关系列表不做具体限定。
在一个本申请实施例中,为了进一步限定及说明,如图4所示,步骤104基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息包括:
1041、计算所述分子结构图像分类结果分别与所述药物分子组成特征、所述药物分子属性特征以及所述药物分子结构之间的相似度;
1042、若所述分子结构图像分类结果与所述药物分子组成特征的第一相似度大于预设第一相似度阈值,则确定包含所述药物分子组成特征的药物特征信息;和/或,
1043、若所述分子结构图像分类结果与所述药物分子属性特征的第二相似度大于预设第二相似度阈值,则确定包含所述药物分子属性特征的药物特征信息;和/或,
1044、若所述分子结构图像分类结果与所述药物分子结构的第三相似度大于预设第三相似度阈值,则确定包含所述药物分子结构的药物特征信息。
为了使分子结构图像分类结果与药物特征处理流程中的各个处理节点进行特征匹配操作,在调取对应的处理节点,如药物分子组成特征处理节点、药物分子属性特征处理节点、药物分子结构处理节点之后,执行对应的匹配操作。本申请实施例中,对于各个处理节点的匹配操作即为按照各处理节点对分子结构图像分类结果进行相似度计算,从而基于得到的相似度确定目标药物的药物特征信息。本申请实施例中,药物分子组成特征处理节点中,通过将分子结构图像分类结果与智能医疗系统中已存储的药物分子组成特征进行相似度计算,由于分子结构图像分类结果包含原子-化学键的图像数据,具体则通过与药物分子组成特征所对应的图像进行相似度计算,从而判断计算得到的相似度值是否大于第一相似度阈值,若是,则确定目标药物包含药物分子组成特征的药物特征信息。药物分子属性特征处理节点中,通过将分子结构图像分类结果与智能医疗系统中已存储的药物分子属性特征进行相似度计算,由于分子结构图像分类结果包含原子-化学键的图像数据,具体则通过与药物分子属性特征所对应的图像进行相似度计算,从而判断计算得到的相似度值是否大于第二相似度阈值,若是,则确定目标药物包含药物分子属性特征的药物特征信息。药物分子结构处理节点通过将分子结构图像分类结果与智能医疗系统中已存储的药物分子结构进行相似度计算,由于分子结构图像分类结果包含原子-化学键的图像数据,具体则通过与药物分子结构所对应的图像进行相似度计算,从而判断计算得到的相似度值是否大于第三相似度阈值,若是,则确定目标药物包含药物分子结构的药物特征信息。
另外,药物分子组成特征用于表征药物中由不同分子所组成的特征内容,例如,包含环芳烃、甲基等分子组成特征,药物分子属性特征用于表征药物中由不同分子所产生的物理或化学属性,例如,甲苯-易氧化等化学属性,药物分子结构用于表征分子之间的化学键结构,例如,芳香烃具有苯环的基本分子结构,因此,可以通过药物分子组成特征、药物分子属性特征、药物分子结构所对应的处理节点对分子结构图像分类结果进行匹配,本申请实施例中,为了提高匹配的准确性,还可以将药物分子组成特征、药物分子属性特征、药物分子结构进行数值化处理,并同时将分子结构图像分类结果进行数值化处理,从而将待进行相似度计算的药物分子组成特征、药物分子属性特征、药物分子结构进行数与分子结构图像分类结果放入一个数据单位进行相似度计算,完成匹配过程。
在一个本申请实施例中,为了进一步限定及说明,步骤基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息之后,所述方法还包括:获取疾病特征数据库,判断所述药物特征信息与所述疾病特征数据库中的各疾病特征信息是否具有对抗性属性;若具有所述对抗性属性,则输出所述疾病特征信息。
本申请实施例中,为了满足智能医疗系统针对目标药物的药物特征信息的使用需求,在得到药物特征信息后,可以将药物特征信息与各疾病特征信息进行判断是否存在对抗性,以确定是否向用户输出与药物特征信息相关的疾病特征信息,从而进行相关疾病的治疗操作。其中,存储于智能医疗系统中的疾病特征数据库中记录有各种已知的疾病所对应的疾病特征信息,此时,疾病特征信息用于表征疾病对人体所产生的病症内容,例如,某疾病会造成血压高于180mmHg,某疾病会造成肾上腺素极具减少等,此时,对抗性属性即为药物特征信息与疾病特征信息之间是否存在相互起到相反作用的属性,例如,药物特征信息是否对疾病特征信息中的血压低起到升压作用,若存在对抗性属性,则说明此目标药物可以作为此疾病特征信息所对应的疾病的治疗或者缓解症状的作用,已经输出给用户进行查看。
在一个本申请实施例中,为了进一步限定及说明,步骤判断所述药物特征与所述疾病特征数据库中的各疾病特征信息是否具有对抗性属性包括:获取对抗性属性列表;基于所述药物特征信息与所述生物学特征信息、所述化学特征信息所对应的标记判断是否具有对抗性属性。
在判断药物特征信息与疾病特征信息之间是否具有对抗性属性时,具体通过智能医疗系统中的对抗性属性列表,所述对抗性属性列表中记录有不同疾病特征信息的生物学特征信息、化学特征信息,分别与不同药物特效信息之间是否具有医疗关联的标记,此标记即为通过医疗实验所确定的具有医疗关联性的对抗性属性,例如,对抗性属性列表中记载药物分子属性特征a与生物学特征信息的提高细胞接收氧原子速度特征之间存在标记,则说明包含有药物分子属性特征a的药物特征信息的目标药物可以用于治疗细胞 接收氧原子速度较慢的疾病。因此,通过药物特征信息与所述生物学特征信息、所述化学特征信息所对应的标记判断是否具有对抗性属性。其中,不同疾病特征信息的生物学特征信息、化学特征信息分别为用于预先录入至智能医疗系统中的信息,作为各疾病的特征内容,生物学特征信息为疾病所展现的生理或生物学上的症状特征,例如,疼痛、发烧、血项值、细胞量等,化学特征信息为疾病所展现的化学成分上的症状特征,激素值等,本申请实施例不做具体限定。
本申请实施例提供了一种基于人工智能的药物特征信息确定方法,与现有技术相比,本申请实施例通过获取目标药物的药物分子结构图像数据;基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;调取与所述分子结构图像分类结果匹配的药物特征处理流程;基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息,实现基于人工智能的药物特征的确定,大大加快了药物特征的识别准确性,减少人为识别的复杂度以及耗时,使得通过药物分子结构的识别来确定药物特征的速度大大加快,从而提高了药物特征匹配病症的有效性,并实现了一种智能化的药物特征确定。
进一步的,作为对上述图1所示方法的实现,本申请实施例提供了一种基于人工智能的药物特征信息确定装置,如图5所示,该装置包括:
获取模块31,用于获取目标药物的药物分子结构图像数据;
处理模块32,用于基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;
调取模块33,用于调取与所述分子结构图像分类结果匹配的药物特征处理流程;
匹配模块34,用于基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息。
进一步地,所述装置还包括:
构建模块,用于基于药物分子结构的节点数、邻接矩阵以及特征矩阵构建图卷积网络;
确定模块,用于确定图拓扑增强的自注意力机制,并将所述自注意力机制引入所述图卷积网络的输入层,以进行分子结构节点特征的挖掘;
解析模块,用于基于所述药物分子结构图像数据中解析得到的图拓扑结构,确定辅助信息,并基于所述辅助信息对所述引入所述自注意力机制的图卷积网络的各网络层进行池化,所述辅助信息包括所述药物分子结构图像数据的全局结构信息以及局部结构信息;
训练模块,用于通过分子结构图像样本数据对完成池化后的图卷积网络进行模型训练,得到图像分类模型。
进一步地,所述确定模块,具体用于基于初等残差对所述图卷积网络的输入层构建跳跃连接,并通过恒等映射对权重矩阵进行单位化处理,将单位化处理后的所述图卷积网络与自注意力机制结合,得到图拓扑增强的自注意力机制。
进一步地,所述装置还包括:
接收模块,接收对所述药物分子结构图像数据触发的分子分类任务,所述分子分类任务用于表征对不同药物分子结构图像数据执行所述药物特征处理流程中对应的处理节点;
所述调取模块包括:
解析单元,用于解析所述分子分类任务中所述药物分子结构图像数据的处理节点,并基于分类处理得到的分子结构图像分类结果与所述处理节点进行匹配,所述处理节点为所述药物特征处理流程中执行药物分子组成特征、药物分子属性特征以及药物分子结构匹配中的至少一个;
调取单元,用于调取与所述分子结构图像分类结果所匹配的至少一个处理节点。
进一步地,所述所述匹配模块包括:
计算单元,用于计算所述分子结构图像分类结果分别与所述药物分子组成特征、所述药物分子属性特征以及所述药物分子结构之间的相似度;
第一确定单元,用于若所述分子结构图像分类结果与所述药物分子组成特征的第一相似度大于预设第一相似度阈值,则确定包含所述药物分子组成特征的药物特征信息;和/或,
第二确定单元,用于若所述分子结构图像分类结果与所述药物分子属性特征的第二相似度大于预设第二相似度阈值,则确定包含所述药物分子属性特征的药物特征信息;和/或,
第三确定单元,用于若所述分子结构图像分类结果与所述药物分子结构的第三相似度大于预设第三相似度阈值,则确定包含所述药物分子结构的药物特征信息。
进一步地,所述装置还包括:
判断模块,用于获取疾病特征数据库,判断所述药物特征信息与所述疾病特征数据库中的各疾病特征信息是否具有对抗性属性;
输出模块,用于若具有所述对抗性属性,则输出所述疾病特征信息。
进一步地,所述判断模块包括:
获取单元,用于获取对抗性属性列表,所述对抗性属性列表中记录有不同疾病特征信息的生物学特征信息、化学特征信息,分别与不同药物特效信息之间是否具有医疗关联的标记;
判断单元,用于基于所述药物特征信息与所述生物学特征信息、所述化学特征信息 所对应的标记判断是否具有对抗性属性。
本申请实施例提供了一种基于人工智能的药物特征信息确定装置,与现有技术相比,本申请实施例通过获取目标药物的药物分子结构图像数据;基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;调取与所述分子结构图像分类结果匹配的药物特征处理流程;基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息,实现基于人工智能的药物特征的确定,大大加快了药物特征的识别准确性,减少人为识别的复杂度以及耗时,使得通过药物分子结构的识别来确定药物特征的速度大大加快,从而提高了药物特征匹配病症的有效性,并实现了一种智能化的药物特征确定。
根据本申请一个实施例提供了一种计算机可读存储介质,所述存储介质存储有至少一可执行指令,该计算机可执行指令可执行上述任意方法实施例中的基于人工智能的药物特征信息确定方法。所述计算机可读存储介质可以是非易失性,也可以是易失性。
图6示出了根据本申请一个实施例提供的一种计算机设备的结构示意图,本申请具体实施例并不对计算机设备的具体实现做限定。
如图6所示,该计算机设备可以包括:处理器(processor)402、通信接口(Communications Interface)404、存储器(memory)406、以及通信总线408。
其中:处理器402、通信接口404、以及存储器406通过通信总线408完成相互间的通信。
通信接口404,用于与其它设备比如客户端或其它服务器等的网元通信。
处理器402,用于执行程序410,具体可以执行上述基于人工智能的药物特征信息确定方法实施例中的相关步骤。
具体地,程序410可以包括程序代码,该程序代码包括计算机操作指令。
处理器402可能是中央处理器CPU,或者是特定集成电路ASIC(Application Specific Integrated Circuit),或者是被配置成实施本申请实施例的一个或多个集成电路。计算机设备包括的一个或多个处理器,可以是同一类型的处理器,如一个或多个CPU;也可以是不同类型的处理器,如一个或多个CPU以及一个或多个ASIC。
存储器406,用于存放程序410。存储器406可能包含高速RAM存储器,也可能还包括非易失性存储器(non-volatile memory),例如至少一个磁盘存储器。
程序410具体可以用于使得处理器402执行以下操作:
获取目标药物的药物分子结构图像数据;
基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;
调取与所述分子结构图像分类结果匹配的药物特征处理流程;
基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息。
显然,本领域的技术人员应该明白,上述的本申请的各模块或各步骤可以用通用的计算装置来实现,它们可以集中在单个的计算装置上,或者分布在多个计算装置所组成的网络上,可选地,它们可以用计算装置可执行的程序代码来实现,从而,可以将它们存储在存储装置中由计算装置来执行,并且在某些情况下,可以以不同于此处的顺序执行所示出或描述的步骤,或者将它们分别制作成各个集成电路模块,或者将它们中的多个模块或步骤制作成单个集成电路模块来实现。这样,本申请不限制于任何特定的硬件和软件结合。
以上所述仅为本申请的优选实施例而已,并不用于限制本申请,对于本领域的技术人员来说,本申请可以有各种更改和变化。凡在本申请的精神和原则之内,所作的任何修改、等同替换、改进等,均应包括在本申请的保护范围之内。

Claims (20)

  1. 一种基于人工智能的药物特征信息确定方法,其中,包括:
    获取目标药物的药物分子结构图像数据;
    基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;
    调取与所述分子结构图像分类结果匹配的药物特征处理流程;
    基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息。
  2. 根据权利要求1所述的方法,其中,所述基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果之前,所述方法还包括:
    基于药物分子结构的节点数、邻接矩阵以及特征矩阵构建图卷积网络;
    确定图拓扑增强的自注意力机制,并将所述自注意力机制引入所述图卷积网络的输入层,以进行分子结构节点特征的挖掘;
    基于所述药物分子结构图像数据中解析得到的图拓扑结构,确定辅助信息,并基于所述辅助信息对所述引入所述自注意力机制的图卷积网络的各网络层进行池化,所述辅助信息包括所述药物分子结构图像数据的全局结构信息以及局部结构信息;
    通过分子结构图像样本数据对完成池化后的图卷积网络进行模型训练,得到图像分类模型。
  3. 根据权利要求2所述的方法,其中,所述确定图拓扑增强的自注意力机制包括:
    基于初等残差对所述图卷积网络的输入层构建跳跃连接,并通过恒等映射对权重矩阵进行单位化处理,将单位化处理后的所述图卷积网络与自注意力机制结合,得到图拓扑增强的自注意力机制。
  4. 根据权利要求1所述的方法,其中,所述调取与所述分子结构图像分类结果匹配的药物特征处理流程之前,所述方法还包括:
    接收对所述药物分子结构图像数据触发的分子分类任务,所述分子分类任务用于表征对不同药物分子结构图像数据执行所述药物特征处理流程中对应的处理节点;
    所述调取与所述分子结构图像分类结果匹配的药物特征处理流程包括:
    解析所述分子分类任务中所述药物分子结构图像数据的处理节点,并基于分类处理得到的分子结构图像分类结果与所述处理节点进行匹配,所述处理节点为所述药物特征处理流程中执行药物分子组成特征、药物分子属性特征以及药物分子结构匹配中的至少一个;
    调取与所述分子结构图像分类结果所匹配的至少一个处理节点。
  5. 根据权利要求4所述的方法,其中,所述基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息包括:
    计算所述分子结构图像分类结果分别与所述药物分子组成特征、所述药物分子属性特征以及所述药物分子结构之间的相似度;
    若所述分子结构图像分类结果与所述药物分子组成特征的第一相似度大于预设第一相似度阈值,则确定包含所述药物分子组成特征的药物特征信息;和/或,
    若所述分子结构图像分类结果与所述药物分子属性特征的第二相似度大于预设第二相似度阈值,则确定包含所述药物分子属性特征的药物特征信息;和/或,
    若所述分子结构图像分类结果与所述药物分子结构的第三相似度大于预设第三相似度阈值,则确定包含所述药物分子结构的药物特征信息。
  6. 根据权利要求1所述的方法,其中,所述基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息之后,所述方法还包括:
    获取疾病特征数据库,判断所述药物特征信息与所述疾病特征数据库中的各疾病特征信息是否具有对抗性属性;
    若具有所述对抗性属性,则输出所述疾病特征信息。
  7. 根据权利要求6所述的方法,其中,所述判断所述药物特征与所述疾病特征数据库中的各疾病特征信息是否具有对抗性属性包括:
    获取对抗性属性列表,所述对抗性属性列表中记录有不同疾病特征信息的生物学特征信息、化学特征信息,分别与不同药物特效信息之间是否具有医疗关联的标记;
    基于所述药物特征信息与所述生物学特征信息、所述化学特征信息所对应的标记判断是否具有对抗性属性。
  8. 一种基于人工智能的药物特征信息确定装置,其中,包括:
    获取模块,用于获取目标药物的药物分子结构图像数据;
    处理模块,用于基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;
    调取模块,用于调取与所述分子结构图像分类结果匹配的药物特征处理流程;
    匹配模块,用于基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息。
  9. 一种计算机可读存储介质,其上存储有计算机可读指令,其中,所述计算机可读指令被处理器执行时实现基于人工智能的药物特征信息确定方法,包括:
    获取目标药物的药物分子结构图像数据;
    基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到 分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;
    调取与所述分子结构图像分类结果匹配的药物特征处理流程;
    基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息。
  10. 根据权利要求9所述的计算机可读存储介质,其中,所述计算机可读指令被处理器执行时实现基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果之前,所述方法还包括:
    基于药物分子结构的节点数、邻接矩阵以及特征矩阵构建图卷积网络;
    确定图拓扑增强的自注意力机制,并将所述自注意力机制引入所述图卷积网络的输入层,以进行分子结构节点特征的挖掘;
    基于所述药物分子结构图像数据中解析得到的图拓扑结构,确定辅助信息,并基于所述辅助信息对所述引入所述自注意力机制的图卷积网络的各网络层进行池化,所述辅助信息包括所述药物分子结构图像数据的全局结构信息以及局部结构信息;
    通过分子结构图像样本数据对完成池化后的图卷积网络进行模型训练,得到图像分类模型。
  11. 根据权利要求10所述的计算机可读存储介质,其中,所述计算机可读指令被处理器执行时实现确定图拓扑增强的自注意力机制包括:
    基于初等残差对所述图卷积网络的输入层构建跳跃连接,并通过恒等映射对权重矩阵进行单位化处理,将单位化处理后的所述图卷积网络与自注意力机制结合,得到图拓扑增强的自注意力机制。
  12. 根据权利要求9所述的计算机可读存储介质,其中,所述计算机可读指令被处理器执行时实现调取与所述分子结构图像分类结果匹配的药物特征处理流程之前,所述方法还包括:
    接收对所述药物分子结构图像数据触发的分子分类任务,所述分子分类任务用于表征对不同药物分子结构图像数据执行所述药物特征处理流程中对应的处理节点;
    所述调取与所述分子结构图像分类结果匹配的药物特征处理流程包括:
    解析所述分子分类任务中所述药物分子结构图像数据的处理节点,并基于分类处理得到的分子结构图像分类结果与所述处理节点进行匹配,所述处理节点为所述药物特征处理流程中执行药物分子组成特征、药物分子属性特征以及药物分子结构匹配中的至少一个;
    调取与所述分子结构图像分类结果所匹配的至少一个处理节点。
  13. 根据权利要求12所述的计算机可读存储介质,其中,所述计算机可读指令被处理器执行时实现基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息包括:
    计算所述分子结构图像分类结果分别与所述药物分子组成特征、所述药物分子属性特征以及所述药物分子结构之间的相似度;
    若所述分子结构图像分类结果与所述药物分子组成特征的第一相似度大于预设第一相似度阈值,则确定包含所述药物分子组成特征的药物特征信息;和/或,
    若所述分子结构图像分类结果与所述药物分子属性特征的第二相似度大于预设第二相似度阈值,则确定包含所述药物分子属性特征的药物特征信息;和/或,
    若所述分子结构图像分类结果与所述药物分子结构的第三相似度大于预设第三相似度阈值,则确定包含所述药物分子结构的药物特征信息。
  14. 根据权利要求9所述的计算机可读存储介质,其中,所述计算机可读指令被处理器执行时实现基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息之后,所述方法还包括:
    获取疾病特征数据库,判断所述药物特征信息与所述疾病特征数据库中的各疾病特征信息是否具有对抗性属性;
    若具有所述对抗性属性,则输出所述疾病特征信息。
  15. 一种计算机设备,包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机可读指令,其中,所述计算机可读指令被处理器执行时实现基于人工智能的药物特征信息确定方法,包括:
    获取目标药物的药物分子结构图像数据;
    基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果,所述图像分类模型为基于自注意力机制挖掘分子结构节点特征并通过图拓扑结构的辅助信息对图卷积网络进行分层池化后,完成训练得到的;
    调取与所述分子结构图像分类结果匹配的药物特征处理流程;
    基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息。
  16. 根据权利要求15所述的计算机设备,其中,所述计算机可读指令被处理器执行时实现基于已完成训练的图像分类模型对所述药物分子结构图像数据进行分类处理,得到分子结构图像分类结果之前,所述方法还包括:
    基于药物分子结构的节点数、邻接矩阵以及特征矩阵构建图卷积网络;
    确定图拓扑增强的自注意力机制,并将所述自注意力机制引入所述图卷积网络的输入层,以进行分子结构节点特征的挖掘;
    基于所述药物分子结构图像数据中解析得到的图拓扑结构,确定辅助信息,并基于所述辅助信息对所述引入所述自注意力机制的图卷积网络的各网络层进行池化,所述辅助信息包括所述药物分子结构图像数据的全局结构信息以及局部结构信息;
    通过分子结构图像样本数据对完成池化后的图卷积网络进行模型训练,得到图像分类模型。
  17. 根据权利要求16所述的计算机设备,其中,所述计算机可读指令被处理器执行时实现确定图拓扑增强的自注意力机制包括:
    基于初等残差对所述图卷积网络的输入层构建跳跃连接,并通过恒等映射对权重矩阵进行单位化处理,将单位化处理后的所述图卷积网络与自注意力机制结合,得到图拓扑增强的自注意力机制。
  18. 根据权利要求15所述的计算机设备,其中,所述计算机可读指令被处理器执行时实现调取与所述分子结构图像分类结果匹配的药物特征处理流程之前,所述方法还包括:
    接收对所述药物分子结构图像数据触发的分子分类任务,所述分子分类任务用于表征对不同药物分子结构图像数据执行所述药物特征处理流程中对应的处理节点;
    所述调取与所述分子结构图像分类结果匹配的药物特征处理流程包括:
    解析所述分子分类任务中所述药物分子结构图像数据的处理节点,并基于分类处理得到的分子结构图像分类结果与所述处理节点进行匹配,所述处理节点为所述药物特征处理流程中执行药物分子组成特征、药物分子属性特征以及药物分子结构匹配中的至少一个;
    调取与所述分子结构图像分类结果所匹配的至少一个处理节点。
  19. 根据权利要求18所述的计算机设备,其中,所述计算机可读指令被处理器执行时实现基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息包括:
    计算所述分子结构图像分类结果分别与所述药物分子组成特征、所述药物分子属性特征以及所述药物分子结构之间的相似度;
    若所述分子结构图像分类结果与所述药物分子组成特征的第一相似度大于预设第一相似度阈值,则确定包含所述药物分子组成特征的药物特征信息;和/或,
    若所述分子结构图像分类结果与所述药物分子属性特征的第二相似度大于预设第二相似度阈值,则确定包含所述药物分子属性特征的药物特征信息;和/或,
    若所述分子结构图像分类结果与所述药物分子结构的第三相似度大于预设第三相似度阈值,则确定包含所述药物分子结构的药物特征信息。
  20. 根据权利要求15所述的计算机设备,其中,所述计算机可读指令被处理器执行时实现基于所述药物特征处理流程对所述分子结构图像分类结果执行特征匹配操作,得到所述目标药物的药物特征信息之后,所述方法还包括:
    获取疾病特征数据库,判断所述药物特征信息与所述疾病特征数据库中的各疾病特征信息是否具有对抗性属性;
    若具有所述对抗性属性,则输出所述疾病特征信息。
PCT/CN2022/089689 2022-01-11 2022-04-27 基于人工智能的药物特征信息确定方法及装置 WO2023134061A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210026435.8A CN114417986A (zh) 2022-01-11 2022-01-11 基于人工智能的药物特征信息确定方法及装置
CN202210026435.8 2022-01-11

Publications (1)

Publication Number Publication Date
WO2023134061A1 true WO2023134061A1 (zh) 2023-07-20

Family

ID=81272180

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/089689 WO2023134061A1 (zh) 2022-01-11 2022-04-27 基于人工智能的药物特征信息确定方法及装置

Country Status (2)

Country Link
CN (1) CN114417986A (zh)
WO (1) WO2023134061A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117391150A (zh) * 2023-12-07 2024-01-12 之江实验室 一种基于分层池化图哈希的图数据检索模型训练方法

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117116384B (zh) * 2023-10-20 2024-01-09 聊城高新生物技术有限公司 一种靶向诱导的医药分子结构生成方法
CN117163355B (zh) * 2023-11-02 2024-01-02 济群医药科技(启东)有限公司 一种药物包装泡罩自动化生成方法及系统

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107657146A (zh) * 2017-09-20 2018-02-02 广州市爱菩新医药科技有限公司 基于三维子结构的药物分子比较方法
CN108304882A (zh) * 2018-02-07 2018-07-20 腾讯科技(深圳)有限公司 一种图像分类方法、装置及服务器、用户终端、存储介质
US20200104729A1 (en) * 2018-09-27 2020-04-02 Babylon Partners Limited Method and system for extracting information from graphs
CN111429977A (zh) * 2019-09-05 2020-07-17 中国海洋大学 一种新的基于图结构注意力的分子相似性搜索算法
CN111860584A (zh) * 2020-06-11 2020-10-30 石家庄铁路职业技术学院 图分类方法和装置
CN113011282A (zh) * 2021-02-26 2021-06-22 腾讯科技(深圳)有限公司 图数据处理方法、装置、电子设备及计算机存储介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107657146A (zh) * 2017-09-20 2018-02-02 广州市爱菩新医药科技有限公司 基于三维子结构的药物分子比较方法
CN108304882A (zh) * 2018-02-07 2018-07-20 腾讯科技(深圳)有限公司 一种图像分类方法、装置及服务器、用户终端、存储介质
US20200104729A1 (en) * 2018-09-27 2020-04-02 Babylon Partners Limited Method and system for extracting information from graphs
CN111429977A (zh) * 2019-09-05 2020-07-17 中国海洋大学 一种新的基于图结构注意力的分子相似性搜索算法
CN111860584A (zh) * 2020-06-11 2020-10-30 石家庄铁路职业技术学院 图分类方法和装置
CN113011282A (zh) * 2021-02-26 2021-06-22 腾讯科技(深圳)有限公司 图数据处理方法、装置、电子设备及计算机存储介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117391150A (zh) * 2023-12-07 2024-01-12 之江实验室 一种基于分层池化图哈希的图数据检索模型训练方法
CN117391150B (zh) * 2023-12-07 2024-03-12 之江实验室 一种基于分层池化图哈希的图数据检索模型训练方法

Also Published As

Publication number Publication date
CN114417986A (zh) 2022-04-29

Similar Documents

Publication Publication Date Title
WO2023134061A1 (zh) 基于人工智能的药物特征信息确定方法及装置
WO2022068196A1 (zh) 跨模态的数据处理方法、装置、存储介质以及电子装置
WO2023029351A1 (zh) 基于自监督学习的药物小分子性质预测方法、装置及设备
CN113707297B (zh) 医疗数据的处理方法、装置、设备及存储介质
CN110162593A (zh) 一种搜索结果处理、相似度模型训练方法及装置
CN110175628A (zh) 一种基于自动搜索与知识蒸馏的神经网络剪枝的压缩算法
WO2021139191A1 (zh) 数据标注的方法以及数据标注的装置
WO2023134062A1 (zh) 基于人工智能的药物靶点作用关系确定方法及装置
CN110765117A (zh) 欺诈识别方法、装置、电子设备及计算机可读存储介质
CN111563192B (zh) 实体对齐方法、装置、电子设备及存储介质
CN112035511A (zh) 基于医学知识图谱的目标数据搜索方法及相关设备
CN108154198A (zh) 知识库实体归一方法、系统、终端和计算机可读存储介质
Crescenzi et al. A framework for learning web wrappers from the crowd
CN113987155B (zh) 一种融合知识图谱与大规模用户日志的会话式检索方法
CN115238173B (zh) 基于大数据的行为分析与医疗服务推送方法、设备及介质
CN110929806A (zh) 基于人工智能的图片处理方法、装置及电子设备
CN114550847A (zh) 基于图卷积神经网络的药物口服利用度及毒性预测方法
CN114358001A (zh) 诊断结果的标准化方法及其相关装置、设备和存储介质
CN113761220A (zh) 信息获取方法、装置、设备及存储介质
CN114822683A (zh) 药物与靶标的相互作用预测方法、装置、设备及存储介质
CN115564017A (zh) 模型数据处理方法、电子设备及计算机存储介质
CN113705698B (zh) 基于点击行为预测的信息推送方法及装置
CN113378938A (zh) 一种基于边Transformer图神经网络的小样本图像分类方法及系统
CN113569018A (zh) 问答对挖掘方法及装置
Xiao et al. Complementary relevance feedback-based content-based image retrieval