CN114724622A - Medicine interaction prediction method and device based on medicine knowledge graph - Google Patents
Medicine interaction prediction method and device based on medicine knowledge graph Download PDFInfo
- Publication number
- CN114724622A CN114724622A CN202210272634.7A CN202210272634A CN114724622A CN 114724622 A CN114724622 A CN 114724622A CN 202210272634 A CN202210272634 A CN 202210272634A CN 114724622 A CN114724622 A CN 114724622A
- Authority
- CN
- China
- Prior art keywords
- subgraph
- graph
- drug
- detected
- medicine
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B15/00—ICT specially adapted for analysing two-dimensional or three-dimensional molecular structures, e.g. structural or functional relations or structure alignment
- G16B15/30—Drug targeting using structural data; Docking or binding prediction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
Abstract
The invention provides a medicine mutual reaction prediction method and a medicine mutual reaction prediction device based on a medicine knowledge graph, wherein the medicine mutual reaction prediction method based on the medicine knowledge graph comprises the following steps: acquiring a drug interaction data set of a drug pair to be detected; wherein the drug interaction dataset is used to represent relationships between different drugs; acquiring a direction subgraph corresponding to the to-be-detected drug pair based on the drug interaction data set and a preset medical knowledge graph; modifying the direction subgraph through graph structure learning to obtain a first subgraph corresponding to the drug pair to be detected; and acquiring an explanatory path between the to-be-detected drug pairs based on the first sub-graph. The method can explain the interaction between the medicines, and can improve the accuracy and the efficiency of the prediction of the interaction between the medicines.
Description
Technical Field
The invention relates to the technical field of knowledge graphs, in particular to a medicine interaction prediction method and device based on a medicine knowledge graph.
Background
Node relation prediction (relation prediction) of a Knowledge graph (Knowledge graph) is a common Knowledge graph using means, and the node relation prediction (relation prediction) can effectively utilize the existing content of the Knowledge graph to predict new Knowledge.
In a knowledge map related to medicine, the labeling of the interaction relationship of the drugs can effectively help to deduce a new interaction relationship of the drugs, however, most of the existing methods for predicting the interaction relationship of the drugs can only predict the specific type of the interaction between the drugs, but cannot explain the interaction between the drugs.
Disclosure of Invention
The invention provides a medicine interaction prediction method and device based on a medicine knowledge graph, which are used for solving the defect that the interaction among medicines cannot be explained when medicine reaction prediction is carried out based on the medicine knowledge graph in the prior art.
The invention provides a medicine interaction prediction method based on a medicine knowledge graph, which comprises the following steps: acquiring a drug interaction data set of a drug pair to be detected; wherein the drug interaction dataset is used to represent relationships between different drugs; acquiring a direction subgraph corresponding to the to-be-detected drug pair based on the drug interaction data set and a preset medical knowledge graph; modifying the direction subgraph through graph structure learning to obtain a first subgraph corresponding to the drug pair to be detected; and acquiring an explanatory path between the to-be-detected drug pairs based on the first sub-graph.
According to the medicine interaction prediction method based on the medicine knowledge graph, the direction subgraph corresponding to the medicine pair to be detected is obtained based on the medicine interaction data set and the preset medicine knowledge graph, and the method comprises the following steps: combining the drug interaction data set and the preset medicine knowledge graph to obtain a second sub-graph; performing graph embedding learning on the second subgraph by using a first graph neural network to obtain an embedded representation of each node in the second subgraph; extracting a 2-hop subgraph corresponding to a first drug in the drug pair to be detected and a 2-hop subgraph corresponding to a second drug in the drug pair to be detected from a second subgraph which completes graph embedding learning; extracting an intersection graph of the 2-hop subgraph corresponding to the first medicine and the 2-hop subgraph corresponding to the second medicine to obtain a third subgraph; and acquiring a direction sub-graph corresponding to the to-be-detected drug pair based on the third sub-graph.
According to the medicine interaction prediction method based on the medicine knowledge graph, the direction subgraph corresponding to the medicine pair to be detected is obtained based on the medicine interaction data set and the preset medicine knowledge graph, and the method comprises the following steps: combining the drug interaction data set and the preset medicine knowledge graph to obtain a second sub-graph; extracting a 2-hop subgraph corresponding to a first drug in the drug pair to be detected and a 2-hop subgraph corresponding to a second drug in the drug pair to be detected from the second subgraph; extracting an intersection graph of the 2-hop subgraph corresponding to the first medicine and the 2-hop subgraph corresponding to the second medicine to obtain a third subgraph; and acquiring a direction sub-graph corresponding to the to-be-detected drug pair based on the third sub-graph.
According to the medicine interaction prediction method based on the medicine knowledge graph, after the direction subgraph corresponding to the medicine pair to be detected is obtained, the method further comprises the following steps: and carrying out graph embedding learning on the directional subgraph by using a first graph neural network to obtain an embedded representation of each node in the directional subgraph.
According to the medicine interaction prediction method based on the medicine knowledge graph, the direction subgraph is modified through graph structure learning to obtain a first subgraph corresponding to the medicine pair to be detected, and the method comprises the following steps: and modifying the direction subgraph through graph structure learning based on the embedded representation of each node in the direction subgraph to obtain a first subgraph corresponding to the to-be-detected drug pair.
According to the medicine interaction prediction method based on the medicine knowledge graph, the direction subgraph is modified through graph structure learning based on the embedded expression of each node in the direction subgraph, and a first subgraph corresponding to the medicine pair to be detected is obtained, and the method comprises the following steps: adding edges between each node of the directional subgraph and the neighbor node of each node; based on the embedded representation of each node in the directional subgraph, calculating the weight of edges between different nodes in the directional subgraph by using a preset edge weight calculation function; modifying the structure of the directional subgraph based on the weight of edges between different nodes and a preset threshold value to obtain a first subgraph corresponding to the drug pair to be detected.
According to the medicine interaction prediction method based on the medicine knowledge graph, the interpretive path between the medicine pair to be detected is obtained based on the first sub-graph, and the method comprises the following steps: performing graph embedding learning on the first subgraph by using a second graph neural network, and updating an embedded representation of each node in the first subgraph; based on the embedded representation of each node in the first subgraph, averaging the embedded representations of all nodes in the first subgraph to obtain the embedded representation of the first subgraph; the embedded representation of the first subgraph, the embedded representation of a node corresponding to a first drug in the drug pair to be detected in the first subgraph and the embedded representation of a node corresponding to a second drug in the drug pair to be detected in the first subgraph are spliced to obtain a splicing result; and predicting the splicing result by utilizing the full-connection layer to obtain an explanatory path between the drug pairs to be detected.
The invention also provides a medicine mutual reaction prediction device based on the medicine knowledge graph, which comprises the following components:
the data acquisition module is used for acquiring a drug interaction data set of the drug pair to be detected; wherein the drug interaction dataset is used to represent relationships between different drugs;
the subgraph extraction module is used for acquiring a direction subgraph corresponding to the to-be-detected medicine pair based on the medicine interaction data set and a preset medicine knowledge graph;
the sub-graph modification module is used for modifying the direction sub-graph through graph structure learning to obtain a first sub-graph corresponding to the to-be-detected drug pair;
and the path acquisition module is used for acquiring an explanatory path between the to-be-detected medicine pairs based on the first sub-graph.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize the medicine interaction prediction method based on the medicine knowledge graph.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method for predicting drug interactions based on a medical knowledge-graph as described in any one of the above.
The present invention also provides a computer program product comprising a computer program which, when executed by a processor, implements a method for predicting drug interactions based on a medical knowledge-graph as described in any of the above.
According to the medicine interaction prediction method and device based on the medicine knowledge graph, firstly, the direction subgraph corresponding to the medicine pair to be detected is obtained through the medicine interaction data set and the preset medicine knowledge graph, some non-joint points in the medicine knowledge graph can be deleted, the calculation speed is accelerated to a certain extent, the prediction efficiency is improved, the direction subgraph is modified through graph structure learning, the first subgraph corresponding to the medicine pair to be detected is obtained, the subgraph structure can be dynamically modified, a better graph structure is obtained, the prediction accuracy is improved, finally, an explanatory path between the medicine pair to be detected is obtained based on the first subgraph, and the interaction between the medicines is explained.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for predicting drug interactions based on a medical knowledge graph according to the present invention;
FIG. 2 is one of the schematic flow charts for obtaining a directional subgraph corresponding to a drug to be detected according to the present invention;
fig. 3 is a second schematic flow chart for obtaining a directional sub-diagram corresponding to a drug to be detected according to the embodiment of the present invention;
FIG. 4 is a schematic flow chart of the present invention for obtaining an explanatory path between pairs of drugs to be tested;
FIG. 5 is a process schematic of the method of drug interaction based on medical knowledge mapping provided by the present invention;
FIG. 6 is a schematic diagram of the structure of the device for predicting drug interactions in a medical knowledge map according to the present invention;
fig. 7 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The terms first, second and the like in the description and in the claims of the present invention are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It will be appreciated that the data so used may be interchanged under appropriate circumstances such that embodiments of the invention may be practiced other than those illustrated or described herein, and that the objects identified as "first," "second," etc. are generally a class of objects and do not limit the number of objects, e.g., a first object may be one or more. In addition, "and/or" in the specification and claims means at least one of connected objects, a character "/" generally means that a preceding and succeeding related objects are in an "or" relationship.
It should be appreciated that reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
The method for predicting drug interactions based on medical knowledge maps according to the present invention is described below with reference to fig. 1 to 4.
FIG. 1 is a flow chart of the method for predicting drug interactions based on the medical knowledge-graph of the present invention. As shown in fig. 1, the method includes:
Wherein the drug interaction data set is used to represent a relationship between different drugs.
It is noted that the drug pair to be detected includes a first drug and a second drug.
The drug interaction data set of the drug pair to be detected provides drug information related to the first drug and the second drug, and can represent various relationships between the first drug and the second drug to a certain extent, but the information provided by the data set is not complete enough in deletion; the drug interaction data set can be represented by a graph-type data structure, which can be represented as a directed graph, with the direction reflecting whether the first drug has an enhancing effect on the second drug or the first drug has an attenuating effect on the second drug.
In the embodiment of the present invention, the final objective is to predict the relationship between the first drug and the second drug, i.e. the type of the interaction between the drug pairs to be detected, and obtain a most explanatory path between the drug pairs to be detected.
It can be appreciated that the electronic device obtains drug interaction data for the drug pair to be detected.
And 102, acquiring a direction subgraph corresponding to the to-be-detected drug pair based on the drug interaction data set and a preset medical knowledge graph.
It should be noted that a knowledge graph is a graph-type data structure, in which each node represents an entity and an edge represents a relationship, and its essence is a directed graph composed of nodes and edges. Typically, a knowledge-graph is maintained as a triplet, i.e., (h, R, t), where h e is the head entity, R e is the relationship, and t e is the tail entity.
The preset medicine knowledge graph is an existing medicine knowledge graph which is comprehensive in medicine information coverage and high in authority, and medicine information related to the first medicine and the second medicine can be found in the medicine knowledge graph.
Generally, the medical knowledge map includes information about drugs, proteins, diseases, genes, and the like.
It should be understood that the directional subgraph is a directed graph, in which the node corresponding to the first drug is taken as a head node and the node corresponding to the second drug is taken as a tail node, where the head node has only an output edge and no input edge, i.e., the in-degree of the head node is 0, and the tail node has only an input edge and no output edge, i.e., the out-degree of the tail node is 0.
It can be understood that the electronic device can acquire the direction subgraph corresponding to the to-be-detected drug pair through the drug interaction data set of the to-be-detected drug pair and the preset drug knowledge graph.
And 103, modifying the direction subgraph through graph structure learning to obtain a first subgraph corresponding to the drug pair to be detected.
It should be noted that Graph Structure Learning (GSL) refers to optimizing and/or refining an original Graph Structure to obtain a better Graph Structure.
In the embodiment of the invention, the graph structure learning model needs to be trained by using a training data set in advance, and the parameters of the graph structure learning model are continuously updated by using the cross entropy loss function until the training is successful. The input of the graph structure learning model is directed subgraph obtained based on the interaction data set of the drug pairs and the preset medical knowledge graph, the output of the graph structure learning model is a new subgraph, and the output subgraph structure is better than that of the directional subgraph.
It should be understood that modifying the direction subgraph through graph structure learning refers to adding or deleting some edges to the direction subgraph through graph structure learning.
The direction subgraph is modified because the medicine information in the direction subgraph is from a preset medicine knowledge graph and/or a medicine interaction data set, the preset medicine knowledge graph is very high in degree, a large number of marked medicine interactions are irrelevant to the prediction of the relation between the medicine pairs to be detected, and irrelevant edges also exist in the direction subgraph, so that how different medicines react can be released by utilizing the direction subgraph in a non-definite way, the edges in the direction subgraph can be deleted or added through graph structure learning, and the modified direction subgraph can better explain the reason of the reaction between different medicines.
It will be appreciated that the first sub-graph is derived from adding or deleting edges on the basis of the directional sub-graph.
It can be understood that the electronic device inputs the direction subgraph corresponding to the drug pair to be detected into the graph structure learning model, and obtains the first subgraph corresponding to the drug pair to be detected through graph structure learning.
And 104, acquiring an explanatory path between the to-be-detected drug pairs based on the first subgraph.
It is understood that the explanatory path between the pair of the drugs to be detected can be used to know why the first drug reacts with the second drug, for example, because the first drug reacts with the second drug due to protein, gene, disease, etc.
It can be understood that the electronic device acquires an explanatory path between the to-be-detected drug pairs based on the first sub-graph.
In the embodiment of the invention, firstly, the direction subgraph corresponding to the drug pair to be detected is obtained through the drug interaction data set and the preset drug knowledge graph, some non-joint points in the drug knowledge graph can be deleted, the calculation speed is accelerated to a certain extent, the prediction efficiency is improved, then the direction subgraph is modified through graph structure learning, the first subgraph corresponding to the drug pair to be detected is obtained, the subgraph structure can be dynamically modified, a better graph structure is obtained, the prediction accuracy is improved, and finally, an explanatory path between the drug pair to be detected is obtained based on the first subgraph, and the interaction between the drugs is explained.
Fig. 2 is one of the schematic flow charts for obtaining a directional subgraph corresponding to a drug to be detected according to the present invention. As shown in fig. 2:
optionally, the step 102 of obtaining a direction sub-graph corresponding to the to-be-detected drug pair based on the drug interaction data set and a preset medical knowledge graph includes: step 1021, step 1022, step 1023, step 1024, and step 1025.
it will be appreciated that the combination of the drug interaction data set and the predetermined medical knowledge map is intended to obtain more comprehensive drug information relating to the first drug and the second drug.
It can be understood that, since the drug interaction dataset is a directed graph, the preset medical knowledge graph is also a directed graph, and then the second sub-graph obtained after merging is also a directed graph.
it should be noted that the embedded representation of the graph nodes, i.e., the representation of the graph nodes, is the key to the success of the task of graph node prediction and edge prediction, and when the second sub-graph does not perform graph embedding learning, the representation of each node is initialized randomly, and the mutual reaction between drug pairs cannot be predicted successfully.
Wherein the first graph neural network may be one of: relational Graph Convolutional Network (RGCN), Graph Convolutional neural Network (GCN), Graph Attention Network (GAT), Graph Isomorphic Network (GIN).
In the embodiment of the invention, the first graph neural network needs to be trained by using a training data set in advance, and the parameters of the first graph neural network are continuously updated by using the cross entropy loss function until the training is successful. The input of the first graph neural network is a graph obtained by combining the drug interaction data set and the preset medicine knowledge graph, and the output of the first graph neural network is an embedded representation of each node in the combined graph.
In one embodiment, the second sub-graph is graph-embedding learned using RGCN, with the formula:
wherein the content of the first and second substances,
an embedded representation representing a level l node v,
representing an embedded representation of layer l-1 node u,
a weight matrix representing the node v itself at the l-th level, the weight matrix being initialized randomly when the first graph neural network is untrained, the weight matrix being adjusted continuously with continuous training of the first graph neural network, NvAll the neighbor nodes of the node v in the subgraph, namely the nodes which are directly connected with the node v by one edge,
a hidden representation representing the relationship between nodes u, v.
It should be understood that after the graph embedding learning is performed on the second sub-graph, each node in the second sub-graph can fuse the information of the neighboring nodes into its own characterization to reflect the connection relationship with the neighboring nodes, but the embedded representation of each node in the second sub-graph is only a preliminary characterization of each node, and since there are also unrelated nodes and edges in the second sub-graph, the interaction between the drug pairs cannot be accurately explained by using the preliminary characterization of the node.
It is to be understood that graph embedding learning does not change the structure of the graph, but only the characterization of the nodes in the graph.
1023, extracting a 2-hop subgraph corresponding to a first drug in the to-be-detected drug pair and a 2-hop subgraph corresponding to a second drug in the to-be-detected drug pair from a second subgraph in which graph embedding learning is completed;
it should be noted that the K-hop subgraph is also referred to as a K-hop subgraph or a K-neighbor subgraph, where K is a positive integer greater than or equal to 1. The K-neighbor operation refers to finding a set of all vertices whose shortest path is K hops (or K steps) from a vertex.
The specific mode is that in a second sub-graph which finishes graph embedding learning, nodes corresponding to the first medicine are taken as vertexes, all nodes directly connected with the vertexes through one edge are found out firstly, namely all neighbor nodes of the vertexes are found out, all nodes directly connected with the neighbor nodes through one edge are found out, namely neighbor nodes of all the neighbor nodes are found out, the vertexes, the neighbor nodes and edges between the vertexes and the neighbor nodes are extracted from the second sub-graph which finishes graph embedding learning, and the 2-hop sub-graph corresponding to the first medicine is obtained.
In a second sub-graph after the graph embedding learning is finished, taking a node corresponding to the second medicine as a vertex, firstly finding out all nodes directly connected with the vertex by an edge, namely finding out all neighbor nodes of the vertex, then finding out all nodes directly connected with the neighbor nodes by an edge, namely finding out all neighbor nodes of all the neighbor nodes, and extracting the vertex, the neighbor nodes of the neighbor nodes and edges among the neighbor nodes from the second sub-graph after the graph embedding learning is finished to obtain the 2-hop sub-graph corresponding to the second medicine.
it should be understood that intersecting the two 2-hop graphs, i.e., only retaining nodes and edges that appear in both 2-hop subgraphs. And (4) forming a closed subgraph, namely a third subgraph, by the intersection graph, the nodes corresponding to the first medicament and the nodes corresponding to the second medicament.
And 1025, acquiring a direction subgraph corresponding to the to-be-detected drug pair based on the third subgraph.
It should be noted that the third sub-graph is also a directed graph, so that the sub-graph is directionally filtered based on the third sub-graph obtaining direction, and the specific manner is as follows: and taking the node corresponding to the first drug in the third subgraph as a head node and the node corresponding to the second drug as a tail node, acquiring all paths pointing to the tail node from the head node and having path lengths between the head node and the tail node smaller than a preset length, merging edges contained in the paths, and acquiring the directional subgraph, wherein the preset length can be defined according to actual needs.
The purpose of obtaining the direction subgraph is to explain information transfer from the node corresponding to the first medicine to the node corresponding to the second medicine, and meanwhile, nodes and edges which are irrelevant to explaining interaction between the first medicine and the second medicine in the second subgraph are deleted, so that the calculation speed is increased to a certain extent, and in general, one third of the nodes without joints can be deleted.
It will be appreciated that the nodes in the directional subgraph still retain the embedded representation of the nodes.
In the embodiment of the invention, firstly, the medicine interaction data set is merged with the preset medicine knowledge graph to obtain a second subgraph, so that more comprehensive medicine information is obtained, then, the first graph neural network is used for graph embedding learning on the second subgraph to obtain the embedded representation of each node in the second subgraph, secondly, 2-hop subgraphs respectively corresponding to the first medicine and the second medicine are extracted, some irrelevant nodes and edges can be deleted preliminarily, the intersection graph of the 2-hop subgraph corresponding to the first medicine and the 2-hop subgraph corresponding to the second medicine is extracted again to obtain a third subgraph, the irrelevant nodes and edges can be deleted continuously in the process, and finally, based on the third subgraph, the third subgraph is filtered in the direction to obtain the direction subgraph corresponding to the medicine to be detected, so that the calculation speed can be increased to a certain degree, and the prediction efficiency is improved.
Fig. 3 is a second schematic flow chart for obtaining a directional diagram corresponding to a drug to be detected according to an embodiment of the present invention. As shown in fig. 3:
in another embodiment, the step 102 of obtaining a direction sub-graph corresponding to the to-be-detected drug pair based on the drug interaction data set and a preset medical knowledge graph includes: step 1026, step 1027, step 1028, and step 1029.
step 1027, extracting a 2-hop subgraph corresponding to a first drug in the drug pair to be detected and a 2-hop subgraph corresponding to a second drug in the drug pair to be detected from the second subgraph;
and step 1029, acquiring a direction subgraph corresponding to the to-be-detected drug pair based on the third subgraph.
It should be understood that, in the process of extracting the corresponding direction subgraph of the detected drug pair based on the second subgraph, graph embedding learning can also be performed on the second subgraph without using the first graph neural network.
In the embodiment of the invention, the drug interaction data set is merged with the preset drug knowledge graph to obtain the second sub-graph, so that more comprehensive drug information is obtained, the 2-hop sub-graphs corresponding to the first drug and the second drug respectively are extracted, some irrelevant nodes and sides can be deleted preliminarily, the intersection graph of the 2-hop sub-graph corresponding to the first drug and the 2-hop sub-graph corresponding to the second drug is extracted again to obtain the third sub-graph, the irrelevant nodes and sides can be deleted continuously in the process, and finally the third sub-graph is filtered in the direction based on the third sub-graph to obtain the direction sub-graph corresponding to the drug to be detected, so that the calculation speed can be accelerated to a certain extent, and the prediction efficiency is improved.
Optionally, after the step 1029 of acquiring the directional subgraph corresponding to the drug pair to be detected, the method further includes:
and carrying out graph embedding learning on the direction subgraph by using a first graph neural network to obtain an embedded representation of each node in the direction subgraph.
It should be noted that, in the process of extracting the direction subgraph corresponding to the drug pair to be detected based on the second subgraph, if the first graph neural network is not used to perform graph embedding learning on the second subgraph, the first graph neural network is used to perform graph embedding learning on the direction subgraph, so as to obtain the embedded representation of each node in the direction subgraph.
It should be understood that the embedded representation of each node in the directional subgraph is only the initial characterization of each node, and the interaction between drug pairs cannot be accurately explained by using the initial characterization of the nodes because irrelevant nodes and edges still exist in the directional subgraph.
In the embodiment of the invention, when the second subgraph is not subjected to graph embedding learning by using the first graph neural network in the process of extracting and detecting the direction subgraph corresponding to the drug pair by using the second subgraph, the direction subgraph is subjected to graph embedding learning by using the first graph neural network, so that data support is provided for subsequent prediction of drug interaction.
Optionally, the modifying the direction subgraph through graph structure learning to obtain a first subgraph corresponding to the to-be-detected drug pair includes:
and modifying the direction subgraph through graph structure learning based on the embedded representation of each node in the direction subgraph to obtain a first subgraph corresponding to the drug to be detected.
It should be noted that, modifying the directional subgraph is adding or deleting some edges to the directional subgraph.
It is to be understood that graph structure learning does not change the representation of nodes in the graph, but only the structure of the graph.
In the embodiment of the invention, irrelevant edges are further deleted by modifying the direction subgraph, so that the calculation speed is accelerated, a better graph structure is obtained, and the prediction accuracy can be improved.
Optionally, the modifying the direction subgraph through graph structure learning based on the embedded representation of each node in the direction subgraph to obtain a first subgraph corresponding to the to-be-detected drug pair includes:
adding edges between each node of the directional subgraph and the neighbor node of each node;
it should be understood that the purpose of adding edges between each node of the directional subgraph and the neighbor node of each node is to make two opposite edges exist between each node and the neighbor node of each node, that is, to add edges between any node u and its neighbor node v in the directional subgraph, so that there are both edges from u to v and from v to u between node u and node v.
Based on the embedded representation of each node in the directional subgraph, calculating the weight of edges between different nodes in the directional subgraph by using a preset edge weight calculation function;
it should be noted that, calculating the weight of the edge between different nodes in the directional subgraph by using the preset edge weight calculation function means calculating the weight between two edges between every two nodes by using the preset edge weight calculation function.
Wherein, the preset edge weight calculation function is as follows:
wherein, the first and the second end of the pipe are connected with each other,
representing the weight, w, of an edge between nodes u, vpIs a weight vector which is initialized randomly before the graph structure learning model is untrained, and is adjusted continuously along with the continuous training of the graph structure learning modelu，hvRespectively representEmbedded representation of nodes u, v.
Alternatively, the first and second electrodes may be,
the preset edge weight calculation function is:
wherein the content of the first and second substances,
representing the weight, h, of the edge between nodes u, vu，hvRepresenting the embedded representation of nodes u, v, respectively.
Modifying the structure of the directional subgraph based on the weight of edges between different nodes and a preset threshold value to obtain a first subgraph corresponding to the drug pair to be detected.
It should be noted that, the structure of the directional subgraph is modified as follows:
wherein, the first and the second end of the pipe are connected with each other,
representing the weight of the edge between the nodes u and v, and gamma representing a preset threshold value which can be customized according to actual needs; when in use
When the value of the Threshold value is lower than a preset Threshold value gamma, the Threshold function resets the weight of the edge between the nodes u and v to 0, namely, the edge between the nodes u and v at the l-th layer is deleted; when in use
When the value of (d) is higher than the preset Threshold value gamma, the Threshold function resets the weight of the edge between the l-th level nodes u, v to 1, i.e. the edge between the l-th level nodes u, v is reserved.
It will be appreciated that modifying the structure of the directional subgraphs results in a new graph structure, i.e. the graph structure of the first subgraph is superior compared to the directional subgraphs.
In the embodiment of the invention, irrelevant edges are further deleted by modifying the direction subgraphs, so that the calculation speed is accelerated, the first subgraph with a better graph structure is obtained, and the prediction accuracy can be improved.
Fig. 4 is a schematic flow chart of obtaining an explanatory path between pairs of drugs to be detected, as shown in fig. 4, the step 104 of obtaining an explanatory path between pairs of drugs to be detected based on the first sub-graph includes: step 1041, step 1042, step 1043 and step 1044.
it should be noted that the second graph neural network may be one of the following: relationship Graph Convolutional Network (RGCN), Graph Convolutional neural Network (GCN), Graph Attention Network (GAT), Graph Isomorphic Network (GIN).
In the embodiment of the invention, the training data set is required to be used for training the second graph neural network in advance, and the parameters of the second graph neural network are continuously updated by using the cross entropy loss function until the training is successful. The input of the second graph neural network is the first subgraph, and the output of the second graph neural network is the embedded representation of each node in the first subgraph.
In one embodiment, the first sub-graph is graph-embedding learned using GCN, with the formula:
wherein the content of the first and second substances,
representing the embedding of the l-th level node v on the first subgraphRepresents;
representing the weight of the edge between nodes u, v on the first subgraph, in this case,
is 0 or 1; w represents the weight matrix of the node v, when the second graph neural network is not trained, the weight matrix is initialized randomly, and the weight matrix is adjusted continuously along with the continuous training of the second graph neural network; n is a radical ofvRepresenting all neighbor nodes of the node v on the first subgraph;
embedded representation of node u.
Step 1042, based on the embedded representation of each node in the first subgraph, averaging the embedded representations of all nodes in the first subgraph to obtain the embedded representation of the first subgraph;
it should be noted that, the embedded representation obtaining process of the first sub-graph is as follows:
wherein h isGAn embedded representation representing the entire first subgraph; l represents the number of layers of the neural network of the second graph,
representing the embedded representation of node e in the first sub-graph after learning by the neural network of the second graph,
the representation averages the embedded representations of all nodes in the first sub-graph, and takes the average as the embedded representation of the entire first sub-graph.
it should be noted that, the calculation process for obtaining the splicing result is as follows:
wherein h represents the splicing result, hGAn embedded representation representing a first sub-graph;
an embedded representation representing a corresponding node of the first drug in the first sub-graph;
an embedded representation of a corresponding node in the first sub-graph representing the second drug.
And step 1044, predicting the splicing result by utilizing the full connection layer to obtain an explanatory path between the to-be-detected drug pairs.
It is understood that the splicing result is predicted by the full connection layer of the Neural Network, wherein the Neural Network may be a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), or the like.
In the embodiment of the invention, the neural network needs to be trained by using a training data set in advance, and the parameters of the neural network are continuously updated by using the cross entropy loss function until the training is successful. The input to the neural network is a stitched graph-embedded representation and the output of the neural network is a type of interaction between the first drug and the second drug.
The prediction result can be expressed as:
wherein h representsA first drug, t represents a second drug, rqRepresenting the relationship between the first drug and the second drug, wqRepresenting a weight vector that is randomly initialized before the neural network is untrained, and continuously adjusting the weight vector as the neural network is continuously trained.
The loss function can be expressed as:
wherein the content of the first and second substances,
yr＝softmaX(f(h,rq,t))
it should be noted that, when the type of the interaction between the first drug and the second drug is obtained, a most explanatory path about how the interaction between the first drug and the second drug occurs can be obtained. For example, a first drug may react with a second drug due to certain proteins, genes, etc.
In the embodiment of the invention, the most explanatory path between the to-be-detected medicine pairs is obtained through the first sub-graph obtained by graph structure learning, and the interaction between the medicines is explained.
FIG. 5 is a process diagram of the method for predicting drug interactions based on medical knowledge-graphs according to the present invention. As shown in fig. 5:
drug1 denotes the first drug in the drug pair to be tested, drug2 denotes the second drug in the drug pair to be tested, GDDIGraph, G, representing drug interaction data set correspondences for pairs of drugs to be testedKGA second sub-graph, shown as G, obtained by combining the interaction data set of the drug to be detected with the predetermined medical knowledge mapKGThe graph represented by the middle ellipse part is the extracted directed subgraph, namely a direction subgraph;
firstly, graph embedding learning is carried out on the directional subgraph by using a first graph neural network RGCN to obtain embedded representation of each node in the directional subgraph;
then, based on the embedded representation of each node in the directional subgraph, performing Graph structure learning (Graph structure learning) on the directional subgraph through a Graph structure learning model, modifying the Graph structure of the directional subgraph to obtain a first subgraph, and deleting the edges marked by X in the directional subgraph as shown in the Graph;
graph embedding learning is then performed on the first sub-graph by using a second graph neural network, and an embedded representation (not shown in FIG. 5) of each node in the first sub-graph is updated;
then, based on the embedded representation of each node in the first subgraph, the embedded representations of all nodes in the first subgraph are averaged to obtain the embedded representation of the first subgraph, namely, the ring graph G in fig. 5r-di(the specific evaluation process is not shown in FIG. 5);
finally, representing h by embedding the corresponding node of the first drug in the drug pair to be detected in the first subgraphhAnd the embedding of the corresponding node of the second drug in the drug pair to be detected in the first subgraph represents htAnd an embedded representation G of the first sub-graphr-diThe splice input to the full-link layer (not shown in detail in FIG. 5) is performed to predict DDI type, i.e., the type of drug interaction.
The medicine interaction prediction device based on the medical knowledge graph provided by the invention is described below, and the medicine interaction prediction device based on the medical knowledge graph described below and the medicine interaction prediction method based on the medical knowledge graph described above can be referred to each other correspondingly.
Fig. 6 is a schematic structural diagram of a device for predicting drug interactions based on a medical knowledge graph according to an embodiment of the present invention, and as shown in fig. 6, the device for predicting drug interactions based on a medical knowledge graph according to an embodiment of the present invention includes:
a data acquisition module 601, configured to acquire a drug interaction dataset of a drug pair to be detected; wherein the drug interaction dataset is used to represent relationships between different drugs;
a sub-graph extraction module 602, configured to obtain a direction sub-graph corresponding to the to-be-detected drug pair based on the drug interaction data set and a preset medical knowledge graph;
a sub-graph modification module 603, configured to modify the direction sub-graph through graph structure learning to obtain a first sub-graph corresponding to the to-be-detected drug pair;
a path obtaining module 604, configured to obtain an explanatory path between the to-be-detected drug pairs based on the first sub-graph.
According to the medicine interaction prediction device based on the medicine knowledge graph, firstly, the direction subgraph corresponding to the medicine pair to be detected is obtained through the medicine interaction data set and the preset medicine knowledge graph, some non-joint points in the medicine knowledge graph can be deleted, the calculation speed is accelerated to a certain extent, the prediction efficiency is improved, the direction subgraph is modified through graph structure learning, the first subgraph corresponding to the medicine pair to be detected is obtained, the subgraph structure can be dynamically modified, a better graph structure is obtained, the prediction accuracy is improved, finally, an explanatory path between the medicine pair to be detected is obtained based on the first subgraph, and the interaction between the medicines is explained.
Optionally, the obtaining a directional sub-graph corresponding to the to-be-detected drug pair based on the drug interaction data set and a preset medical knowledge graph includes:
combining the drug interaction data set and the preset medicine knowledge graph to obtain a second sub-graph;
performing graph embedding learning on the second subgraph by using a first graph neural network to obtain an embedded representation of each node in the second subgraph;
extracting a 2-hop subgraph corresponding to a first drug in the drug pair to be detected and a 2-hop subgraph corresponding to a second drug in the drug pair to be detected from a second subgraph which completes graph embedding learning;
extracting an intersection graph of the 2-hop subgraph corresponding to the first medicine and the 2-hop subgraph corresponding to the second medicine to obtain a third subgraph;
and acquiring a direction sub-graph corresponding to the to-be-detected drug pair based on the third sub-graph.
Optionally, the obtaining a directional sub-graph corresponding to the to-be-detected drug pair based on the drug interaction data set and a preset medical knowledge graph includes:
combining the drug interaction data set and the preset medicine knowledge graph to obtain a second sub-graph;
extracting a 2-hop subgraph corresponding to a first drug in the drug pair to be detected and a 2-hop subgraph corresponding to a second drug in the drug pair to be detected from the second subgraph;
extracting an intersection graph of the 2-hop subgraph corresponding to the first medicine and the 2-hop subgraph corresponding to the second medicine to obtain a third subgraph;
and acquiring a direction sub-graph corresponding to the to-be-detected drug pair based on the third sub-graph.
Optionally, after the obtaining of the directional subgraph corresponding to the drug pair to be detected, the method further includes:
and carrying out graph embedding learning on the direction subgraph by using a first graph neural network to obtain an embedded representation of each node in the direction subgraph.
Optionally, the modifying the direction subgraph through graph structure learning to obtain a first subgraph corresponding to the to-be-detected drug pair includes:
and modifying the direction subgraph through graph structure learning based on the embedded representation of each node in the direction subgraph to obtain a first subgraph corresponding to the to-be-detected drug pair.
Optionally, the modifying the direction subgraph through graph structure learning based on the embedded representation of each node in the direction subgraph to obtain a first subgraph corresponding to the drug pair to be detected includes:
adding edges between each node of the directional subgraph and the neighbor node of each node;
based on the embedded representation of each node in the directional subgraph, calculating the weight of edges between different nodes in the directional subgraph by using a preset edge weight calculation function;
modifying the structure of the directional subgraph based on the weight of edges between different nodes and a preset threshold value to obtain a first subgraph corresponding to the drug pair to be detected.
Optionally, the obtaining an explanatory path between the pair of to-be-detected drugs based on the first sub-graph includes:
performing graph embedding learning on the first subgraph by using a second graph neural network, and updating an embedded representation of each node in the first subgraph;
based on the embedded representation of each node in the first subgraph, averaging the embedded representations of all nodes in the first subgraph to obtain the embedded representation of the first subgraph;
the embedded representation of the first subgraph, the embedded representation of a node corresponding to a first drug in the drug pair to be detected in the first subgraph and the embedded representation of a node corresponding to a second drug in the drug pair to be detected in the first subgraph are spliced to obtain a splicing result;
and predicting the splicing result by utilizing a full connecting layer to obtain an explanatory path between the to-be-detected drug pairs.
It should be noted that, the device for predicting drug interactions based on a medical knowledge base provided in the embodiment of the present invention can implement all the method steps implemented by the embodiment of the method for predicting drug interactions based on a medical knowledge base, and can achieve the same technical effects, and detailed descriptions of the same parts and beneficial effects as those in the embodiment of the method are omitted here.
Fig. 7 illustrates a physical structure diagram of an electronic device, and as shown in fig. 7, the electronic device may include: a processor (processor)710, a communication Interface (Communications Interface)720, a memory (memory)730, and a communication bus 740, wherein the processor 710, the communication Interface 720, and the memory 730 communicate with each other via the communication bus 740. Processor 710 may invoke logic instructions in memory 730 to perform a method for medical-knowledge-map-based drug interaction prediction, the method comprising: acquiring a drug interaction data set of a drug pair to be detected; wherein the drug interaction dataset is used to represent relationships between different drugs; acquiring a direction subgraph corresponding to the to-be-detected drug pair based on the drug interaction data set and a preset medical knowledge graph; modifying the direction subgraph through graph structure learning to obtain a first subgraph corresponding to the drug pair to be detected; and acquiring an explanatory path between the to-be-detected drug pairs based on the first sub-graph.
In addition, the logic instructions in the memory 730 can be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk, and various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product comprising a computer program, the computer program being stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is capable of executing the method for predicting drug interactions based on a medical knowledge graph provided by the above methods, the method comprising: acquiring a drug interaction data set of a drug pair to be detected; wherein the drug interaction dataset is used to represent relationships between different drugs; acquiring a direction subgraph corresponding to the to-be-detected drug pair based on the drug interaction data set and a preset medical knowledge graph; modifying the direction subgraph through graph structure learning to obtain a first subgraph corresponding to the drug pair to be detected; and acquiring an explanatory path between the to-be-detected medicine pairs based on the first sub-graph.
In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method for predicting drug interactions based on a medical knowledge-graph provided by the above methods, the method comprising: acquiring a drug interaction data set of a drug pair to be detected; wherein the drug interaction dataset is used to represent relationships between different drugs; acquiring a direction subgraph corresponding to the to-be-detected drug pair based on the drug interaction data set and a preset medical knowledge graph; modifying the direction subgraph through graph structure learning to obtain a first subgraph corresponding to the drug pair to be detected; and acquiring an explanatory path between the to-be-detected drug pairs based on the first sub-graph.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A medicine interaction prediction method based on a medicine knowledge graph is characterized by comprising the following steps:
acquiring a drug interaction data set of a drug pair to be detected; wherein the drug interaction dataset is used to represent relationships between different drugs;
acquiring a direction subgraph corresponding to the to-be-detected drug pair based on the drug interaction data set and a preset medical knowledge graph;
modifying the direction subgraph through graph structure learning to obtain a first subgraph corresponding to the drug pair to be detected;
and acquiring an explanatory path between the to-be-detected drug pairs based on the first sub-graph.
2. The method for predicting drug interactions based on a medical knowledge graph of claim 1, wherein the obtaining of the orientation subgraph corresponding to the drug pair to be detected based on the drug interaction dataset and a preset medical knowledge graph comprises:
combining the drug interaction data set and the preset medicine knowledge graph to obtain a second sub-graph;
performing graph embedding learning on the second subgraph by using a first graph neural network to obtain an embedded representation of each node in the second subgraph;
extracting a 2-hop subgraph corresponding to a first drug in the to-be-detected drug pair and a 2-hop subgraph corresponding to a second drug in the to-be-detected drug pair from a second subgraph in which graph embedding learning is completed;
extracting an intersection graph of the 2-hop subgraph corresponding to the first medicine and the 2-hop subgraph corresponding to the second medicine to obtain a third subgraph;
and acquiring a direction sub-graph corresponding to the to-be-detected medicine pair based on the third sub-graph.
3. The method for predicting drug interactions based on a medical knowledge graph of claim 1, wherein the obtaining of the orientation subgraph corresponding to the drug pair to be detected based on the drug interaction dataset and a preset medical knowledge graph comprises:
combining the drug interaction data set and the preset medicine knowledge graph to obtain a second sub-graph;
extracting a 2-hop subgraph corresponding to a first drug in the drug pair to be detected and a 2-hop subgraph corresponding to a second drug in the drug pair to be detected from the second subgraph;
extracting an intersection graph of the 2-hop subgraph corresponding to the first medicine and the 2-hop subgraph corresponding to the second medicine to obtain a third subgraph;
and acquiring a direction sub-graph corresponding to the to-be-detected drug pair based on the third sub-graph.
4. The method for predicting drug interactions based on the medical knowledge graph of claim 3, wherein after obtaining the orientation subgraph corresponding to the drug pair to be detected, the method further comprises:
and carrying out graph embedding learning on the directional subgraph by using a first graph neural network to obtain an embedded representation of each node in the directional subgraph.
5. The method for predicting drug interactions based on a medical knowledge graph of any one of claims 2 to 4, wherein the modifying the orientation subgraph through graph structure learning to obtain a first subgraph corresponding to the drug pair to be detected comprises:
and modifying the direction subgraph through graph structure learning based on the embedded representation of each node in the direction subgraph to obtain a first subgraph corresponding to the drug to be detected.
6. The method for predicting drug interactions based on a medical knowledge graph of claim 5, wherein the modifying the direction subgraph through graph structure learning based on the embedded representation of each node in the direction subgraph to obtain the first subgraph corresponding to the drug pair to be detected comprises:
adding edges between each node of the directional subgraph and the neighbor node of each node;
based on the embedded representation of each node in the directional subgraph, calculating the weight of edges between different nodes in the directional subgraph by using a preset edge weight calculation function;
and modifying the structure of the direction subgraph based on the weight of the edges between different nodes and a preset threshold value to obtain a first subgraph corresponding to the drug pair to be detected.
7. The method for predicting drug interactions based on a medical knowledge graph of claim 1 or 6, wherein the step of obtaining an explanatory path between the pairs of the drugs to be detected based on the first sub-graph comprises:
performing graph embedding learning on the first subgraph by using a second graph neural network, and updating an embedded representation of each node in the first subgraph;
based on the embedded representation of each node in the first subgraph, averaging the embedded representations of all nodes in the first subgraph to obtain the embedded representation of the first subgraph;
the embedded representation of the first subgraph, the embedded representation of a node corresponding to a first drug in the drug pair to be detected in the first subgraph and the embedded representation of a node corresponding to a second drug in the drug pair to be detected in the first subgraph are spliced to obtain a splicing result;
and predicting the splicing result by utilizing a full connecting layer to obtain an explanatory path between the to-be-detected drug pairs.
8. A device for predicting drug interactions based on a medical knowledge graph, comprising:
the data acquisition module is used for acquiring a drug interaction data set of the drug pair to be detected; wherein the drug interaction dataset is used to represent relationships between different drugs;
the subgraph extraction module is used for acquiring a direction subgraph corresponding to the to-be-detected medicine pair based on the medicine interaction data set and a preset medicine knowledge graph;
the sub-graph modification module is used for modifying the direction sub-graph through graph structure learning to obtain a first sub-graph corresponding to the to-be-detected drug pair;
and the path acquisition module is used for acquiring an explanatory path between the to-be-detected medicine pairs based on the first sub-graph.
9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor, when executing the program, implements a method of drug interaction prediction for a medical knowledge graph as claimed in any one of claims 1 to 7.
10. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements a method for drug interaction prediction of a medical knowledge-graph as claimed in any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210272634.7A CN114724622A (en) | 2022-03-18 | 2022-03-18 | Medicine interaction prediction method and device based on medicine knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210272634.7A CN114724622A (en) | 2022-03-18 | 2022-03-18 | Medicine interaction prediction method and device based on medicine knowledge graph |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114724622A true CN114724622A (en) | 2022-07-08 |
Family
ID=82237257
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210272634.7A Pending CN114724622A (en) | 2022-03-18 | 2022-03-18 | Medicine interaction prediction method and device based on medicine knowledge graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114724622A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117079762A (en) * | 2023-09-25 | 2023-11-17 | 腾讯科技（深圳）有限公司 | Drug effect prediction model training method, drug effect prediction method and device thereof |
| CN117438104A (en) * | 2023-12-21 | 2024-01-23 | 成都市第一人民医院 | Intelligent medicine early warning method, electronic equipment and computer storage medium |
-
2022
- 2022-03-18 CN CN202210272634.7A patent/CN114724622A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117079762A (en) * | 2023-09-25 | 2023-11-17 | 腾讯科技（深圳）有限公司 | Drug effect prediction model training method, drug effect prediction method and device thereof |
| CN117079762B (en) * | 2023-09-25 | 2024-01-23 | 腾讯科技（深圳）有限公司 | Drug effect prediction model training method, drug effect prediction method and device thereof |
| CN117438104A (en) * | 2023-12-21 | 2024-01-23 | 成都市第一人民医院 | Intelligent medicine early warning method, electronic equipment and computer storage medium |
| CN117438104B (en) * | 2023-12-21 | 2024-03-22 | 成都市第一人民医院 | Intelligent medicine early warning method, electronic equipment and computer storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chu et al. | Neural turtle graphics for modeling city road layouts | |
| CN114724622A (en) | Medicine interaction prediction method and device based on medicine knowledge graph | |
| CN110674869B (en) | Classification processing and graph convolution neural network model training method and device | |
| EP3951713A1 (en) | Method and apparatus for training image semantic segmentation network, device, and storage medium | |
| CN111522962B (en) | Sequence recommendation method, device and computer readable storage medium | |
| CN104461693B (en) | Virtual machine update method and system under a kind of desktop cloud computing environment | |
| CN110213164B (en) | Method and device for identifying network key propagator based on topology information fusion | |
| CN112214775B (en) | Injection attack method, device, medium and electronic equipment for preventing third party from acquiring key diagram data information and diagram data | |
| CN110825884A (en) | Embedded representation processing method and device based on artificial intelligence and electronic equipment | |
| Geiß et al. | Best match graphs | |
| JP7457125B2 (en) | Translation methods, devices, electronic equipment and computer programs | |
| CN107003834B (en) | Pedestrian detection device and method | |
| CN112905801A (en) | Event map-based travel prediction method, system, device and storage medium | |
| CN104200133A (en) | Read and distance distribution based genome De novo sequence splicing method | |
| JP7328331B2 (en) | Image coloring completion method, its device and its computer program, artificial neural network learning method, its device and its computer program | |
| CN111626489B (en) | Shortest path planning method and device based on time sequence difference learning algorithm | |
| CN102708327A (en) | Network community discovery method based on spectrum optimization | |
| CN111967972A (en) | Financial product recommendation method and device | |
| CN112100450A (en) | Graph calculation data segmentation method, terminal device and storage medium | |
| Wu et al. | Sharing deep neural network models with interpretation | |
| CN111126578B (en) | Joint data processing method, device and system for model training | |
| Yuan et al. | Solving job shop scheduling problems via deep reinforcement learning | |
| US20210404835A1 (en) | Method, apparatus, and computer program for generating road network data for autonomous driving vehicle | |
| CN110008304A (en) | The difference visible processing method and device of behavior tree | |
| CN114897161B (en) | Mask-based graph classification backdoor attack defense method and system, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |