CN111798934A - Molecular property prediction method based on graph neural network - Google Patents
Molecular property prediction method based on graph neural network Download PDFInfo
- Publication number
- CN111798934A CN111798934A CN202010578320.0A CN202010578320A CN111798934A CN 111798934 A CN111798934 A CN 111798934A CN 202010578320 A CN202010578320 A CN 202010578320A CN 111798934 A CN111798934 A CN 111798934A
- Authority
- CN
- China
- Prior art keywords
- information
- module
- vector
- graph
- neural network
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 23
- 239000013598 vector Substances 0.000 claims abstract description 38
- 230000008569 process Effects 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 14
- 230000005540 biological transmission Effects 0.000 claims abstract description 9
- 230000004927 fusion Effects 0.000 claims abstract description 9
- 239000000126 substance Substances 0.000 claims abstract description 5
- 230000006870 function Effects 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 9
- 230000009466 transformation Effects 0.000 claims description 9
- 238000012546 transfer Methods 0.000 claims description 7
- 230000014509 gene expression Effects 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 claims description 3
- 238000012216 screening Methods 0.000 abstract description 7
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000000605 extraction Methods 0.000 abstract description 2
- 239000003814 drug Substances 0.000 description 4
- 229940079593 drug Drugs 0.000 description 3
- 238000012827 research and development Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004883 computer application Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 229940043263 traditional drug Drugs 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/50—Molecular design, e.g. of drugs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/70—Machine learning, data mining or chemometrics
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
Abstract
The invention provides a molecular property prediction method based on a graph neural network, which comprises a training process and an application process based on the training process, wherein a chemical molecular formula is converted into a directed graph structure or an undirected graph structure for storage, all nodes are connected to information by using a computer data structure for description and storage, the discretized data are changed into continuous values by an embedding module, the graph structure and node embedding vectors jointly enter an information transmission module to obtain fusion information, the fusion information is sent to a reading module for further extraction, and the obtained result is finally sent to a judgment module for prediction. The invention has the beneficial effects that: the graph neural network is applied to molecular property prediction, thereby realizing large-scale molecular screening.
Description
Technical Field
The invention belongs to the technical field of computer application, and particularly relates to a molecular property prediction method based on a graph neural network.
Background
In the field of biomedical research and development, the molecular screening process plays an important role in determining the cost of medicine research and development.
The traditional drug molecule screening technology is composed of a set of complex framework, and each step needs a great deal of engineering and experience through a plurality of manual design and fussy steps. For complex molecular structure information and in the face of large-scale drug molecule screening, the ideal effect can not be achieved by only applying the traditional method.
In view of this, how to analyze complex molecular structure information and predict molecular properties, and to implement large-scale drug molecule screening is a problem that is urgently faced at present.
Disclosure of Invention
In order to solve the defects of the prior art, the invention provides a molecular property prediction method based on a graph neural network.
The purpose of the invention is realized by the following technical scheme:
a molecular property prediction method based on a graph neural network is applied to prediction of molecular properties.
Preferably, the molecular property prediction method based on the graph neural network comprises a training process and an application process based on the training process, wherein the training process comprises the following steps,
s1, converting the chemical molecular formula into an undirected graph or a directed graph structure, connecting all nodes in the converted mechanism to information by using a computer data structure for description and storage, wherein all stored information is represented by a certain integer, and the integer is an ID (identity) for representing the information;
s2, sending the information stored in S1 to an embedded vector module, wherein the embedded vector module converts the received information into an embedded vector which can be expressed by a computer, namely a process of converting ID into the embedded vector;
s3, the information stored after being converted in S1 and the embedded vector information after being converted in S2 are transmitted to a transfer module for fusion, and the information transfer module fuses the node information and the graph connection information to be used as middle hidden layer information;
s4, sending the intermediate hidden layer information formed by fusing the information in the S3 into a reading module, and further abstracting and fusing the obtained information to obtain a vector with fixed dimension;
s5, sending the vector obtained in S4 to a judgment module, and analyzing the vector to obtain a judgment result;
and S6, forming a loss function by the network output result after the judgment of the S5 and the standard answer together for judging the quality of the network output result, updating parameters by using a gradient return algorithm after the loss function is calculated, repeating the steps S1-S6, continuously updating until the loss function is reduced to be below a set threshold value, and indicating that the network training is finished.
Preferably, the application process comprises the following steps:
and S7, converting any new molecular structure into an undirected graph or directed graph structure, sequentially entering an embedding vector expression module, an information transmission module, a vector reading module and a judgment module, and finally obtaining answers given by the network.
Preferably, the fusion of the delivery module in S3 includes word embedding multi-layer long-and-short-term memory model and/or convolutional layer stacking processing, and full-connection nonlinear transformation processing, and sequentially fuses node information and structural connection information described in the graph structure by circularly traversing the layer.
The neural network layer applied in the transmission module can be formed by mixing one or more of a long-time memory layer, a convolutional layer or a residual error structure. Of course, the specific components included in the neural network layer can be adjusted according to the actual application.
Preferably, the vector in S4 is a reading vector, and the dimension can be freely set according to actual conditions.
Preferably, the processing step of the S4 reading module includes matrix merging and adding.
Preferably, the step of discriminating in S6 includes a step of multi-layer nonlinear transformation and linear transformation.
Preferably, the selection of the loss function in S6 may be changed according to a classification problem or a regression problem, where the classification problem uses cross entropy and the regression problem uses least squares or average absolute error.
The invention has the beneficial effects that: the chemical molecular formula is converted into a form of an undirected graph or a directed graph structure, extraction and analysis are carried out by combining a neural network, and the neural network of the graph is applied to molecular property prediction, so that large-scale molecular screening is realized.
Drawings
FIG. 1: the invention discloses a frame structure schematic diagram of a molecular property prediction method based on a graph neural network.
Detailed Description
The invention discloses a molecular property prediction method based on a graph neural network, which is combined with a graph 1 and comprises a training process and an application process based on the training process, wherein the training process comprises the following steps,
s1, converting the chemical molecular formula into an undirected graph or a directed graph structure, connecting all nodes in the converted mechanism to information by using a computer data structure for description and storage, wherein all stored information is represented by a certain integer, and the integer is an ID (identity) for representing the information;
s2, sending the information stored in S1 to an embedded vector module, wherein the embedded vector module converts the received information into an embedded vector which can be expressed by a computer, namely a process of converting ID into the embedded vector;
s3, the information stored after being converted in S1 and the embedded vector information after being converted in S2 are transmitted to a transfer module for fusion, and the information transfer module fuses the node information and the graph connection information to be used as middle hidden layer information; the fusion of the transmission module comprises word embedding multi-layer long-time memory model and/or convolution layer stacking processing and full-connection nonlinear transformation processing, and node information and structure connection information described in the graph structure are sequentially fused by circularly traversing the layers. The method mainly aims to perform fusion between vector expressions on a plurality of equal node information and connection information through circular information transmission. The advantage of multiple transmission is that the network is subjected to deep and continuous information abstraction, so that the neural network has strong generalization capability.
The neural network layer applied in the transmission module can be formed by mixing one or more of a long-time memory layer, a convolutional layer or a residual error structure. Of course, the specific components included in the neural network layer can be adjusted according to the actual application.
S4, sending the intermediate hidden layer information formed by fusing the information in the S3 into a reading module, and further abstracting and fusing the obtained information to obtain a vector with fixed dimension; the vector is a reading vector, and the dimensionality can be freely set according to actual conditions. The processing steps of the reading module comprise matrix combination and addition.
S5, sending the vector obtained in S4 to a judgment module, and analyzing the vector to obtain a judgment result; the data type of the result is determined by the specific application, and can be used for classification and regression.
And S6, forming a loss function by the network output result after the judgment of the S5 and the standard answer together for judging the quality of the network output result, updating parameters by using a gradient return algorithm after the loss function is calculated, repeating the steps S1-S6, continuously updating until the loss function is reduced to be below a set threshold value, and indicating that the network training is finished. The discriminating step includes a multi-layer nonlinear transformation and a linear transformation step. The selection of the loss function can be changed according to classification or regression problems, wherein the classification problems adopt cross entropy, and the regression problems adopt least square or average absolute error.
The application process comprises the following steps:
and S7, converting any new molecular structure into an undirected graph or directed graph structure, sequentially entering an embedding vector expression module, an information transmission module, a vector reading module and a judgment module, and finally obtaining answers given by the network.
The graph neural network is a novel artificial intelligence technology, the principle of the graph neural network is that input data are described into a directed graph or undirected graph structure, and data fitting is carried out on complex data distribution through a deep neural network, so that a final target is achieved. Compared with the traditional method, the deep neural network has the characteristics of big data analysis, autonomous learning, parallel computation, low cost and the like, is introduced and applied to analyzing complex molecular structure information and predicting molecular properties, and therefore, large-scale drug molecule screening is achieved.
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 (8)
1. A molecular property prediction method based on a graph neural network, characterized in that the graph neural network is applied to the prediction of molecular properties.
2. The method of claim 1, wherein the method comprises: comprises a training process and an application process based on the training process, wherein the training process comprises the following steps,
s1, converting the chemical molecular formula into an undirected graph or a directed graph structure, connecting all nodes in the converted mechanism to information by using a computer data structure for description and storage, wherein all stored information is represented by a certain integer, and the integer is an ID (identity) for representing the information;
s2, sending the information stored in S1 to an embedded vector module, wherein the embedded vector module converts the received information into an embedded vector which can be expressed by a computer, namely a process of converting ID into the embedded vector;
s3, the information stored after being converted in S1 and the embedded vector information after being converted in S2 are transmitted to a transfer module for fusion, and the information transfer module fuses the node information and the graph connection information to be used as middle hidden layer information;
s4, sending the intermediate hidden layer information formed by fusing the information in the S3 into a reading module, and further abstracting and fusing the obtained information to obtain a vector with fixed dimension;
s5, sending the vector obtained in S4 to a judgment module, and analyzing the vector to obtain a judgment result;
and S6, forming a loss function by the network output result after the judgment of the S5 and the standard answer together for judging the quality of the network output result, updating parameters by using a gradient return algorithm after the loss function is calculated, repeating the steps S1-S6, continuously updating until the loss function is reduced to be below a set threshold value, and indicating that the network training is finished.
3. The method of claim 1, wherein the method comprises: the application process comprises the following steps:
and S7, converting any new molecular structure into an undirected graph or directed graph structure, sequentially entering an embedding vector expression module, an information transmission module, a vector reading module and a judgment module, and finally obtaining answers given by the network.
4. The method of claim 2, wherein the method comprises: the fusion of the transfer modules in S3 includes word embedding multi-layer long-and-short term memory models and/or convolutional layer stacking processing, and full-connection nonlinear transformation processing, and sequentially fuses node information and structural connection information described in the graph structure by circularly traversing the layers.
5. The method of claim 2, wherein the method comprises: the vector in S4 is a read vector, and the dimension can be freely set according to actual conditions.
6. The method of claim 2, wherein the method comprises: the processing steps of the S4 reading module comprise matrix combination and addition.
7. The method of claim 2, wherein the method comprises: the discriminating step in S6 includes a multi-layer nonlinear transformation and a linear transformation step.
8. The method of claim 2, wherein the method comprises: the selection of the loss function in S6 may be changed according to a classification problem using cross entropy or a regression problem using least squares or average absolute errors.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010578320.0A CN111798934B (en) | 2020-06-23 | 2020-06-23 | Molecular property prediction method based on graph neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010578320.0A CN111798934B (en) | 2020-06-23 | 2020-06-23 | Molecular property prediction method based on graph neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111798934A true CN111798934A (en) | 2020-10-20 |
CN111798934B CN111798934B (en) | 2023-11-14 |
Family
ID=72804579
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010578320.0A Active CN111798934B (en) | 2020-06-23 | 2020-06-23 | Molecular property prediction method based on graph neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111798934B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113241130A (en) * | 2021-06-08 | 2021-08-10 | 西南交通大学 | Molecular structure prediction method based on graph convolution network |
WO2022226940A1 (en) * | 2021-04-29 | 2022-11-03 | Huawei Cloud Computing Technologies Co., Ltd. | Method and system for generating task-relevant structural embeddings from molecular graphs |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110009093A (en) * | 2018-12-07 | 2019-07-12 | 阿里巴巴集团控股有限公司 | For analyzing the nerve network system and method for relational network figure |
US20190272468A1 (en) * | 2018-03-05 | 2019-09-05 | The Board Of Trustees Of The Leland Stanford Junior University | Systems and Methods for Spatial Graph Convolutions with Applications to Drug Discovery and Molecular Simulation |
CN110348573A (en) * | 2019-07-16 | 2019-10-18 | 腾讯科技(深圳)有限公司 | The method of training figure neural network, figure neural network unit, medium |
US20200027528A1 (en) * | 2017-09-12 | 2020-01-23 | Massachusetts Institute Of Technology | Systems and methods for predicting chemical reactions |
WO2020020088A1 (en) * | 2018-07-23 | 2020-01-30 | 第四范式(北京)技术有限公司 | Neural network model training method and system, and prediction method and system |
CN110782015A (en) * | 2019-10-25 | 2020-02-11 | 腾讯科技(深圳)有限公司 | Training method and device for network structure optimizer of neural network and storage medium |
CN110909868A (en) * | 2019-12-04 | 2020-03-24 | 支付宝(杭州)信息技术有限公司 | Node representation method and device based on graph neural network model |
CN110970099A (en) * | 2019-12-10 | 2020-04-07 | 北京大学 | Medicine molecule generation method based on regularization variational automatic encoder |
CN110993037A (en) * | 2019-10-28 | 2020-04-10 | 浙江工业大学 | Protein activity prediction device based on multi-view classification model |
CN111063398A (en) * | 2019-12-20 | 2020-04-24 | 吉林大学 | Molecular discovery method based on graph Bayesian optimization |
CN111223532A (en) * | 2019-11-14 | 2020-06-02 | 腾讯科技(深圳)有限公司 | Method, apparatus, device, medium for determining a reactant of a target compound |
-
2020
- 2020-06-23 CN CN202010578320.0A patent/CN111798934B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20200027528A1 (en) * | 2017-09-12 | 2020-01-23 | Massachusetts Institute Of Technology | Systems and methods for predicting chemical reactions |
US20190272468A1 (en) * | 2018-03-05 | 2019-09-05 | The Board Of Trustees Of The Leland Stanford Junior University | Systems and Methods for Spatial Graph Convolutions with Applications to Drug Discovery and Molecular Simulation |
WO2020020088A1 (en) * | 2018-07-23 | 2020-01-30 | 第四范式(北京)技术有限公司 | Neural network model training method and system, and prediction method and system |
CN110009093A (en) * | 2018-12-07 | 2019-07-12 | 阿里巴巴集团控股有限公司 | For analyzing the nerve network system and method for relational network figure |
CN110348573A (en) * | 2019-07-16 | 2019-10-18 | 腾讯科技(深圳)有限公司 | The method of training figure neural network, figure neural network unit, medium |
CN110782015A (en) * | 2019-10-25 | 2020-02-11 | 腾讯科技(深圳)有限公司 | Training method and device for network structure optimizer of neural network and storage medium |
CN110993037A (en) * | 2019-10-28 | 2020-04-10 | 浙江工业大学 | Protein activity prediction device based on multi-view classification model |
CN111223532A (en) * | 2019-11-14 | 2020-06-02 | 腾讯科技(深圳)有限公司 | Method, apparatus, device, medium for determining a reactant of a target compound |
CN110909868A (en) * | 2019-12-04 | 2020-03-24 | 支付宝(杭州)信息技术有限公司 | Node representation method and device based on graph neural network model |
CN110970099A (en) * | 2019-12-10 | 2020-04-07 | 北京大学 | Medicine molecule generation method based on regularization variational automatic encoder |
CN111063398A (en) * | 2019-12-20 | 2020-04-24 | 吉林大学 | Molecular discovery method based on graph Bayesian optimization |
Non-Patent Citations (2)
Title |
---|
李淑怡: "基于符号图卷积网络的药物互作用关系预测", 《现代计算机》, pages 12 - 15 * |
郑睿刚: "图卷积算法的研究进展", 《中山大学学报(自然科学版)》, vol. 59, no. 2, pages 1 - 13 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2022226940A1 (en) * | 2021-04-29 | 2022-11-03 | Huawei Cloud Computing Technologies Co., Ltd. | Method and system for generating task-relevant structural embeddings from molecular graphs |
CN113241130A (en) * | 2021-06-08 | 2021-08-10 | 西南交通大学 | Molecular structure prediction method based on graph convolution network |
CN113241130B (en) * | 2021-06-08 | 2022-04-22 | 西南交通大学 | Molecular structure prediction method based on graph convolution network |
Also Published As
Publication number | Publication date |
---|---|
CN111798934B (en) | 2023-11-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108874914B (en) | Information recommendation method based on graph convolution and neural collaborative filtering | |
CN113905391B (en) | Integrated learning network traffic prediction method, system, equipment, terminal and medium | |
Fang et al. | A new sequential image prediction method based on LSTM and DCGAN | |
CN111798934A (en) | Molecular property prediction method based on graph neural network | |
CN113361559B (en) | Multi-mode data knowledge information extraction method based on deep-width combined neural network | |
CN116628597B (en) | Heterogeneous graph node classification method based on relationship path attention | |
CN112381179A (en) | Heterogeneous graph classification method based on double-layer attention mechanism | |
CN112861976A (en) | Sensitive image identification method based on twin graph convolution hash network | |
CN116049467A (en) | Non-supervision image retrieval method and system based on label visual joint perception | |
CN113761250A (en) | Model training method, merchant classification method and device | |
Dai et al. | Hybrid deep model for human behavior understanding on industrial internet of video things | |
CN114020891A (en) | Double-channel semantic positioning multi-granularity attention mutual enhancement video question-answering method and system | |
CN114067162A (en) | Image reconstruction method and system based on multi-scale and multi-granularity feature decoupling | |
CN114817773A (en) | Time sequence prediction system and method based on multi-stage decomposition and fusion | |
CN116485839A (en) | Visual tracking method based on attention self-adaptive selection of transducer | |
CN111832637A (en) | Distributed deep learning classification method based on alternative direction multiplier method ADMM | |
CN110020379B (en) | Link prediction method based on deep dynamic network embedded representation model | |
CN112966706A (en) | Multilayer network representation learning method based on graph attention machine mechanism | |
CN113918711B (en) | Academic paper-oriented classification method based on multi-view multi-layer attention | |
CN115457657A (en) | Method for identifying channel characteristic interaction time modeling behaviors based on BERT model | |
CN114386600A (en) | Network characterization method based on self-adaptive structure and position coding | |
CN114943016A (en) | Cross-granularity joint training-based graph comparison representation learning method and system | |
CN114529908A (en) | Offline handwritten chemical reaction type image recognition technology | |
CN111798933B (en) | Molecular docking discrimination method based on deep learning | |
CN112561174B (en) | LSTM and MLP-based superimposed neural network prediction geothermal energy production method |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |