CN117393075A - Model training method and task execution method based on molecular energy information - Google Patents

Abstract

The specification discloses a model training method and a task execution method based on molecular energy information. The model training method comprises the following steps: acquiring characterization data of a specified compound molecule, and determining three-dimensional molecular map information capable of specifying the compound molecule; inputting the three-dimensional molecular diagram information into a prediction model, and determining the three-dimensional molecular diagram characteristics corresponding to the specified compound molecules; predicting molecular energy information corresponding to the specified compound molecules according to the three-dimensional molecular diagram characteristics; the model is trained with the objective of optimizing the minimized deviation between the predicted molecular energy information and the actual molecular energy information corresponding to the specified compound molecule. In the subsequent process of predicting the energy of the whole molecule by the molecular structure characteristics, the whole energy of the compound molecule under any structure can be rapidly and accurately predicted by the prediction model, and the accuracy of molecular energy prediction is improved.

Description

A model training method and task execution method based on molecular energy information

Technical field

This description relates to the field of computer technology, and in particular, to a model training method and a task execution method based on molecular energy information.

Background technique

With the development of science and technology, molecular research has become an important part of the social and scientific development process. By constructing the structure-energy relationship of compound molecules, downstream tasks related to the configuration changes of compound molecules (molecular regulation, molecular design, etc.) ) research provides favorable technical support, which makes efficient and accurate prediction of the energy of compound molecules a new exploration direction in the process of molecular research.

However, currently, simple shallow neural networks are usually used to characterize compound molecules. The accuracy of the compound molecule energy predicted by the model is low, making it difficult to be used to perform downstream molecular research tasks.

Therefore, how to accurately determine the energy information of compound molecules is an urgent problem to be solved.

Contents of the invention

This specification provides a model training method and a task execution method based on molecular energy information to partially solve the above problems existing in the existing technology.

This manual adopts the following technical solutions:

This manual provides a method for model training, including:

Obtain characterization data of the specified compound molecule, the characterization data being used to characterize the position information and attribute information of each atom in the specified compound molecule;

Process the characterization data to determine the three-dimensional molecular map information of the designated compound molecule;

The three-dimensional molecular graph information is input into a prediction model to be trained, so that through the prediction model, based on the equivariance between the position information and the embedded features corresponding to the position information, and the attribute information and the attribute The information corresponds to the invariance between the embedded features, and the three-dimensional molecular graph features corresponding to the specified compound molecules are determined based on the three-dimensional molecular graph information;

Predict the molecular energy information corresponding to the specified compound molecule according to the three-dimensional molecular graph characteristics;

The prediction model is trained with an optimization goal of minimizing the deviation between the predicted molecular energy information and the actual molecular energy information corresponding to the specified compound molecule.

Optionally, obtain the characterization data of the specified compound molecule, including:

Select the initial data of the designated compound molecules from the data set of molecular compounds;

The characterization data is determined based on initial data for the designated compound molecule.

Optionally, the position information includes: the coordinates of each atom in the specified compound molecule in a specified coordinate system;

The attribute information includes: the type of each atom in the designated compound molecule, the direction vector between any two atoms in the designated compound molecule, and the connection information between any two atoms in the designated compound molecule.

Optionally, a graph attention mechanism network is provided in the prediction model;

Determining the three-dimensional molecular map characteristics corresponding to the specified compound molecule based on the three-dimensional molecular map information specifically includes:

Determine the attention weight corresponding to the specified compound molecule through the graph attention mechanism network;

According to the attention weight and each embedded feature determined by the graph attention mechanism network based on the three-dimensional molecular graph information, determine the invariant features and equal variable features corresponding to the designated compound molecule;

The three-dimensional molecular graph features are determined based on the invariant features and the equivariable features.

This instruction sheet provides a task execution method based on molecular energy information, including:

Receive a task execution request for the original compound and obtain characterization data of the original compound molecule;

Process the characterization data of the original compound molecules to determine the three-dimensional molecular map information corresponding to the original compound molecules;

The three-dimensional molecular graph information of the original compound molecule is input into the pre-trained prediction model, so that the three-dimensional molecular graph characteristics corresponding to the original compound molecule are determined through the prediction model, and the three-dimensional molecular graph corresponding to the original compound molecule is determined based on the three-dimensional molecular graph information of the original compound molecule. graph features to determine the molecular energy information corresponding to the original compound molecule, wherein the prediction model is trained by the above model training method;

Perform tasks based on the molecular energy information corresponding to the original compound molecules.

Optionally, the tasks include: molecular regulation tasks or molecular design tasks;

Perform tasks based on the molecular energy information corresponding to the original compound molecules, specifically including:

Determine the corresponding relationship between the molecular structure and molecular energy of the original compound molecule according to the molecular energy information corresponding to the original compound molecule and the characterization data corresponding to the original compound molecule;

The molecular design task or the molecular regulation task is performed based on the corresponding relationship.

This instruction manual provides a model training device, including:

An acquisition module, used to obtain characterization data of a designated compound molecule, where the characterization data is used to characterize the position information and attribute information of each atom in the designated compound molecule;

a determination module, which processes the characterization data and determines the three-dimensional molecular diagram information of the specified compound molecule;

An input module for inputting the three-dimensional molecular graph information into a prediction model to be trained, so that through the prediction model, based on the equivariance between the position information and the embedded features corresponding to the position information, and the attributes The invariance between the information and the embedded features corresponding to the attribute information, and determining the three-dimensional molecular graph features corresponding to the specified compound molecules based on the three-dimensional molecular graph information;

A prediction module, used to predict the molecular energy information corresponding to the specified compound molecule based on the three-dimensional molecular graph characteristics;

A training module, configured to train the prediction model with an optimization goal of minimizing the deviation between the predicted molecular energy information and the actual molecular energy information corresponding to the specified compound molecule.

This description provides a task execution device based on molecular energy information, including:

The receiving module is used to receive the task execution request for the original compound and obtain the characterization data of the original compound molecule;

A building module for processing the characterization data of the original compound molecules and determining the three-dimensional molecular diagram information corresponding to the original compound molecules;

A determination module, configured to input the three-dimensional molecular graph information of the original compound molecule into a pre-trained prediction model, so as to determine the three-dimensional molecular graph characteristics corresponding to the original compound molecule through the prediction model, and determine the characteristics of the three-dimensional molecular graph corresponding to the original compound molecule according to the original compound molecule. The three-dimensional molecular graph characteristics corresponding to the molecules determine the molecular energy information corresponding to the original compound molecules, wherein the prediction model is trained by the above model training method;

An execution module is used to execute tasks according to the molecular energy information corresponding to the original compound molecules.

This specification provides a computer-readable storage medium. The storage medium stores a computer program. When the computer program is executed by a processor, the above-mentioned model training method or task execution method based on molecular energy information is implemented.

This specification provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the above-mentioned model training method is implemented or based on molecular energy information. task execution method.

At least one of the above technical solutions adopted in this manual can achieve the following beneficial effects:

In the model training method provided in this manual, the characterization data of the specified compound molecule is obtained, and the three-dimensional molecular graph information of the specified compound molecule is constructed; the three-dimensional molecular graph information is input into the prediction model to determine the three-dimensional molecular graph characteristics corresponding to the specified compound molecule. ; Predict the molecular energy information corresponding to the specified compound molecule based on the characteristics of the three-dimensional molecular graph; train the model with the optimization goal of minimizing the deviation between the predicted molecular energy information and the actual molecular energy information corresponding to the specified compound molecule. In the subsequent process of predicting the energy of the entire molecule based on the molecular structure characteristics, this prediction model can be used to quickly and accurately predict the overall energy of the compound molecule under any structure, improving the accuracy of molecular energy prediction.

It can be seen from the above method that this solution can construct three-dimensional molecular graph information through the characterization data of compound molecules, fully characterizing the structure and properties of each atom in the compound molecule. After inputting the three-dimensional molecular graph into the prediction model, it can be based on Feature extraction is performed by using the isovariance of position information and the invariance of attribute information to accurately express the characteristics of compound molecules, fully ensuring the original characteristics of compound molecules in terms of attributes and structure, and then extracting features based on the extracted features. The energy information of compound molecules is predicted and the model is trained based on the prediction results, which fully improves the accuracy of the energy information of compound molecules predicted by the prediction model after training, and provides effective guarantee for downstream molecular research tasks.

Description of the drawings

The drawings described here are used to provide a further understanding of this specification and constitute a part of this specification. The illustrative embodiments and descriptions of this specification are used to explain this specification and do not constitute an improper limitation of this specification. In the attached picture:

Figure 1 is a schematic flow chart of a model training method provided in this specification;

Figure 2 is a schematic flow chart of a task execution method based on molecular energy information provided in this specification;

Figure 3 is a schematic diagram of the architecture of a system for exploring the molecular structure-energy relationship of compounds provided in this specification;

Figure 4 is a schematic diagram of a model training device provided in this specification;

Figure 5 is a schematic diagram of a task execution device based on molecular energy information provided in this specification;

FIG. 6 is a schematic structural diagram of an electronic device corresponding to FIG. 1 or 2 provided in this specification.

Detailed ways

In order to make the purpose, technical solutions and advantages of this specification more clear, the technical solutions of this specification will be clearly and completely described below in conjunction with specific embodiments of this specification and the corresponding drawings. Obviously, the described embodiments are only some of the embodiments of this specification, but not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of this specification.

The technical solutions provided by each embodiment of this specification will be described in detail below with reference to the accompanying drawings.

Figure 1 is a schematic flow chart of a model training method provided in this manual, including the following steps:

S101: Obtain the characterization data of the specified compound molecule. The characterization data is used to characterize the position information and attribute information of each atom in the specified compound molecule.

S102: Process the characterization data to determine the three-dimensional molecular map information of the specified compound molecule.

The execution subject of the model training method provided in this manual can be a terminal device such as a desktop computer or a laptop computer, or of course a server. For the convenience of explanation, the following only takes the terminal device as the execution subject as an example to describe the model training provided. The method is explained.

In this specification, the terminal device can obtain raw data on the structure of various specified compound molecules to build a data set for training the model. These data can be obtained by crawling from the external network or through files. Enter the form to obtain.

After obtaining the above data set, the terminal device needs to search for data suitable for training the prediction model. In other words, not all the specified compound molecules included in the data set are suitable for training as samples, and some data may not be very suitable for training. Good label molecule energy information, some data can be classified as "dirty data".

Therefore, the terminal device needs to search for designated compound molecules suitable as training samples from the data set, and then determine the initial data of the designated compound molecules. The specific implementation method can be to create an extensible three-dimensional molecular graph structure data generator to clean, reconstruct and optimize the molecular compound data of the specified compound class, thereby searching for the specified compound molecules as training samples and determining the specified compound. Three-dimensional molecular map information of molecules.

In this specification, after the terminal device determines the initial data of the specified compound molecule from the above data set, it can further determine the characterization data of the specified compound molecule. The characterization data mentioned here is used to characterize each atom in the specified compound molecule. Position information and attribute information. The position information includes: specifying the coordinates of each atom in the compound molecule in the specified coordinate system. Attribute information includes: specifying the type of each atom in the compound molecule, and specifying the direction between any two atoms in the compound molecule. Vector and connection information between any two atoms in the specified compound molecule. Of course, other characterization data may also be included, such as chemical formula, molecular weight, chemical bonds, structure, physical properties, etc., which are not specifically limited in this specification.

After determining the characterization data of the specified compound molecule, the terminal device can reconstruct and optimize part of the characterization data (including the above attribute information and position information), thereby obtaining the three-dimensional molecular map information corresponding to the specified compound molecule. graph data.

Specifically, the coordinates of each atom in the specified compound molecule in the specified coordinate system can be determined with reference to international standards, and the category vector of each atom in the specified compound molecule adopts a one-hot form of vector encoding.

The direction vector between any two atoms in a compound molecule and the connection information between any two atoms in a specified compound molecule can be expressed as matrix data.

The direction vector between two atoms of a specified compound molecule can be expressed as a directional vector obtained by fully connecting all atoms in the molecule. The vector dimension can be 3*N (N= , where n is the number of atoms in the molecule). Among them, the terminal device can define different connection information according to different atoms. For example, the connection information between two atoms AA can be represented by "0", and the connection information between two atoms AB can be represented by "1".

The distance between atoms in a specified compound molecule and the distance unit vector used to characterize it can be determined by the following formula:

.

As can be seen from the above content, since this specification fully considers the attributes and positions of each atom in the specified compound molecule when determining the three-dimensional molecular map information of the designated compound molecule, it is guaranteed that the finally determined three-dimensional molecular map information can be comprehensive Characterize the characteristics of the specified compound molecule, thereby ensuring the accuracy and rationality of the prediction results of the subsequent prediction model.

S103: Input the three-dimensional molecular graph into the prediction model to be trained, so that through the prediction model, based on the equivariance between the position information and the embedded features corresponding to the position information, and the attribute information and the The attribute information corresponds to the invariance between the embedded features, and the three-dimensional molecular graph features corresponding to the specified compound molecules are determined based on the molecular graph information of the three-dimensional molecular graph.

S104: Predict the molecular energy information corresponding to the specified compound molecule based on the three-dimensional molecular graph features.

The terminal device can input the three-dimensional molecular diagram of the specified compound molecule into the prediction model to be trained, so that through the prediction model, based on the above-mentioned position information and the equivariance between the position information corresponding embedding features, and the attribute information and the attribute information corresponding embedding Invariance between features, the embedded features corresponding to the position information and attribute information of the specified compound molecules are determined based on the molecular graph information of the three-dimensional molecular graph, and then the three-dimensional molecular graph features corresponding to the specified compound molecules are determined based on the embedded features.

The specific description of equivariability and invariance is as follows:

for functions ,where x/> X, f (x)/> Y, there is an E(3) equivariant group G acting on X and Y, if for all x/> X and all g/> G, if f (G(x))=G(f (x)), then the function f is said to be equivariant with respect to G, if for all x/> X and all g/> G, if f (G(x))=f (x), then the function f is said to be invariant with respect to G.

In this specification, the above prediction model can be equipped with a graph attention mechanism network and a multi-layer perceptron based on three-dimensional equivariance, and the prediction model finally predicts the energy information of the specified compound molecule based on the input three-dimensional molecular graph information. The process can actually be viewed as a graph embedding-feature regression process.

Specifically, after the terminal device inputs the three-dimensional molecular graph information of the specified compound molecule into the prediction model to be trained, the graph attention mechanism network can determine the atom type and coordinate information within the compound molecule, as well as the inter-atom relationship based on the three-dimensional molecular graph information. The embedded features corresponding to the direction vector and connection information are then determined in each embedded feature to determine the invariant features and equal variable features corresponding to the compound molecule, and then the three-dimensional molecular graph features of the specified compound molecule are determined based on the invariant features and equal variable features. .

The terminal device can then input the three-dimensional molecular graph features into the multi-layer perceptron, and determine the energy information corresponding to the specified compound molecule through the multi-layer perceptron.

Among them, when determining the embedded features of the molecular graph of the specified compound (including atomic coordinates and types, embedded features corresponding to the direction vectors and connection information between two atoms), it can be determined in the following ways:

Among them, a and b represent different atoms, f is the embedded eigenvalue, C is the Clebsch-Gordan coefficient, Y is the real spherical harmonic function, and R is the learnable function about distance. The embedding method of the above molecules can make the embedded features have SE(3) invariance for atomic coordinates, and satisfy SE(3) invariance for atom types, direction vectors and connection information between any two atoms.

Of course, in practical applications, the above-mentioned embedded features can also be determined through other implementable methods, and this specification will not illustrate them one by one.

After obtaining the above embedded features, the embedded features can actually be further processed through the attention mechanism to obtain more accurate three-dimensional molecular graph features. Specifically, the graph attention mechanism network can determine the attention weight for the specified compound molecule, and then, based on the attention weight and the determined embedding features, determine the invariant features and equivariable features for the specified compound molecule, and then based on This invariant feature and the equivariable feature determine the three-dimensional molecular graph features of the specified compound molecule.

The above-mentioned process of determining the three-dimensional molecular graph characteristics of a specified compound molecule through the attention mechanism will be described below with a specific example.

The embedding form of query (Q) and key (K) can be obtained through the three-dimensional molecular graph attention mechanism as follows:

in, Used to represent the Q characteristic of atom a in the specified compound molecule,/> Used to represent the K characteristics of atoms a←b in the specified compound molecule. Among them,/> for/> abbreviation expression,/> for abbreviation expression. At this time, the attention function is expressed as:

An attention mechanism is introduced into the molecular graph embedding formula.

Furthermore, the terminal device can predict the molecular energy information corresponding to the specified compound molecules through the pooling function of the prediction network and the multi-layer perceptron.

Then, with the optimization goal of minimizing the deviation between the predicted molecular energy information and the actual molecular energy information corresponding to the specified compound molecule, the prediction model is trained. This process can be expressed as:

argmin{sum(ΔE=|E _p -E _t |)}

In this way, in the subsequent process of predicting the molecular structure, the prediction model can be used to quickly and accurately predict the molecular energy, thereby improving the efficiency and accuracy of exploring the molecular structure-energy relationship.

In this specification, the three-dimensional molecular graph characteristics of a specified compound molecule can be determined through the graph attention mechanism. As mentioned above, the process of predicting energy based on the three-dimensional molecular graph characteristics of a specified compound molecule can be regarded as a multi-layer perceptron that predicts energy based on the invariant characteristics and equivariable characteristics of each atom in the specified compound molecule. Predict new atoms or new molecular structures that can be connected to the specified compound molecule. Therefore, when determining the characteristics of the above three-dimensional molecular graph, you can actually obtain the updated value of each atom in the specified compound molecule through the connection of the perceptron. Invariant features and equal variable features are used to predict energy.

It should be noted that in this specification, the above prediction models can actually be trained together using joint training. That is, after the molecular coordinate embedding characteristics of the compound are determined, the equivariant graph attention mechanism neural network and the multi-layer perceptron are adjusted together with minimizing the loss value as the optimization goal. Among them, this loss value can actually be determined by the deviation between the target predicted energy information and the real energy information corresponding to the specified compound molecule.

After training the above prediction model, the molecular structure information can be predicted through the trained prediction model to achieve recommendation of molecular structure information. The specific process is shown in the figure below.

Figure 2 is a schematic flow chart of a task execution method based on molecular energy information provided in this specification, including the following steps:

S201: Receive a task execution request for an original compound, and obtain characterization data of the original compound molecule.

S202: Process the characterization data of the original compound molecules to determine the three-dimensional molecular diagram information corresponding to the original compound molecules.

After the prediction model meets the training goal (such as reaching a preset training number or converging to a preset range), the terminal device can deploy it to perform downstream tasks.

In this specification, after receiving a task execution request for the original compound, the terminal device can obtain the prediction instruction of the molecular structure input by the user, so as to obtain the characterization data of the original designated compound molecule based on the prediction instruction.

Afterwards, the characterization data is reconstructed and optimized to obtain the three-dimensional molecular graph information of the specified compound molecule. The determination of the three-dimensional molecular graph information here is basically the same as the determination process of the three-dimensional molecular graph information in the above model training. Here I won’t go into details. The terminal devices mentioned here may refer to desktop computers, laptops and other devices.

S203: Input the three-dimensional molecular graph information of the original compound molecule into the pre-trained prediction model, so as to determine the three-dimensional molecular graph characteristics corresponding to the original compound molecule through the prediction model, and according to the corresponding three-dimensional molecular graph characteristics of the original compound molecule Three-dimensional molecular graph features are used to determine the molecular energy information corresponding to the original compound molecules, where the prediction model is trained by the above model training method.

The terminal device can input the three-dimensional molecular graph information of the original designated compound molecule into the prediction model deployed in the terminal device, and the prediction model will output energy information that is combined with the original designated compound molecule to form a molecule with preset functions.

It should be pointed out that since in the above model training process, the prediction model has been trained through supervised training, so in the actual application process, the reinforcement learning model can no longer be used, that is, the molecules output by the prediction model The final energy information does not require further screening.

S204: Perform a task according to the molecular energy information corresponding to the original compound molecule.

After determining the molecular energy information of the specified compound molecule, the terminal device can determine the corresponding relationship between the molecular structure and molecular energy of the original compound molecule based on the molecular energy information corresponding to the original compound molecule and the characterization data corresponding to the original compound molecule. Then, the molecular design task or the molecular regulation task is performed based on the corresponding relationship.

Of course, the terminal device can also recommend the corresponding relationship between the molecular structure and molecular energy of the original compound molecule as the information to be recommended to the user.

Furthermore, this specification also provides a system for exploring the molecular structure-energy relationship of compounds, as shown in Figure 3.

Figure 3 is a schematic diagram of the architecture of a system for exploring the molecular structure-energy relationship of compounds provided in this specification.

As can be seen from Figure 3, the system mainly consists of the following parts:

The storage subsystem is used to store the above-mentioned data sets, as well as the molecular energy information and given structural information predicted by the prediction model in practical applications.

The control subsystem is used to predict the energy information of the compound molecules based on the three-dimensional molecular graph information of the original specified compound molecules input into the subsystem.

The control subsystem contains three units: a molecular feature extraction unit, a molecular energy prediction training unit, and a molecular energy prediction output unit. These three units are used in turn to obtain three-dimensional molecular graph features and molecular structure-prediction model training result information. , and obtain molecular energy information based on the newly input molecular structure information.

It can be seen from the above methods that compared with existing solutions, the prediction model in this application will greatly improve the efficiency and accuracy of relationship construction. With the continuous development of fields such as artificial intelligence and machine learning and the trend towards molecular science and technology, With continuous penetration, the performance prediction and design of molecules have become more complex than the earlier fully connected networks. Molecular structure-energy relationship prediction models based on artificial intelligence or machine learning rely on the representation of molecular characteristics. Using molecular graph representation, atoms and chemical bonds within the molecule can be regarded as nodes and edges to describe the structural characteristics of the molecule. By combining the above deep network model methods, developers can effectively construct the structure-energy relationship of compound molecules to provide powerful technical methods and means for downstream tasks such as studying molecular energy-related dynamic properties and designing molecules with certain functions according to user requirements.

However, the methods currently used cannot more comprehensively describe the characteristics of molecules and cannot efficiently predict the molecular structure-energy relationship information of a specified compound.

Therefore, in the process of determining the three-dimensional molecular diagram information of the specified compound molecule, the model training method provided in this specification comprehensively refers to the characterization data of various molecular structures of the specified compound molecule, which can make the final determined three-dimensional molecule Graph information can comprehensively characterize the molecular structural characteristics of a specified compound molecule.

Moreover, when determining the three-dimensional molecular graph characteristics of a designated compound molecule, it is determined based on the invariant characteristics and equivariable characteristics of the designated compound molecule. In this way, the molecular structure characteristics of the designated compound molecule can be fully characterized. , thereby ensuring that the prediction model can subsequently predict accurate and reasonable energy information by specifying the three-dimensional molecular graph characteristics of compound molecules.

The above are one or more implementation methods of this specification. Based on the same idea, this specification also provides a schematic diagram of a corresponding model training device and a task execution device based on molecular energy information, as shown in Figures 4 and 5.

Figure 4 is a schematic diagram of a model training device provided in this specification, including:

The acquisition module 401 is used to obtain the characterization data of the specified compound molecule, where the characterization data is used to characterize the position information and attribute information of each atom in the specified compound molecule;

The determination module 402 is used to process the characterization data and determine the three-dimensional molecular diagram information of the specified compound molecule;

Input module 403 is used to input the three-dimensional molecular graph information into a prediction model to be trained, so that through the prediction model, based on the equivariance between the position information and the embedded features corresponding to the position information, and the The invariance between the attribute information and the embedded features corresponding to the attribute information, and determining the three-dimensional molecular graph features corresponding to the specified compound molecules based on the three-dimensional molecular graph information;

The prediction module 404 is used to predict the molecular energy information corresponding to the specified compound molecules based on the three-dimensional molecular graph features;

The training module 405 is used to train the prediction model with the optimization goal of minimizing the deviation between the predicted molecular energy information and the actual molecular energy information corresponding to the specified compound molecule.

Optionally, the acquisition module 401 is specifically configured to select the initial data of the designated compound molecule from the data set of molecular compounds; and determine the characterization data based on the initial data of the designated compound molecule.

Optionally, the position information includes: the coordinates of each atom in the specified compound molecule in a specified coordinate system;

The attribute information includes: the type of each atom in the designated compound molecule, the direction vector between any two atoms in the designated compound molecule, and the connection information between any two atoms in the designated compound molecule.

Optionally, a graph attention mechanism network is provided in the prediction model;

The input module 403 is specifically configured to determine the attention weight corresponding to the specified compound molecule through the graph attention mechanism network; based on the attention weight and the graph attention mechanism network based on the three-dimensional molecular graph Each embedded feature determined by the information determines the invariant features and equivariable features corresponding to the designated compound molecule; based on the invariant features and the equivariable features, the three-dimensional molecular graph features are determined.

Figure 5 is a schematic diagram of a task execution device based on molecular energy information provided in this specification, including:

The receiving module 501 is used to receive a task execution request for the original compound and obtain the characterization data of the original compound molecule;

The construction module 502 is used to process the characterization data of the original compound molecules and determine the three-dimensional molecular map information corresponding to the original compound molecules;

The determination module 503 is used to input the three-dimensional molecular graph information of the original compound molecule into a pre-trained prediction model, so as to determine the three-dimensional molecular graph characteristics corresponding to the original compound molecule through the prediction model, and determine the three-dimensional molecular graph characteristics of the original compound molecule according to the original compound molecule. The three-dimensional molecular graph characteristics corresponding to the compound molecules determine the molecular energy information corresponding to the original compound molecules, wherein the prediction model is trained by the above-mentioned model training method;

The execution module 504 is used to execute tasks according to the molecular energy information corresponding to the original compound molecules.

Optionally, the tasks include: molecular regulation tasks or molecular design tasks;

The execution module 504 is specifically configured to determine the corresponding relationship between the molecular structure and molecular energy of the original compound molecule according to the molecular energy information corresponding to the original compound molecule and the characterization data corresponding to the original compound molecule; based on The corresponding relationship performs the molecular design task or the molecular regulation task.

This specification also provides a computer-readable storage medium that stores a computer program. The computer program can be used to perform a model training method provided in Figure 1 or a task based on molecular energy information provided in Figure 2 execution method.

This specification also provides a schematic structural diagram of the electronic device shown in FIG. 6 corresponding to FIG. 1 or FIG. 2 . As shown in Figure 6, at the hardware level, the electronic device includes a processor, internal bus, network interface, memory and non-volatile memory, and of course may also include other hardware required for business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to implement the model training method described in Figure 1 or the recommended method of molecular structure information described in Figure 2.

The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or A combination of any of these devices.

For the convenience of description, when describing the above device, the functions are divided into various units and described separately. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and/or hardware.

Those skilled in the art will understand that embodiments of the present specification may be provided as methods, systems, or computer program products. Thus, the present description may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment that combines software and hardware aspects. Furthermore, the present description may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

This specification is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the specification. It will be understood that each process and/or block in the flowchart illustrations and/or block diagrams, and combinations of processes and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce a use A device for realizing the functions specified in a process or processes in a flowchart and/or a block or blocks in a block diagram.

These computer program instructions may also be stored in a computer-readable memory that causes a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including the instruction means, the instructions The device implements the functions specified in a process or processes in the flowchart and/or in a block or blocks in the block diagram.

These computer program instructions may also be loaded onto a computer or other programmable data processing device, causing a series of operating steps to be performed on the computer or other programmable device to produce computer-implemented processing, thereby executing on the computer or other programmable device. Instructions provide steps for implementing the functions specified in a process or processes of a flowchart diagram and/or a block or blocks of a block diagram.

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

Memory may include non-permanent storage in computer-readable media, random access memory (RAM), and/or non-volatile memory in the form of read-only memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

Computer-readable media includes both persistent and non-volatile, removable and non-removable media that can be implemented by any method or technology for storage of information. Information may be computer-readable instructions, data structures, modules of programs, or other data.

It should also be noted that the terms "comprises," "comprises," or any other variation thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements not only includes those elements, but also includes Other elements are not expressly listed or are inherent to the process, method, article or equipment. Without further limitation, an element defined by the statement "comprises a..." does not exclude the presence of additional identical elements in a process, method, article, or device that includes the stated element.

Those skilled in the art will appreciate that embodiments of the present specification may be provided as methods, systems, or computer program products. Thus, the present description may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment that combines software and hardware aspects. Furthermore, the present description may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

This specification may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present description may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through communications networks. In a distributed computing environment, program modules may be located in both local and remote computer storage media including storage devices.

Each embodiment in this specification is described in a progressive manner. The same and similar parts between the various embodiments can be referred to each other. Each embodiment focuses on its differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple. For relevant details, please refer to the partial description of the method embodiment.

The above descriptions are only examples of this specification and are not intended to limit this specification. Various modifications and variations may occur to those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this specification shall be included in the scope of the claims of this specification.

Claims (10)

1. A model training method, characterized by comprising: Obtain characterization data of the specified compound molecule, the characterization data being used to characterize the position information and attribute information of each atom in the specified compound molecule; Process the characterization data to determine the three-dimensional molecular map information of the designated compound molecule; The three-dimensional molecular graph information is input into a prediction model to be trained, so that through the prediction model, based on the equivariance between the position information and the embedded features corresponding to the position information, and the attribute information and the attribute The information corresponds to the invariance between the embedded features, and the three-dimensional molecular graph features corresponding to the specified compound molecules are determined based on the three-dimensional molecular graph information; Predict the molecular energy information corresponding to the specified compound molecule according to the three-dimensional molecular graph characteristics; The prediction model is trained with an optimization goal of minimizing the deviation between the predicted molecular energy information and the actual molecular energy information corresponding to the specified compound molecule. 2. The method of claim 1, wherein obtaining characterization data of designated compound molecules specifically includes: Select the initial data of the designated compound molecules from the data set of molecular compounds; The characterization data is determined based on initial data for the designated compound molecule. 3. The method of claim 1, wherein the position information includes: the coordinates of each atom in the designated compound molecule in a designated coordinate system; The attribute information includes: the type of each atom in the designated compound molecule, the direction vector between any two atoms in the designated compound molecule, and the connection information between any two atoms in the designated compound molecule. 4. The method of claim 1, wherein a graph attention mechanism network is provided in the prediction model; Determining the three-dimensional molecular map characteristics corresponding to the specified compound molecule based on the three-dimensional molecular map information specifically includes: Determine the attention weight corresponding to the specified compound molecule through the graph attention mechanism network; According to the attention weight and each embedded feature determined by the graph attention mechanism network based on the three-dimensional molecular graph information, determine the invariant features and equal variable features corresponding to the designated compound molecule; The three-dimensional molecular graph features are determined based on the invariant features and the equivariable features. 5. A task execution method based on molecular energy information, characterized by including: Receive a task execution request for the original compound and obtain characterization data of the original compound molecule; Process the characterization data of the original compound molecules to determine the three-dimensional molecular map information corresponding to the original compound molecules; The three-dimensional molecular graph information of the original compound molecule is input into the pre-trained prediction model, so that the three-dimensional molecular graph characteristics corresponding to the original compound molecule are determined through the prediction model, and the three-dimensional molecular graph corresponding to the original compound molecule is determined based on the three-dimensional molecular graph information of the original compound molecule. graph features to determine the molecular energy information corresponding to the original compound molecules, wherein the prediction model is trained by the method described in any one of the above claims 1 to 4; Perform tasks based on the molecular energy information corresponding to the original compound molecules. 6. The method of claim 5, wherein the task includes: a molecular regulation task or a molecular design task; Perform tasks based on the molecular energy information corresponding to the original compound molecules, specifically including: Determine the corresponding relationship between the molecular structure and molecular energy of the original compound molecule according to the molecular energy information corresponding to the original compound molecule and the characterization data corresponding to the original compound molecule; The molecular design task or the molecular regulation task is performed based on the corresponding relationship. 7. A model training device, characterized by comprising: An acquisition module, used to obtain characterization data of a designated compound molecule, where the characterization data is used to characterize the position information and attribute information of each atom in the designated compound molecule; A determination module, used to process the characterization data and determine the three-dimensional molecular diagram information of the specified compound molecule; An input module for inputting the three-dimensional molecular graph information into a prediction model to be trained, so that through the prediction model, based on the equivariance between the position information and the embedded features corresponding to the position information, and the attributes The invariance between the information and the embedded features corresponding to the attribute information, and determining the three-dimensional molecular graph features corresponding to the specified compound molecules based on the three-dimensional molecular graph information; A prediction module, used to predict the molecular energy information corresponding to the specified compound molecule based on the three-dimensional molecular graph characteristics; A training module, configured to train the prediction model with an optimization goal of minimizing the deviation between the predicted molecular energy information and the actual molecular energy information corresponding to the specified compound molecule. 8. A task execution device based on molecular energy information, characterized by including: The receiving module is used to receive the task execution request for the original compound and obtain the characterization data of the original compound molecule; A building module for processing the characterization data of the original compound molecules and determining the three-dimensional molecular diagram information corresponding to the original compound molecules; A determination module, configured to input the three-dimensional molecular graph information of the original compound molecule into a pre-trained prediction model, so as to determine the three-dimensional molecular graph characteristics corresponding to the original compound molecule through the prediction model, and determine the characteristics of the three-dimensional molecular graph corresponding to the original compound molecule according to the original compound molecule. The three-dimensional molecular graph characteristics corresponding to the molecules are used to determine the molecular energy information corresponding to the original compound molecules, wherein the prediction model is trained by the method described in any one of the above claims 1 to 4; An execution module is used to execute tasks according to the molecular energy information corresponding to the original compound molecules. 9. A computer-readable storage medium, characterized in that the storage medium stores a computer program, and when the computer program is executed by a processor, the method of any one of claims 1 to 6 is implemented. 10. An electronic device, including a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that when the processor executes the program, any one of claims 1 to 6 is realized. method described in the item.