Drug molecule screening method and system

Through the drug molecular screening method based on molecular directed information descriptors and self-attention mechanism, the problems of insufficient flexibility and information redundancy in molecular information processing are solved, and more accurate molecular description and prediction effects are achieved.

CN114530210BActive Publication Date: 2024-11-08SHANDONG NORMAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210013833.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-06
Publication Date
2024-11-08
Estimated Expiration
2042-01-06

AI Technical Summary

Technical Problem

The existing molecular fingerprint and graph convolutional neural networks have problems such as insufficient flexibility and information redundancy in molecular information processing, resulting in reduced molecular description capabilities and inaccurate predictions.

Method used

A drug molecular screening method based on molecular orientation information descriptors and self-attention mechanism is adopted. By acquiring molecular orientation information descriptors and combining convolutional neural networks and self-attention modules, it avoids information cyclic collection and extracts more comprehensive molecular information features.

Benefits of technology

It improves the flexibility and accuracy of molecular description, enhances the classification ability of the model, avoids information redundancy, and improves the accuracy of molecular properties prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114530210B_ABST
    Figure CN114530210B_ABST
Patent Text Reader

Abstract

The present invention provides a drug molecule screening method and system, which belongs to the field of biological information processing technology, including: obtaining drug molecules to be screened; obtaining molecular orientation information descriptors based on the molecular formula of the drug molecules; wherein the molecular orientation information descriptors are composed of multiple one-dimensional vectors characterizing the molecular bond characteristics of the drug molecules; using a pre-trained classification model to process the molecular orientation information descriptors to obtain molecular property classification results. The present invention can flexibly encode molecules, and molecular information avoids circular collection, thereby improving the accuracy of molecular description; a convolutional neural network combined with a self-attention module is used to process molecular descriptors, extracting more and more comprehensive molecular information features, and improving the model classification capability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of biological information processing, and in particular to a drug molecule screening method and system based on molecule-oriented information descriptors and a self-attention mechanism. Background Art

[0002] Molecular property prediction is a chemical information processing technology. The development of deep learning networks in recent years has greatly improved the effect of molecular property prediction. There are two types of models that have shown good molecular description capabilities and have produced good results in molecular property prediction. One is a neural network used to calculate molecular fingerprints or molecular descriptors made by experts, and the other is a graph convolutional neural network that learns molecular features based on the molecular graph structure.

[0003] Representatives of molecular fingerprints include RDKit molecular fingerprint, Morgan molecular fingerprint, etc. Representatives of graph convolutional neural networks include Gated Graph Neural Networks (GG-NN), Molecular Graph Convolutions, Neural Message Passing for Quantum Chemistry, etc.

[0004] Although the above two types of model methods perform well, they each have their own defects. For example, when a fixed molecular fingerprint is used to describe a molecule, the work of collecting molecular information is not flexible; molecular graph convolutional neural networks mostly use atoms in molecules as the center to transmit molecular information, and atomic information will be collected cyclically, resulting in repeated and redundant information, thereby reducing the ability to describe molecules; molecular property classification networks mostly use feed-forward neural networks (FFN), Random Forest (RF) and other networks with simpler structures, which are insufficient in extracting key molecular information and are not accurate in predicting structures. Summary of the invention

[0005] The purpose of the present invention is to provide a drug molecule screening method and system based on molecular directional information descriptors and self-attention mechanisms, which avoids the cyclic collection of molecular information and extracts more comprehensive molecular key information, so as to solve at least one technical problem existing in the above-mentioned background technology.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] In one aspect, the present invention provides a method for screening drug molecules, comprising:

[0008] Obtain drug molecules to be screened;

[0009] Based on the molecular formula of the drug molecule, a molecular orientation information descriptor is obtained; wherein the molecular orientation information descriptor is composed of multiple one-dimensional vectors characterizing the molecular bond characteristics of the drug molecule;

[0010] The molecular orientation information descriptor is processed using a pre-trained classification model to obtain a molecular property classification result.

[0011] Preferably, obtaining the molecular orientation information descriptor includes: starting from the initial molecular bond, collecting molecular information along a fixed direction, and updating the collected molecular bond information.

[0012] Preferably, the pre-trained classification model uses a 1DCNN network cascading multiple convolutional layers, multiple maximum pooling layers and multiple dropout layers.

[0013] Preferably, the molecular orientation information descriptor is first input into a cascade of two one-dimensional convolutional layers and activated with a tanh function; after the output of the last discard layer, it passes through a flattening layer and is input into two fully connected layers, and the molecular properties are classified by a keras classifier.

[0014] Preferably, a self-attention module is embedded before the flattening layer to capture key information in the molecular structure and increase the weight of key information in the molecular property classification process.

[0015] Preferably, the number of neurons in the first fully connected layer is 540, and the activation function is the relu function; the number of neurons in the second layer is 2, and the activation function is the softmax function; the second fully connected layer is used for binary classification operations, and finally outputs the classification probability of each category of molecular properties.

[0016] In a second aspect, the present invention provides a drug molecule screening system, comprising:

[0017] An acquisition module, used to acquire drug molecules to be screened;

[0018] A conversion module, used to obtain a molecular orientation information descriptor based on the molecular formula of the drug molecule; wherein the molecular orientation information descriptor is composed of multiple one-dimensional vectors characterizing the molecular bond characteristics of the drug molecule;

[0019] The classification module is used to process the molecular orientation information descriptor using a pre-trained classification model to obtain a molecular property classification result.

[0020] In a third aspect, the present invention provides a non-transitory computer-readable storage medium, which is used to store computer instructions. When the computer instructions are executed by a processor, the drug molecule screening method as described above is implemented.

[0021] In a fourth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed on one or more processors, is used to implement the drug molecule screening method as described above.

[0022] In a fifth aspect, the present invention provides an electronic device comprising: a processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes instructions for implementing the drug molecule screening method as described above.

[0023] The beneficial effects of the present invention are as follows: molecules can be flexibly encoded, and molecular information avoids circular collection, thereby improving the accuracy of molecular description; a convolutional neural network is used in combination with a self-attention module to process molecular descriptors, thereby extracting more comprehensive molecular information features and improving the model classification capability.

[0024] Additional aspects and advantages of the present invention will be given in part in the following description, which will become obvious from the following description, or may be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0026] Figure 1 This is a diagram of the network process architecture for drug molecule screening based on molecule-oriented information descriptors and self-attention mechanisms according to an embodiment of the present invention.

[0027] Figure 2 This is a schematic diagram of generating a molecular orientation information descriptor according to an embodiment of the present invention by converting a data set into molecules.

[0028] Figure 3 The figure is a schematic diagram of the molecular orientation information descriptor information collection process according to an embodiment of the present invention.

[0029] Figure 4 This is a ROC curve diagram of the BBBP data set running results described in an embodiment of the present invention.

[0030] Figure 5 This is a comparison chart of test running results using different models described in an embodiment of the present invention. DETAILED DESCRIPTION

[0031] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below by the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be interpreted as limiting the present invention.

[0032] It should be understood by those skilled in the art that unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art to which this invention belongs.

[0033] It should also be understood that terms, such as those defined in commonly used dictionaries, should be understood to have a meaning consistent with that in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless as defined herein.

[0034] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements and / or groups thereof.

[0035] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0036] To facilitate understanding of the present invention, the present invention is further explained below with reference to specific embodiments in conjunction with the accompanying drawings, and the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0037] Those skilled in the art should understand that the drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily necessary for implementing the present invention.

[0038] Example 1

[0039] like Figure 1 As shown, this embodiment 1 provides a drug molecule screening system, which includes:

[0040] An acquisition module, used to acquire drug molecules to be screened;

[0041] A conversion module, used to obtain a molecular orientation information descriptor based on the molecular formula of the drug molecule; wherein the molecular orientation information descriptor is composed of multiple one-dimensional vectors characterizing the molecular bond characteristics of the drug molecule;

[0042] The classification module is used to process the molecular orientation information descriptor using a pre-trained classification model to obtain a molecular property classification result.

[0043] In this embodiment 1, a drug molecule screening method is implemented using the above system, including:

[0044] Use the acquisition module to obtain drug molecules to be screened;

[0045] Using a conversion module, based on the molecular formula of the drug molecule, a molecular orientation information descriptor is obtained; wherein the molecular orientation information descriptor is composed of multiple one-dimensional vectors characterizing the molecular bond characteristics of the drug molecule;

[0046] The molecular orientation information descriptor is processed using a classification module based on a pre-trained classification model to obtain a molecular property classification result.

[0047] Acquiring the molecular orientation information descriptor includes: taking the initial molecular bond as a starting point, collecting molecular information along a fixed direction, and updating the collected molecular bond information.

[0048] Specifically, a directed message passing network is used to process the data set, and the molecular formula in the data set is converted into a 300-bit directed molecular information descriptor, which is a 300-bit one-dimensional vector. The directed molecular information descriptor is obtained as follows:

[0049] First, the hidden features of the molecular bond connecting atoms a and b in their original state

[0050]

[0051] in, represents the hidden features of the molecular bond connecting atoms a and b in the original state, ε represents the activation function of the activation function, and Z i represents the learning matrix, x a represents the characteristic information of a atom, l ab represents the characteristic information of the molecular bond (ab bond) from atom a to atom b, c(x a ,l ab ) represents the connection of the characteristics of a atom and the characteristics of ab bond to form a matrix.

[0052] Next, get the information about the molecular bond from atom a to atom b:

[0053]

[0054]

[0055] Among them, the c atom represents all atoms connected to the atom except the b atom. represents the characteristic information of the molecular bond from atom c to atom a in the tth iteration, x c Represents the characteristic information of c atoms, It represents the information transmitted to atom a by all atoms connected to atom a except atom b. After the information is accumulated, it needs to be transmitted from atom a to atom b along the ab bond. g represents the learning matrix, Q(a) is the set of atoms around a. t stands for the information transfer function, which is an arbitrary function that depends on the parameters. Note that The same as the previous iteration of the experimental operation It doesn't matter.

[0056] Finally, each molecule needs to be represented by atoms. Taking atom a as an example, the molecular bond characteristic information of atom a needs to be accumulated and summed:

[0057]

[0058] K a =ε(Z j c(x a ,g a ));

[0059] g a Represents the cumulative result of all the information transmitted to the outside by atom a, Z j represents the learning matrix, c(x a ,g a ) will combine the characteristics of atom a and g a The information contained in the two is connected to form a matrix, K a Represents all the characteristic information contained in atom a.

[0060] To obtain the characteristic information of a molecule, it is necessary to sum the characteristic information of all atoms:

[0061] Among them, v represents all the atoms that make up the molecule, K represents the sum of all the atomic information, and K x Represents the information of an atom in a molecule.

[0062] Ultimately, what is obtained is a 300-bit directional molecular information descriptor for each target molecule.

[0063] The pre-trained classification model uses a 1DCNN network to cascade multiple convolutional layers, multiple maximum pooling layers, and multiple dropout layers.

[0064] The molecular orientation information descriptor is first input into a cascade of two one-dimensional convolutional layers and activated with a tanh function; after the output of the last layer is discarded, it passes through a flattening layer and is input into two fully connected layers, and the molecular properties are classified by a keras classifier.

[0065] A self-attention module is embedded before the flattening layer to capture the key information in the molecular structure and increase the weight of the key information in the molecular property classification process.

[0066] The number of neurons in the first fully connected layer is 540, and the activation function is the relu function. The number of neurons in the second layer is 2, and the activation function is the softmax function. The second fully connected layer is used for binary classification operations and finally outputs the classification probability of each category of molecular properties.

[0067] Specifically, in this embodiment 1, when processing the directional molecular information descriptor, a convolutional neural network is used to process it. The network model is as follows: Figure 1 As shown in Figure 2, the 1DCNN network used in the model cascades 4 one-dimensional convolutional layers, 3 maximum pooling layers, and 2 dropout layers.

[0068] In this embodiment 1, the first part of the neural network consists of two one-dimensional convolutional layers and a maximum pooling layer. The data is first input into the cascade of two one-dimensional convolutional layers. The number of convolution kernels in the first one-dimensional convolutional layer is 20, and the convolution kernel size is 3; the number of convolution kernels in the second one-dimensional convolutional layer is 60, and the convolution kernel size is 3; the tanh function is selected as the activation method of the two one-dimensional convolutional layers, and the step size is set to 1. Then, the output of the second one-dimensional convolutional layer is input into the maximum pooling layer with a pooling window size of 3. The purpose of this layer is to retain the main features to prevent overfitting and improve the operation rate.

[0069] The second and third parts of the neural network are composed of a one-dimensional convolutional layer, a maximum pooling layer and a dropout layer. After each maximum pooling layer, a dropout layer is connected to prevent overfitting.

[0070] The one-dimensional convolutional layers in the second and third parts have 60 convolution kernels, a convolution kernel size of 3, a tanh function as the activation method, and a step size of 1; the window parameters of the two maximum pooling layers are set to 3; the drop probability of the drop layer in the second part is set to 0.5, and the drop probability of the drop layer in the third part is set to 0.2.

[0071] After the data is output by the third part of the discard layer, it is input to the self-attention module. After being processed by the self-attention module, the output of the self-attention module is connected to the fully connected layer using the flattening layer. There are two fully connected layers in this part. The number of neurons in the first fully connected layer is 540, and the activation function is the relu function. The number of neurons in the second layer is 2, and the activation function is the softmax function. The second fully connected layer is used for binary classification operations, and finally outputs the classification probability of each category, thereby outputting the predicted label of the molecule. For example: After the network predicts molecule A, the probability of 1 is 0.75, and the probability of 0 is 0.25, and then it can be concluded that the predicted label of molecule A is 1.

[0072] In this embodiment 1, the loss function for experimental model training is the cross entropy loss function (categorical cross entropy), the optimizer is the Adam optimizer, and the evaluation indicator is the AUC value.

[0073] The present invention conducted experiments on BBBP (Blood-brain barrier penetration) dataset, ClinTox dataset, SIDER (Side Effect Resource) dataset and Tox21 (Toxicology in the 21st Century) dataset. The molecular formula in each dataset was processed by a directed message passing network to obtain 300 directional molecular information descriptors, which were input into the network classification model proposed in this embodiment, and processed jointly by a convolutional neural network and a self-attention module to finally classify the molecular properties.

[0074] In this embodiment 1, the labeled data set is first trained, the best performing model parameters are selected, and then the best performing model is saved to predict molecules with unknown labels.

[0075] Example 2

[0076] In this embodiment 2, a drug molecule screening algorithm based on molecule-oriented information descriptors and self-attention mechanism is provided. The overall architecture of the algorithm is as follows: Figure 1 As shown, it includes two parts: the generation of molecular orientation information descriptors and the operation of convolutional neural networks. The specific steps are as follows:

[0077] The molecular orientation information descriptor is converted from the molecular formula in the data set. The conversion process is as follows: Figure 2 The information collection process of the molecular orientation information descriptor is shown in Figure 3 The specific acquisition method is as follows:

[0078] First, the hidden features of the atomic bond connecting atoms a and b in their original state

[0079]

[0080] in, represents the hidden features of the molecular bond connecting atoms a and b in the original state, ε represents the activation function of the activation function, and Z i represents the learning matrix, x a represents the characteristic information of a atom, l ab represents the characteristic information of the molecular bond (ab bond) from atom a to atom b, c(x a ,l ab ) represents the connection of the characteristics of a atom and the characteristics of ab bond to form a matrix.

[0081] Next, get the information about the molecular bond from atom a to atom b:

[0082]

[0083]

[0084] Among them, the c atom represents all atoms connected to the atom except the b atom. represents the characteristic information of the molecular bond from atom c to atom a in the tth iteration, x c Represents the characteristic information of c atoms, It represents the information transmitted to atom a by all atoms connected to atom a except atom b. After the information is accumulated, it needs to be transmitted from atom a to atom b along the ab bond. g represents the learning matrix, Q(a) is the set of atoms around a. t stands for the information transfer function, which is an arbitrary function that depends on the parameters. Note that The same as the previous iteration of the experimental operation It doesn't matter.

[0085] Finally, each molecule is represented by an atom, and all the molecular bond characteristics pointing to atom a need to be accumulated and summed:

[0086]

[0087] K a =ε(Z j c(x a ,g a ));

[0088] g aRepresents the cumulative result of all the information transmitted to the outside by atom a, Z j represents the learning matrix, c(x a ,g a ) will combine the characteristics of atom a and g a The information contained in the two is connected to form a matrix, K a Represents all the characteristic information contained in atom a.

[0089] To obtain the characteristic information of a molecule, it is necessary to sum the characteristic information of all atoms:

[0090] Among them, v represents all the atoms that make up the molecule, K represents the sum of all the atomic information, and K x Represents information about an atom in a molecule.

[0091] Ultimately, what is obtained is a 300-bit directional molecular information descriptor for each target molecule.

[0092] In the second part, the obtained molecular orientation information descriptors are input into the proposed classification model for processing.

[0093] The data is first input into the first part of the classification model, which consists of two convolutional layers and a maximum pooling layer in cascade. The number of convolution kernels in the first one-dimensional convolutional layer is 20, and the convolution kernel size is 3; the number of convolution kernels in the second one-dimensional convolutional layer is 60, and the convolution kernel size is 3; the tanh function is selected as the activation method of the two one-dimensional convolutional layers, and the step size is set to 1. Then, the output of the second one-dimensional convolutional layer is input into the maximum pooling layer, and the pooling window size is 3. The output of the first part is then sequentially input into the second part of the model, and the output of the second part is then input into the third part for processing. Both the second and third parts are composed of a one-dimensional convolutional layer, a maximum pooling layer, and a dropout layer in cascade, and a dropout layer is connected after each maximum pooling layer. The one-dimensional convolution layer of the second and third parts has 60 convolution kernels, a convolution kernel size of 3, an activation method using the tanh function, and a step size of 1; the window parameter of the maximum pooling layer of the two parts is set to 3; the discard probability of the discard layer of the second part is set to 0.5, and the discard probability of the discard layer of the third part is set to 0.2. After the third part completes the data processing, the output of the third part needs to be input into the self-attention module. After being processed by the self-attention module, the self-attention module is connected to the fully connected layer using the flattening layer. There are two layers of fully connected layers in this part. The number of neurons in the first fully connected layer is 540, and the activation function is the relu function. The number of neurons in the second layer is 2, and the activation function is the softmax function. The second fully connected layer performs molecular classification operations and finally outputs the classification probability of each category, thereby judging the label of the current molecule. For example: After the network predicts that the probability of molecule A is 1 is 0.75, and the probability of 0 is 0.25, and then the label of the current molecule A is judged to be 1. Finally, the AUC value was calculated for the molecular part of the entire BBBP dataset that was classified as the prediction set to verify the overall prediction effect. Figure 4 The ROC curves of the proposed method and the D-MPNN method on the BBBP dataset and the AUC values ​​calculated from the ROC curves are shown.

[0094] In order to verify the advantages of the algorithm of Example 2 for molecular property screening, molecular property prediction experiments were conducted on the BBBP (Blood-brain barrier penetration) dataset, ClinTox dataset, SIDER (Side Effect Resource) dataset, and Tox21 (Toxicology in the 21st Century) dataset. The experimental results are as follows: Figure 5 As shown in the figure, due to the reasons of the dataset itself, it is meaningless to test the RF method on the ClinTox dataset, so no comparison was made. Figure 5It can be seen that the drug molecule screening method based on molecular directional information descriptor and self-attention mechanism established in this embodiment has achieved good results in molecular property prediction, and the AUC value is significantly higher than other methods. The higher the AUC value, the stronger the classification ability, and the prediction of molecular properties is accurate and effective.

[0095] Example 3

[0096] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the above-mentioned drug molecule screening method is implemented, and the method includes:

[0097] Obtain drug molecules to be screened;

[0098] Based on the molecular formula of the drug molecule, a molecular orientation information descriptor is obtained; wherein the molecular orientation information descriptor is composed of multiple one-dimensional vectors characterizing the molecular bond characteristics of the drug molecule;

[0099] The molecular orientation information descriptor is processed using a pre-trained classification model to obtain a molecular property classification result.

[0100] Example 4

[0101] Embodiment 4 of the present invention provides a computer program (product), including a computer program, wherein when the computer program is run on one or more processors, the computer program is used to implement the above-mentioned drug molecule screening method, the method comprising:

[0102] Obtain drug molecules to be screened;

[0103] Based on the molecular formula of the drug molecule, a molecular orientation information descriptor is obtained; wherein the molecular orientation information descriptor is composed of multiple one-dimensional vectors characterizing the molecular bond characteristics of the drug molecule;

[0104] The molecular orientation information descriptor is processed using a pre-trained classification model to obtain a molecular property classification result.

[0105] Example 5

[0106] Embodiment 5 of the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes instructions for implementing the above-mentioned drug molecule screening method, the method comprising:

[0107] Obtain drug molecules to be screened;

[0108] Based on the molecular formula of the drug molecule, a molecular orientation information descriptor is obtained; wherein the molecular orientation information descriptor is composed of multiple one-dimensional vectors characterizing the molecular bond characteristics of the drug molecule;

[0109] The molecular orientation information descriptor is processed using a pre-trained classification model to obtain a molecular property classification result.

[0110] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0111] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0112] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0113] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0114] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Technical personnel in the relevant field should understand that various modifications or variations that can be made by those skilled in the art without creative work on the basis of the technical solution disclosed in the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for screening drug molecules, It is characterized in that include: Obtain drug molecules to be screened; Based on the molecular formula of the drug molecule, a molecular orientation information descriptor is obtained; wherein the molecular orientation information descriptor is composed of multiple one-dimensional vectors characterizing the molecular bond characteristics of the drug molecule; Using a pre-trained classification model, processing the molecular orientation information descriptor to obtain a molecular property classification result; Acquiring the molecular orientation information descriptor includes: starting from the initial molecular bond, collecting molecular information along a fixed direction, and updating the collected molecular bond information; The pre-trained classification model uses a 1DCNN network cascaded with multiple convolutional layers, multiple maximum pooling layers and multiple dropout layers; The molecular orientation information descriptor is first input into a cascade of two one-dimensional convolutional layers and activated with a tanh function; after the output of the last layer is discarded, it passes through a flattening layer and is input into two fully connected layers, and the molecular properties are classified by a keras classifier.

2. The drug molecule screening method according to claim 1, It is characterized in that A self-attention module is embedded before the flattening layer to capture the key information in the molecular structure and increase the weight of the key information in the molecular property classification process.

3. The drug molecule screening method according to claim 2, It is characterized in that The number of neurons in the first fully connected layer is 540, and the activation function is the relu function. The number of neurons in the second layer is 2, and the activation function is the softmax function. The second fully connected layer is used for binary classification operations and finally outputs the classification probability of each category of molecular properties.

4. A drug molecule screening system, It is characterized in that include: An acquisition module, used to acquire drug molecules to be screened; A conversion module, used to obtain a molecular orientation information descriptor based on the molecular formula of the drug molecule; wherein the molecular orientation information descriptor is composed of multiple one-dimensional vectors characterizing the molecular bond characteristics of the drug molecule; A classification module, used to process the molecular orientation information descriptor using a pre-trained classification model to obtain a molecular property classification result; Acquiring the molecular orientation information descriptor includes: starting from the initial molecular bond, collecting molecular information along a fixed direction, and updating the collected molecular bond information; The pre-trained classification model uses a 1DCNN network cascaded with multiple convolutional layers, multiple maximum pooling layers and multiple dropout layers; The molecular orientation information descriptor is first input into a cascade of two one-dimensional convolutional layers and activated with a tanh function; after the output of the last layer is discarded, it passes through a flattening layer and is input into two fully connected layers, and the molecular properties are classified by a keras classifier.

5. A non-transitory computer-readable storage medium, It is characterized in that The non-transitory computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by the processor, the drug molecule screening method according to any one of claims 1 to 3 is implemented.

6. A computer program product, It is characterized in that The invention comprises a computer program, which is used to implement the drug molecule screening method according to any one of claims 1 to 3 when running on one or more processors.

7. An electronic device, It is characterized in that include: A processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes instructions for implementing the drug molecule screening method as described in any one of claims 1 to 3.