Organic reaction classification method and device, electronic equipment and storage medium
By classifying the precursors and products of organic reactions using deep learning models, the problem of relying on manual operation and reaction fingerprints in existing technologies is solved, and rapid and efficient organic reaction classification is achieved.
Patent Information
- Application Number
- CN202211542890.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-12-02
- Publication Date
- 2025-12-30
- Estimated Expiration
- 2042-12-02
AI Technical Summary
In existing technologies, the classification of organic reactions relies on manual operations, which are cumbersome and time-consuming. Machine learning methods heavily depend on reaction fingerprints, making it difficult to classify reactions quickly and efficiently.
A deep learning model is used to classify the precursors and products of organic reactions. By learning reaction features through graph isomorphic neural networks and cross-attention layers, rapid classification without reaction fingerprints is achieved.
It reduces the burden of manual sorting, improves sorting speed and efficiency, has a wider range of applications, and realizes simple and convenient organic reaction sorting.
Smart Images
Figure CN115954058B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of deep learning technology, and in particular to an organic reaction classification method, apparatus, electronic device, and storage medium. Background Technology
[0002] Organic reaction classification, a fundamental exercise in chemistry, has long been a subject of in-depth research and exploration. Chemists use reaction categories to analyze the similarities between reactions within the same category and to infer optimal reaction conditions. They also use reaction categories to predict products and assess reaction quality. Reaction classification can also be used to navigate large reaction databases. However, manually classifying reactions is a tedious and challenging task, requiring the observation of differences between precursors and products based on domain-specific knowledge. As the amount of data grows exponentially, it can become increasingly time-consuming and expensive. How to achieve reaction classification quickly and efficiently has become a pressing problem for many researchers.
[0003] The emergence of machine learning methods has alleviated the burden of manual classification and eliminated the need for professionally written classification rules. Machine learning methods perform well in reaction classification, but they rely heavily on reaction fingerprints and cannot learn the deeper meaning of reactions from simple features. Summary of the Invention
[0004] This application provides an organic reaction classification method, apparatus, electronic device, and storage medium. By classifying organic reactions using a deep learning model, it can not only reduce the burden of manual classification but also be faster and more efficient because it does not rely on reaction fingerprints.
[0005] In a first aspect, embodiments of this application provide a method for classifying organic reactions, the method comprising:
[0006] To obtain the precursors and products of organic reactions;
[0007] The precursor and the product are respectively input into a pre-trained deep learning model;
[0008] The organic reaction is classified based on the precursor and the product using the deep learning model.
[0009] Secondly, embodiments of this application also provide an organic reaction classification device, the device comprising: an acquisition module, an input module, and a classification module; wherein,
[0010] The acquisition module is used to acquire the precursors and products of organic reactions;
[0011] The input module is used to input the precursor and the product into a pre-trained deep learning model, respectively.
[0012] The classification module is used to determine the classification of the organic reaction based on the precursor and the product using the deep learning model.
[0013] Thirdly, embodiments of this application provide an electronic device, including:
[0014] One or more processors;
[0015] Memory, used to store one or more programs.
[0016] When the one or more programs are executed by the one or more processors, the one or more processors implement the organic reaction classification method described in any embodiment of this application.
[0017] Fourthly, embodiments of this application provide a storage medium storing a computer program thereon, which, when executed by a processor, implements the organic reaction classification method described in any embodiment of this application.
[0018] This application proposes an organic reaction classification method, apparatus, electronic device, and storage medium. First, the precursors and products of the organic reaction are obtained. Then, the precursors and products are input into a pre-trained deep learning model. Finally, the deep learning model determines the classification of the organic reaction based on the precursors and products. In other words, the technical solution of this application can determine the classification of organic reactions using a deep learning model. In contrast, existing technologies classify organic reactions manually or using machine learning-based methods. Manual classification is labor-intensive, and machine learning-based methods heavily rely on reaction fingerprints. Therefore, compared to existing technologies, the organic reaction classification method, apparatus, electronic device, and storage medium proposed in this application, which classifies organic reactions using a deep learning model, not only reduces the burden of manual classification but also does not rely on reaction fingerprints, resulting in faster speed and higher efficiency. Furthermore, the technical solution of this application is simple and convenient to implement, easy to popularize, and has a wider range of applications. Attached Figure Description
[0019] Figure 1 This is a schematic diagram of the first process of the organic reaction classification method provided in the embodiments of this application;
[0020] Figure 2 This is a schematic diagram of the second process of the organic reaction classification method provided in the embodiments of this application;
[0021] Figure 3 A schematic diagram of the third process of the organic reaction classification method provided in the embodiments of this application;
[0022] Figure 4 This is a schematic diagram of the organic reaction sorting device provided in the embodiments of this application;
[0023] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0024] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the application and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present application, not the entire structure.
[0025] Example 1
[0026] Figure 1 This is a first flowchart illustrating an organic reaction classification method provided in an embodiment of this application. The method can be executed by an organic reaction classification device or electronic device, which can be implemented in software and / or hardware, and can be integrated into any smart device with network communication capabilities. Figure 1 As shown, the organic reaction classification method may include the following steps:
[0027] S101, Obtain the precursors and products of organic reactions.
[0028] In this step, the electronic device can acquire the precursor and product of the organic reaction. The precursor of the organic reaction in the embodiments of this application may include one or more organic or inorganic substances; the product of the organic reaction in the embodiments of this application may include one or more organic or inorganic substances.
[0029] S102. Input the precursor and product into the pre-trained deep learning model respectively.
[0030] In this step, the electronic device can input the precursor and product into a pre-trained deep learning model, respectively. Specifically, this step can input the names of the precursor and product of the organic reaction into the pre-trained deep learning model, respectively.
[0031] S103. Classify organic reactions based on precursors and products using deep learning models.
[0032] In this step, the electronic device can determine the classification of organic reactions based on precursors and products using a deep learning model. Specifically, the electronic device can first obtain graphical representations of the precursors and products using a graph isomorphic neural network in the deep learning model; then, the graphical representations of the precursors and products are respectively input into the cross-attention layer in the deep learning model; the cross-attention layer then learns the correspondence between precursors and products, and determines the classification of organic reactions based on the correspondence between precursors and products.
[0033] The organic reaction classification method proposed in this application first obtains the precursors and products of the organic reaction; then, the precursors and products are respectively input into a pre-trained deep learning model; finally, the classification of the organic reaction is determined by the deep learning model based on the precursors and products. In other words, the classification of organic reactions can be determined by a deep learning model in this application. In contrast, existing technologies classify organic reactions manually or using machine learning methods; manual classification is labor-intensive, and machine learning methods heavily rely on reaction fingerprints. Therefore, compared with existing technologies, the organic reaction classification method, apparatus, electronic device, and storage medium proposed in this application, which classifies organic reactions using a deep learning model, not only reduces the burden of manual classification but also does not rely on reaction fingerprints, resulting in faster speed and higher efficiency; furthermore, the technical solution of this application is simple and convenient to implement, easy to popularize, and has a wider range of applications.
[0034] Example 2
[0035] Figure 2 This is a schematic diagram of the second process of the organic reaction classification method provided in this application embodiment. Further optimizations and extensions can be made based on the above technical solution, and it can be combined with the various optional implementation methods described above. For example... Figure 2 As shown, the organic reaction classification method may include the following steps:
[0036] S201, Obtain the precursors and products of organic reactions.
[0037] S202. Input the precursor and product into the pre-trained deep learning model respectively.
[0038] S203. Obtain the graphical representation of the precursor and the graphical representation of the product through the graph isomorphic neural network in the deep learning model.
[0039] In this step, the electronic device can obtain the graphical representations of the precursor and the product through a graph isomorphic neural network in a deep learning model. In this embodiment, the graphical representation of the precursor refers to the precursor represented graphically; the graphical representation of the product refers to the product represented graphically.
[0040] S204. Input the graphical representations of the precursor and the product into the cross-attention layer of the deep learning model, respectively.
[0041] In this step, the electronic device can input the graphical representations of the precursor and the product into the cross-attention layer of the deep learning model, respectively. The deep learning model in this embodiment may include: a graph isomorphic neural network, a cross-attention layer, a graph representation layer, and a multilayer perceptron. This application first obtains the graphical representations of the precursor and the product through the graph isomorphic neural network in the deep learning model; then inputs these representations into the cross-attention layer; the cross-attention layer learns the correspondence between the precursor and the product, and inputs this correspondence into the graph representation layer of the deep learning model; the graph representation layer extracts the features of the precursor and the product; the features of the precursor and the product are compared and learned to obtain the comparison learning result; the comparison learning result is input into the multilayer perceptron in the deep learning model; the multilayer perceptron determines the embedding of the organic reaction, and the organic reaction is classified based on its embedding.
[0042] S205. Learn the correspondence between precursors and products through cross-attention layers, and determine the classification of organic reactions based on the correspondence between precursors and products.
[0043] In this step, the electronic device can learn the correspondence between precursors and products through a cross-attention layer, and determine the classification of organic reactions based on the correspondence between precursors and products. Specifically, the electronic device can first input the correspondence between precursors and products into the graph representation layer of the deep learning model; then extract the features of the precursors and products through the graph representation layer; and finally determine the classification of organic reactions based on the features of the precursors and products.
[0044] The organic reaction classification method proposed in this application first obtains the precursors and products of the organic reaction; then, the precursors and products are respectively input into a pre-trained deep learning model; finally, the classification of the organic reaction is determined by the deep learning model based on the precursors and products. In other words, the classification of organic reactions can be determined by a deep learning model in this application. In contrast, existing technologies classify organic reactions manually or using machine learning methods; manual classification is labor-intensive, and machine learning methods heavily rely on reaction fingerprints. Therefore, compared with existing technologies, the organic reaction classification method, apparatus, electronic device, and storage medium proposed in this application, which classifies organic reactions using a deep learning model, not only reduces the burden of manual classification but also does not rely on reaction fingerprints, resulting in faster speed and higher efficiency; furthermore, the technical solution of this application is simple and convenient to implement, easy to popularize, and has a wider range of applications.
[0045] Example 3
[0046] Figure 3 This is a schematic diagram of the third process of the organic reaction classification method provided in this application embodiment. Further optimizations and extensions based on the above technical solution are possible, and it can be combined with the various optional implementation methods described above. For example... Figure 3 As shown, the organic reaction classification method may include the following steps:
[0047] S301, Obtain the precursors and products of organic reactions.
[0048] S302. Input the precursor and product into the pre-trained deep learning model respectively.
[0049] S303. Obtain the graphical representation of the precursor and the graphical representation of the product through the graph isomorphic neural network in the deep learning model.
[0050] S304. Input the graphical representations of the precursor and the product into the cross-attention layer of the deep learning model, respectively.
[0051] S305. The correspondence between the precursor and the product is learned through the cross-attention layer, and the correspondence between the precursor and the product is input into the graph representation layer of the deep learning model.
[0052] S306. Extract the features of the precursor and the product through the graphical representation layer.
[0053] In this step, the electronic device can extract features of the precursor and the product through the graphical representation layer.
[0054] S307. Classify organic reactions based on the characteristics of the precursors and products.
[0055] In this step, the electronic device can determine the classification of an organic reaction based on the characteristics of the precursor and the product. Specifically, the electronic device can first perform comparative learning on the characteristics of the precursor and the product to obtain the comparative learning results; then, it can determine the classification of the organic reaction based on the comparative learning results of the precursor and product characteristics. Furthermore, the electronic device can first input the comparative learning results of the precursor and product characteristics into a multilayer perceptron in a deep learning model; then, it can determine the embedding of the organic reaction through the multilayer perceptron, and determine the classification of the organic reaction based on the embedding.
[0056] In a specific embodiment of this application, the deep learning model can be pre-trained before obtaining the precursors and products of the organic reaction. Specifically, if the deep learning model does not meet the pre-set convergence conditions, the electronic device can extract a sample from a pre-constructed training sample set as the current sample; then, it can use the current sample to train the deep learning model; this process is repeated until the deep learning model meets the convergence conditions. Further, the electronic device can calculate the loss function of the deep learning model based on the current sample; then, it can train the deep learning model based on the loss function.
[0057] In a specific embodiment of this application, when classifying reactions, the characteristics of precursors and products are first observed, and then atoms are mapped to find the atom pairs that change before and after the reaction. Graph neural networks (GNNs) have been widely applied to tasks such as molecular property prediction, graph classification, and node classification due to their powerful processing capabilities for graph data. In recent years, many variants of GNNs have been proposed, with different aggregation functions, node update functions, and graph-level pooling. Within the GNN framework, GIN is considered the most expressive. GIN utilizes learnable MLP layers to model neighborhood aggregation and message passing functions, and obtains a graphical representation through the READOUT function. This model proposes ContraGIN. Features of precursors and products are extracted using RDKit. To learn the respective features of precursors and products, the model inputs both into a GIN block with shared parameters; then, a cross-attention layer is used to learn the atomic correspondences between precursors and products. Next, the reaction embeddings are obtained through a readout block and a contrast layer. More specifically, sumpooling is used to integrate the features of all nodes in the readout block. In the contrast layer, the embedding of precursors and products is simply added; finally, a multilayer perceptron is applied to classify the reaction embeddings. This application processes precursors and products independently, with two parallel cross-attention modules. Atoms of precursors and products can communicate with each other. Based on the comparative learning of precursor and product features, the characteristics of the reaction are learned, and the organic reactions are classified.
[0058] This application uses three metrics: ACC (accuracy), CEN (confusion entropy of the confusion matrix), and MCC (Matthews correlation coefficient). Comparing the results of this model with those of other models shows that ACC and MCC increased from 98.9% to 99.3%, while CEN decreased from 0.006 to 0.001. Furthermore, the two models in this application... - Contra + Contra - Compared to Contra +More stable; comparing the time taken by different models to process each reaction, the results show that the model in this application is 10 times faster than the BERT model.
[0059] The organic reaction classification method proposed in this application first obtains the precursors and products of the organic reaction; then, the precursors and products are respectively input into a pre-trained deep learning model; finally, the classification of the organic reaction is determined by the deep learning model based on the precursors and products. In other words, the classification of organic reactions can be determined by a deep learning model in this application. In contrast, existing technologies classify organic reactions manually or using machine learning methods; manual classification is labor-intensive, and machine learning methods heavily rely on reaction fingerprints. Therefore, compared with existing technologies, the organic reaction classification method proposed in this application, which uses a deep learning model to classify organic reactions, not only reduces the burden of manual classification but also does not rely on reaction fingerprints, resulting in faster speed and higher efficiency; furthermore, the technical solution of this application is simple and convenient to implement, easy to popularize, and has a wider range of applications.
[0060] Example 4
[0061] Figure 4 This is a schematic diagram of the organic reaction sorting device provided in an embodiment of this application. Figure 4 As shown, the organic reaction classification device includes: an acquisition module 401, an input module 402, and a classification module 403; wherein,
[0062] The acquisition module 401 is used to acquire the precursors and products of organic reactions;
[0063] The input module 402 is used to input the precursor and the product into a pre-trained deep learning model, respectively.
[0064] The classification module 403 is used to determine the classification of the organic reaction based on the precursor and the product using the deep learning model.
[0065] The above-described organic reaction sorting device can perform the method provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects for performing the method. Technical details not described in detail in this embodiment can be found in the organic reaction sorting method provided in any embodiment of this application.
[0066] Example 5
[0067] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 5 A block diagram of an exemplary electronic device suitable for implementing embodiments of the present application is shown. Figure 5The electronic device 12 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.
[0068] like Figure 5 As shown, the electronic device 12 is represented in the form of a general-purpose computing device. The components of the electronic device 12 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and bus 18 connecting different system components (including system memory 28 and processing unit 16).
[0069] Bus 18 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.
[0070] Electronic device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by electronic device 12, including volatile and non-volatile media, removable and non-removable media.
[0071] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Electronic device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (… Figure 5 Not shown; usually referred to as a "hard drive"). Although Figure 5 Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. Memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of this application.
[0072] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of this application.
[0073] Electronic device 12 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable a user to interact with electronic device 12, and / or with any device that enables electronic device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via input / output (I / O) interface 22. Furthermore, electronic device 12 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 20. As shown, network adapter 20 communicates with other modules of electronic device 12 via bus 18. It should be understood that, although... Figure 5 As not shown, other hardware and / or software modules may be used in conjunction with electronic device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.
[0074] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the organic reaction classification method provided in the embodiments of this application.
[0075] Example 6
[0076] This application provides a computer storage medium.
[0077] The computer-readable storage medium of this application embodiment can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be—but is not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.
[0078] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.
[0079] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.
[0080] Computer program code for performing the operations of this application can be written in one or more programming languages or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, and C++, and conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0081] Note that the above are merely preferred embodiments and the technical principles employed in this application. Those skilled in the art will understand that this application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of this application, the scope of which is determined by the scope of the appended claims.
Claims
1. An organic reaction classification method characterized by, The method comprises: obtaining a precursor and a product of an organic reaction; inputting the precursor and the product into a pre-trained deep learning model respectively; obtaining a graph representation corresponding to the precursor and a graph representation corresponding to the product through a graph isomorphism neural network in the deep learning model; inputting the graph representation corresponding to the precursor and the graph representation corresponding to the product into a cross-attention layer in the deep learning model respectively; learning a corresponding relationship between the precursor and the product through the cross-attention layer, and inputting the corresponding relationship between the precursor and the product into a graph representation layer of the deep learning model; extracting features of the precursor and features of the product through the graph representation layer; performing contrastive learning on the features of the precursor and the features of the product to obtain a contrastive learning result of the features of the precursor and the features of the product; inputting the contrastive learning result of the features of the precursor and the features of the product into a multilayer perceptron in the deep learning model; determining an embedding of the organic reaction through the multilayer perceptron, and determining a classification of the organic reaction according to the embedding of the organic reaction.
2. The method of claim 1, wherein, Before obtaining the precursor and the product of the organic reaction, the method further comprises: if the deep learning model does not satisfy a pre-set convergence condition, extracting a sample from a pre-constructed training sample set as a current sample; training the deep learning model using the current sample; and repeatedly performing the above operations until the deep learning model satisfies the convergence condition.
3. The method of claim 2, wherein, Training the deep learning model using the current sample comprises: calculating a loss function of the deep learning model according to the current sample; and training the deep learning model based on the loss function.
4. An organic reaction classification apparatus characterized by comprising: The device comprises an obtaining module, an inputting module and a classifying module; wherein, the obtaining module is configured to obtain a precursor and a product of an organic reaction; the inputting module is configured to input the precursor and the product into a pre-trained deep learning model respectively; the classifying module is configured to obtain a graph representation corresponding to the precursor and a graph representation corresponding to the product through a graph isomorphism neural network in the deep learning model; input the graph representation corresponding to the precursor and the graph representation corresponding to the product into a cross-attention layer in the deep learning model respectively; learn a corresponding relationship between the precursor and the product through the cross-attention layer, and input the corresponding relationship between the precursor and the product into a graph representation layer of the deep learning model; extract features of the precursor and features of the product through the graph representation layer; perform contrastive learning on the features of the precursor and the features of the product to obtain a contrastive learning result of the features of the precursor and the features of the product; input the contrastive learning result of the features of the precursor and the features of the product into a multilayer perceptron in the deep learning model; determine an embedding of the organic reaction through the multilayer perceptron, and determine a classification of the organic reaction according to the embedding of the organic reaction.
5. An electronic device, comprising: comprise: one or more processors; a memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the organic reaction classification method as claimed in any one of claims 1 to 3.
6. A storage medium having stored thereon a computer program, characterized in that The program, when executed by a processor, implements the organic reaction classification method as claimed in any one of claims 1 to 3. The program, when executed by a processor, implements the organic reaction classification method as claimed in any one of claims 1 to 3.