A method and related equipment for processing molecular document information
By generating reaction datasets containing both two-dimensional and three-dimensional molecular information, the problem of relying solely on two-dimensional graph structures in existing technologies is solved, enabling more efficient and accurate molecular research and chemical reaction prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI
- Filing Date
- 2023-11-01
- Publication Date
- 2026-05-26
AI Technical Summary
In existing technologies, methods for characterizing chemical reactions using molecular data mainly rely on two-dimensional graph structures, which cannot provide three-dimensional structural information of molecules. This limits the understanding of reaction mechanisms and results in low research efficiency and accuracy.
By creating reaction plate classes and reaction classes, reading molecular file data, extracting three-dimensional and two-dimensional structural information, generating a reaction dataset containing two-dimensional and three-dimensional molecular data, and applying deep learning for prediction.
It improves the efficiency and accuracy of molecular research, provides more data support for deep learning, and enables a better understanding and prediction of the interactions and reaction outcomes between different molecules in chemical reactions.
Smart Images
Figure CN117995291B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence technology, and in particular to a molecular file information processing method, system, terminal, and computer-readable storage medium. Background Technology
[0002] When applying computer technology to the fields of chemistry and biology, researchers often need to process large amounts of molecular file information for in-depth analysis and modeling. This molecular file information contains key information about molecules, including their structure and properties.
[0003] In recent years, with the rapid development of deep learning and artificial intelligence technologies, researchers have begun to explore how to utilize these technologies to better understand and predict the interactions between different molecules and reaction yields in chemical reactions. Specifically, deep learning has been applied to the molecular field to infer the relationship between relevant molecules (such as ligands, catalysts, etc.) and reaction yields in chemical reactions. This application involves learning molecular characteristics to understand how they affect the outcome of the reaction.
[0004] However, previous research in this field has some limitations, mainly including:
[0005] Two-dimensional structural data: Previous studies have mainly focused on processing two-dimensional structural data of molecules, including the connection relationships between atoms and eigenvectors. Although this data is very important for understanding the structure and properties of molecules, it often cannot provide detailed information about the three-dimensional structure and atomic coordinates of molecules, which may limit a deeper understanding of reaction mechanisms in some cases.
[0006] Manual fine-tuning: In previous studies, researchers often needed to manually fine-tune the data to generate different types of response datasets. This process was time-consuming and prone to human error, which reduced the efficiency of the research.
[0007] For example, some studies have arranged reaction-related molecular data in experiments to generate reaction datasets containing two-dimensional molecular infographics. This process requires intervention and manual operation by researchers, which may introduce subjectivity and inconsistency, limiting the reproducibility and scalability of the research.
[0008] Therefore, existing technologies still need to be improved and developed. Summary of the Invention
[0009] The main objective of this invention is to provide a molecular file information processing method, system, terminal, and computer-readable storage medium. This invention aims to address the problem that existing data processing methods for characterizing chemical reactions using molecular data generally rely on the two-dimensional graph structure of molecules, which limits the in-depth understanding of reaction mechanisms and leads to low efficiency and accuracy in molecular research.
[0010] To achieve the above objectives, the present invention provides a molecular file information processing method, which includes the following steps:
[0011] Create reaction plate classes and reaction classes, and construct and generate chemical reactions with different reaction conditions based on the types of reaction molecules;
[0012] Read the molecular file data in the target folder. Each molecular file data includes multiple SDF files. Read each SDF file and extract the structural information of the molecule.
[0013] In three dimensions, the text within a specified region of the SDF file is read sequentially, segmented, and the three-dimensional coordinates and atomic number of each atom in the molecule are extracted to obtain the three-dimensional structural information.
[0014] In two dimensions, the SDF file is converted into a Smiles expression, and the atomic attribute vectors, the connection relationships between atoms, and the feature vectors of the edges are extracted to obtain the two-dimensional structural information.
[0015] The three-dimensional structural information and the two-dimensional structural information are spliced together to obtain two-dimensional and three-dimensional molecular data;
[0016] The generated reaction plate is searched to obtain the molecules contained in each reaction. The two-dimensional and three-dimensional molecular data are combined and arranged according to the reaction plate to obtain reaction data with two-dimensional and three-dimensional molecular information.
[0017] The actual yield of each reaction in the molecular file data is read, the actual yield is added to the reaction data, and the reaction data is cleaned to remove the reactions with an actual yield of 0 from the dataset, thus obtaining the target dataset.
[0018] Optionally, in the molecular file information processing method, each reaction includes all reaction molecule categories, which include halides, catalysts, ligands, and additives, and different chemical reactions are generated by arranging and combining different types of molecules.
[0019] Optionally, in the molecular file information processing method, reading each SDF file specifically involves reading each SDF file using Python's Rdkit library.
[0020] Optionally, the molecular file information processing method, wherein the step of sequentially reading and segmenting the text within a specified region of the SDF file in three dimensions to extract the three-dimensional coordinates and atomic number of each atom in the molecule to obtain three-dimensional structural information, further includes:
[0021] The three-dimensional structural information is converted into a Data format that can be directly read by the pytorch_geometric library.
[0022] Optionally, in the molecular file information processing method, the step of converting the SDF file into a Smiles expression in two dimensions and extracting atomic attribute vectors, the connection relationships between atoms, and the feature vectors of edges to obtain two-dimensional structural information specifically involves:
[0023] In two dimensions, the SDF file is first converted into a Smiles expression using the Chem.MolToSmiles function provided by the Rdkit library, and then the atomic attribute vectors, the connection relationships between atoms, and the feature vectors of the edges are extracted using the torch_geometric library to obtain the two-dimensional structural information.
[0024] Optionally, the molecular file information processing method, wherein in two dimensions, the SDF file is converted into a Smiles expression, and atomic attribute vectors, the connection relationships between atoms, and the feature vectors of edges are extracted to obtain two-dimensional structural information, further includes:
[0025] The two-dimensional structural information is converted into a Data format that can be directly read by the pytorch_geometric library.
[0026] Optionally, in the molecular file information processing method, the target dataset is a two-dimensional and three-dimensional molecular reaction dataset.
[0027] Furthermore, to achieve the above objectives, the present invention also provides a molecular document information processing system, wherein the molecular document information processing system comprises:
[0028] The chemical reaction building module is used to create reaction plate classes and reaction classes, and to build and generate chemical reactions with different reaction conditions based on the types of reaction molecules.
[0029] The structure information extraction module is used to read molecular file data from the target folder. Each molecular file data includes multiple SDF files. The module reads each SDF file and extracts the structural information of the molecule.
[0030] The three-dimensional structure information extraction module is used to read the text in a specified area within an SDF file in three dimensions, segment it, and extract the three-dimensional coordinates and atomic number of each atom in the molecule to obtain the three-dimensional structure information.
[0031] The two-dimensional structure information extraction module is used to convert SDF files into Smiles expressions in two dimensions and extract atomic attribute vectors, connection relationships between atoms, and feature vectors of edges to obtain two-dimensional structure information.
[0032] The structural information splicing module is used to splice the three-dimensional structural information and the two-dimensional structural information to obtain two-dimensional and three-dimensional molecular data;
[0033] The reaction data acquisition module is used to search the generated reaction plate to obtain the molecules contained in each reaction, and to combine and arrange the two-dimensional and three-dimensional molecular data according to the reaction plate to obtain the reaction data with two-dimensional and three-dimensional molecular information.
[0034] The dataset processing module is used to read the actual yield of each reaction in the molecular file data, add the actual yield to the reaction data, clean the reaction data, remove the reactions with an actual yield of 0 from the dataset, and obtain the target dataset.
[0035] In addition, to achieve the above objectives, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a molecular file information processing program stored in the memory and executable on the processor, wherein when the molecular file information processing program is executed by the processor, it implements the steps of the molecular file information processing method as described above.
[0036] In addition, to achieve the above objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a molecular file information processing program, which, when executed by a processor, implements the steps of the molecular file information processing method as described above.
[0037] In this invention, reaction plate classes and reaction classes are created. Chemical reactions of molecules under different reaction conditions are constructed and generated based on the categories of reaction molecules. Molecular file data from a target folder is read; each molecular file includes multiple SDF files. The structural information of the molecules is extracted from each SDF file. In three dimensions, text within a specified region of the SDF file is read sequentially and segmented to extract the three-dimensional coordinates and atomic numbers of each atom in the molecule, obtaining three-dimensional structural information. In two dimensions, the SDF file is converted into a SMILE expression, and atomic attribute vectors, connections between atoms, and edge feature vectors are extracted to obtain two-dimensional structural information. The three-dimensional and two-dimensional structural information are concatenated to obtain two-dimensional and three-dimensional molecular data. The generated reaction plate is searched to obtain the molecules contained in each reaction. The two-dimensional and three-dimensional molecular data are combined and arranged according to the reaction plate to obtain reaction data with two-dimensional and three-dimensional molecular information. The actual yield of each reaction in the molecular file data is read, added to the reaction data, and the reaction data is cleaned, removing reactions with an actual yield of 0 from the dataset to obtain the target dataset. This invention processes molecular file information to generate a reaction dataset of conditional molecules containing two-dimensional and three-dimensional structural information. By applying this two-dimensional information to deep learning to predict reaction yields, the efficiency and accuracy of molecular research are improved, providing more data support for deep learning to better understand and predict the interactions and reaction outcomes between different molecules in chemical reactions. Attached Figure Description
[0038] Figure 1 This is a flowchart of a preferred embodiment of the molecular file information processing method of the present invention;
[0039] Figure 2 This is a schematic diagram illustrating the automated processing of molecular file information and the generation of a reaction dataset containing two-dimensional and three-dimensional structural information in a preferred embodiment of the molecular file information processing method of the present invention.
[0040] Figure 3 This is a schematic diagram of a preferred embodiment of the molecular document information processing system of the present invention;
[0041] Figure 4 This is a schematic diagram of the operating environment of a preferred embodiment of the terminal of the present invention. Detailed Implementation
[0042] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
[0043] Existing data processing methods that involve characterizing chemical reactions using molecular data generally utilize the two-dimensional graph structure of molecules. They do not simultaneously utilize the information of both the two-dimensional and three-dimensional structures of the original molecular data, preventing researchers from using its unique structural information in subsequent studies.
[0044] This invention relates to a method for processing molecular file information, which can process input molecular file data to generate a reaction dataset of conditional molecules containing two-dimensional and three-dimensional structural information. This dataset can provide researchers with a new direction, namely, applying the two-dimensional information to deep learning to predict reaction yields.
[0045] The molecular file information processing method described in the preferred embodiment of the present invention, such as... Figure 1 and Figure 2 As shown, the molecular file information processing method includes the following steps:
[0046] Step S10: Create reaction plate classes and reaction classes, and construct and generate chemical reactions with different reaction conditions based on the categories of reaction molecules.
[0047] Specifically, a reaction plate class and a reaction class are created to manage and manipulate reaction conditions. The reaction plate is represented as a two-dimensional layout, allowing different combinations of molecule classes to fill different regions with reaction conditions, simulating different experimental settings on a lab plate. The reaction class reads the combination of reaction plates and provides methods for printing specific reactions. It automatically constructs and generates chemical reactions with different reaction conditions based on the categories of reaction molecules, such as... Figure 2 As shown, each reaction contains all categories of reactant molecules, namely halides, catalysts, ligands, and additives, which combine different types of molecules to produce different chemical reactions.
[0048] Step S20: Read the molecular file data of the target folder. Each molecular file data includes multiple SDF files. Read each SDF file and extract the structural information of the molecule.
[0049] Specifically, molecular file data (such as...) in the target folder can be read using the SDF data extraction module. Figure 2 As shown, for example, it includes molecular file a, molecular file b, etc., and each molecular file data includes multiple SDF files (each molecule corresponds to one SDF file). The structure information of the molecule is extracted by reading each SDF file through the Rdkit library of Python.
[0050] Step S30: In three dimensions, read the text in the specified area of the SDF file in sequence, segment it, and extract the three-dimensional coordinates and atomic number of each atom in the molecule to obtain the three-dimensional structure information.
[0051] Specifically, such as Figure 2 As shown, in three dimensions, text within a specified region of an SDF file is read sequentially and segmented to extract the three-dimensional coordinates and atomic numbers of each atom in the molecule, thus obtaining three-dimensional structural information. This information is then converted into a Data format that can be directly read by the pytorch_geometric library, facilitating subsequent research using deep learning. Specifically, the three-dimensional structural information includes the three-dimensional coordinates and atomic numbers of the atoms.
[0052] Step S40: In two dimensions, the SDF file is converted into a Smiles expression, and the atomic attribute vectors, the connection relationships between atoms, and the feature vectors of the edges are extracted to obtain the two-dimensional structure information.
[0053] Specifically, such as Figure 2 As shown, in two dimensions, the SDF file is first converted into a Smiles expression using the Chem.MolToSmiles function provided by the Rdkit library. Then, the torch_geometric library is used to extract the atomic attribute vectors (i.e., the feature vectors of atoms), the connection relationships between atoms, and the feature vectors of edges, thus obtaining the two-dimensional structure information. This information is then converted into a Data format that the pytorch_geometric library can directly read. In other words, the two-dimensional structure information includes atomic attribute vectors, the connection relationships between atoms, and the feature vectors of edges.
[0054] Step S50: The three-dimensional structural information and the two-dimensional structural information are spliced together to obtain two-dimensional and three-dimensional molecular data.
[0055] Specifically, such as Figure 2 As shown, molecular data in two dimensions are spliced together (the splicing method is a combination, that is, the new data contains both two-dimensional and three-dimensional data), that is, the three-dimensional structural information and the two-dimensional structural information are spliced together to obtain two-dimensional and three-dimensional molecular data that can be directly read by the pytorch_geometric library.
[0056] Step S60: Search the generated reaction plate to obtain the molecules contained in each reaction, and combine and arrange the two-dimensional and three-dimensional molecular data according to the reaction plate to obtain reaction data with two-dimensional and three-dimensional molecular information.
[0057] Specifically, such as Figure 2As shown, the molecules contained in each reaction can be obtained by searching the generated reaction plate through the reaction generation module. The two-dimensional and three-dimensional molecular data obtained in step S50 are combined and arranged according to the reaction plate (different types of molecules can be combined) to obtain reaction data containing two-dimensional molecular information (forming a reaction data containing four different molecules).
[0058] Step S70: Read the actual yield of each reaction in the molecular file data, add the actual yield to the reaction data, clean the reaction data, remove the reactions with an actual yield of 0 from the dataset, and obtain the target dataset.
[0059] Specifically, such as Figure 2 As shown, the information from the CSV file in the SDF file is read (the molecular file data in the target folder only contains information about each molecule itself, and the CSV file records the actual yield corresponding to each reaction). The actual yield data of each reaction is added to the reaction data, and the reaction data is cleaned to remove reactions with an actual yield of 0 from the dataset, thus obtaining the target dataset (the target dataset is a two-dimensional and three-dimensional molecular reaction dataset, such as reaction x, reaction y, reaction z, etc.).
[0060] This invention first classifies different reaction molecules by constructing a reaction plate, and then combines and screens each type of molecule to generate different reactions. Then, for a specific molecule, the program extracts information from different dimensions of the molecule file and stores it in the same array, and also adds the yield of the reaction to the array. Finally, the generated reaction dataset is converted into a new file, thus verifying the effectiveness of this invention.
[0061] This invention processes the same molecular file data to obtain a set of two-dimensional graph structure information and three-dimensional coordinate information of a molecule. Different types of condition molecules are automatically combined through a reaction plate to obtain a reaction containing multi-dimensional molecular information and yield, and a dataset is constructed.
[0062] The reaction dataset (i.e., the target dataset) generated by this invention includes not only two-dimensional molecular information, such as feature vectors and connectivity relationships, but also three-dimensional molecular information, including three-dimensional coordinates and atomic numbers. This allows researchers to gain a more comprehensive understanding of molecular structure and properties, which is highly helpful for understanding the impact on reaction mechanisms and yields. Scalability: Compared to previous methods that required manual fine-tuning to generate different types of reaction datasets, the method of this invention has higher scalability. It can automatically generate diverse reaction datasets according to different research needs and reaction types, providing more possibilities for further research. Universality: The dataset format processed by this invention can be directly read by the PyTorch and PyTorch Geometric libraries, making it suitable for direct use within these libraries.
[0063] This invention was applied to the Buchwald-Hartwig CN coupling reaction molecular dataset and successfully obtained a dataset containing two-dimensional and three-dimensional molecular information, proving the feasibility of the results.
[0064] Furthermore, considering adding support for multiple molecular file formats to this invention, rather than being limited to SDF files, would increase the versatility of the method, enabling it to handle molecular data from a wider range of sources. Integration of this invention with other chemical information processing tools and software could also be considered to provide more comprehensive molecular information analysis, allowing users to perform a broader range of analytical tasks within an integrated environment.
[0065] Furthermore, such as Figure 3 As shown, based on the above-described molecular file information processing method, the present invention also provides a molecular file information processing system, wherein the molecular file information processing system includes:
[0066] The chemical reaction construction module 51 is used to create reaction plate classes and reaction classes, and to construct and generate chemical reactions with different reaction conditions based on the type of reaction molecules.
[0067] The structure information extraction module 52 is used to read molecular file data in the target folder. Each molecular file data includes multiple SDF files. The module reads each SDF file and extracts the structural information of the molecule.
[0068] The three-dimensional structure information extraction module 53 is used to read the text in a specified area of the SDF file in three dimensions, segment it, extract the three-dimensional coordinates of each atom in the molecule and the atomic number of the atom, and obtain the three-dimensional structure information.
[0069] The two-dimensional structure information extraction module 54 is used to convert the SDF file into a Smiles expression in two dimensions, and extract the atomic attribute vectors, the connection relationships between atoms and the feature vectors of the edges to obtain the two-dimensional structure information.
[0070] The structural information splicing module 55 is used to splice the three-dimensional structural information and the two-dimensional structural information to obtain two-dimensional and three-dimensional molecular data;
[0071] The reaction data acquisition module 56 is used to search the generated reaction plate to obtain the molecules contained in each reaction, and to combine and arrange the two-dimensional and three-dimensional molecular data according to the reaction plate to obtain reaction data with two-dimensional and three-dimensional molecular information.
[0072] The dataset processing module 57 is used to read the actual yield of each reaction in the molecular file data, add the actual yield to the reaction data, clean the reaction data, remove the reactions with an actual yield of 0 from the dataset, and obtain the target dataset.
[0073] Furthermore, such as Figure 4 As shown, based on the above-mentioned molecular file information processing method and system, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 4 Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.
[0074] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code installed on the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores a molecular file information processing program 40, which can be executed by the processor 10 to implement the molecular file information processing method of this application.
[0075] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the molecular file information processing method.
[0076] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The components 10-30 of the terminal communicate with each other via a system bus.
[0077] In one embodiment, when the processor 10 executes the molecular file information processing program 40 in the memory 20, the following steps are performed:
[0078] Create reaction plate classes and reaction classes, and construct and generate chemical reactions with different reaction conditions based on the types of reaction molecules;
[0079] Read the molecular file data in the target folder. Each molecular file data includes multiple SDF files. Read each SDF file and extract the structural information of the molecule.
[0080] In three dimensions, the text within a specified region of the SDF file is read sequentially, segmented, and the three-dimensional coordinates and atomic number of each atom in the molecule are extracted to obtain the three-dimensional structural information.
[0081] In two dimensions, the SDF file is converted into a Smiles expression, and the atomic attribute vectors, the connection relationships between atoms, and the feature vectors of the edges are extracted to obtain the two-dimensional structural information.
[0082] The three-dimensional structural information and the two-dimensional structural information are spliced together to obtain two-dimensional and three-dimensional molecular data;
[0083] The generated reaction plate is searched to obtain the molecules contained in each reaction. The two-dimensional and three-dimensional molecular data are combined and arranged according to the reaction plate to obtain reaction data with two-dimensional and three-dimensional molecular information.
[0084] The actual yield of each reaction in the molecular file data is read, the actual yield is added to the reaction data, and the reaction data is cleaned to remove the reactions with an actual yield of 0 from the dataset, thus obtaining the target dataset.
[0085] Each reaction includes all categories of reactant molecules, including halides, catalysts, ligands, and additives, which generate different chemical reactions through the arrangement and combination of different types of molecules.
[0086] Specifically, reading each SDF file involves using Python's Rdkit library to read each SDF file.
[0087] The process of sequentially reading and segmenting text within a specified region of an SDF file in three dimensions to extract the three-dimensional coordinates and atomic number of each atom in the molecule, thereby obtaining three-dimensional structural information, further includes:
[0088] The three-dimensional structural information is converted into a Data format that can be directly read by the pytorch_geometric library.
[0089] Specifically, in two dimensions, the SDF file is converted into a Smiles expression, and the atomic attribute vectors, the connection relationships between atoms, and the feature vectors of edges are extracted to obtain the two-dimensional structural information.
[0090] In two dimensions, the SDF file is first converted into a Smiles expression using the Chem.MolToSmiles function provided by the Rdkit library, and then the atomic attribute vectors, the connection relationships between atoms, and the feature vectors of the edges are extracted using the torch_geometric library to obtain the two-dimensional structural information.
[0091] Specifically, in two dimensions, the SDF file is converted into a Smiles expression, and atomic attribute vectors, the connection relationships between atoms, and the feature vectors of edges are extracted to obtain two-dimensional structural information. This process also includes:
[0092] The two-dimensional structural information is converted into a Data format that can be directly read by the pytorch_geometric library.
[0093] The target dataset is a two-dimensional and a three-dimensional molecular reaction dataset.
[0094] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a molecular file information processing program, which, when executed by a processor, implements the steps of the molecular file information processing method described above.
[0095] In summary, this invention provides a molecular file information processing method and related equipment. The method includes: creating reaction plate classes and reaction classes; constructing and generating chemical reactions of molecules under different reaction conditions according to the categories of reaction molecules; reading molecular file data from a target folder, where each molecular file data includes multiple SDF files; reading each SDF file and extracting the structural information of the molecules; in three dimensions, sequentially reading and segmenting the text within a specified region of the SDF file to extract the three-dimensional coordinates and atomic number of each atom in the molecule, thereby obtaining three-dimensional structural information; and in two dimensions, converting the SDF file into a Smi... The LES expression is used to extract atomic attribute vectors, inter-atomic connections, and edge feature vectors to obtain two-dimensional structural information. The three-dimensional structural information is then concatenated with the two-dimensional structural information to obtain two-dimensional and three-dimensional molecular data. A reaction plate is generated to identify the molecules contained in each reaction. The two-dimensional and three-dimensional molecular data are then combined and arranged according to the reaction plate to obtain reaction data with two-dimensional and three-dimensional molecular information. The actual yield of each reaction in the molecular file data is read, added to the reaction data, and the data is cleaned, removing reactions with an actual yield of 0 to obtain the target dataset. This invention processes molecular file information to generate a reaction dataset of conditional molecules containing two-dimensional and three-dimensional structural information. Applying this two-dimensional information to deep learning for reaction yield prediction improves the efficiency and accuracy of molecular research, provides more data support for deep learning, and enables a better understanding and prediction of the interactions and reaction outcomes between different molecules in chemical reactions.
[0096] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.
[0097] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The computer-readable storage medium can be a memory, magnetic disk, optical disk, etc.
[0098] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.
Claims
1. A method of processing information of a molecular file, characterized by, The molecular file information processing method includes: Create reaction plate classes and reaction classes, and construct and generate chemical reactions with different reaction conditions based on the types of reaction molecules; Read the molecular file data in the target folder. Each molecular file data includes multiple SDF files. Read each SDF file and extract the structural information of the molecule. In three dimensions, the text within a specified region of the SDF file is read sequentially, segmented, and the three-dimensional coordinates and atomic number of each atom in the molecule are extracted to obtain the three-dimensional structural information. In two dimensions, the SDF file is converted into a Smiles expression, and the atomic attribute vectors, the connection relationships between atoms, and the feature vectors of the edges are extracted to obtain the two-dimensional structural information. The three-dimensional structural information and the two-dimensional structural information are spliced together to obtain two-dimensional and three-dimensional molecular data; The generated reaction plate is searched to obtain the molecules contained in each reaction. The two-dimensional and three-dimensional molecular data are combined and arranged according to the reaction plate to obtain reaction data with two-dimensional and three-dimensional molecular information. The actual yield of each reaction in the molecular file data is read, the actual yield is added to the reaction data, and the reaction data is cleaned to remove reactions with an actual yield of 0 from the dataset, thus obtaining the target dataset. The two dimensions of information are then applied to deep learning to predict the reaction yield.
2. The method of claim 1, wherein Each reaction includes all categories of reactant molecules, including halides, catalysts, ligands, and additives, which generate different chemical reactions through the arrangement and combination of different types of molecules.
3. The method of claim 1, wherein the step of processing the molecular file information is performed by a computer program. The reading of each SDF file specifically refers to reading each SDF file using Python's Rdkit library.
4. The method of claim 1, wherein The process involves sequentially reading and segmenting text within a specified region of an SDF file in three dimensions to extract the three-dimensional coordinates and atomic number of each atom in the molecule, thereby obtaining three-dimensional structural information. This process further includes: The three-dimensional structural information is converted into a Data format that can be directly read by the pytorch_geometric library.
5. The method of claim 1, wherein the step of processing the molecular file information is performed by a computer program. In two dimensions, the SDF file is converted into a Smiles expression, and the atomic attribute vectors, the connection relationships between atoms, and the feature vectors of edges are extracted to obtain the two-dimensional structural information, specifically: In two dimensions, the SDF file is first converted into a Smiles expression using the Chem.MolToSmiles function provided by the Rdkit library, and then the atomic attribute vectors, the connection relationships between atoms, and the feature vectors of the edges are extracted using the torch_geometric library to obtain the two-dimensional structural information.
6. The method of claim 1 or 5, wherein In two dimensions, the SDF file is converted into a Smiles expression, and atomic attribute vectors, the connection relationships between atoms, and the feature vectors of edges are extracted to obtain two-dimensional structural information. This process also includes: The two-dimensional structural information is converted into a Data format that can be directly read by the pytorch_geometric library.
7. The method of claim 1, wherein the step of processing the molecular file information is performed by a computer program. The target dataset is a two-dimensional and three-dimensional molecular reaction dataset.
8. A system for processing information of a molecular file, characterized by comprising: The molecular document information processing system includes: The chemical reaction building module is used to create reaction plate classes and reaction classes, and to build and generate chemical reactions with different reaction conditions based on the types of reaction molecules. The structure information extraction module is used to read molecular file data from the target folder. Each molecular file data includes multiple SDF files. The module reads each SDF file and extracts the structural information of the molecule. The three-dimensional structure information extraction module is used to read the text in a specified area within an SDF file in three dimensions, segment it, and extract the three-dimensional coordinates and atomic number of each atom in the molecule to obtain the three-dimensional structure information. The two-dimensional structure information extraction module is used to convert SDF files into Smiles expressions in two dimensions and extract atomic attribute vectors, connection relationships between atoms, and feature vectors of edges to obtain two-dimensional structure information. The structural information splicing module is used to splice the three-dimensional structural information and the two-dimensional structural information to obtain two-dimensional and three-dimensional molecular data; The reaction data acquisition module is used to search the generated reaction plate to obtain the molecules contained in each reaction, and to combine and arrange the two-dimensional and three-dimensional molecular data according to the reaction plate to obtain the reaction data with two-dimensional and three-dimensional molecular information. The dataset processing module is used to read the actual yield of each reaction in the molecular file data, add the actual yield to the reaction data, clean the reaction data, remove reactions with an actual yield of 0 from the dataset, obtain the target dataset, and apply the two dimensions of information to deep learning to predict the reaction yield.
9. A terminal, characterized by comprising: The terminal includes: a memory, a processor, and a molecular file information processing program stored in the memory and executable on the processor. When the molecular file information processing program is executed by the processor, it implements the steps of the molecular file information processing method as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a molecular file information processing program, which, when executed by a processor, implements the steps of the molecular file information processing method as described in any one of claims 1-7.