Reverse molecular design method and system of organic framework material COFs based on machine learning and quantitative calculation technology

By employing machine learning and quantum chemical computation techniques, a mapping relationship between the structure and properties of COFs is established, solving the problems of low efficiency and high cost in traditional methods. This enables efficient and accurate design and screening of COF materials, and is applicable to the research of COFs and other framework materials.

CN121011270APending Publication Date: 2025-11-25HAINAN NORMAL UNIV
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202510286253.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-03-12
Publication Date
2025-11-25

AI Technical Summary

Technical Problem

Traditional methods are inefficient and costly in designing and screening novel COFs materials with high photoelectric properties, and it is difficult to systematically study the relationship between structure and performance.

Method used

Using machine learning and quantum chemical computation techniques, a mapping relationship between the structure and properties of COFs is established, and the optoelectronic properties of COF materials are screened and predicted through reverse molecular design methods.

Benefits of technology

It enables efficient and accurate prediction and directional design of COFs materials, reduces R&D costs, improves material design efficiency, and is applicable to the research of COFs and other similar framework materials.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121011270A_ABST
    Figure CN121011270A_ABST
Patent Text Reader

Abstract

The invention discloses a reverse molecular design method and system of an organic framework material COFs based on a machine learning and quantitative calculation technology, and relates to the technical field of material science and computer simulation. The photoelectric property of the COFs material can be accurately predicted, and a new material can be directionally synthesized under guidance. A COFs structure database is established, Gaussian is used for representing the fragment structure and electronic characteristics of the COFs, a molecular descriptor and a machine learning algorithm are used for establishing a structure-performance mapping relation, quantitative calculation software is used for obtaining an important target quantity for describing the photoelectric characteristics of the COFs, SHAP is used for carrying out importance analysis, and finally a machine learning model for predicting the photoelectric properties of the COFs is obtained. According to the model, through theoretical prediction of a novel COFs structure, a mapping relation between the COFs structure and photoelectric specificity can be established, and a novel COFs hypothetical structure can be reset according to important molecular fragments or word structures of the important molecular fragments; theoretical guidance can be provided for application in the fields of photoelectric catalysts and the like.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the fields of materials science and computer simulation technology, specifically to a method for reverse molecular design of organic framework materials (COFs) based on machine learning and quantitative calculation techniques. Background Technology

[0002] Professor Omar M. Yaghi, a pioneer in "network chemistry," has developed a series of novel long-range ordered porous materials, such as metal-organic frameworks (MOFs), zeolite imidazolate frameworks (ZIFs), and covalent organic frameworks (COFs), through the precise design and control of molecular bonding mechanisms and network topology. Among these, COFs are porous organic polymers composed of light elements such as C, O, and N linked by strong covalent bonds, exhibiting high crystallinity, high porosity, and excellent thermal and chemical stability. They have broad application prospects in photocatalysis, gas separation, and drug delivery. Due to their unique photoelectric properties, COFs show great application potential in photocatalytic water splitting for hydrogen production and photoelectrocatalysis. These materials have broad application prospects and are leading new directions in materials science research.

[0003] However, designing and screening novel COF materials with high photoelectric properties from scratch remains a tedious and time-consuming process, typically relying on traditional experimental screening methods, which are inefficient and costly. Traditional experimental synthesis methods are also time-consuming and labor-intensive, and struggle to systematically study the relationship between the structure and properties of COFs. Existing research methods still have limitations in terms of descriptor screening, model accuracy, and prediction range. Therefore, developing a method for controllably generating COFs based on computer simulations and predicting their photoelectric properties is of great significance.

[0004] In recent years, computational technologies such as computer simulation and machine learning have developed rapidly. The application of materials informatics and data science has provided new ideas for the research, rational design, and screening of novel COF materials. Machine learning-assisted material design has become a rapidly developing trend. By utilizing the structural information and property data of existing materials, a "structure-property relationship" can be established. By combining quantum chemical calculations and machine learning methods, the design process of COF materials can be effectively accelerated, the photoelectric properties of COFs can be efficiently predicted, and the design and experimental synthesis of new COF materials can be effectively guided.

[0005] Against this background, the present invention aims to develop a method for reverse molecular design of organic framework materials (COFs) based on machine learning and quantitative calculation techniques. By using machine learning techniques to construct a mapping relationship between COF structure and optoelectronic properties, the method can achieve efficient prediction and screening of the properties of novel COF materials, thereby accelerating the research and development process of COF materials, reducing research and development costs, and opening up new avenues for the application of novel COF materials. Summary of the Invention

[0006] The purpose of this invention is to provide a method for reverse molecular design of organic framework materials (COFs) based on machine learning and quantitative computation techniques, and to develop a database of machine learning models and predictions of their photoelectric properties. This method establishes a database of known COF structural parameters and properties, combines quantum chemical calculations and machine learning algorithms to construct a "structure-property" mapping relationship, and achieves accurate prediction and targeted design of COF photoelectric properties.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] This invention proposes a reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative computation techniques. The reverse molecular design method includes the following steps:

[0009] Step S1: Establish a COFs structure database;

[0010] Step S2: Select the organic ligand and node types, and mark the docking atoms when the ligand and node are connected in the CIF structure dataset of the organic ligand and node respectively;

[0011] Step S3: Based on the topology type, dock the labeled organic ligands with the nodes to construct a covalent organic framework (COF) model, and combine the organic ligands with the nodes to generate the CIF structure dataset of the COF model.

[0012] Step S4: Obtain the minimum repeating structural unit of the unit cell from the CIF structure dataset of the COFs model, and use it as a fragment of COFs for modeling. Then, perform DFT and TD-DFT calculations on the modeled COFs model to obtain the relevant numerical values ​​of the COFs model, the geometric and electronic structural parameters of the COFs, and the photoelectric property descriptors.

[0013] Step S5: Based on the relevant numerical values ​​of the COFs model, the geometric and electronic structural parameters of COFs, and the descriptors of photoelectric properties, a structure-performance mapping model of COFs is established using machine learning algorithms, and key descriptors and important segments and substructures that affect material properties are selected.

[0014] Step S6: Use the SHAP model to rank the descriptors by importance and use PCA to cluster the COFs structures to identify the material categories with high photocatalytic activity;

[0015] Step S7: Based on the descriptors ranked by importance, obtain the important segments and substructures that affect photoelectric properties, randomly recombine the important segments and substructures, and reverse-engineer novel COF structures; predict the relevant parameters of the hypothetical COFs through regression equations, and use the relevant parameters as theoretical basis to guide experimental synthesis.

[0016] Furthermore, the aforementioned organic framework materials include, but are not limited to, COFs, MOFs, or ZIFs.

[0017] Furthermore, step S1 above specifically includes:

[0018] Step S11: Obtain the structural information of existing COFs and perform 3D modeling based on Materials Studio software;

[0019] Step S12: Extract crystal structure parameters based on 3D modeling and establish a COFs structure database.

[0020] Furthermore, the aforementioned photoelectric property descriptors include HOMO and LUMO energy levels, band gap values, UV-Vis absorption spectra, and charge transfer path parameters.

[0021] Furthermore, the aforementioned machine learning algorithms include random forest, SVM, GBDT, and neural network models; key descriptors are screened using Pearson correlation coefficients.

[0022] Furthermore, descriptors corresponding to Pearson correlation coefficients with values ​​greater than 0.9 are considered redundant, while descriptors corresponding to Pearson correlation coefficients with values ​​no greater than 0.9 are retained.

[0023] Furthermore, cross-validation correlation coefficient R2 and error MAE and RMSE analyses were used to evaluate the model performance of each machine learning model on the test set.

[0024] The reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative computation techniques described in this invention can be entirely implemented using computer software. Therefore, correspondingly, this invention also provides a reverse molecular design system for organic framework materials (COFs) based on machine learning and quantitative computation techniques, the system comprising:

[0025] Storage device used to build a COF structure database;

[0026] Storage devices for selecting organic ligand and node types, and for marking docking atoms when ligands are connected to nodes in the CIF structure dataset of organic ligands and nodes respectively.

[0027] A storage device for connecting labeled organic ligands to nodes based on topology type to construct a covalent organic framework (COF) model, and combining organic ligands with nodes to generate a CIF structure dataset of the COF model.

[0028] This is used to obtain the minimum repeating structural unit of the unit cell from the CIF structure dataset of the COFs model, and use it as a fragment of COFs for modeling. Then, DFT and TD-DFT calculations are performed on the modeled COFs model to obtain the relevant numerical values ​​of the COFs model, the geometric and electronic structural parameters of the COFs, and the storage device of the photoelectric property descriptor.

[0029] This is used to establish a structure-performance mapping model of COFs based on relevant numerical values ​​of COFs models, geometric and electronic structural parameters of COFs, and descriptors of photoelectric properties, and to screen out key descriptors and important segments and substructures that affect material properties for storage devices.

[0030] This is used to rank descriptors by importance using the SHAP model and cluster COFs structures using PCA to identify storage devices with high photocatalytic activity.

[0031] Descriptors are used to sort by importance to obtain important segments and substructures that affect photoelectric properties. These important segments and substructures are then randomly recombined to reverse-engineer novel COF structures. Furthermore, regression equations are used to predict the relevant parameters of hypothetical COFs, which are then used as theoretical basis to guide the experimental synthesis of storage devices.

[0032] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative calculation techniques as described above.

[0033] The present invention also provides a computer device, which includes a memory and a processor. The memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes the reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative calculation techniques described above.

[0034] The beneficial effects of this invention are as follows:

[0035] 1. This invention establishes a COFs structure database and combines quantum chemical calculations with machine learning algorithms to achieve accurate prediction and targeted design from structural information to photoelectric properties. This method can efficiently screen COFs material models with excellent photoelectric properties, avoiding the extensive trial-and-error synthesis required by traditional experimental methods, greatly saving time and costs, and improving the efficiency of materials research and development.

[0036] 2. The method of the present invention has good universality. It is not only applicable to the performance prediction and optimization of known COFs materials, but can also be extended to the research of other similar framework materials. It provides a general and efficient solution for material design and performance prediction in related fields, which is of great significance to promoting the development and application of organic framework materials.

[0037] The present invention proposes a reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative calculation technology, which can predict the photoelectric properties of 4,000 designed COFs and their application prospects in photocatalytic water splitting reaction. Attached Figure Description

[0038] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0039] Figure 1 This is a flowchart of a reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative calculation technology proposed in this invention.

[0040] Figure 2 This is a schematic diagram of the modeling of some COFs structures described in this invention;

[0041] Figure 3 This is a cross-validation diagram of the machine learning algorithm and quantization calculation method described in this invention. Detailed Implementation

[0042] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known machine learning model construction methods and related usage methods are omitted to avoid unnecessary detail from hindering the description of this application.

[0043] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0044] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention.

[0045] Implementation Method 1, see [link] Figure 1 This embodiment aims to provide a method for reverse molecular design of organic framework materials (COFs) based on machine learning and quantitative computation techniques, developing a machine learning model and a database to predict their photoelectric properties. This method establishes a database of known COF structural parameters and properties, combines quantum chemical calculations and machine learning algorithms to construct a "structure-property" mapping relationship, achieving accurate prediction and targeted design of COF photoelectric properties.

[0046] like Figure 1 As shown, the reverse molecular design method includes the following steps:

[0047] Step S1: Establish a COFs structure database;

[0048] Step S2: Select the organic ligand and node types, and mark the docking atoms when the ligand and node are connected in the CIF structure dataset of the organic ligand and node respectively;

[0049] Step S3: Based on the topology type, dock the labeled organic ligands with the nodes to construct a covalent organic framework (COF) model, and combine the organic ligands with the nodes to generate the CIF structure dataset of the COF model.

[0050] Step S4: Quantum chemical calculations, including: obtaining the smallest repeating unit cell from the CIF structure dataset of the COFs model, using it as a fragment of COFs for modeling, and then performing DFT and TD-DFT calculations on the modeled COFs model to obtain the relevant numerical values ​​of the COFs model, the geometric and electronic structural parameters of COFs, and the photoelectric property descriptors.

[0051] Step S5: Based on the relevant numerical values ​​of the COFs model, the geometric and electronic structural parameters of COFs, and the descriptors of photoelectric properties, a structure-performance mapping model of COFs is established using machine learning algorithms, and key descriptors and important segments and substructures that affect material properties are selected.

[0052] Step S6: Use the SHAP model to rank the descriptors by importance and use PCA to cluster the COFs structures to identify the material categories with high photocatalytic activity;

[0053] Step S7: Based on the descriptors ranked by importance, obtain the important segments and substructures that affect photoelectric properties, randomly recombine the important segments and substructures, and reverse-engineer novel COF structures; predict the relevant parameters of the hypothetical COFs through regression equations, and use the relevant parameters as theoretical basis to guide experimental synthesis.

[0054] Implementation Method 2, see below Figure 2 and Figure 3 This embodiment describes in detail the method for reverse molecular design of organic framework materials (COFs) based on machine learning and quantitative calculation technology as described in Embodiment 1 above.

[0055] Step S1: Establish a COFs structure database;

[0056] Specifically:

[0057] Structural information of known COFs was obtained from literature and experimental data. Three-dimensional modeling was performed using Materials Studio software, and unit cell parameters were extracted. Crystal structure parameters such as XRD were fitted to construct a COF structure database.

[0058] Step S2: Select the organic ligand and node types, and mark the docking atoms when the ligand and node are connected in the CIF structure dataset of the organic ligand and node respectively;

[0059] Specifically:

[0060] Organic ligands and node types can be selected, for example, 40 organic fragments such as benzene rings, biphenyls, bipyridines, conjugated alkenes, triphenylamine, aniline, triazine, or benzothiazole can be chosen as ligands. By selecting appropriate organic ligands, the pore size, porosity, and functional properties of COFs can be controlled. For example, COFs with different porosities and pore sizes can be designed; highly crystalline COFs can also be designed to improve the stability and reproducibility of the material; COFs with specific dimensions can also be designed. Selecting appropriate node types can control the connection mode and overall structure of COFs, thereby controlling the stability and function of COFs. The docking atoms when ligands connect to nodes are marked in the CIF structure datasets of organic ligands and nodes, respectively. The CIF structure dataset is a file format used to describe crystal structures, which contains information such as the positions of all atoms and chemical bonds in the crystal. This embodiment, by marking the docking atoms when organic ligands connect to nodes, can clearly show the connection mode and structural characteristics of COFs, which is helpful for further structural analysis and performance prediction.

[0061] Step S3: Based on the topology type, dock the labeled ligands with the nodes to construct a covalent organic framework (COF) model, such as... Figure 2 As shown; and organic ligands are combined with nodes to generate a CIF structure dataset for the COFs model;

[0062] Specifically, the crystal structure is constructed using experimental data, the unit cell parameters are fitted, the relationship between COF nodes and organic ligands is obtained, the unit cell box is simulated, and a nearest neighbor list is used to record the connection information between nodes and organic ligands. Specifically, the number of the organic ligand connected to each node is recorded, and a nearest neighbor list is generated by traversing all nodes.

[0063] Step S4: Obtain the minimum repeating structural unit of the unit cell from the CIF structure dataset of the COFs model, and use it as a fragment of COFs for modeling. Then, perform DFT and TD-DFT calculations on the modeled COFs model to obtain the relevant numerical values ​​of the COFs model, the geometric and electronic structural parameters of the COFs, and the photoelectric property descriptors.

[0064] Specifically:

[0065] The smallest repeating unit cells were obtained from the CIF structural data and used as fragments for modeling COFs. Quantum chemical calculations were performed using Gaussian software to conduct density functional theory (DFT) and time-dependent density functional theory (TD-DFT) calculations on the COF models. This yielded values ​​for HOMO and LUMO energy levels, band gap values, UV-Vis absorption spectra, number of transferred electrons, hole-electron parameters, differential density, polarity, dipole moment, planarity, electronic transition paths, charge separation parameters, electron affinity, ionization energy, electron extraction energy, electron lifetime, work function, hole energy, excitation energy, dihedral angle, bond length, bond order, electronegativity, substituent types, elemental composition, symmetry, and topological connectivity. Descriptors for the geometric and electronic structure parameters and photoelectric properties of the COFs were then extracted.

[0066] Step S5: Based on the relevant numerical values ​​of the COFs model, the geometric and electronic structural parameters of COFs, and the descriptors of photoelectric properties, a structure-performance mapping model of COFs is established using machine learning algorithms, and key descriptors and important segments and substructures that affect material properties are selected.

[0067] Specifically:

[0068] Machine learning algorithms are applied to perform machine learning modeling. These algorithms include random forests, support vector machines (SVM), gradient boosting trees (GBDT), Gaussian regression, Bayesian algorithms, decision trees, neural networks, and deep learning. These algorithms are used to establish the structure-performance mapping relationship of COFs and to screen key descriptors and important segments and substructures that affect material properties.

[0069] In practical applications, the COFs database can be divided into training, validation, and test sets. The training set is used to train various machine learning models through grid search. During the iterative training of each machine learning model, the hyperparameters are optimized using ten-fold cross-validation based on the validation set. The performance of each machine learning model on the test set is evaluated, and the best machine learning model is selected.

[0070] Step S6: Use the SHAP model to rank the descriptors by importance and use PCA to cluster the COFs structures to identify the material categories with high photocatalytic activity;

[0071] Specifically:

[0072] Principal component analysis (PCA) and SHAP models were used to rank the importance of descriptors, analyze the influence of structural classification and group effects on photoelectric properties, and predict the application potential of COFs in photocatalysis. A correlation heatmap was used to determine important descriptors; chemical descriptors with a Pearson correlation coefficient greater than 0.9 were considered redundant; chemical descriptors with a Pearson correlation coefficient less than or equal to 0.9 were retained as initially screened chemical descriptors. In the step of training based on the COFs database and the screened chemical descriptors using the machine learning method described in claim 5, optimizing hyperparameters through 10-fold cross-validation and predicting model performance to obtain the optimal machine learning model, the cross-validation correlation coefficient R² and error MAE and RMSE analyses were used to evaluate the model performance of each machine learning model on the test set. Figure 3 As shown;

[0073]

[0074] In the process of optimizing hyperparameters, mean absolute error (MAE) and root mean square error (RMSE) are used to evaluate the loss of the function.

[0075] The mean absolute error (MAE) is:

[0076]

[0077] The root mean square error (RMSE) is:

[0078]

[0079] Step S7: Based on the descriptors ranked by importance, obtain the important segments and substructures that affect photoelectric properties, randomly recombine the important segments and substructures, and reverse-engineer novel COF structures; predict the relevant parameters of the hypothetical COFs through regression equations, and use the relevant parameters as theoretical basis to guide experimental synthesis.

[0080] Specifically:

[0081] A novel COF design library is generated based on machine learning models. By resetting fragments and substructures, novel hypothetical COF structures are obtained. Through regression equations, a series of parameters such as the ultraviolet absorption spectrum, electron transfer path, photogenerated electron efficiency, and electron lifetime of the hypothetical COFs are predicted, providing a theoretical basis for experiments, narrowing the experimental synthesis range, and accelerating the material development cycle.

[0082] Implementation Method 3: This implementation method illustrates a reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative calculation techniques, as proposed in the above implementation methods.

[0083] Step 1: Establish a COFs structure database: Collect structural information of 150 known COFs from the literature, use Materials Studio software to perform 3D modeling, extract unit cell parameters, and fit crystal structure parameters such as XRD.

[0084] Step 2: Selecting Organic Ligands and Node Types: Forty organic fragments, including benzene rings, biphenyl, bipyridine, conjugated alkenes, triphenylamine, aniline, triazine, and benzothiazole, were selected as ligands. Based on 200 topological networks, ligands and nodes were labeled and docking was performed to construct a COFs model. The organic ligands and nodes were combined using TOBACCO software to generate the CIF structure dataset of the COFs model.

[0085] Step 3: Quantum Chemical Calculations: Obtain the smallest repeating structural unit cell from the CIF structural data, using it as a fragment of COFs for modeling. Use Gaussian software to perform DFT and TD-DFT calculations on the COF model to obtain values ​​such as HOMO and LUMO energy levels, band gap values ​​and UV-Vis absorption spectra, number of transferred electrons, hole-electron parameters, differential density, polarity, dipole moment, planarity, electronic transition paths, charge separation parameters, electron affinity, ionization energy, electron extraction energy, electron lifetime, work function, hole energy, excitation energy, dihedral angle, bond length, bond order, electronegativity, substituent types, elemental composition, symmetry, and topological connectivity. Extract descriptors for the geometric and electronic structure parameters and photoelectric properties of the COFs.

[0086] Step 4: Machine learning modeling: The random forest algorithm is used to establish the "structure-performance" mapping relationship, independent descriptors are selected by Pearson correlation coefficient, and hyperparameters are optimized to improve the model prediction accuracy.

[0087] Step 5: Importance Analysis: Use the SHAP model to resolve the contribution of descriptors to photoelectric performance, and combine principal component analysis to cluster COF structures to identify material categories with high photocatalytic activity.

[0088] Step 6: Reverse design: From the importance analysis of SHAP, obtain the important segments and substructures that affect its photoelectric properties, and randomly recombine the segments and substructures to obtain novel COF structures.

[0089] Step 7: Prediction and Screening: A library of 4,000 novel COFs is generated based on machine learning models. A series of parameters such as the ultraviolet absorption spectrum, electron transfer path, photogenerated electron efficiency, and electron lifetime of the hypothetical COFs are predicted by regression equations, providing a theoretical basis for experiments, narrowing the experimental synthesis range, and accelerating the material development cycle.

[0090] Implementation Method 4: The reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative calculation techniques proposed in the above implementation methods can be fully implemented using computer software. Therefore, this implementation method provides a reverse molecular design system for organic framework materials (COFs) based on machine learning and quantitative calculation techniques, the system comprising:

[0091] Storage device used to build a COF structure database;

[0092] Storage devices for selecting organic ligand and node types, and for marking docking atoms when ligands are connected to nodes in the CIF structure dataset of organic ligands and nodes respectively.

[0093] A storage device for connecting labeled organic ligands to nodes based on topology type to construct a covalent organic framework (COF) model, and combining organic ligands with nodes to generate a CIF structure dataset of the COF model.

[0094] This is used to obtain the minimum repeating structural unit of the unit cell from the CIF structure dataset of the COFs model, and use it as a fragment of COFs for modeling. Then, DFT and TD-DFT calculations are performed on the modeled COFs model to obtain the relevant numerical values ​​of the COFs model, the geometric and electronic structural parameters of the COFs, and the storage device of the photoelectric property descriptor.

[0095] This is used to establish a structure-performance mapping model of COFs based on relevant numerical values ​​of COFs models, geometric and electronic structural parameters of COFs, and descriptors of photoelectric properties, and to screen out key descriptors and important segments and substructures that affect material properties for storage devices.

[0096] This is used to rank descriptors by importance using the SHAP model and cluster COFs structures using PCA to identify storage devices with high photocatalytic activity.

[0097] Descriptors are used to sort by importance to obtain important segments and substructures that affect photoelectric properties. These important segments and substructures are then randomly recombined to reverse-engineer novel COF structures. Furthermore, regression equations are used to predict the relevant parameters of hypothetical COFs, which are then used as theoretical basis to guide the experimental synthesis of storage devices.

[0098] Implementation Method 5: This implementation method provides a computer-readable storage medium storing a computer program. When the computer program is run by a processor, it executes the reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative calculation techniques described in any of the above implementation methods.

[0099] Implementation Method Six: This implementation method provides a computer device, which includes a memory and a processor. The memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes the reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative calculation technology as described in any of the above implementation methods.

[0100] This embodiment provides a computer device, the hardware of which is a general-purpose model and is not shown in the figure. The system includes a processor and a memory, which can be connected via a bus or other means. The memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs and modules, as well as corresponding program instructions / modules. The processor executes various functional applications and data processing by running the non-transitory software programs, instructions and modules stored in the memory, so as to realize the reverse molecular design method and steps of organic framework materials (COFs) based on machine learning and quantitative calculation technology in the above method embodiment.

[0101] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0102] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative computation techniques, characterized in that, The method is as follows: S1: Establish a COFs structure database; S2: Select the organic ligand and node types, and mark the docking atoms when the ligand and node are connected in the CIF structure dataset of the organic ligand and node respectively; S3: Based on the topology type, the labeled organic ligands and nodes are docked to construct a covalent organic framework (COF) model, and the organic ligands and nodes are combined to generate the CIF structure dataset of the COF model. S4: Obtain the minimum repeating unit cell from the CIF structure dataset of the COFs model, and use it as a fragment of COFs for modeling. Then, perform DFT and TD-DFT calculations on the modeled COFs model to obtain the relevant numerical values ​​of the COFs model, the geometric and electronic structural parameters of COFs, and the photoelectric property descriptors. S5: Based on the relevant numerical values ​​of the COFs model, the geometric and electronic structural parameters of COFs, and the descriptors of photoelectric properties, a structure-performance mapping model of COFs is established using machine learning algorithms, and key descriptors and important segments and substructures that affect material properties are selected. S6: The SHAP model is used to rank the descriptors by importance, and PCA is used to cluster the COFs structures to identify the material categories with high photocatalytic activity. S7: Based on the descriptors ranked by importance, obtain the important segments and substructures that affect photoelectric properties, randomly recombine the important segments and substructures, and reverse-engineer novel COF structures; predict the relevant parameters of the hypothetical COFs through regression equations, and use the relevant parameters as theoretical basis to guide experimental synthesis.

2. The reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative computation technology according to claim 1, characterized in that, Organic framework materials include, but are not limited to, COFs, MOFs, or ZIFs.

3. The reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative computation technology according to claim 1, characterized in that, S1 specifically refers to: Obtain the structural information of existing COFs and perform 3D modeling based on Materials Studio software; Crystal structure parameters were extracted based on 3D modeling to establish a COF structure database.

4. The reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative computation technology according to claim 1, characterized in that, Photoelectric property descriptors include HOMO and LUMO energy levels, band gap values, UV-Vis absorption spectra, and charge transfer path parameters.

5. The reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative computation technology according to claim 1, characterized in that, Machine learning algorithms include random forest, SVM, GBDT, and neural network models; key descriptors are screened using Pearson correlation coefficients.

6. The reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative computation technology according to claim 5, characterized in that, Descriptors corresponding to Pearson correlation coefficients with values ​​greater than 0.9 are considered redundant, while descriptors corresponding to Pearson correlation coefficients with values ​​no greater than 0.9 are retained.

7. The reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative computation technology according to claim 5, characterized in that, Cross-validation correlation coefficient R 2 Error MAE and RMSE analyses were used to evaluate the performance of each machine learning model on the test set.

8. A reverse molecular design system for organic framework materials (COFs) based on machine learning and quantitative computation techniques, characterized in that, The system includes a storage device for performing the method and steps of claim 1.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative computation techniques as described in any one of claims 1-7.

10. A computer device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes the reverse molecular design method for organic framework materials (COFs) based on machine learning and quantitative calculation technology as described in any one of claims 1-7.

Citation Information

Cited By

  • A method for reverse design of surfactant molecules for oil displacement

    CN122598843A