Artificial intelligence guided molecular screening for coordination framework compounds

Abstract

Described herein are methods and systems for proposing coordination framework compounds, such as crystalline porous materials, crystalline open frameworks, reticular chemicals, metal-organic framework (MOF) compounds, covalent organic framework (COF) compounds, zeolitic imidazole framework (ZIF) compounds, and combinations thereof. Also described herein are coordination framework compounds produced by the methods and systems. The methods and systems described herein combine machine learning and chemistry to propose chemically effective and improved performance coordination framework compounds that meet the different objectives of material discovery.

Description

[0001] Government Support

[0002] This invention was made with government support under HR001121C0020 awarded by the Defense Advanced Research Projects Agency (DARPA). The government has certain rights in the invention. BACKGROUND

[0003] The field of the present disclosure relates generally to methods and systems for proposing coordination framework compounds, such as crystalline porous materials, crystalline open frameworks, reticular chemical substances, metal-organic framework (MOF) compounds, covalent organic framework (COF) compounds, zeolitic imidazolate framework (ZIF) compounds, and combinations thereof. The field of the present disclosure also relates to coordination framework compounds made therefrom.

[0004] Coordination framework compounds, such as MOF, ZIF, and COF compounds, can be useful for a variety of purposes. For example, they can be particularly useful in carbon capture sorbent systems or methane capture sorbent systems, such as for post-combustion air capture and direct air capture of carbon dioxide (CO2). As another example, they are particularly useful in atmospheric water extraction (AWE), where water is generated remotely and on-demand. However, carbon capture and AWE processes each require constant and iterative improvements to materials.

[0005] Existing iterative trial-and-error approaches to coordination framework compound discovery based on chemical knowledge require significant investment and effort to achieve innovative and effective materials. While pure machine learning (ML) approaches can be helpful in discovery, they lack deep chemical substance insight. For example, while a model developed by Xie et al. (“Crystal Diffusion Variational Autoencoder for Periodic Material Generation.” arXiv preprint arXiv:2110.06197 (2021)) improves machine learning capabilities applied to small crystals, it lacks chemical knowledge and guidance. The model also only applies to small crystals, and does not prove larger structures, such as MOF compounds.

[0006] Due to the shortcomings of existing approaches, it is desirable to develop an approach that combines the advantages of machine learning-based and chemical knowledge-based approaches in order to more effectively propose and discover coordination framework compounds. SUMMARY

[0007] In one aspect, provided herein is a method of proposing coordination framework compounds. In exemplary embodiments, the method is performed using a coordination framework compound proposal (CFCP) computing device comprising a processor coupled to a memory device. The exemplary method comprises: generating, using a machine learning model of the CFCP computing device, an initial set of coordination framework compounds; subjecting at least one of the coordination framework compounds of the initial set of coordination framework compounds to an examination of at least one chemical property; and generating, using the CFCP computing device, a preliminary set of coordination framework compounds based on the initial set of coordination framework compounds and the examination of the at least one chemical property of the at least one of the coordination framework compounds of the initial set of coordination framework compounds with the machine learning model. The method further comprises proposing, using the CFCP computing device, a coordination framework compound.

[0008] In another aspect, provided herein is a coordination framework compound proposal (CFCP) computing device comprising: a memory; and a processor communicatively coupled to the memory, wherein the processor is programmed to: generate, with a machine learning model, an initial set of coordination framework compounds; and subject at least one of the coordination framework compounds of the initial set of coordination framework compounds to an examination of at least one chemical property. The processor is further programmed to generate, with the machine learning model, a preliminary set of coordination framework compounds based on the initial set of coordination framework compounds and the examination of the at least one chemical property of the at least one of the coordination framework compounds of the initial set of coordination framework compounds; and propose a coordination framework compound.

[0009] In yet another aspect, provided herein is a non-transitory computer-readable storage medium having computer-executable instructions embodied thereon, wherein when executed by a coordination framework compound proposal (CFCP) computing device comprising at least one processor in communication with a memory, the computer-readable instructions cause the CFCP computing device to: generate, with a machine learning model, an initial set of coordination framework compounds; and subject at least one of the coordination framework compounds of the initial set of coordination framework compounds to an examination of at least one chemical property. The computer-readable instructions also cause the CFCP computing device to generate, with the machine learning model, a preliminary set of coordination framework compounds based on the initial set of coordination framework compounds and the examination of the at least one chemical property of the at least one of the coordination framework compounds of the initial set of coordination framework compounds; and propose a coordination framework compound. BRIEF DESCRIPTION OF DRAWINGS

[0010] These and other features, aspects, and advantages of the present disclosure will become better understood when the following detailed description is read with reference to the accompanying drawings in which all changes in and relative sizes and proportions of the elements illustrated therein are intended to convey a particular understanding and are not being meant in a limiting sense or as an indication that such changes are not suggested, wherein like characters designate like parts throughout the drawings and wherein: a.

[0011] Figure 1A is an example method flowchart according to the present disclosure;

[0012] Figure 1B is an example method flowchart according to the present disclosure;

[0013] Figure 2 is another example method flowchart according to the present disclosure;

[0014] Figure 3 is a further example method flowchart according to the present disclosure;

[0015] Figure 4 is yet another example method flowchart according to the present disclosure;

[0016] Figure 5 is a block diagram of an example computer system according to the present disclosure;

[0017] Figure 6 is an example configuration of a server system (such as the computer system of Figure 1A

[0018] Figure 7 is an example configuration of a client system shown in Figure 1A

[0019] Figure 8 is an example method according to the present disclosure;

[0020] Figure 9 is an example coordination framework compound obtained from an ML model according to the present disclosure; and

[0021] Figure 10 includes example molecules used in replacement linkers in coordination framework compounds according to the present disclosure.

[0022] The drawings provided herein are intended to illustrate features of embodiments of the present disclosure. It is believed that these features are applicable to a wide variety of systems including one or more embodiments of the present disclosure. As such, the drawings are not intended to be inclusive of all of the features that would be required by persons of ordinary skill in the art to practice embodiments of the present disclosure described herein. DETAILED DESCRIPTION

[0023] Embodiments described herein overcome at least some of the shortcomings of known methods of proposing coordination framework compounds. The present embodiments combine machine learning and chemistry to propose chemically valid and performance-improved coordination framework compounds that meet different objectives of materials discovery. Machine learning provides materials structure imagination and generation, and chemistry guides proposing modifications and improvements to the materials generated by machine learning.​​

[0024] As used herein, coordination framework compounds include nodes, which are points of structural connectivity and give rise to secondary building units (SBUs), one or more linkers, and a topology that defines how the SBUs and linkers are connected together. Changing any of these aspects, alone or in combination, results in new and different coordination framework compounds. The methods described herein are applicable to the discovery of any crystal structure.

[0025] Examples of coordination framework compounds include metal-organic framework (MOF) compounds, covalent organic framework (COF) compounds, zeolitic imidazolate framework (ZIF) compounds, crystalline porous materials, crystalline open frameworks, reticular chemical substances, and combinations thereof. MOF compounds have strong bonds between metal atoms and charged ligands, and can include, but are not limited to, only one or more types of metal atoms in the SBU and linker composition, while COF compounds have strong covalent bonds between light elements (e.g., B, C, N, O) and include, but are not limited to, only one or more types of organic structures as SBUs. ZIFs are a sub-class of MOF compounds that are composed of tetrahedrally coordinated transition metal ions (e.g., Fe, Co, Cu, Zn) connected by imidazolate salt linkers that are topologically isomorphic to zeolites.

[0026] In general, coordination framework compounds are not limited to having only one SBU or one linker. Rather, they can have more complex structures composed of multiple types of nodes and linkers. In other words, the components of coordination framework compounds can be combined in a variety of ways. For example, a coordination framework compound can include two linkers and one metal node.

[0027] The example embodiments described herein include a method of proposing coordination framework compounds, wherein the method is performed using a coordination framework compound proposal (CFCP) computing device that includes a processor coupled to a memory device. The method includes: generating an initial set of coordination framework compounds using a machine learning model of the CFCP computing device; subjecting at least one of the coordination framework compounds of the initial set of coordination framework compounds to an examination of at least one chemical property; generating a preliminary set of coordination framework compounds using the machine learning model based on the initial set of coordination framework compounds and the examination of the at least one chemical property of at least one of the coordination framework compounds of the initial set of coordination framework compounds using the CFCP computing device; and proposing coordination framework compounds using the CFCP computing device.

[0028] The method can further include one or more additional process steps adapted to facilitate the methods described herein. In some embodiments, the method further includes subjecting at least one coordination framework compound of the preliminary set of coordination framework compounds to a review of at least one chemical property. In some embodiments, the method further includes validating at least one coordination framework compound of the preliminary set of coordination framework compounds. In some embodiments, the method further includes performing at least one iteration of a sequence comprising: generating a further preliminary set of coordination framework compounds with the machine learning model based on the at least one generated set of coordination framework compounds and the review of at least one chemical property of at least one coordination framework compound of the at least one generated set of coordination framework compounds; optionally subjecting at least one coordination framework compound of the further preliminary set of coordination framework compounds to a review of at least one chemical property; and optionally validating at least one coordination framework compound of the further preliminary set of coordination framework compounds.

[0029] Exemplary embodiments also include a coordination framework compound proposal (CFCP) computing device comprising a memory and a processor communicatively coupled to the memory, wherein the processor is programmed to implement the method embodiments described herein.

[0030] The processor can be further programmed to perform one or more additional steps adapted to facilitate the methods described herein. In some embodiments, the processor is further programmed to subject at least one coordination framework compound of the preliminary set of coordination framework compounds to a review of at least one chemical property. In some embodiments, the processor is further programmed to validate at least one coordination framework compound of the preliminary set of coordination framework compounds. In some embodiments, the processor is further programmed to perform at least one iteration of a sequence comprising: generating a further preliminary set of coordination framework compounds with the machine learning model based on the at least one generated set of coordination framework compounds and the review of at least one chemical property of at least one coordination framework compound of the at least one generated set of coordination framework compounds; optionally subjecting at least one coordination framework compound of the further preliminary set of coordination framework compounds to a review of at least one chemical property; and optionally validating at least one coordination framework compound of the further preliminary set of coordination framework compounds.

[0031] Exemplary embodiments also include a non-transitory computer- readable storage medium having computer-executable instructions embodied thereon, wherein when executed by a coordination framework compound proposal (CFCP) computing device comprising at least one processor in communication with a memory. The computer-readable instructions cause the coordination framework compound proposal computing device to implement the method embodiments described herein.

[0032] The computer-readable instructions can include one or more additional steps adapted to facilitate the methods described herein. In some embodiments, the computer-readable instructions further cause the CFCP computing device to subject at least one of the coordination framework compounds of the preliminary set of coordination framework compounds to a review of at least one chemical property. In some embodiments, the computer-readable instructions further cause the CFCP computing device to validate at least one of the coordination framework compounds of the preliminary set of coordination framework compounds. In some embodiments, the computer-readable instructions further cause the CFCP computing device to perform at least one iteration of a sequence comprising: generating a further preliminary set of coordination framework compounds based on the at least one generated set of coordination framework compounds and the review of at least one chemical property of at least one of the coordination framework compounds of the at least one generated set of coordination framework compounds; optionally subjecting at least one of the coordination framework compounds of the further preliminary set of coordination framework compounds to a review of at least one chemical property; and optionally validating at least one of the coordination framework compounds of the further preliminary set of coordination framework compounds.

[0033] Generally, any of the methods according to the present disclosure can be implemented using a coordination framework compound proposal (CFCP) computing device. As used herein, a coordination framework compound proposal computing device includes any suitable computing device known in the art that facilitates the methods described herein. Suitable computing devices can include, but are not limited to, a computer, a desktop computer, a handheld computer, or a smartphone alone.

[0034] Generally, a coordination framework compound can be any suitable coordination framework compound implemented or proposed by the exemplary methods described herein. In some embodiments, the coordination framework compound is a coordination framework compound of an initial set of coordination framework compounds, a coordination framework compound of a preliminary set of coordination framework compounds, or a coordination framework compound of a further preliminary set of coordination framework compounds. In other words, the coordination framework compound can be a hypothetical coordination framework compound, a coordination framework compound implemented by chemical insight into the hypothetical coordination framework compound, or a coordination framework compound implemented by iterative analysis of coordination framework compounds.

[0035] In some embodiments, the machine learning model is trained with existing coordination framework compounds, new coordination framework compounds, and combinations thereof. In some embodiments, the machine learning model is trained with existing coordination framework compounds.

[0036] Generally, the machine learning model can use any suitable technique that facilitates the exemplary methods described herein. In some embodiments, the machine learning model uses at least one of the following techniques: latent space, inverse search, variational autoencoder (VAE), crystal diffusion variational autoencoder (CDVAE), inverse search of VAE latent space, graph neural network (GNN), neural network, optimization, and combinations thereof. In some embodiments, the machine learning model is programmed to learn a latent space via artificial intelligence, which is configured to reconstruct coordination framework compound crystal structures and accurately predict associated target properties.

[0037] In some embodiments, the machine learning model is programmed to learn via a VAE (e.g., CDVAE) that maps coordination framework compound structures (e.g., crystal structures of MOFs) to a point in a latent space, which then reconstructs the point from the latent space to the original coordination framework compound structure. In this case, any point in the latent space corresponds to its correct crystal structure, and a search can be performed in the latent space to propose new coordination framework compound structures. Furthermore, since each coordination framework compound will have certain target chemical properties, such as water isotherms, this latent space can also be mapped to accurately predict the target chemical properties of each coordination framework compound. Thus, each point in the latent space will correspond to its correct coordination framework compound as well as its target chemical properties. Reverse search and target property optimization techniques can be performed to find new coordination framework compounds with more desirable target chemical properties. This VAE method can be combined with GNN techniques that map crystal structures to machine-readable graphs that can be read and trained by the machine learning model.

[0038] In some embodiments, the at least one chemical property is selected from one or more of the following: adsorbate uptake capacity, adsorbate uptake kinetics, adsorbate weight productivity, adsorbate volume productivity, adsorbate isotherms and isosteres, pore size, pore volume, heat of adsorption, isosteric heat of adsorption, chemical stability, thermal stability, mechanical stability, synthetic feasibility, zeta potential, surface energy, hydrophobicity, hydrophilicity, chemical performance, chemical modification for improved, and combinations thereof.

[0039] In some embodiments, validating the at least one of the coordination framework compounds includes experimentally validating the structure and / or properties of the at least one of the coordination framework compounds.

[0040] In some embodiments, experimental verification of the structure and / or properties of at least one coordination framework compound includes experimental verification using one or more techniques selected from the following: powder X-ray diffraction (PXRD), single-crystal X-ray diffraction, solid-state nuclear magnetic resonance (SS-NMR), digested NMR, and combinations thereof.

[0041] In some embodiments, experimental verification of the structure and / or properties of at least one coordination framework compound includes using first-principles calculations to relax (e.g., electronic relaxation and / or optimization) the structure to ensure the stability of the coordination framework compound structure. The pore volume and pore size of the coordination framework compound are then calculated. Water stability and hydrophilicity are analyzed to verify the compound's potential to adsorb more water molecules. Subsequently, the calculated water isotherm of the proposed coordination framework compound is estimated using Gibbs Ensemble Monte Carlo (GEMC) calculations to estimate the compound's water adsorption capacity. The proposed coordination framework compound is then verified through synthetic experiments and experimentally measured chemical properties, such as the water isotherm.

[0042] In some implementations, the machine learning model generates preliminary coordination framework compounds based on at least one of the following inputs: the crystal structure of existing coordination framework compounds, target properties, target chemical properties, adsorption isotherms, water adsorption isotherms, pore volume, pore size, water stability, hydrophilicity, and combinations thereof.

[0043] In some embodiments, the review of at least one chemical property is performed by a machine, a person, or a combination thereof. In some embodiments, the review of at least one chemical property is performed by a person. In some embodiments, a CFCP computing device is used to perform the review of at least one chemical property.

[0044] Generally speaking, there are several challenges to applying ML to coordination framework compounds. These challenges include the reversibility of representing the periodic crystal structure as an input to the ML, the invariance of the crystal structure, and the large number of unit cells and atoms.

[0045] An exemplary periodic structure can be represented as M = (A, X, L), where N is the number of atoms; A ∈ A N Atom types; X∈R N×3 Atomic position; L∈R 3×3 Periodic lattice; c∈R | A | Composition. Graphically, periodic structures can be represented using atoms as nodes and bonds as edges.

[0046] Regarding the invariance of materials, substitution invariance includes exchanging the exponents of any pair of atoms. Translation invariance includes translating X by any vector. Rotation invariance includes rotating X and L together by any rotation matrix. Periodicity invariance includes an infinite number of unit cells with different shapes and sizes.

[0047] ML training can be implemented as follows: First, a periodic GNN encoder encodes M as z. Second, a property predictor predicts c, L, and N of M and predicts target properties, such as water isotherms or pore volume (PV), based on z. Third, a periodic GNN decoder... Denoising is performed, conditioned on z. Fourth, This was obtained by adding different levels of noise to X and A.

[0048] Material optimization can be achieved as follows: First, start with an existing structure M (e.g., MOF-303) and encode M as z. Second, optimize PV relative to z to find z' that gives the desired target properties (such as the desired water isotherm or a larger PV). Third, use z' to predict c, L, and N. Fourth, randomly initialize the initial M0. Fifth, decode M0 to update the optimized material M' that gives the desired target properties, such as the desired water isotherm or a larger PV.

[0049] In some implementations, variational autoencoder-based ML models (such as crystal diffusion variational autoencoders (CDVAE)) are used to generate target-oriented novel coordination framework compounds. In some preferred implementations, CDVAEs are trained only on existing coordination framework compounds to generate target-oriented novel MOFs.

[0050] In some embodiments, the code library is used to perform deconstruction, modification, and / or reconstruction of ML-generated coordination framework compounds. Generally, the code library is compatible with any coordination framework compound. In some embodiments, the code library is selected from the group consisting of: ToBaCCo-based code libraries, MOFid-based code libraries, molfunc-based code libraries, and combinations thereof. In some embodiments, the ToBaCCo-based code library is used to perform deconstruction-modification-reconstruction of ML-generated coordination framework compounds having rod-shaped secondary building units (SBUs).

[0051] In many embodiments, the proposed coordination framework compound can be used for any suitable purpose known in the art. In some embodiments, the proposed coordination framework compound is used in adsorbent systems. In some embodiments, the proposed coordination framework compound is used in carbon capture adsorbent systems. In some embodiments, the proposed coordination framework compound is used in moisture adsorbent systems. In some embodiments, the proposed coordination framework compound is used for gas capture. In some embodiments, the proposed coordination framework compound is used for post-combustion capture of CO2 and / or direct air capture of CO2. In some embodiments, the proposed coordination framework compound is used for atmospheric water extraction.

[0052] Now turn to the attached image. Figure 1A This is an exemplary method flowchart 110. Method flowchart 110 depicts exemplary steps of an embodiment of the method described herein and is not intended to limit the method embodiment. In an exemplary embodiment, an initial set of coordination framework compounds 112 is generated using a machine learning model of a CFCP computing device. At least one coordination framework compound in the initial set of coordination framework compounds is subjected to a review of at least one chemical property 114. Using a CFCP computing device, based on the initial set of coordination framework compounds and the review of at least one chemical property of at least one coordination framework compound in the initial set of coordination framework compounds, a preliminary set of coordination framework compounds 116 is generated using a machine learning model. Coordination framework compounds 118 are proposed using a CFCP computing device.

[0053] Figure 1A The method flowchart 110 may be recursive and / or iterative, such that any exemplary step may be performed more than once and / or used to notify another exemplary step. One such exemplary implementation is... Figure 1BThe mixed ML-chemistry coordination framework compound proposal workflow is shown. In this active learning workflow implementation, method flowchart 120 depicts exemplary steps of the method implementation described herein and is not intended to limit the method implementation. In the exemplary implementation, machine learning model 122 outputs 124 a proposal of coordination framework compounds. At least one of the proposed coordination framework compounds is subjected to 126 chemical calculations and / or experiments. The results of the chemical calculations and / or experiments may lead to a proposal 134 for modification of the proposed coordination framework compound. This modification proposal 134 may be directed to machine learning model 122 or as the proposed coordination framework compound output 124. The experimental feasibility and calculated and / or measured chemical properties of the coordination framework compound are determined 128. The proposed compound with the desired properties may be labeled 136 as a target for laboratory-scale synthesis. Other proposed compounds may be compared with experimental and hypothetical coordination framework compounds 130. The comparison may further include structural and isotherm inputs 132. The learning of the method may be reapplied to the machine learning model until satisfactory results are achieved.

[0054] Figure 2 This is an exemplary flowchart 210 for proposing new coordination framework compounds based on known coordination framework compounds. Object 212 includes starting with a known coordination framework compound. Object 214 includes modifying the SBU, connector, and / or topology of a known coordination framework compound. Object 216 includes proposing new coordination framework compound structures based on variants of known coordination framework compounds. This approach is more targeted and allows for the immediate suggestion of coordination framework compounds. However, it requires basic chemical insight.

[0055] Figure 3This is an exemplary target flowchart 310. Object 312 includes inputs for coordination framework compounds. Object 314 includes an ML model, such as a VAE, updated with the inputs. Object 316 includes predictions of coordination framework compound properties, such as a water isotherm. Object 318 includes a backward search of the latent space to suggest and validate new coordination framework compounds. These new coordination framework compounds can be used as inputs in object 312 in subsequent iterations of the exemplary target flowchart 310. The exemplary target flowchart 310 allows learning coordination framework compound construction information based on known coordination framework compounds. It includes a well-defined latent space representing the space of coordination framework compounds. It requires a large number of coordination framework compound structures with chemical properties, such as pore volume, H-bonding site density, and a water isotherm. In some implementations, the model may include a large number of coordination framework compound structures, where water isotherms calculated by DFT / GCMC or DFT / GEMC are computed using techniques such as density functional theory (DFT) and / or grand canonical Monte Carlo (GCMC) calculations and / or Gibbs ensemble Monte Carlo (GEMC) calculations. It can explore coordination framework compounds without rod-like metallic nodes.

[0056] Figure 4 This is an exemplary ML model flowchart 410. Object 412 includes the input of the coordination framework compound. Object 414 includes a low-dimensional latent space compared to the input, but capable of searching for desired coordination framework compound properties, as well as reconstructing and mapping to new coordination framework compounds. Object 416 includes a first output, which includes the reconstructed coordination framework compound. Object 418 includes a second output, which includes coordination framework compound properties, such as water isotherms.

[0057] Figure 5 This is a block diagram of an exemplary embodiment of a computer system 500 for proposing coordination framework compounds, including a computing device 502, according to an exemplary embodiment of the present disclosure. The computing device 502 may also be referred to herein as a coordination framework compound proposal (CFCP) computing device. In the exemplary embodiment, system 500 is used to propose coordination framework compounds as described herein. Computer system 500 can be used to implement one or more of the methods described herein.

[0058] More specifically, in an exemplary embodiment, system 500 includes a computing device 502 and multiple client subsystems connected to the computing device 502, also referred to as client systems 504. In one embodiment, client system 504 is a computer including a web browser, enabling the computing device 502 to be accessed by the client system 504 using the Internet and / or network 506. Client system 504 is interconnected to the Internet via a number of interfaces, including network 506, such as a local area network (LAN) or wide area network (WAN), dial-up connection, cable modem, dedicated high-speed Integrated Services Digital Network (ISDN) line, and RDT network. Client system 504 may include external systems for storing data. The computing device 502 also uses network 506 to communicate with one or more data sources 514. Additionally, client system 504 may additionally use network 506 to communicate with data source 514. Furthermore, in some embodiments, one or more client systems 504 may serve as data source 514, as described herein. Client system 504 may be any device capable of interconnecting to the Internet, including web-based telephones, PDAs, or other web-based connectable devices.

[0059] Database server 508 connects to database 512, which contains information about various matters, as described in more detail below. In one embodiment, centralized database 512 is stored on device 502 and can be accessed by potential users at one of client systems 504 who log in to computing device 502 through one of client systems 504. In another embodiment, database 512 is stored remotely from device 502 and can be decentralized. Database 512 can be a database configured to store information used by computing device 502, including, for example, transaction records, as described herein.

[0060] Database 512 may include a single database with separate sections or partitions, or it may include multiple databases, each separate from the others. Database 512 may store data received from data source 514 and generated by computing device 502. For example, database 512 may store coordination framework compound data, as described in detail herein.

[0061] In an exemplary embodiment, client system 504 may be associated with any party capable of using system 500 as described herein. In an exemplary embodiment, at least one of client systems 504 includes a user interface 510. For example, user interface 510 may include an interactive graphical user interface such that coordination framework compound data and proposals transferred from computing device 502 to client system 504 can be displayed in a graphical format. A user of client system 504 can interact with user interface 510 to view, explore, and otherwise interact with the displayed information.

[0062] In an exemplary implementation, computing device 502 receives data from multiple data sources 514 and aggregates and analyzes the received data (e.g., using machine learning) to propose coordination framework compounds, as described in detail herein.

[0063] Figure 6 An exemplary configuration of a server system 602 (such as a computing device) according to an exemplary embodiment of the present disclosure is shown. The server system 602 can be used to implement one or more methods described herein. The server system 602 may also include, but is not limited to, a database server (not shown). In an exemplary embodiment, the server system 602 proposes coordination framework compounds as described herein.

[0064] Server system 602 includes a processor 606 for executing instructions. For example, instructions may be stored in memory region 610. Processor 606 may include one or more processing units for executing instructions (e.g., in a multi-core configuration). Instructions can run on various operating systems on server system 602 (such as UNIX, LINUX, Microsoft...). Execution within (etc.). It should also be understood that when a computer-based method is started, various instructions may be executed during initialization. Some operations may be required to execute one or more of the processes described herein, while others may be more general and / or specific to a particular programming language (e.g., C, C#, C++, Java, or other suitable programming languages, etc.).

[0065] Processor 606 is operatively coupled to communication interface 604, enabling server system 602 to communicate with remote devices, such as user systems or another server system 602. For example, communication interface 604 may receive requests from client systems via the Internet (not shown).

[0066] Processor 606 may also be operatively coupled to storage device 612. Storage device 612 is hardware suitable for any computer operation of storing and / or retrieving data. In some embodiments, storage device 612 is integrated into server system 602. For example, server system 602 may include one or more hard disk drives as storage device 612. In other embodiments, storage device 612 is external to server system 602 and can be accessed by multiple server systems 602. For example, storage device 612 may include multiple storage units, such as hard disks or solid-state drives in a redundant array of inexpensive disks (RAID) configuration. Storage device 612 may include storage area network (SAN) and / or attached network storage (NAS) systems.

[0067] In some implementations, processor 606 is operatively coupled to storage device 612 via storage interface 608. Storage interface 608 is any component capable of providing processor 606 with access to storage device 612. Storage interface 608 may include, for example, an Advanced Technology Attachment (ATA) adapter, a Serial ATA (SATA) adapter, a Small Computer System Interface (SCSI) adapter, a RAID controller, a SAN adapter, a network adapter, and / or any component that provides processor 606 with access to storage device 612.

[0068] Memory region 610 may include, but is not limited to, random access memory (RAM) (such as dynamic RAM (DRAM) or static RAM (SRAM)), read-only memory (ROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), and non-volatile RAM (NVRAM). The above memory types are merely examples and are therefore not limited to the types of memory that can be used to store computer programs.

[0069] Figure 7 An exemplary configuration of a client computing device 702 is shown. The client computing device 702 can be used to implement one or more methods described herein. The client computing device 702 may include, but is not limited to, a client system (“client computing device”) 504. The client computing device 702 includes a processor 704 for executing instructions. In some embodiments, the executable instructions are stored in a memory region 706. The processor 704 may include one or more processing units (e.g., in a multi-core configuration). The memory region 706 is any device that allows storage and retrieval of information such as executable instructions and / or other data. The memory region 706 may include one or more computer-readable media.

[0070] The client computing device 702 also includes at least one media output component 708 for presenting information to the user 714. The media output component 708 is any component capable of conveying information to the user 714. In some embodiments, the media output component 708 includes an output adapter, such as a video adapter and / or an audio adapter. The output adapter is operatively coupled to the processor 704 and operatively coupled to an output device, such as a display device (e.g., a liquid crystal display (LCD), an organic light-emitting diode (OLED) display, a cathode ray tube (CRT), or an "electronic ink" display) or an audio output device (e.g., a speaker or headphones).

[0071] In some implementations, the client computing device 702 includes an input device 710 for receiving input from the user 714. The input device 710 may include, for example, a keyboard, pointing device, mouse, stylus, touch-sensitive control panel (e.g., touchpad or touchscreen), camera, gyroscope, accelerometer, position detector, and / or audio input device. A single component (such as a touchscreen) may serve as both an output device for the media output component 708 and an input device 710.

[0072] The client computing device 702 may also include a communication interface 712, which is communicatively coupled to a remote device such as server system 301 or a web server. The communication interface 712 may include, for example, a wired or wireless network adapter or a wireless data transceiver for use with mobile phone networks (e.g., Global System for Mobile Communications (GSM), 3G, 4G, 5G, or Bluetooth) or other mobile data networks (e.g., Global System for Microwave Access Interoperability (WIMAX)).

[0073] Storing in memory area 706 are computer-readable instructions, for example, for providing a user interface to user 714 via media output component 708 and optionally receiving and processing input from input device 710. The user interface may include web browsers and client applications, etc. A web browser enables user 714 to display and interact with media and other information typically embedded in web pages or websites from a web server. A client application allows user 714 to interact with a server application. The user interface, via one or both of a web browser and a client application, facilitates the display of information provided by computing device 502. The client application may be able to operate in both online mode (where the client application communicates with computing device 502) and offline mode (where the client application does not communicate with computing device 502).

[0074] Figure 8 This is an exemplary embodiment of a proposed novel coordination framework compound based on known coordination framework compounds. It is... Figure 2 This is an example of a method, illustrating how chemical insights can be incorporated into the method of this invention. Figure 8In this study, a known MOF, MOF 303; Al(OH)(1H-pyrazole-3,5-dicarboxylic acid ester), was considered. This known MOF has a rod-shaped SBU with alternating cis-trans angles sharing aluminum octahedrons. The SBU was set as a bent aluminum-SBU, and the topology was maintained. Changing the joints yielded new MOFs, which are variants of the known MOF-303 MOF. In this example, changing any one of the SBU, joints, or topology would produce a new MOF. This approach is more targeted and provides on-the-fly MOF recommendations, but requires chemical awareness and knowledge. However, in this method, data requirements can be reduced, where one variable is limited, such as by focusing on MOFs with Al-SBUs.

[0075] Figure 9 These are exemplary coordination framework compounds, specifically MOFs, obtained from ML models. ML models analyzed the structure and composition of MOFs and suggested the use of mixed short-to-long linkers. A chemical review of this suggestion determined that replacing the biphenyl-4,4'-dicarboxylate linker with a shorter linker would improve hydrophilicity, and that replacing the biphenyl-4,4'-dicarboxylate linker with a longer but more hydrophilic linker would increase pore volume and potentially improve the water isotherm.

[0076] Figure 10 This includes exemplary molecules used in the substitutional linkers of coordination framework compounds according to this disclosure. It can be seen that each of the fifteen substitutional base molecules has two carboxylic acid ester sites. Preliminary isotherms were obtained for five substitutional base molecules.

[0077] Other aspects of this disclosure are provided by the subject matter of the following provisions:

[0078] 1. A method for proposing a coordination framework compound, the method being performed using a coordination framework compound proposal (CFCP) computing device, the CFCP computing device including a processor coupled to a memory device, the method comprising:

[0079] The machine learning model of the CFCP computing device is used to generate an initial set of coordination framework compounds;

[0080] At least one coordination framework compound in the initial set of coordination framework compounds is subjected to examination for at least one chemical property.

[0081] Using the CFCP computing device, based on the initial set of coordination framework compounds and the review of at least one chemical property of at least one coordination framework compound in the initial set of coordination framework compounds, a preliminary set of coordination framework compounds is generated using the machine learning model; and

[0082] The coordination framework compound was proposed using the CFCP computing device.

[0083] 2. The method according to the foregoing clause, the method further comprising subjecting at least one coordination framework compound in the preliminary set of coordination framework compounds to examination for at least one chemical property.

[0084] 3. The method according to any of the foregoing clauses, the method further comprising verifying at least one coordination framework compound in the preliminary set of coordination framework compounds.

[0085] 4. The method according to any of the preceding clauses, further comprising at least one iteration of an execution sequence, the sequence comprising:

[0086] Based on the review of at least one chemical property of at least one coordination framework compound in at least one set of generated coordination framework compounds and at least one set of generated coordination framework compounds, a preliminary set of further coordination framework compounds is generated using the machine learning model.

[0087] Optionally, at least one coordination framework compound in the preliminary set of the further coordination framework compounds is subjected to examination for at least one chemical property; and

[0088] Optionally, at least one coordination framework compound in the preliminary set of the further coordination framework compounds may be verified.

[0089] 5. The method according to any of the preceding clauses, wherein the coordination framework compound is a coordination framework compound of an initial set of the coordination framework compounds, a coordination framework compound of a preliminary set of the coordination framework compounds, or a coordination framework compound of a preliminary set of the further coordination framework compounds.

[0090] 6. The method according to any of the preceding clauses, wherein the coordination framework compound comprises secondary building units (SBUs), connectors, and topologies.

[0091] 7. The method according to any of the preceding clauses, wherein the coordination framework compound is a metal-organic framework (MOF) compound or a covalent organic framework (COF) compound.

[0092] 8. The method according to any of the preceding clauses, wherein the machine learning model is trained using existing coordination framework compounds.

[0093] 9. The method according to any of the preceding clauses, wherein the machine learning model uses at least one technique selected from the group consisting of: latent space, inverse search, variational autoencoder (VAE), crystal diffusion variational autoencoder (CDVAE), inverse search of VAE latent space, graph neural network (GNN), neural network, optimization, and combinations thereof.

[0094] 10. The method according to any of the preceding clauses, wherein the machine learning model is trained to learn to be configured to reconstruct the crystal structure of the coordination framework compound and accurately predict the latent space of the associated target properties.

[0095] 11. The method according to any of the preceding clauses, wherein the at least one chemical property is selected from the group consisting of: adsorbate absorption capacity, adsorbate absorption kinetics, adsorbate gravimetric productivity, adsorbate volumetric productivity, adsorbate isotherms and isobars, pore size, pore volume, heat of adsorption, isothermal heat of adsorption, chemical stability, thermal stability, mechanical stability, synthetic feasibility, zeta potential, surface energy, hydrophobicity, hydrophilicity, chemical properties, chemical modifications for improvement, and combinations thereof.

[0096] 12. The method according to any of the preceding clauses, wherein the machine learning model generates the preliminary coordination framework compound based on at least one input selected from the group consisting of: the crystal structure of existing coordination framework compounds, target properties, target chemical properties, adsorption isotherms, water adsorption isotherms, pore volume, pore size, water stability, hydrophilicity, and combinations thereof.

[0097] 13. The method according to any of the preceding clauses, wherein the examination of at least one chemical property is performed by a machine, a person, or a combination thereof.

[0098] 14. A framework compound proposal (CFCP) computing device, the CFCP computing device comprising:

[0099] Memory; and

[0100] A processor communicatively coupled to the memory, the processor being programmed to:

[0101] Use machine learning models to generate an initial set of coordination framework compounds;

[0102] At least one coordination framework compound in the initial set of coordination framework compounds is subjected to examination for at least one chemical property.

[0103] Based on the initial set of coordination framework compounds and the review of at least one chemical property of at least one coordination framework compound in the initial set of coordination framework compounds, an initial set of coordination framework compounds is generated using the machine learning model; and

[0104] The coordination framework compound is proposed.

[0105] 15. The CFCP computing device according to the foregoing clause, wherein the processor is further programmed to subject at least one coordination framework compound in the preliminary set of coordination framework compounds to examination of at least one chemical property.

[0106] 16. The CFCP computing device according to any of the preceding clauses, wherein the processor is further programmed to verify at least one coordination framework compound in the preliminary set of coordination framework compounds.

[0107] 17. A CFCP computing device according to any of the preceding clauses, wherein the processor is further programmed to perform at least one iteration of a sequence comprising:

[0108] Based on the review of at least one chemical property of at least one coordination framework compound in at least one set of generated coordination framework compounds and at least one set of generated coordination framework compounds, a preliminary set of further coordination framework compounds is generated using the machine learning model.

[0109] Optionally, at least one coordination framework compound in the preliminary set of the further coordination framework compounds is subjected to examination for at least one chemical property; and

[0110] Optionally, at least one coordination framework compound in the preliminary set of the further coordination framework compounds may be verified.

[0111] 18. The CFCP computing device according to any of the preceding clauses, wherein the coordination framework compound is a coordination framework compound of an initial set of the coordination framework compounds, a coordination framework compound of a preliminary set of the coordination framework compounds, or a coordination framework compound of a preliminary set of the further coordination framework compounds.

[0112] 19. The CFCP computing device according to any of the preceding clauses, wherein the coordination framework compound includes secondary building units (SBUs), connectors, and topologies.

[0113] 20. The CFCP computing device according to any of the preceding clauses, wherein the coordination framework compound is a metal-organic framework (MOF) compound or a covalent organic framework (COF) compound.

[0114] 21. The CFCP computing device according to any of the preceding clauses, wherein the machine learning model is trained using existing coordination framework compounds.

[0115] 22. The CFCP computing device according to any of the preceding clauses, wherein the machine learning model uses at least one technique selected from the group consisting of: latent space, inverse search, variational autoencoder (VAE), crystal diffuse variational autoencoder (CDVAE), inverse search of VAE latent space, graph neural network (GNN), neural network, optimization, and combinations thereof.

[0116] 23. The CFCP computing device according to any of the preceding clauses, wherein the machine learning model is trained to learn the latent space configured to reconstruct the crystal structure of the coordination framework compound and accurately predict the associated target properties.

[0117] 24. The CFCP computing apparatus according to any of the preceding clauses, wherein the at least one chemical property is selected from the group consisting of: adsorbate absorption capacity, adsorbate absorption kinetics, adsorbate gravimetric productivity, adsorbate volumetric productivity, adsorbate isotherms and isobars, pore size, pore volume, heat of adsorption, isothermal heat of adsorption, chemical stability, thermal stability, mechanical stability, synthetic feasibility, zeta potential, surface energy, hydrophobicity, hydrophilicity, chemical properties, chemical modifications for improvement, and combinations thereof.

[0118] 25. The CFCP computing device according to any of the preceding clauses, wherein the machine learning model generates the preliminary coordination framework compound based on at least one input selected from the group consisting of: the crystal structure of existing coordination framework compounds, target properties, target chemical properties, adsorption isotherms, water adsorption isotherms, pore volume, pore size, water stability, hydrophilicity, and combinations thereof.

[0119] 26. The CFCP computing device according to any of the preceding clauses, wherein the review of at least one chemical property is performed by a machine, a person, or a combination thereof.

[0120] 27. A non-transitory computer-readable storage medium having computer-executable instructions embodied thereon, wherein when executed by a Coordination Framework Compound Proposal (CFCP) computing device including at least one processor communicating with memory, the computer-readable instructions cause the CFCP computing device to:

[0121] Use machine learning models to generate an initial set of coordination framework compounds;

[0122] At least one coordination framework compound in the initial set of coordination framework compounds is subjected to examination for at least one chemical property.

[0123] Based on the initial set of coordination framework compounds and the review of at least one chemical property of at least one coordination framework compound in the initial set of coordination framework compounds, an initial set of coordination framework compounds is generated using the machine learning model; and

[0124] The coordination framework compound is proposed.

[0125] 28. The non-transitory computer-readable storage medium according to the foregoing clause, wherein the computer-readable instructions further cause: the CFCP computing device to subject at least one coordination framework compound of the preliminary set of coordination framework compounds to examination of at least one chemical property.

[0126] 29. The non-transitory computer-readable storage medium according to the foregoing clause, wherein the computer-readable instructions further enable the CFCP computing device to verify at least one coordination framework compound in the preliminary set of coordination framework compounds.

[0127] 30. The non-transitory computer-readable storage medium according to the foregoing clause, wherein the computer-readable instructions further cause the CFCP computing device to perform at least one iteration of a sequence, the sequence comprising:

[0128] Based on the review of at least one chemical property of at least one coordination framework compound in at least one set of generated coordination framework compounds and at least one set of generated coordination framework compounds, a preliminary set of further coordination framework compounds is generated using the machine learning model.

[0129] Optionally, at least one coordination framework compound in the preliminary set of the further coordination framework compounds is subjected to examination for at least one chemical property; and

[0130] Optionally, at least one coordination framework compound in the preliminary set of the further coordination framework compounds may be verified.

[0131] 31. The non-transitory computer-readable storage medium according to any of the preceding clauses, wherein the coordination framework compound is a coordination framework compound of an initial set of the coordination framework compounds, a coordination framework compound of a preliminary set of the coordination framework compounds, or a coordination framework compound of a preliminary set of the further coordination framework compounds.

[0132] 32. The non-transitory computer-readable storage medium according to any of the preceding clauses, wherein the coordination framework compound includes secondary building units (SBUs), connectors, and topologies.

[0133] 33. The non-transitory computer-readable storage medium according to any of the preceding clauses, wherein the coordination framework compound is a metal-organic framework (MOF) compound or a covalent organic framework (COF) compound.

[0134] 34. The non-transitory computer-readable storage medium according to any of the preceding clauses, wherein the machine learning model is trained using existing coordination framework compounds.

[0135] 35. The non-transitory computer-readable storage medium according to any of the preceding clauses, wherein the machine learning model uses at least one technique selected from the group consisting of: latent space, inverse search, variational autoencoder (VAE), crystal diffuse variational autoencoder (CDVAE), inverse search of VAE latent space, graph neural network (GNN), neural network, optimization, and combinations thereof.

[0136] 36. The non-transitory computer-readable storage medium according to any of the preceding clauses, wherein the machine learning model is trained to learn a latent space configured to reconstruct the crystal structure of a coordination framework compound and accurately predict associated target properties.

[0137] 37. The non-transitory computer-readable storage medium according to any of the preceding clauses, wherein the at least one chemical property is selected from the group consisting of: adsorbate absorption capacity, adsorbate absorption kinetics, adsorbate gravimetric productivity, adsorbate volumetric productivity, adsorbate isotherms and isobars, pore size, pore volume, heat of adsorption, isothermal heat of adsorption, chemical stability, thermal stability, mechanical stability, synthetic feasibility, zeta potential, surface energy, hydrophobicity, hydrophilicity, chemical properties, chemical modifications for improvement, and combinations thereof.

[0138] 38. A non-transitory computer-readable storage medium according to any of the preceding clauses, wherein the machine learning model generates the preliminary coordination framework compound based on at least one input selected from the group consisting of: crystal structure, target properties, target chemical properties, adsorption isotherms, water adsorption isotherms, pore volume, pore size, water stability, hydrophilicity, and combinations thereof of existing coordination framework compounds.

[0139] 39. A non-transitory computer-readable storage medium according to any of the preceding clauses, wherein the examination of at least one chemical property is performed by a machine, a person, or a combination thereof.

[0140] The use of "some implementations" in the above description is not intended to be construed as excluding the existence of additional implementations that also include the described features.

[0141] Example

[0142] Without further detailed description, it is believed that those skilled in the art will be able to utilize the invention to its fullest extent using the foregoing description. Therefore, the following embodiments are to be interpreted as illustrative only and are not intended to limit the disclosure in any way. The starting materials used in the following embodiments need not be prepared by a specific preparation run, the procedure of which is described in other embodiments. It should also be understood that any numerical range described herein includes all values ​​from the lower limit to the upper limit. For example, if the range is stated as 10 to 50, values ​​such as 12 to 30, 20 to 40, or 30 to 50 are intended to be explicitly listed in this specification. These are merely embodiments of particular intent, and all possible combinations of values ​​between and including the listed minimum and maximum values ​​are considered to be explicitly stated in this application. The following embodiments may be performed using the CFCP computing device and / or non-transitory computer-readable storage medium described herein.

[0143] Example 1 .

[0144] As detailed in this article, follow the proposed workflow of hybrid ML-chemistry MOF.

[0145] Initially, the CDVAE model was provided with training data from the MOF database, including the MOF-303 variant. Starting with the MOF-303 variant, the MOF structure was optimized to maximize the pore volume (PV). Approximately 10,000 MOFs were generated. After algorithmic filtering, approximately 100 MOFs remained. Figure 9 An example MOF from an ML model is shown in the figure.

[0146] Next, chemical analysis was performed on the filtered MOF. CDVAE proposed using a mixture of long and short joints. This proposal was further explored by performing Perdew-Burke-Ernzerhof (PBE) density functional theory in the Vienna Ab Initio Simulation Package (VASP) program. A 520 eV energy cutoff was used for plane waves; soft PAW pseudopotentials were used for the atoms (H, C, N, O) available to them. The k-point grid included only the Γ point. The convergence threshold of the self-consistent field loop was 10. -6 eV; the force threshold for geometry optimization is Thirty-nine density functional theory (DFT) candidates were identified. Different SBUs and different linkers were examined. Chemical analysis showed that replacing the biphenyl linker with a shorter linker improved the hydrophilicity of the MOF.

[0147] Following this chemical analysis, fifteen different compounds, optimized using density functional theory, were proposed as MOF compounds. These fifteen different compounds include shorter linkers and were chemically evaluated and proposed as MOF compounds. The molecules used to replace the linkers in these fifteen compounds are depicted as follows: Figure 10 The corresponding pore volumes are shown in Table 1 below. As described below, MOF-303 corresponds to Al(OH)(1H-pyrazole-3,5-dicarboxylic acid ester). Finally, several hypothetical MOFs with larger pore volumes than MOF-303 were identified, and these compounds potentially exhibit improved water isotherms compared to MOF-303.

[0148] Table 1. Pore volumes of mixed linker MOFs .

[0149]

[0150] Unless otherwise indicated, approximate language used herein, such as “generally,” “substantially,” and “about,” indicates, as will be recognized by one of ordinary skill in the art, that such modified terms may apply only to approximations, not absolute or perfect degrees. Therefore, a value modified by one or more terms (such as “about,” “approximately,” and “substantially”) is not limited to the specified precise value. In at least some cases, approximate language may correspond to the precision of the instrument used to measure the value. Furthermore, unless otherwise indicated, the terms “first,” “second,” etc., are used herein merely as labels and are not intended to impose any order, position, or ranking requirements on the items referred to by these terms. Moreover, for example, a reference to a “second” item does not require or exclude the existence of an item such as a “first” or lower-numbered item, or a “third” or higher-numbered item.

[0151] While specific features of various embodiments of the invention may be shown in some figures and not in others, this is merely for convenience. Furthermore, the reference to "some embodiments" in the above description is not intended to exclude the existence of additional embodiments that also include the described features. Based on the principles of the invention, any feature of any other figure may be referenced and / or claimed in conjunction with any feature of any other figure.

[0152] This written description uses examples to disclose the invention, including the best mode, and also enables any person skilled in the art to practice the invention, including making and using any device or system and performing any combination of methods. The patentable scope of the invention is defined by the claims and may include other examples that would occur to a person skilled in the art. Such other examples are contemplated within the scope of the claims if they have structural elements that are not different from the literal language of the claims, or if they include equivalent structural elements that are not substantially different from the literal language of the claims.

Claims

1. A method for proposing a coordination framework compound, the method being performed using a coordination framework compound proposal (CFCP) computing device, the CFCP computing device including a processor coupled to a memory device, the method comprising: The machine learning model of the CFCP computing device is used to generate an initial set of coordination framework compounds; At least one coordination framework compound in the initial set of coordination framework compounds is subjected to examination for at least one chemical property. Using the CFCP computing device, based on the initial set of coordination framework compounds and the review of at least one chemical property of at least one coordination framework compound in the initial set of coordination framework compounds, a preliminary set of coordination framework compounds is generated using the machine learning model. as well as The coordination framework compound was proposed using the CFCP computing device.

2. The method of claim 1, further comprising subjecting at least one coordination framework compound in the preliminary set of coordination framework compounds to examination for at least one chemical property.

3. The method of claim 1, further comprising verifying at least one coordination framework compound in the preliminary set of coordination framework compounds.

4. The method of claim 1, further comprising performing at least one iteration of a sequence, the sequence comprising: Based on the review of at least one chemical property of at least one coordination framework compound in at least one set of generated coordination framework compounds and at least one set of generated coordination framework compounds, a preliminary set of further coordination framework compounds is generated using the machine learning model. Optionally, at least one of the coordination framework compounds in the preliminary set of the further coordination framework compounds is subjected to examination of at least one chemical property. and Optionally, at least one coordination framework compound in the preliminary set of the further coordination framework compounds may be verified.

5. The method of claim 1, wherein the coordination framework compound comprises secondary building units (SBUs), connectors, and topologies.

6. The method according to claim 1, wherein the coordination framework compound is a metal-organic framework (MOF) compound or a covalent organic framework (COF) compound.

7. The method of claim 1, wherein the machine learning model is trained using existing coordination framework compounds.

8. The method of claim 1, wherein the machine learning model uses at least one technique selected from the group consisting of: latent space, inverse search, variational autoencoder (VAE), crystal diffusion variational autoencoder (CDVAE), inverse search of VAE latent space, graph neural network (GNN), neural network, optimization, and combinations thereof.

9. The method of claim 1, wherein the machine learning model is trained to learn a latent space configured to reconstruct the crystal structure of the coordination framework compound and accurately predict the associated target properties.

10. The method according to claim 1, wherein the at least one chemical property is selected from the group consisting of: adsorbate absorption capacity, adsorbate absorption kinetics, adsorbate gravimetric productivity, adsorbate volumetric productivity, adsorbate isotherms and isobars, pore size, pore volume, heat of adsorption, isothermal heat of adsorption, chemical stability, thermal stability, mechanical stability, synthetic feasibility, zeta potential, surface energy, hydrophobicity, hydrophilicity, chemical properties, chemical modifications for improvement, and combinations thereof.

11. The method of claim 1, wherein the machine learning model generates an initial coordination framework compound based on at least one input selected from the group consisting of: the crystal structure of existing coordination framework compounds, target properties, target chemical properties, adsorption isotherms, water adsorption isotherms, pore volume, pore size, water stability, hydrophilicity, and combinations thereof.

12. The method of claim 1, wherein the examination of at least one chemical property is performed by a machine, a person, or a combination thereof.

13. A framework compound proposal (CFCP) computing device, the CFCP computing device comprising: Memory; as well as A processor communicatively coupled to the memory, the processor being programmed to: Use machine learning models to generate an initial set of coordination framework compounds; At least one coordination framework compound in the initial set of coordination framework compounds is subjected to examination for at least one chemical property. Based on the initial set of coordination framework compounds and the review of at least one chemical property of at least one coordination framework compound in the initial set of coordination framework compounds, an initial set of coordination framework compounds is generated using the machine learning model; and The coordination framework compound is proposed.

14. The CFCP computing device of claim 13, wherein the processor is further programmed to subject at least one coordination framework compound in the preliminary set of coordination framework compounds to examination of at least one chemical property.

15. The CFCP computing device of claim 13, wherein the processor is further programmed to verify at least one coordination framework compound in the preliminary set of coordination framework compounds.

16. The CFCP computing device of claim 13, wherein the processor is further programmed to execute at least one iteration of a sequence comprising: Based on the review of at least one chemical property of at least one coordination framework compound in at least one set of generated coordination framework compounds and at least one set of generated coordination framework compounds, a preliminary set of further coordination framework compounds is generated using the machine learning model. Optionally, at least one of the coordination framework compounds in the preliminary set of the further coordination framework compounds is subjected to examination of at least one chemical property. and Optionally, at least one coordination framework compound in the preliminary set of the further coordination framework compounds may be verified.

17. A non-transitory computer-readable storage medium having computer-executable instructions embodied thereon, wherein when executed by a Coordination Framework Compound Proposal (CFCP) computing device including at least one processor in communication with memory, the computer-readable instructions cause the CFCP computing device to: Use machine learning models to generate an initial set of coordination framework compounds; At least one coordination framework compound in the initial set of coordination framework compounds is subjected to examination for at least one chemical property. Based on the initial set of coordination framework compounds and the review of at least one chemical property of at least one coordination framework compound in the initial set of coordination framework compounds, an initial set of coordination framework compounds is generated using the machine learning model; and The coordination framework compound is proposed.

18. The non-transitory computer-readable storage medium of claim 17, wherein the computer-readable instructions further cause: the CFCP computing device to subject at least one coordination framework compound in the preliminary set of coordination framework compounds to examination of at least one chemical property.

19. The non-transitory computer-readable storage medium of claim 17, wherein the computer-readable instructions further enable the CFCP computing device to verify at least one coordination framework compound in the preliminary set of coordination framework compounds.

20. The non-transitory computer-readable storage medium of claim 17, wherein the computer-readable instructions further cause the CFCP computing device to perform at least one iteration of a sequence, the sequence comprising: Based on the review of at least one chemical property of at least one coordination framework compound in at least one set of generated coordination framework compounds and at least one set of generated coordination framework compounds, a preliminary set of further coordination framework compounds is generated using the machine learning model. Optionally, at least one of the coordination framework compounds in the preliminary set of the further coordination framework compounds is subjected to examination of at least one chemical property. and Optionally, at least one coordination framework compound in the preliminary set of the further coordination framework compounds may be verified.

Images