FAIR-MOFs: Structure-centred synthesis inference from three-dimensional structures of metal-organic frameworks
Thomas Heine, Dinga A.D. Wonanke, Antonio Longa, Asha Pankajakshan, Ogar Joseph, Lauri Himanen, Alvin Ladine, José P. Márquez et al.
Research Square·2026
Metal-organic frameworks (MOFs) hold great promise for applications ranging from CO2 capture to atmospheric water harvesting, yet their discovery is hindered by limited access to reproducible and scalable synthesis. Here, we introduce FAIR-MOFs, a data-driven framework that directly links 3D crystal structures to experimentally validated synthetic routes. FAIR-MOFs combines a graph neural network that predicts essential synthesis precursors from structure alone with a retrosynthetic recommender that learns reagent co-occurrence patterns from literature. The framework is built on over 47,000 curated experimentally synthesised structures integrated through automated structural curation and building-unit recognition to maintain data integrity and FAIR principles. We demonstrate its predictive power by experimentally synthesising three hypothetical MOFs using conditions proposed by the model. Overall, FAIR-MOFs introduces a structure-centred approach to synthetic planning by demonstrating that synthetic precursors can be inferred directly from three-dimensional crystal structures of metalorganic frameworks.
Sahina Khatun, A. D.Dinga Wonanke, Matthew A. Addicoat, Sanhita Maity, Anirban Pradhan
Journal of Materials Chemistry A·2026
First chrysene and pyrene based porous graphene electrocatalyst, CY-Pyre-PG, for hydrogen evolution from a conjugated microporous polymer, CY-Pyre-CMP is used as a metal-free electrocatalyst for hydrogen evolution reaction through water splitting.
A. D. Dinga Wonanke, Antonio Longa, Asha Pankajakshan, Lauri Himanen, Alvin Noe Ladines, J.A. Marquez, Matthew A. Addicoat, Deborah L. Crittenden et al.
ChemRxiv·2025
Metal-organic frameworks (MOFs) are a versatile class of materials with applications in gas storage, separations, and catalysis. Despite extensive research, one of the key factors hindering their broader deployment is the absence of reproducible and scalable synthetic protocols. To address this, we introduce FAIR-MOFs, a database designed to be Findable, Accessible, Interoperable, and Reusable (FAIR), comprising 45{,}700 curated experimental structures, 33{,}361 geometry-optimised structures, and 4{,}161 entries linked to reported synthesis conditions. Analysis of the dataset showed that the propensity of open-metal sites in MOFs is statistically associated with reaction temperature, metal salts, ligands, topology, and metal secondary building unit. Furthermore, we developed a retrosynthetic recommender that captures literature co-usage patterns among solvents, metal salts, and ligands. For any given component, the system suggests compatible reagents and retrieves MOFs prepared under similar conditions. Finally, we trained a graph-based neural network integrated with our MOF deconstruction module to predict most probable metal salts, ligands and solvents directly from 3D structures of experimental or hypothetical MOFs. Using this model, we successfully synthesised MOFs randomly selected from hypothetical MOF databases illustrating the potential of FAIR-MOFs to accelerate the discovery and enable data-driven synthesis of MOFs.
Mehrdad Jalali, A. D. Dinga Wonanke, Pascal Friederich, Christof Wöll
Journal of Chemical Information and Modeling·2025
The expansion of large-scale materials databases has facilitated the development of graph-based representations, encoding structural and functional similarities as edges in data-driven networks. These enable machine learning models to leverage both local features and global relationships. However, densely connected datasets often introduce redundancy and noise, escalating computational complexity without improving performance. Here, we introduce the Black Hole Strategy, a gravity-based representative sampling method that constructs compact, informative subsets from large materials datasets while preserving essential structural and property diversity. Using metal-organic frameworks (MOFs) as a case study, we demonstrate that graph neural networks (GraphSAGE, GCN, and GAT) trained on Black Hole-sparsified datasets achieve comparable or superior classification and regression performance compared to full-dataset models, despite utilizing significantly fewer data points and reduced memory and training time requirements. Analysis of class-level confusion matrices confirms that critical structure-property relationships─such as pore-limiting diameter (PLD)─persist under substantial sparsification. An ablation study on gravity score weights validates the balanced formulation and robustness of the approach. Topological and efficiency benchmarks further demonstrate that the method preserves modularity, diversity, and connectivity across sparsification levels. These findings establish the Black Hole Strategy as a principled and frugal approach for machine learning in materials science, enabling efficient, interpretable, and scalable discovery workflows. Importantly, this work contributes to the objectives of the FAIRmat consortium, which aims to develop a FAIR data infrastructure for condensed matter physics and materials science. Our approach advances FAIR (Findable, Accessible, Interoperable, Reusable) data practices through optimized sampling techniques that enhance data management, reusability, and interoperability in materials informatics.
A.H. Mirza, Nawaf Alampara, Sreekanth Kunchapu, Martiño Ríos-García, Benedict Emoekabu, Aswanth Krishnan, Tanya Gupta, Mara Schilling-Wilhelmi et al.
Nature Chemistry·2025
Large language models (LLMs) have gained widespread interest owing to their ability to process human language and perform tasks on which they have not been explicitly trained. However, we possess only a limited systematic understanding of the chemical capabilities of LLMs, which would be required to improve models and mitigate potential harm. Here we introduce ChemBench, an automated framework for evaluating the chemical knowledge and reasoning abilities of state-of-the-art LLMs against the expertise of chemists. We curated more than 2,700 question-answer pairs, evaluated leading open- and closed-source LLMs and found that the best models, on average, outperformed the best human chemists in our study. However, the models struggle with some basic tasks and provide overconfident predictions. These findings reveal LLMs' impressive chemical capabilities while emphasizing the need for further research to improve their safety and usefulness. They also suggest adapting chemistry education and show the value of benchmarking frameworks for evaluating LLMs in specific domains.
Eunji Jin, Volodymyr Bon, Shubhajit Das, A. D. Dinga Wonanke, Martin Etter, Martin A. Karlsen, Ankita De, Nadine Bönisch et al.
Journal of the American Chemical Society·2025
Metal-organic polyhedra (MOPs) are versatile supramolecular building blocks for the design of highly porous frameworks by reticular assembly because of their diverse geometries, multiple degrees of freedom regarding functionalization, and accessible metal sites. Lipophilic functionalization is demonstrated to enable the rational assembly and crystallization with photoactive N-donor ligands in an aliphatic solvent to achieve multiaxially aligned photoresponsive diarylethene (DTE) moieties in 3D frameworks (DUT-210(M), M = Cu and Rh) featuring cooperative switchability. Combined experimental and theoretical investigations based on in situ PXRD, UV-vis spectroscopy, and density functional theory calculations demonstrate deliberate kinetic engineering of photoswitchability based on variations in metal-ligand bond strengths. The novel porous frameworks are an important step toward the knowledge-based development of photon-driven motors, actuators, and release systems.
Mehrdad Jalali, A. D. Dinga Wonanke, Christof Wöll
Journal of Cheminformatics·2023
Metal-organic frameworks (MOFs), are porous crystalline structures comprising of metal ions or clusters intricately linked with organic entities, displaying topological diversity and effortless chemical flexibility. These characteristics render them apt for multifarious applications such as adsorption, separation, sensing, and catalysis. Predominantly, the distinctive properties and prospective utility of MOFs are discerned post-manufacture or extrapolation from theoretically conceived models. For empirical researchers unfamiliar with hypothetical structure development, the meticulous crystal engineering of a high-performance MOF for a targeted application via a bottom-up approach resembles a gamble. For example, the precise pore limiting diameter (PLD), which determines the guest accessibility of any MOF cannot be easily inferred with mere knowledge of the metal ion and organic ligand. This limitation in bottom-up conceptual understanding of specific properties of the resultant MOF may contribute to the cautious industrial-scale adoption of MOFs.Consequently, in this study, we take a step towards circumventing this limitation by designing a new tool that predicts the guest accessibility-a MOF key performance indicator-of any given MOF from information on only the organic linkers and the metal ions. This new tool relies on clustering different MOFs in a galaxy-like social network, MOFGalaxyNet, combined with a Graphical Convolutional Network (GCN) to predict the guest accessibility of any new entry in the social network. The proposed network and GCN results provide a robust approach for screening MOFs for various host-guest interaction studies.
Santu Ruidas, Avik Chowdhury, Anirban Ghosh, Avik Ghosh, Sujan Mondal, A. D. Dinga Wonanke, Matthew A. Addicoat, Abhijit K. Das et al.
Langmuir·2023
Exploring a covalent organic framework (COF) material as an efficient metal-free photocatalyst and as an adsorbent for the removal of pollutants from contaminated water is very challenging in the context of sustainable chemistry. Herein, we report a new porous crystalline COF, C 6 -TRZ-TPA COF, via segregation of donor–acceptor moieties through the extended Schiff base condensation between tris(4-formylphenyl)amine and 4,4′,4″-(1,3,5-triazine-2,4,6-triyl)trianiline. This COF displayed a Brunauer–Emmett–Teller (BET) surface area of 1058 m 2 g –1 with a pore volume of 0.73 cc g –1 . Again, extended π-conjugation, the presence of heteroatoms throughout the framework, and a narrow band gap of 2.2 eV, all these features collectively work for the environmental remediation in two different perspectives: it could harness solar energy for environmental clean-up, where the COF has been explored as a robust metal-free photocatalyst for wastewater treatment and as an adsorbent for iodine capture. In our endeavor of wastewater treatment, we have conducted the photodegradation of rose bengal (RB) and methylene blue (MB) as model pollutants since these are extremely toxic, are health hazard, and bioaccumulative in nature. The catalyst C 6 -TRZ-TPA COF showed a very high catalytic efficiency of 99% towards the degradation of 250 parts per million (ppm) of RB solution in 80 min under visible light irradiation with the rate constant of 0.05 min –1 . Further, C 6 -TRZ-TPA COF is found to be an excellent adsorbent as it efficiently adsorbed radioactive iodine from its solution as well as from the vapor phase. The material exhibits a very rapid iodine capturing tendency with an outstanding iodine vapor uptake capacity of 4832 mg g –1 .
A. D. Dinga Wonanke, Deborah L. Crittenden
Journal of Molecular Spectroscopy·2022
Composite approaches in which clustering and encapsulation effects are modelled at different levels of theory are capable of reproducing experimental infrared (IR) spectroscopic band centres to within 23 cm−1 (mean absolute deviation), with maximum absolute errors less than 65 cm−1. Anharmonic fundamentals for water stretching modes within isolated clusters may be computed by applying complexation shifts to experimental gas phase water fundamentals, or by empirically scaling harmonic fundamentals computed using “medium accuracy” quantum chemical methods such as dispersion-corrected generalised gradient approximation density functional theories (DFT) or second-order Møller–Plesset perturbation theory (MP2). Environmental effects are modelled using density functional tight binding (DFTB) theories as the difference between harmonic fundamentals in the crystalline environment and in the gas phase. This approach affords a significant improvement in accuracy over conventional approaches in which IR band centres are modelled at a single level of theory. DFT or MP2 calculations on isolated chloride hydrate clusters yield mean and maximum absolute errors of 36 cm−1 and 144 cm−1, respectively. Directly predicting vibrational frequencies in the condensed phase using DFTB models is less accurate again, incurring mean and maximum absolute errors of 93 cm−1 and 204 cm−1, respectively.
Robert McNair, Sushil Kumar, A. D. Dinga Wonanke, Matthew A. Addicoat, Robert A. W. Dryfe, György Székely
Desalination·2022
Membrane capacitive deionization (MCDI) is a promising technique to achieve desalination of low-salinity water resources. The primary requirements for developing and designing materials for MCDI applications are large surface area, high wettability to water, high conductivity, and efficient ion-transport pathways. Herein, we synthesized ionic covalent organic nanosheets (iCONs) containing guanidinium units that carry a positive charge. A series of quaternized polybenzimidazole (QPBI)/iCON (iCON@QPBI) nanocomposite membranes was fabricated using solution casting. The surface, thermal, wettability, and electrochemical properties of the iCON@QPBI nanocomposite membranes were evaluated. The iCON@QPBI anion-exchange membranes achieved a salt adsorption capacity as high as 15.6 mg g−1 and charge efficiency of up to 90%, which are 50% and 20% higher than those of the pristine QPBI membrane, respectively. The performance improvement was attributed to the increased ion-exchange capacity (2.4 mmol g−1), reduced area resistance (5.4 Ω cm2), and enhanced hydrophilicity (water uptake = 32%) of the iCON@QPBI nanocomposite membranes. This was due to the additional quaternary ammonium groups and conductive ion transport networks donated by the iCON materials. The excellent desalination performance of the iCON@polymer nanocomposite membranes demonstrated their potential for use in MCDI applications and alternative electromembrane processes
Digambar Shinde, Li Cao, Sushil Kumar, Zongyao Zhou, IChun Chen, Dinga Wonanke, Matthew Addicoat, Zhiping Lai
Journal of Membrane Science and Research·2022
Two-dimensional covalent organic framework (COFs) membranes have shown promise for organic solvent nanofiltration applications. However, the ability to modulate the chemical properties of the membranes and their effects on the molecular transport process has not yet been explored. Here, we demonstrate the synthesis of two COF membranes (TFP-MPOHF and TFP-MPF) with the same scaffold structures but different internal chemical properties. The presence of hydroxyl groups in the TFP-MPOHF membranes resulted in a significant improvement in polar solvent permeability. In contrast, the hydrophobic TFP-MPF membranes offered excellent permeability to nonpolar solvents, which was 130 – 235% higher than the TFP-MPOHF and commercial polymeric membranes. In addition, both COF membranes exhibited precise molecular sieving capacity with an apparent molecular weight cut-off (MWCO) of 800 g mol-1 and excellent stability. A deviation from the pore-flow model was observed for the TFP-MPOHF membranes, which was due to the specific interactions between solvent molecules and polar channel walls.
Sushil Kumar, Mahmoud A. Abdulhamid, A. D. Dinga Wonanke, Matthew A. Addicoat, György Székely
Nanoscale·2022
separation efficiency was investigated, and the results revealed that ND-COF-1 is more selective than ND-COF-2, which could be attributed to the less hindered pathway offered to methane gas molecules by the framework pore.
A. D. Dinga Wonanke, Matthew A. Addicoat
Physical Chemistry Chemical Physics·2022
Elucidating the precise stacking configuration of a covalent organic framework, COF, is critical to fully understand their various applications. Unfortunately, most COFs form powder crystals whose atomic characterisations are possible only through powder X-ray diffraction (PXRD) analysis. However, this analysis has to be coupled with computational simulations, wherein computed PXRD patterns for different stacking configurations are compared with experimental patterns to predict the precise stacking configuration. This task is often computationally challenging firstly because, computation of these systems mostly rely on the use of semi-empirical methods that need to be adequately parametrised for the system being studied and secondly because some of these compounds possess guest molecules, which are not often taken into account during computation. COF-1 is an extreme case in which the presence of the guest molecule plays a critical role in predicting the precise stacking configuration. Using this as a case study, we mapped out a full PES for the stacking configuration in the guest free and guest containing system using the GFN-xTB semi-empirical method followed by a periodic energy decomposition analysis using first-principles Density Functional Theory (DFT). Our results showed that the presence of the guest molecule leads to multiple low energy stacking configurations with significantly different lateral offsets. Also, the semi-empirical method does not precisely predict DFT low energy configurations, however, it accurately accounts for dispersion. Finally, our quantum-mechanical analysis demonstrates that electrostatic-dispersion model suggested Hunter and Sanders accurately describes the stacking in 2D COFs as opposed to the newly suggested Pauli-dispersion model.
Digambar Balaji Shinde, Li Cao, Xiaowei Liu, Dinga A.D. Wonanke, Zongyao Zhou, Mohamed Nejib Hedhili, Matthew A. Addicoat, Kuo‐Wei Huang et al.
Journal of Membrane Science Letters·2021
Three crystalline truxene-based β-ketoenamine COF membranes (TFP-HETTA, TFP-HBTTA and TFP-HHTTA) are fabricated via a de novo monomer design approach to understand the fundamental correlations between pore structure and molecular separation performance. By introducing bulky alkyl groups into the truxene framework, the pore size of TFP-HETTA, TFP-HBTTA, and TFP-HHTTA are systematically tuned from 1.08 to 0.72 nm. Accordingly, the TFP-HETTA showed good water permeance of 47 L m−2 h−1 bar−1 along with a prominent rejection rate of Reactive Blue (RB, 800 Da) but less than 10% rejection rate of inorganic salts. In contrast, the TFP-HHTTA membrane with pore size of 0.72 nm can reject small dye molecules such as Safranin O (SO, 350 Da) and trivalent salts but with a moderate water permeance of 19 L m−2 h−1 bar−1. The pore-flow model rooted from the viscous flow could well fit the observed organic solvent nanofiltration results of all three COF membranes.
Yogendra Nailwal, A. D. Dinga Wonanke, Matthew A. Addicoat, Santanu Kumar Pal
Macromolecules·2021
Covalent organic frameworks (COFs) offer great potential for various advanced applications such as photocatalysis, sensing, and so on because of their fully conjugated, porous, and chemically stable unique structural architecture. In this work, we have designed and developed a truxene-based ultrastable COF ( Tx-COF-2 ) by Schiff-base condensation between 1,3,5-tris(4-aminophenyl)benzene (TAPB) and 5,5,10,10,15,15-hexamethyl-10,15-dihydro-5 H -diindeno(1,2- a:1′,2′- c )fluorene-2,7,12-tricarbaldehyde (Tx-CHO) for the first time. The resulting COF possesses excellent crystallinity, permanent porosity, and high Brunauer–Emmett–Teller (BET) surface areas (up to 1137 m 2 g –1 ). The COF was found to be a heterogeneous, recyclable photocatalyst for efficient conversion of arylboronic acids to phenols under visible-light irradiation, an environmentally friendly alternative approach to conventional metal-based photocatalysis. Besides, Tx-COF-2 provides an immediate naked-eye color change (<1 s) and fluorescence “turn-on” phenomena upon exposure to HCl. The response is highly sensitive, with an ultralow detection limit of up to 4.5 nmol L –1 .
Subbareddy Mekapothula, A. D. Dinga Wonanke, Matthew A. Addicoat, David J. Boocock, John D. Wallis, Gareth W. V. Cave
International Journal of Molecular Sciences·2021
A silica-bound C-butylpyrogallol[4]arene chromatographic stationary phase was prepared and characterised by thermogravimetric analysis, scanning electron microscopy, NMR and mass spectrometry. The chromatographic performance was investigated by using C60 and C70 fullerenes in reverse phase mode via flash column and high-pressure liquid chromatography (HPLC). The resulting new stationary phase was observed to demonstrate size-selective molecular recognition as postulated from our in-silico studies. The silica-bound C-butylpyrogallol[4]arene flash and HPLC stationary phases were able to separate a C60- and C70-fullerene mixture more effectively than an RP-C18 stationary phase. The presence of toluene in the mobile phase plays a significant role in achieving symmetrical peaks in flash column chromatography.
A. D. Dinga Wonanke, Poppy Bennett, Lewis Caldwell, Mathew Addicoat
ChemRxiv·2021
Metal-organic frameworks, MOFs, offer an effective templet for polymerisation of polymers with precisely controlled structures within the sub-nanometre scales. However, synthetic difficulties such as monomer infiltration, detailed understanding of polymerisation mechanisms within the MOF nano-channels and the mechanism for removing the MOF template post polymerisation have prevented wide scale implementation of polymerisation in MOFs. This is partly due to the significant lack in understanding of the energetic and atomic-scale intermolecular interactions between the monomers and the MOFs. Consequently in this study, we explore the interaction of varied concentration of styrene, and EDOT, at the surface and in the nano-channel of Zn2(1,4- ndc)2(dabco), where 1,4-ndc = 1,4-naphthalenedicarboxylate and dabco = 1,4-diazabicyclo[2.2.2]octane. Our results showed that the interactions between monomers are stronger in the nano-channels than at the surfaces of the MOF. Moreover, the MOF-monomer interactions are strongest in the nano-channels and increases with increase in the number of monomers. However, as the number of monomer increases, the monomers turn to bind more strongly at the surface leading to a potential agglomeration of the monomers at the surface.
Subbareddy Mekapothula, A. D. Dinga Wonanke, Matthew A. Addicoat, John D. Wallis, David J. Boocock, Gareth W. V. Cave
New Journal of Chemistry·2020
The chromatographic separation of proteomic standards via a silica immobilized pillararene cavitand has been designed in silico using host–guest binding energy studies and realized experimentally to selectively interact with peptides.
Shanshan Tao, Lipeng Zhai, A. D. Dinga Wonanke, Matthew A. Addicoat, Qiuhong Jiang, Donglin Jiang
Nature Communications·2020
Development of porous materials combining stability and high performance has remained a challenge. This is particularly true for proton-transporting materials essential for applications in sensing, catalysis and energy conversion and storage. Here we report the topology guided synthesis of an imine-bonded (C=N) dually stable covalent organic framework to construct dense yet aligned one-dimensional nanochannels, in which the linkers induce hyperconjugation and inductive effects to stabilize the pore structure and the nitrogen sites on pore walls confine and stabilize the H 3 PO 4 network in the channels via hydrogen-bonding interactions. The resulting materials enable proton super flow to enhance rates by 2–8 orders of magnitude compared to other analogues. Temperature profile and molecular dynamics reveal proton hopping at low activation and reorganization energies with greatly enhanced mobility.
Digambar Balaji Shinde, Li Cao, A. D. Dinga Wonanke, Xiang Li, Sushil Kumar, Xiaowei Liu, Mohamed Nejib Hedhili, Abdul‐Hamid Emwas et al.
Chemical Science·2020
) but different lengths of carbon chains aiming to rationally control the pore size. The membrane permeation results in the applications of organic solvent nanofiltration and molecular sieving of organic dyes showed a systematic shift of the membrane flux and molecular weight cut-off correlated to the pore size change. These results enhanced our fundamental understanding of transport through uniform channels at nanometer scales. Pore engineering of the covalent organic framework membranes was demonstrated for the first time.
A. D. Dinga Wonanke, Jayne L. Ferguson, Christopher M. Fitchett, Deborah L. Crittenden
Chemistry - An Asian Journal·2019
Photochemical oxidative cyclodehydrogenation reactions are a versatile class of aromatic ring-forming reactions. They are tolerant to functional group substitution and heteroatom inclusion, so can be used to form a diverse range of extended polyaromatic systems by fusing existing ring substituents. However, despite their undoubted synthetic utility, there are no existing models-computational or heuristic-that predict the outcome of photocyclisation reactions across all possible classes of reactants. This can be traced back to the fact that "negative" results are rarely published in the synthetic literature and the lack of a general conceptual framework for understanding how photoexcitation affects reactivity. In this work, we address both of these issues. We present experimental data for a series of aromatically substituted pyrroles and indoles, and show that quantifying induced atomic forces upon photoexcitation provides a powerful predictive model for determining whether a given reactant will photoplanarise and hence proceed to photocyclised product under appropriate reaction conditions. The propensity of a molecule to photoplanarise is related to localised changes in charge distribution around the putative forming ring upon photoexcitation. This is promoted by asymmetry in molecular structures and/or charge distributions, inclusion of heteroatoms and ethylene bridging and well-separated or isolated photocyclisation sites.
A. D. Dinga Wonanke, Deborah L. Crittenden
Australian Journal of Chemistry·2018
The Mallory (photocyclization) and Scholl (thermal cyclohydrogenation) reactions are widely used in the synthesis of extended conjugated p systems of high scientific interest and technological importance, including molecular wires, semiconducting polymers, and nanographenes. While simple electrocyclization reactions obey the Woodward-Hoffman rules, no such simple, general, and powerful model is available for eliminative cyclization reactions due to their increased mechanistic complexity. In this work, detailed mechanistic investigations of prototypical reactions reveal that there is no single rate-determining step for thermal oxidative dehydrogenation reactions, but they are very sensitive to the presence and distribution of heteroatoms around the photocyclizing ring system. Key aspects of reactivity are correlated to the constituent ring oxidation potentials. For photocyclization reactions, planarization occurs readily and/or spontaneously following photo-excitation, and is promoted by heteroatoms within 5-membered ring adjacent to the photocyclizing site. Oxidative photocyclization requires intersystem crossing to proceed to products, while reactants configured to undergo purely eliminative photocyclization could proceed to products entirely in the excited state. Overall, oxidative photocyclization seems to strike the optimal balance between synthetic convenience (ease of preparation of reactants, mild conditions, tolerant to chemical diversity in reactants) and favourable kinetic and thermodynamic properties.
