Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Prediction of water stability of metal–organic frameworks using machine learning

Abstract

Owing to their highly tunable structures, metal–organic frameworks (MOFs) are considered suitable candidates for a range of applications, including adsorption, separation, sensing and catalysis. However, MOFs must be stable in water vapour to be considered industrially viable. It is currently challenging to predict water stability in MOFs; experiments involve time-intensive MOF synthesis, while modelling techniques do not reliably capture the water stability behaviour. Here, we build a machine learning-based model to accurately and instantly classify MOFs as stable or unstable depending on the target application, or the amount of water exposed. The model is trained using an empirically measured dataset of water stabilities for over 200 MOFs, and uses a comprehensive set of chemical features capturing information about their constituent metal node, organic ligand and metal–ligand molar ratios. In addition to screening stable MOF candidates for future experiments, the trained models were used to extract a number of simple water stability trends in MOFs. This approach is general and can also be used to screen MOFs for other design criteria.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.

from$8.99

All prices are NET prices.

Fig. 1: Workflow adopted to build ML models of water stability in MOFs.
Fig. 2: MOF water stability training data.
Fig. 3: Performance of the classification models to predict water stability in MOFs.
Fig. 4: Mined chemical trends.

Data availability

The MOF water-stability data (illustrated in Fig. 2) used to train the models were obtained from ref. 13. The water-stability data used for validation (recent 10 MOFs) and screening (88 new MOFs) were obtained from the literature as cited in the Article. These datasets, including MOF features, are deposited at https://doi.org/10.5281/zenodo.4014333. Source data are provided with this paper.

Code availability

The machine learning training and prediction codes underlying this work are freely available for general use under GNU General Public Licence v3.0 and are deposited at https://doi.org/10.5281/zenodo.4014333.

References

  1. 1.

    Yoon, J. W. et al. Selective nitrogen capture by porous hybrid materials containing accessible transition metal ion sites. Nat. Mater. 16, 526–531 (2017).

    Google Scholar 

  2. 2.

    Adil, K. et al. Gas/vapour separation using ultra-microporous metal–organic frameworks: insights into the structure/separation relationship. Chem. Soc. Rev. 46, 3402–3430 (2017).

    Google Scholar 

  3. 3.

    Mason, J. A., Veenstra, M. & Long, J. R. Evaluating metal–organic frameworks for natural gas storage. Chem. Sci. 5, 32–51 (2014).

    Google Scholar 

  4. 4.

    Furukawa, H., Cordova, K. E., O’Keeffe, M. & Yaghi, O. M. The chemistry and applications of metal–organic frameworks. Science 341, 1230444 (2013).

    Google Scholar 

  5. 5.

    Dusselier, M. & Davis, M. E. Small-pore zeolites: synthesis and catalysis. Chem. Rev. 118, 5265–5329 (2018).

    Google Scholar 

  6. 6.

    Yang, D. & Gates, B. C. Catalysis by metal–organic frameworks: perspective and suggestions for future research. ACS Catal. 9, 1779–1798 (2019).

    Google Scholar 

  7. 7.

    Furukawa, H. et al. Ultrahigh porosity in metal–organic frameworks. Science 329, 424–428 (2010).

    Google Scholar 

  8. 8.

    Li, H., Eddaoudi, M., O’Keeffe, M. & Yaghi, O. M. Design and synthesis of an exceptionally stable and highly porous metal–organic framework. Nature 402, 276–279 (1999).

    Google Scholar 

  9. 9.

    Cohen, S. M. Postsynthetic methods for the functionalization of metal–organic frameworks. Chem. Rev. 112, 970–1000 (2011).

    Google Scholar 

  10. 10.

    Zhang, Y.-B. et al. Introduction of functionality, selection of topology and enhancement of gas adsorption in multivariate metal–organic framework-177. J. Am. Chem. Soc. 137, 2641–2650 (2015).

    Google Scholar 

  11. 11.

    Kaye, S. S., Dailly, A., Yaghi, O. M. & Long, J. R. Impact of preparation and handling on the hydrogen storage properties of Zn4O(1,4-benzenedicarboxylate)3 (MOF-5). J. Am. Chem. Soc. 129, 14176–14177 (2007).

    Google Scholar 

  12. 12.

    Ma, D., Li, Y. & Li, Z. Tuning the moisture stability of metal–organic frameworks by incorporating hydrophobic functional groups at different positions of ligands. Chem. Commun. 47, 7377–7379 (2011).

    Google Scholar 

  13. 13.

    Burtch, N. C., Jasuja, H. & Walton, K. S. Water stability and adsorption in metal–organic frameworks. Chem. Rev. 114, 10575–10612 (2014).

    Google Scholar 

  14. 14.

    Schoenecker, P. M., Carson, C. G., Jasuja, H., Flemming, C. J. & Walton, K. S. Effect of water adsorption on retention of structure and surface area of metal–organic frameworks. Ind. Eng. Chem. Res. 51, 6513–6519 (2012).

    Google Scholar 

  15. 15.

    Bosch, M., Zhang, M. & Zhou, H.-C. Increasing the stability of metal-organic frameworks. Adv. Chem. 2014, 182327 (2014).

    Google Scholar 

  16. 16.

    Rieth, A. J., Wright, A. M. & Dinca, M. Kinetic stability of metal–organic frameworks for corrosive and coordinating gas capture. Nat. Rev. Mater 4, 708–725 (2019).

    Google Scholar 

  17. 17.

    ul Qadir, N., Said, S. A. & Bahaidarah, H. M. Structural stability of metal–organic frameworks in aqueous media–controlling factors and methods to improve hydrostability and hydrothermal cyclic stability. Micropor. Mesopor. Mater. 201, 61–90 (2015).

    Google Scholar 

  18. 18.

    Plessius, R. et al. Highly selective water adsorption in a lanthanum metal–organic framework. Chem. Eur. J. 20, 7922–7925 (2014).

    Google Scholar 

  19. 19.

    Qin, L. et al. A water-stable metal–organic framework of a zwitterionic carboxylate with dysprosium: a sensing platform for Ebolavirus RNA sequences. Chem. Commun. 52, 132–135 (2016).

    Google Scholar 

  20. 20.

    Liu, T.-F. et al. Topology-guided design and syntheses of highly stable mesoporous porphyrinic zirconium metal–organic frameworks with high surface area. J. Am. Chem. Soc. 137, 413–419 (2014).

    Google Scholar 

  21. 21.

    Zhang, J.-P., Zhu, A.-X., Lin, R.-B., Qi, X.-L. & Chen, X.-M. Pore surface tailored SOD-type metal–organic zeolites. Adv. Mater. 23, 1268–1271 (2011).

    Google Scholar 

  22. 22.

    Nijem, N. et al. Water cluster confinement and methane adsorption in the hydrophobic cavities of a fluorinated metal–organic framework. J. Am. Chem. Soc. 135, 12615–12626 (2013).

    Google Scholar 

  23. 23.

    Yang, C. et al. Fluorous metal–organic frameworks with superior adsorption and hydrophobic properties toward oil spill cleanup and hydrocarbon storage. J. Am. Chem. Soc. 133, 18094–18097 (2011).

    Google Scholar 

  24. 24.

    Shih, Y.-H. et al. A simple approach to enhance the water stability of a metal–organic framework. Chem. Eur. J. 23, 42–46 (2017).

    Google Scholar 

  25. 25.

    Taylor, J. M., Vaidhyanathan, R., Iremonger, S. S. & Shimizu, G. K. Enhancing water stability of metal–organic frameworks via phosphonate monoester linkers. J. Am. Chem. Soc. 134, 14338–14340 (2012).

    Google Scholar 

  26. 26.

    Canivet, J., Fateeva, A., Guo, Y., Coasne, B. & Farrusseng, D. Water adsorption in MOFs: fundamentals and applications. Chem. Soc. Rev. 43, 5594–5617 (2014).

    Google Scholar 

  27. 27.

    OpenSMILES; http://opensmiles.org

  28. 28.

    Kim, C., Chandrasekaran, A., Huan, T. D., Das, D. & Ramprasad, R. Polymer genome: a data-powered polymer informatics platform for property predictions. J. Phys. Chem. C 122, 17575–17585 (2018).

    Google Scholar 

  29. 29.

    Mannodi-Kanakkithodi, A. et al. Scoping the polymer genome: a roadmap for rational polymer dielectrics design and beyond. Mater. Today 21, 785–796 (2018).

    Google Scholar 

  30. 30.

    Huan, T. D., Mannodi-Kanakkithodi, A. & Ramprasad, R. Accelerated materials property predictions and design using motif-based fingerprints. Phys. Rev. B 92, 014106 (2015).

    Google Scholar 

  31. 31.

    Nantasenamat, C., Isarankura-Na-Ayudhya, C. & Prachayasittikul, V. Advances in computational methods to predict the biological activity of compounds. Expert Opin. Drug Discov. 5, 633–654 (2010).

    Google Scholar 

  32. 32.

    RDKit Open Source Toolkit for Cheminformatics; http://www.rdkit.org/ (accessed 3 September 2019).

  33. 33.

    Jha, A., Chandrasekaran, A., Kim, C. & Ramprasad, R. Impact of dataset uncertainties on machine learning model predictions: the example of polymer glass transition temperatures. Model. Simul. Mater. Sci. Eng. (2018); https://doi.org/10.1088/1361-651X/aaf8ca

  34. 34.

    Shannon, R. D. Revised effective ionic radii and systematic studies of interatomic distances in halides and chalcogenides. Acta Crystallogr. A 32, 751–767 (1976).

    Google Scholar 

  35. 35.

    Haynes, W. M. CRC Handbook of Chemistry and Physics (CRC Press, 2014).

  36. 36.

    Pauling, L. The nature of the chemical bond. IV. The energy of single bonds and the relative electronegativity of atoms. J. Am. Chem. Soc. 54, 3570–3582 (1932).

    MATH  Google Scholar 

  37. 37.

    Guyon, I., Weston, J., Barnhill, S. & Vapnik, V. Gene selection for cancer classification using support vector machines. Mach. Learn. 46, 389–422 (2002).

    MATH  Google Scholar 

  38. 38.

    Pedregosa, F. et al. Scikit-learn: machine learning in Python. J. Mach. Learn. Res. 12, 2825–2830 (2011).

    MathSciNet  MATH  Google Scholar 

  39. 39.

    Xie, L., Liu, D., Huang, H., Yang, Q. & Zhong, C. Efficient capture of nitrobenzene from waste water using metal–organic frameworks. Chem. Eng. J. 246, 142–149 (2014).

    Google Scholar 

  40. 40.

    Wang, D., Zhang, L., Li, G., Huo, Q. & Liu, Y. Luminescent MOF material based on cadmium(ii) and mixed ligands: application for sensing volatile organic solvent molecules. RSC Adv. 5, 18087–18091 (2015).

    Google Scholar 

  41. 41.

    Liao, P.-Q. et al. Drastic enhancement of catalytic activity via post-oxidation of a porous Mnii triazolate framework. Chem. Eur. J. 20, 11303–11307 (2014).

    Google Scholar 

  42. 42.

    Jing, F. et al. Mil-68(Fe) as an efficient visible-light-driven photocatalyst for the treatment of a simulated waste-water contain Cr(vi) and malachite green. Appl. Catal. B Environ. 206, 9–15 (2017).

    Google Scholar 

  43. 43.

    Cadiau, A. et al. Design of hydrophilic metal organic framework water adsorbents for heat reallocation. Adv. Mater. 27, 4775–4780 (2015).

    Google Scholar 

  44. 44.

    Bazaga-Garcia, M. et al. Tuning proton conductivity in alkali metal phosphonocarboxylates by cation size-induced and water-facilitated proton transfer pathways. Chem. Mater. 27, 424–435 (2015).

    Google Scholar 

  45. 45.

    Gutov, O. V. et al. Water-stable zirconium-based metal–organic framework material with high-surface area and gas-storage capacities. Chem. Eur. J. 20, 12389–12393 (2014).

    Google Scholar 

  46. 46.

    Duan, J., Jin, W. & Krishna, R. Natural gas purification using a porous coordination polymer with water and chemical stability. Inorg. Chem. 54, 4279–4284 (2015).

    Google Scholar 

  47. 47.

    Nguyen, K. T., Blum, L. C., Van Deursen, R. & Reymond, J.-L. Classification of organic molecules by molecular quantum numbers. ChemMedChem 4, 1803–1805 (2009).

    Google Scholar 

  48. 48.

    Lin, R.-B. et al. Molecular sieving of ethylene from ethane using a rigid metal–organic framework. Nat. Mater. 17, 1128–1133 (2018).

    Google Scholar 

  49. 49.

    Sun, Y. & Han, H. A novel 3D Agi cationic metal–organic framework based on 1,2,4,5-tetra(4-pyridyl) benzene with selective adsorption of CO2 over CH4, H2O over C2H5OH, and trapping Cr2O72−. J. Mol. Struct. 1194, 73–77 (2019).

    Google Scholar 

Download references

Acknowledgements

This work was supported as part of the Center for Understanding and Control of Acid Gas-Induced Evolution of Materials for Energy (UNCAGE-ME), an Energy Frontier Research Center funded by the US Department of Energy, Office of Science, Basic Energy Sciences under award no. DE-SC0012577. C.C. gratefully acknowledges a fellowship from the Achievement Rewards for College Scientists (ARCS) Foundation. R.B. acknowledges insightful discussions with D.S. Sholl.

Author information

Affiliations

Authors

Contributions

R.B. and R.R. initiated this research project. R.B. developed and analysed the ML models. C.C. and T.G.E. contributed to data collection. All co-authors contributed to the model analysis, discussions and writing of the manuscript.

Corresponding author

Correspondence to Rampi Ramprasad.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Statistics on water stability in MOFs.

Distribution of MOFs into 4 categories of water stability based on the constituting metal node.

Source data

Extended Data Fig. 2 Performance comparison of ML algorithms for 2-class model.

Performance comparison of SVM, RF and GB methods for the 2-class model (’S’, stable and ’U’, unstable MOFs) using the RFE based reduced feature set. Left panel shows the overall class-weighted accuracies, while the right two panels show the per-class test scores, that is F1, area under the ROC curve (AUC), precision (P) and recall (R), for the RF and SVM models. The RF model can be seen to outperform in all accounts and was selected as the 2-class model in this work.

Source data

Extended Data Fig. 3 Performance comparison of ML algorithms for 3-class model.

Performance comparison of SVM, RF and GB methods for the 3-class model (’S’, stable, ’HK’, high kinetic stable, and ’U’, unstable MOFs) using the RFE based reduced feature set. Left panel shows the overall class-weighted accuracies, while the right two panels respectively show the per-class F1 and recall scores, for the RF and SVM models. The RF model can be seen to have poor performance for the underrepresented stable (S) class, although it was trained to maximize the class-weighted accuracy. Similar results were found for GB algorithm as well. Thus, SVM with best performance for all classes was selected as the 3-class model in this work.

Source data

Extended Data Fig. 4 Important MOF water stability descriptors.

Relative feature importance as extracted from the random forest (RF) 2-class model. The feature importance in case of RF is based on the concept of mean decrease in impurity (MDI), as explained here (G. Louppe, Understanding Random Forests: From Theory to Practice, PhD Thesis, U. of Liege, 2014). The features with relatively high importance were selected to mine important chemical trends of water stability in MOFs. The first letter of the descriptor, that is, M or L, denotes the metal or the ligand associated features, respectively (see main article for details). Features with high importance were used to derive important stability trends as discussed in the main article.

Source data

Extended Data Fig. 5 Correlation between MOF water stability and its descriptors.

A subset of post-RFE features were analyzed to see if linear correlations between MOF water stability for the case with two classes (S+HK and U+LK) and the features values could be used to derive some chemical trends. This figure suggests that the presence of certain chemical motifs, especially those containing N or ketone groups, and 5-member rings, tend to enhance the water stability in MOFs. Each marker in the figure represents a MOF from the Burtch data set. See Supplementary Information for details on the different descriptors.

Source data

Supplementary information

Supplementary Information

Supplementary Tables 1 and 2 discussing the reduced feature set and model predictions on 88 new MOFS, respectively.

Reporting Summary

Source data

Source Data Fig. 3

Statistical source data.

Source Data Fig. 5

Statistical source data.

Source Data Extended Data Fig. 1

Statistical source data.

Source Data Extended Data Fig. 2

Statistical source data.

Source Data Extended Data Fig. 3

Statistical source data.

Source Data Extended Data Fig. 4

Statistical source data.

Source Data Extended Data Fig. 5

Statistical source data.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Batra, R., Chen, C., Evans, T.G. et al. Prediction of water stability of metal–organic frameworks using machine learning. Nat Mach Intell 2, 704–710 (2020). https://doi.org/10.1038/s42256-020-00249-z

Download citation

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing