Machine learning-based discovery of molecular descriptors that control polymer gas permeation

Shastry, Tejus; Basdogan, Yasemin; Wang, Zhen-Gang; Kumar, Sanat K.; Carbone, Matthew R.

doi:10.1016/j.memsci.2024.122563

Machine learning-based discovery of molecular descriptors that control polymer gas permeation

Journal Article · Mon Feb 19 00:00:00 EST 2024 · Journal of Membrane Science

DOI:https://doi.org/10.1016/j.memsci.2024.122563· OSTI ID:2338112

Shastry, Tejus ^[1]; Basdogan, Yasemin ^[2]; Wang, Zhen-Gang ^[2]; Kumar, Sanat K. ^[1]; ^[3]

Columbia University, New York, NY (United States)
California Institute of Technology (CalTech), Pasadena, CA (United States)
Brookhaven National Laboratory (BNL), Upton, NY (United States)

While machine learning has found increasing use in predicting the properties of polymeric materials with only a knowledge of chain architecture, determining the molecular factors underpinning properties (“interpretable AI”) has remained less well explored. We show that encoding chain chemistry in commonly employed formats, e.g., binary-valued fingerprints, leads to uniqueness issues during the hashing process to save storage space. This is because the hashing algorithm can map several chemical moieties into the same bit. These issues carry over into the ML algorithms, especially for “inverse” design and interpretable AI, and cannot be avoided by changing the length of the fingerprint. Using MACCS key featurizations of monomer repeats resolves some of these issues, and we show that a few substructures consistently appear in top features for maximizing permeability across several gases and ML models. These are carbon–carbon double bonds (as in polyacetylenes) especially when they are associated with methyl groups (found in branching architectures). Here these results, derived from the limited data set of ~ 500 polymers with experimental gas permeation data, are in agreement with physical insight and thus provide a robust foundation which could further enable study of these material classes through detailed experiments and simulations.

View Accepted Manuscript (DOE)

Research Organization:: Brookhaven National Laboratory (BNL), Upton, NY (United States)

Sponsoring Organization:: King Abdullah University of Science and Technology; USDOE Laboratory Directed Research and Development (LDRD) Program; USDOE Office of Science (SC), Basic Energy Sciences (BES)

Grant/Contract Number:: SC0008772; SC0012704

OSTI ID:: 2338112

Alternate ID(s):: OSTI ID: 2369909

Report Number(s):: BNL--225505-2024-JAAM

Journal Information:: Journal of Membrane Science, Journal Name: Journal of Membrane Science Vol. 697; ISSN 0376-7388

Publisher:: ElsevierCopyright Statement

Country of Publication:: United States

Language:: English

References (23)

Data from: Machine learning-based discovery of molecular descriptors that control polymer gas permeation Kumar, Sanat; Shastry, Tejus; Carbone, Matthew Dryad https://doi.org/10.5061/dryad.5x69p8dbm	dataset	January 2024
Molecularly engineering polymeric membranes for H ₂ / CO ₂ separation at 100–300 °C Hu, Leiqing; Pal, Sankhajit; Nguyen, Hien Journal of Polymer Science, Vol. 58, Issue 18 https://doi.org/10.1002/pol.20200220	journal	July 2020
A group contribution approach to predict permeability and permselectivity of aromatic polymers Robeson, L. Journal of Membrane Science, Vol. 132, Issue 1 https://doi.org/10.1016/S0376-7388(97)00031-8	journal	August 1997
Machine learning strategies for the structure-property relationship of copolymers Tao, Lei; Byrnes, John; Varshney, Vikas iScience, Vol. 25, Issue 7 https://doi.org/10.1016/j.isci.2022.104585	journal	July 2022
Highly Polar but Amorphous Polymers with Robust Membrane CO2/N2 Separation Performance Liu, Junyi; Zhang, Shaoze; Jiang, De-en Joule, Vol. 3, Issue 8 https://doi.org/10.1016/j.joule.2019.07.003	journal	August 2019
The upper bound revisited Robeson, Lloyd M. Journal of Membrane Science, Vol. 320, Issue 1-2, p. 390-400 https://doi.org/10.1016/j.memsci.2008.04.030	journal	July 2008
An empirical correlation of gas permeability and permselectivity in polymers and its theoretical basis Robeson, L. M.; Freeman, B. D.; Paul, D. R. Journal of Membrane Science, Vol. 341, Issue 1-2 https://doi.org/10.1016/j.memsci.2009.06.005	journal	September 2009
Molecular fingerprint similarity search in virtual screening Cereto-Massagué, Adrià; Ojeda, María José; Valls, Cristina Methods, Vol. 71 https://doi.org/10.1016/j.ymeth.2014.08.005	journal	January 2015
Representing Polymers as Periodic Graphs with Learned Descriptors for Accurate Polymer Property Predictions Antoniuk, Evan R.; Li, Peggy; Kailkhura, Bhavya Journal of Chemical Information and Modeling, Vol. 62, Issue 22 https://doi.org/10.1021/acs.jcim.2c00875	journal	October 2022
Effect of Branch Length on the Structural and Separation Properties of Hyperbranched Poly(1,3-dioxolane) Huang, Liang; Guo, Wenji; Mondal, Himangshu Macromolecules, Vol. 55, Issue 1 https://doi.org/10.1021/acs.macromol.1c02045	journal	December 2021
Maximizing Ether Oxygen Content in Polymers for Membrane CO ₂ Removal from Natural Gas Liu, Junyi; Zhang, Gengyi; Clark, Krysta ACS Applied Materials & Interfaces, Vol. 11, Issue 11 https://doi.org/10.1021/acsami.9b01079	journal	February 2019
Chemistry-Informed Machine Learning for Polymer Electrolyte Discovery Bradford, Gabriel; Lopez, Jeffrey; Ruza, Jurgis ACS Central Science, Vol. 9, Issue 2 https://doi.org/10.1021/acscentsci.2c01123	journal	January 2023
BigSMILES: A Structurally-Based Line Notation for Describing Macromolecules Lin, Tzyy-Shyang; Coley, Connor W.; Mochigase, Hidenobu ACS Central Science, Vol. 5, Issue 9 https://doi.org/10.1021/acscentsci.9b00476	journal	September 2019
Random Forest Predictor for Diblock Copolymer Phase Behavior Arora, Akash; Lin, Tzyy-Shyang; Rebello, Nathan J. ACS Macro Letters, Vol. 10, Issue 11 https://doi.org/10.1021/acsmacrolett.1c00521	journal	October 2021
Extended-Connectivity Fingerprints Rogers, David; Hahn, Mathew Journal of Chemical Information and Modeling, Vol. 50, Issue 5 https://doi.org/10.1021/ci100050t	journal	April 2010
Basis of Permeability/Selectivity Tradeoff Relations in Polymeric Gas Separation Membranes Freeman, Benny D. Macromolecules, Vol. 32, Issue 2 https://doi.org/10.1021/ma9814548	journal	January 1999
Random Forests Breiman, Leo Machine Learning, Vol. 45, Issue 1, p. 5-32 https://doi.org/10.1023/A:1010933404324	journal	January 2001
Seven chemical separations to change the world Sholl, David S.; Lively, Ryan P. Nature, Vol. 532, Issue 7600 https://doi.org/10.1038/532435a	journal	April 2016
From local explanations to global understanding with explainable AI for trees Lundberg, Scott M.; Erion, Gabriel; Chen, Hugh Nature Machine Intelligence, Vol. 2, Issue 1 https://doi.org/10.1038/s42256-019-0138-9	journal	January 2020
Model agnostic generation of counterfactual explanations for molecules Wellawatte, Geemi P.; Seshadri, Aditi; White, Andrew D. Chemical Science, Vol. 13, Issue 13 https://doi.org/10.1039/D1SC05259D	journal	January 2022
Designing exceptional gas-separation polymer membranes using machine learning Barnett, J. Wesley; Bilchak, Connor R.; Wang, Yiwen Science Advances, Vol. 6, Issue 20 https://doi.org/10.1126/sciadv.aaz4301	journal	May 2020
Machine learning enables interpretable discovery of innovative polymers for gas separation membranes Yang, Jason; Tao, Lei; He, Jinlong Science Advances, Vol. 8, Issue 29 https://doi.org/10.1126/sciadv.abn9545	journal	July 2022
Maximizing the right stuff: The trade-off between membrane permeability and selectivity Park, Ho Bum; Kamcev, Jovan; Robeson, Lloyd M. Science, Vol. 356, Issue 6343 https://doi.org/10.1126/science.aab0530	journal	June 2017

Similar Records

Fingerprinting Interactions between Proteins and Ligands for Facilitating Machine Learning in Drug Discovery

Journal Article · Thu Jan 04 19:00:00 EST 2024 · Biomolecules · OSTI ID:2471231

BeyondFingerprinting: AI-guided discovery of robust materials & processes

Technical Report · Sun Sep 01 00:00:00 EDT 2024 · OSTI ID:2462990

Related Subjects

36 MATERIALS SCIENCE
explainable AI
gas transport
machine learning
membranes
polymers

Machine learning-based discovery of molecular descriptors that control polymer gas permeation

Citation Formats

References (23)

Similar Records

Related Subjects