Machine learning with persistent homology and chemical word embeddings improves prediction accuracy and interpretability in metal-organic frameworks

Abstract Machine learning has emerged as a powerful approach in materials discovery. Its major challenge is selecting features that create interpretable representations of materials, useful across multiple prediction tasks. We introduce an end-to-end machine learning model that automatically generat...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Aditi S. Krishnapriyan, Joseph Montoya, Maciej Haranczyk, Jens Hummelshøj, Dmitriy Morozov
Formato:	article
Lenguaje:	EN
Publicado:	Nature Portfolio 2021
Materias:	Medicine R Science Q
Acceso en línea:	https://doaj.org/article/00a270034e4b4f5cb9d506b1c33ddda6
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:00a270034e4b4f5cb9d506b1c33ddda6
record_format	dspace
spelling	oai:doaj.org-article:00a270034e4b4f5cb9d506b1c33ddda62021-12-02T17:15:16ZMachine learning with persistent homology and chemical word embeddings improves prediction accuracy and interpretability in metal-organic frameworks10.1038/s41598-021-88027-82045-2322https://doaj.org/article/00a270034e4b4f5cb9d506b1c33ddda62021-04-01T00:00:00Zhttps://doi.org/10.1038/s41598-021-88027-8https://doaj.org/toc/2045-2322Abstract Machine learning has emerged as a powerful approach in materials discovery. Its major challenge is selecting features that create interpretable representations of materials, useful across multiple prediction tasks. We introduce an end-to-end machine learning model that automatically generates descriptors that capture a complex representation of a material’s structure and chemistry. This approach builds on computational topology techniques (namely, persistent homology) and word embeddings from natural language processing. It automatically encapsulates geometric and chemical information directly from the material system. We demonstrate our approach on multiple nanoporous metal–organic framework datasets by predicting methane and carbon dioxide adsorption across different conditions. Our results show considerable improvement in both accuracy and transferability across targets compared to models constructed from the commonly-used, manually-curated features, consistently achieving an average 25–30% decrease in root-mean-squared-deviation and an average increase of 40–50% in R2 scores. A key advantage of our approach is interpretability: Our model identifies the pores that correlate best to adsorption at different pressures, which contributes to understanding atomic-level structure–property relationships for materials design.Aditi S. KrishnapriyanJoseph MontoyaMaciej HaranczykJens HummelshøjDmitriy MorozovNature PortfolioarticleMedicineRScienceQENScientific Reports, Vol 11, Iss 1, Pp 1-11 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	Medicine R Science Q
spellingShingle	Medicine R Science Q Aditi S. Krishnapriyan Joseph Montoya Maciej Haranczyk Jens Hummelshøj Dmitriy Morozov Machine learning with persistent homology and chemical word embeddings improves prediction accuracy and interpretability in metal-organic frameworks
description	Abstract Machine learning has emerged as a powerful approach in materials discovery. Its major challenge is selecting features that create interpretable representations of materials, useful across multiple prediction tasks. We introduce an end-to-end machine learning model that automatically generates descriptors that capture a complex representation of a material’s structure and chemistry. This approach builds on computational topology techniques (namely, persistent homology) and word embeddings from natural language processing. It automatically encapsulates geometric and chemical information directly from the material system. We demonstrate our approach on multiple nanoporous metal–organic framework datasets by predicting methane and carbon dioxide adsorption across different conditions. Our results show considerable improvement in both accuracy and transferability across targets compared to models constructed from the commonly-used, manually-curated features, consistently achieving an average 25–30% decrease in root-mean-squared-deviation and an average increase of 40–50% in R2 scores. A key advantage of our approach is interpretability: Our model identifies the pores that correlate best to adsorption at different pressures, which contributes to understanding atomic-level structure–property relationships for materials design.
format	article
author	Aditi S. Krishnapriyan Joseph Montoya Maciej Haranczyk Jens Hummelshøj Dmitriy Morozov
author_facet	Aditi S. Krishnapriyan Joseph Montoya Maciej Haranczyk Jens Hummelshøj Dmitriy Morozov
author_sort	Aditi S. Krishnapriyan
title	Machine learning with persistent homology and chemical word embeddings improves prediction accuracy and interpretability in metal-organic frameworks
title_short	Machine learning with persistent homology and chemical word embeddings improves prediction accuracy and interpretability in metal-organic frameworks
title_full	Machine learning with persistent homology and chemical word embeddings improves prediction accuracy and interpretability in metal-organic frameworks
title_fullStr	Machine learning with persistent homology and chemical word embeddings improves prediction accuracy and interpretability in metal-organic frameworks
title_full_unstemmed	Machine learning with persistent homology and chemical word embeddings improves prediction accuracy and interpretability in metal-organic frameworks
title_sort	machine learning with persistent homology and chemical word embeddings improves prediction accuracy and interpretability in metal-organic frameworks
publisher	Nature Portfolio
publishDate	2021
url	https://doaj.org/article/00a270034e4b4f5cb9d506b1c33ddda6
work_keys_str_mv	AT aditiskrishnapriyan machinelearningwithpersistenthomologyandchemicalwordembeddingsimprovespredictionaccuracyandinterpretabilityinmetalorganicframeworks AT josephmontoya machinelearningwithpersistenthomologyandchemicalwordembeddingsimprovespredictionaccuracyandinterpretabilityinmetalorganicframeworks AT maciejharanczyk machinelearningwithpersistenthomologyandchemicalwordembeddingsimprovespredictionaccuracyandinterpretabilityinmetalorganicframeworks AT jenshummelshøj machinelearningwithpersistenthomologyandchemicalwordembeddingsimprovespredictionaccuracyandinterpretabilityinmetalorganicframeworks AT dmitriymorozov machinelearningwithpersistenthomologyandchemicalwordembeddingsimprovespredictionaccuracyandinterpretabilityinmetalorganicframeworks
_version_	1718381240838520832

Machine learning with persistent homology and chemical word embeddings improves prediction accuracy and interpretability in metal-organic frameworks

Ejemplares similares