MOFSimplify, machine learning models with extracted stability data of three thousand metal-organic frameworks.

Area of Science:

• Materials Science
• Computational Chemistry
• Data Science

Background:

• Metal-organic frameworks (MOFs) are advanced porous materials with diverse applications.
• Characterizing MOF stability (solvent removal, thermal decomposition) is crucial but labor-intensive.
• Existing literature contains vast, unstructured data on MOF properties.

Purpose of the Study:

• To create an automated workflow for extracting MOF stability data from scientific literature.
• To build machine learning models for predicting MOF stability.
• To develop a user-friendly platform for accessing data and predictions.

Main Methods:

• Natural Language Processing (NLP) for text mining of MOF literature.
• Extraction of solvent removal stability and thermal decomposition temperatures.
• Development of artificial neural network (ANN) models using graph and pore-geometry features.
• Validation against a hand-labeled dataset.

Main Results:

• Successfully extracted over 2,000 solvent removal stability measures and 3,000 thermal decomposition temperatures.
• Validated NLP methods and data extraction accuracy.
• Developed ML models capable of predicting MOF stability with quantified uncertainty.

Conclusions:

• Automated data extraction via NLP significantly accelerates the acquisition of MOF stability data.
• Machine learning models can reliably predict MOF stability, aiding materials design.
• The MOFSimplify platform democratizes access to MOF data and predictive tools, fostering community-driven research.