Machine Learning in Polymer Research.

Area of Science:

• Polymer Chemistry
• Materials Science
• Computational Chemistry

Background:

• Machine learning (ML) is increasingly used in polymer chemistry to correlate chemical structures with macroscopic properties.
• Challenges include limited/poorly curated datasets, broad molecular weight distributions, and irregular polymer configurations.
• Off-the-shelf ML models often require refinement for specific polymer applications.

Purpose of the Study:

• To address dataset curation hurdles in polymer chemistry for ML applications.
• To highlight advances in polymer synthesis and modeling that improve data availability.
• To survey ML approaches for predicting polymer properties and explore emerging applications.

Main Methods:

• Review of current literature on ML in polymer chemistry.
• Discussion of dataset curation strategies and advances in polymer synthesis.
• Survey of ML models applied to predict solid-state properties, solution behavior, and composite performance.
• Exploration of ML in emerging areas like drug delivery and the polymer-biology interface.

Main Results:

• Identified key challenges in applying ML to polymer chemistry data.
• Highlighted advances in polymer synthesis and data management enhancing ML applicability.
• Surveyed diverse ML applications, from property prediction to drug delivery.
• Emphasized the need for collaboration between chemists and mathematicians.

Conclusions:

• Successful ML application in polymer chemistry requires addressing data limitations and refining models.
• FAIR data principles and integration of polymer theory with data are crucial.
• The machine-human interface is essential for effective collaboration and innovation.