Predicting polymerization reactions via transfer learning using chemical language models

• DOI: 10.1038/s41524-024-01304-8
• Publication Date: 2024-06-04T15:28:02Z
• Authors: Brenda S. Ferrari; Matteo Manica; Ronaldo Giro; Teodoro Laino; Mathias B. Steiner
• Keywords: Transfer of learning

Abstract Polymers are candidate materials for a wide range of sustainability applications such as carbon capture and energy storage. However, computational polymer discovery lacks automated analysis reaction pathways stability assessment through retro-synthesis. Here, we report an extension transformer-based language models to polymerization both retrosynthesis tasks. To that end, have curated dataset vinyl polymers covering reactions representative homo-polymers co-polymers. Overall, obtain forward model Top-4 accuracy 80% backward 60%. We further analyze the performance with examples evaluate its prediction quality from science perspective. enable validation reuse, made our data available in public repositories.