SynCat: molecule-level attention graph neural network for precise reaction classification
SynCat is a cross-attention GNN that embeds molecules, detects reagents, and estimates reaction-center embeddings via permutation-invariant pairwise aggregation. It delivers improved reaction-classification performance compared with RXNFP and DRFP.
Authors
- Peter F. Stadler (ORCID: https://orcid.org/0000-0002-5016-5191)
- Thomas Gärtner (ORCID: https://orcid.org/0000-0001-5985-9213)
- Nguyen Bao Vy Tran (ORCID: https://orcid.org/0000-0003-3768-2589)
- Tuyen Ngoc Truong (ORCID: https://orcid.org/0000-0002-0952-1633)
- Tieu-Long Phan (ORCID: https://orcid.org/0000-0002-3532-2064)
- Daniel Merkle (ORCID: https://orcid.org/0000-0001-7792-375X)
- Van-Thinh To (ORCID: https://orcid.org/0000-0002-7640-0807)
- Phuoc-Chung Van Nguyen (ORCID: https://orcid.org/0009-0000-1825-2781)
Institutions
- University of Applied Sciences Technikum Wien (AT)
- University of Vienna (AT)
- University of Southern Denmark (DK)
- Copenhagen Business School (DK)
- Bielefeld University (DE)
- University of Medicine and Pharmacy at Ho Chi Minh City (VN)
- Universidad Nacional de Colombia (CO)
- Telstra (Australia) (AU)
- Max Planck Institute for Mathematics in the Sciences (DE)
Publication Details
- Journal
- Digital Discovery
- Published
- 2025-11-06
- DOI
- https://doi.org/10.1039/d5dd00367a
- Primary Topic
- Machine Learning in Materials Science
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Korea International Cooperation Agency
- Novo Nordisk
- Novo Nordisk Fonden
- HORIZON EUROPE Framework Programme