GraphXForm: graph transformer for computer-aided molecular design

2025/01/01 by Jonathan Pirnay, Jan G. Rittig, Alexander Wolf +4 · 1 voice · 1 citation
Computer Science · Decision Sciences · #Distributed and Parallel Computing Systems #Scientific Computing and Data Management

paper · pdf · doi:10.1039/d4dd00339j

openalex publication_date 2025/01/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/12

Abstract

GraphXForm is a graph transformer that iteratively builds molecules by adding atoms and bonds. A pretrained model is quickly fine-tuned for downstream tasks while respecting structural constraints, making it powerful for drug and solvent design.

Citations

• Los carnavales de La Bañeza: Aspectos antropológicos
• Constructions in combinatorics via neural networks

Cited by

• Molecular Machine Learning in Chemical Process Design

Discussions

• Pirnay, Rittig, Wolf et al. @dominikgrimm.bsky.social present GraphXForm, a graph-based molecule generator using transformers for constrained molecular design. It outperforms string-based methods in d [bsky, 2 points, 0 comments]

Related

• JTreeformer: Graph-Transformer via Latent-Diffusion Model for Molecular Generation
• Graph Data Modeling: Molecules, Proteins, & Chemical Processes
• A Graph VAE and Graph Transformer Approach to Generating Molecular Graphs
• Transformers are efficient hierarchical chemical graph learners
• The Graph Grammar Library - a generic framework for chemical graph rewrite systems