Flow-based fragment identification via binding site-specific latent representations

Journal article

Fragment-based drug design is a promising strategy leveraging the binding of small chemical moieties that can efficiently guide drug discovery. The initial step of fragment identification remains challenging, as fragments often bind weakly and non-specifically. We developed a protein-fragment encoder that relies on a contrastive learning approach to map both molecular fragments and protein surfaces in a shared latent space. The encoder captures interaction-relevant features and achieves strong discrimination between binding and non-binding regions, reaching ROC–area under the curve values of 0.92 on pocket surfaces and enrichment factors of 22.85 across full protein surfaces. Building on this representation, our generative method LatentFrag produces chemically realistic fragment identities and positions conditioned on the protein surface. LatentFrag improves fragment recovery over docking-based virtual screening, achieving a sampling hit rate more than four times higher at a fraction of its computational cost providing a valuable starting point for fragment hit discovery. We further show the practical utility of LatentFrag and extend the workflow to full ligand design tasks. Together, these approaches contribute to advancing fragment identification and provide valuable tools for fragment-based drug discovery.

Authors: Neeser, RM; Igashov, I; Schneuing, A; Bronstein, M; Schwaller, P

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: IOP Publishing
• Journal: Machine Learning: Science and Technology
• Volume: 7
• Issue: 3
• Pages: 035015
• Article number: 035015
• Publication date: 2026-05-14
• Acceptance date: 2026-04-09
• DOI: 10.1088/2632-2153/ae5d85
• EISSN: 2632-2153
• ISSN: 2632-2153
• Language: English
• Keywords: fragment identification; machine learning; drug design; fragment based drug design; representation learning; generative modelling