A Unified 3D Generative Model for Synthesizable Structure-Based Drug Design
- 1Institute of Bioengineering, EPFL, Switzerland
Abstract
Traditional screening-based drug discovery is inherently limited by the astronomical scale of the chemical space. Generative modelling offers a compelling alternative to the classical search paradigm and enables rational, bottom-up design of novel and target-specific small molecules. However, its impact has been hampered by challenges in synthetic accessibility of the designed compounds and lack of large-scale experimental validation. Here, we introduce LDDM (Large Drug Discovery Model), a generative framework that supports a range of drug discovery tasks, including constrained and unconstrained docking, fragment linking and growing, and de novo design. We further introduce a programmable design algorithm that enables accurate design of synthetically accessible compounds satisfying various fine-grained objectives. We experimentally validated the designed or optimised ligands for five therapeutically relevant protein targets. In all cases, LDDM achieved high success rates, allowing us to identify molecules with confirmed binding affinity while synthesizing only a small number of generated compounds. The best designs were structurally characterised through NMR spectroscopy and X-ray crystallography, demonstrating high prediction accuracy. Overall, LDDM provides a scalable and flexible platform for the rapid and tailored design of small molecules and non-natural peptides for therapeutic applications.
Related
- scFLAME: a unified generative model for interpretable clustering, hierarchical structure discovery and marker-gene identification in single-cell RNA-seq data
- Synthesis-aware generative design in trillion-scale chemical spaces for automated drug discovery
- Structure-Based Drug Design of Novel Inhibitors Targeting the Thiamine Biosynthesis Pathway in Bacteria
- Scaffold-Constrained Generative Design Overcomes Failure Modes in Computational Protein Engineering
- Ourotide: decoding the hierarchical peptide recognition for generative design
BibTeX
@misc{igashov2026unified,
title = {{A Unified 3D Generative Model for Synthesizable Structure-Based Drug Design}},
author = {I. Igashov and A. Schneuing and A. W. Dobbelstein and I. Morozova and R. M. Neeser and K. Zielinski and L. A. Abriata and A. S. Petruzzella and D. R. Pavel Iosub and O. Gampp and A. Y. Lyubimov and E. Elizarova and I. Ferrara and P. M. F. Sousa and A. R. Lemos and F. Testori and P. A. Miranda Herrera and L. Kanis and J. Schmidt and M. K. E. Braza and R. E. Amaro and N. Thoma and D. M. Ferraris and R. Riek and J. S. Fraser and P. Schwaller and M. Bronstein and B. Correia},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.15.751537},
url = {https://www.biorxiv.org/content/10.64898/2026.09.15.751537v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - A Unified 3D Generative Model for Synthesizable Structure-Based Drug Design
AU - I. Igashov
AU - A. Schneuing
AU - A. W. Dobbelstein
AU - I. Morozova
AU - R. M. Neeser
AU - K. Zielinski
AU - L. A. Abriata
AU - A. S. Petruzzella
AU - D. R. Pavel Iosub
AU - O. Gampp
AU - A. Y. Lyubimov
AU - E. Elizarova
AU - I. Ferrara
AU - P. M. F. Sousa
AU - A. R. Lemos
AU - F. Testori
AU - P. A. Miranda Herrera
AU - L. Kanis
AU - J. Schmidt
AU - M. K. E. Braza
AU - R. E. Amaro
AU - N. Thoma
AU - D. M. Ferraris
AU - R. Riek
AU - J. S. Fraser
AU - P. Schwaller
AU - M. Bronstein
AU - B. Correia
PY - 2026
DA - 2026/09/15
PB - bioRxiv
DO - 10.64898/2026.09.15.751537
UR - https://www.biorxiv.org/content/10.64898/2026.09.15.751537v1
AB - Traditional screening-based drug discovery is inherently limited by the astronomical scale of the chemical space. Generative modelling offers a compelling alternative to the classical search paradigm and enables rational, bottom-up design of novel and target-specific small molecules. However, its impact has been hampered by challenges in synthetic accessibility of the designed compounds and lack of large-scale experimental validation. Here, we introduce LDDM (Large Drug Discovery Model), a generative framework that supports a range of drug discovery tasks, including constrained and unconstrained docking, fragment linking and growing, and de novo design. We further introduce a programmable design algorithm that enables accurate design of synthetically accessible compounds satisfying various fine-grained objectives. We experimentally validated the designed or optimised ligands for five therapeutically relevant protein targets. In all cases, LDDM achieved high success rates, allowing us to identify molecules with confirmed binding affinity while synthesizing only a small number of generated compounds. The best designs were structurally characterised through NMR spectroscopy and X-ray crystallography, demonstrating high prediction accuracy. Overall, LDDM provides a scalable and flexible platform for the rapid and tailored design of small molecules and non-natural peptides for therapeutic applications.
ER -