Synthesis-aware generative design in trillion-scale chemical spaces for automated drug discovery
- 1Department of Chemistry, KAIST, Daejeon, Republic of Korea; HITS Inc., Seoul, Republic of Korea
Abstract
Autonomous drug discovery via generative molecular design is critically bottlenecked by the production of chemically intractable structures. Multi-trillion-scale make-on-demand libraries guarantee synthetic feasibility, but conventional virtual screening cannot efficiently navigate these vast spaces or address multi-parameter optimization. We introduce Hyper Screening X, a generative AI framework integrating structure-based design with automated synthesis hardware. By encoding deterministic reaction logic in a generative flow network, the framework evaluates only 10 million candidates to navigate an 11-trillion-compound space, jointly optimizing physicochemical properties while ensuring compatibility with automated synthesis. Prospective validation against an SLC1A5 variant with a cryptic interface yielded a 96% synthesis success rate and a 50% functional hit rate, identifying two first-in-class lead compounds. This synthesis-aware strategy advances autonomous, closed-loop drug discovery.
Related
- Generative discovery of chemical reaction networks
- scFLAME: a unified generative model for interpretable clustering, hierarchical structure discovery and marker-gene identification in single-cell RNA-seq data
- Generative Access to Make-on-Demand Chemical Space Enables Ultra-Large Virtual Screening
- Automated High-Throughput Microfluidic Platform for Cell Pairing and Chemical Fusion
- Scaling genetic discovery through automated post-GWAS interpretation
BibTeX
@misc{seo2026synthesisaware,
title = {{Synthesis-aware generative design in trillion-scale chemical spaces for automated drug discovery}},
author = {S. Seo and Y. Sung and S.-Y. Hwang and M. Kang and J. Lee and S. Bordi and L. Raguz and J. Choi and D. Melnichenko and W. Namkung and S. Lee and J. Lim and B. M. Wanner and J. M. Han and W. Y. Kim},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.01.755933},
url = {https://www.biorxiv.org/content/10.64898/2026.10.01.755933v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Synthesis-aware generative design in trillion-scale chemical spaces for automated drug discovery
AU - S. Seo
AU - Y. Sung
AU - S.-Y. Hwang
AU - M. Kang
AU - J. Lee
AU - S. Bordi
AU - L. Raguz
AU - J. Choi
AU - D. Melnichenko
AU - W. Namkung
AU - S. Lee
AU - J. Lim
AU - B. M. Wanner
AU - J. M. Han
AU - W. Y. Kim
PY - 2026
DA - 2026/10/01
PB - bioRxiv
DO - 10.64898/2026.10.01.755933
UR - https://www.biorxiv.org/content/10.64898/2026.10.01.755933v1
AB - Autonomous drug discovery via generative molecular design is critically bottlenecked by the production of chemically intractable structures. Multi-trillion-scale make-on-demand libraries guarantee synthetic feasibility, but conventional virtual screening cannot efficiently navigate these vast spaces or address multi-parameter optimization. We introduce Hyper Screening X, a generative AI framework integrating structure-based design with automated synthesis hardware. By encoding deterministic reaction logic in a generative flow network, the framework evaluates only 10 million candidates to navigate an 11-trillion-compound space, jointly optimizing physicochemical properties while ensuring compatibility with automated synthesis. Prospective validation against an SLC1A5 variant with a cryptic interface yielded a 96% synthesis success rate and a 50% functional hit rate, identifying two first-in-class lead compounds. This synthesis-aware strategy advances autonomous, closed-loop drug discovery.
ER -