Synthesis-aware generative design in trillion-scale chemical spaces for automated drug discovery
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…
A groundbreaking generative AI framework named Hyper Screening X has emerged to tackle the issue of chemically unfeasible structures in autonomous drug discovery. The challenge lies in producing multi-trillion-scale make-on-demand libraries, which conventional virtual screening methods fail to efficiently manage or optimize due to their vast scale.
Hyper Screening X resolves this issue by integrating structure-based design with automated synthesis hardware. The framework encodes deterministic reaction logic within a generative flow network, enabling it to evaluate just 10 million candidates against an 11-trillion-compound space. This synthesis-aware strategy optimizes both physicochemical properties and compatibility with automated synthesis simultaneously, significantly reducing the number of compounds to be synthesized.
In a validation test, Hyper Screening X was applied to an SLC1A5 variant with a cryptic interface. The results were impressive, with a 96% synthesis success rate and a 50% functional hit rate. Crucially, this strategy led to the discovery of two first-in-class lead compounds, marking a significant advancement in the field of autonomous, closed-loop drug discovery.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
