Open-Source Engine Accelerates AI Drug Discovery

Accurately modeling biomolecular interactions is a critical bottleneck in both fundamental biology and therapeutic discovery. This research introduces the Open Drug Discovery Engine (OpenDDE), an innovative open-source, all-atom biomolecular foundation model designed to address this challenge. OpenDDE leverages co-folding as its primary entry point, establishing a shared structural reasoning layer for understanding sequence-structure-function relationships across complex biomolecular systems. Unlike approaches that treat structure prediction as an isolated goal, OpenDDE is built as a foundational system capable of supporting a wide array of drug discovery tasks, including de novo design, affinity estimation, and structure-conditioned optimization. The engine integrates state-of-the-art advancements in all-atom architecture, atomic latent reasoning, and inference optimization, coupled with large-scale data processing, to achieve IsoDDE-level co-folding accuracy within a fully reproducible and openly accessible framework. The researchers also identified two key scaling-law directions for co-folding models, outlining practical pathways for continuous improvement through scaling data, model size, inference capabilities, and training processes. By releasing its training code, inference pipelines, checkpoints, and benchmarks, OpenDDE aims to democratize access to frontier biomolecular intelligence, foster global collaboration, and establish an open foundation for the next generation of AI-driven drug discovery systems.