Core Insights - The article discusses the breakthrough of ODesign, a universal molecular design model that allows for precise and controllable design of various biological ligands, marking a significant advancement in AI-enabled drug development [3][4][12]. Group 1: ODesign Overview - ODesign was developed by a collaboration of institutions including Shanghai AI Laboratory and Harvard University, and it represents the first universal molecular design model [3]. - The model allows scientists to specify target sites on any type of target and achieve one-click design of proteins, peptides, nucleic acids, small molecules, and metal ions [11][12]. - ODesign significantly outperforms existing models like RFDiffusion and BindCraft in multiple industry-standard test sets, indicating a shift from "single-point breakthroughs" to "general intelligence" in generative AI drug development [4][12]. Group 2: Technological Advancements - ODesign achieves a nearly 50-fold increase in design efficiency compared to similar models, reducing the complete design cycle from days to hours [12]. - The model incorporates a new structural generation architecture with five core modules that enable multi-level representation of different molecular modalities and flexible control of conditions [16][20]. - It utilizes a cross-modal shared generative language to unify various molecular types into a common molecular generation space, allowing for collaborative construction based on atomic interactions [20]. Group 3: Performance Validation - ODesign has been tested across 11 molecular design tasks covering proteins, nucleic acids, and small molecules, demonstrating superior capabilities in protein design, including complex structures and functional optimization [23][26]. - In nucleic acid design, ODesign achieved approximately 60% and 20% RMSD success rates for 60nt RNA and DNA monomer design tasks, respectively [29]. - The model also excels in small molecule design, achieving about four times the success rate compared to mainstream models in targeting RNA [29][31]. Group 4: Practical Applications - The ODesign team has launched an online trial system for researchers and industry users, enabling rapid generation and visualization of high-quality molecular candidates [32][34]. - This platform aims to facilitate the transition from a research tool to a creative platform for AI-driven molecular creation, opening new avenues in drug development [32].