AI“零样本”发现新抗体，人工智能驱动的药物正加速走向临床

Group 1 - The core viewpoint is that drug molecular design is transitioning from "probabilistic collisions" to "atomic-level precision engineering," ushering in a faster, more accurate, and smarter era of molecular design [1][2] - AI-driven drug discovery has made significant advancements, with Chai Discovery's AI model Chai-2 achieving a historic success rate of 16% in "de novo" antibody design, which is over 100 times more effective than previous methods [1][2] - The success of Chai-2 is attributed to its multimodal generative architecture, which integrates full-atom structure prediction and generative models, addressing the low success rates of traditional methods that often rely on extensive experimental screening [1] Group 2 - Despite the advancements, AI-discovered drugs entering late-stage clinical trials remain limited, primarily due to the early development stage of AI in drug discovery and the presence of misleading claims about AI capabilities [2] - A recent study published in Nature Medicine highlights a significant milestone, showing that an AI-discovered drug for idiopathic pulmonary fibrosis demonstrated safety and efficacy in a randomized phase II clinical trial [2][3] - The AI-powered drug Rentosertib, developed by Insilico Medicine, showed good safety and tolerability, with dose-dependent efficacy observed in patients, marking a successful demonstration of the company's generative AI platform [3]