MIT AI-assisted antibiotic project yields nine candidates
MIT Technology Review China reports that MIT’s Antibiotics-AI Project has used AI to virtually screen more than 70 billion molecules, test more than 38,000 for antimicrobial activity and screen more than 100,000 in human-cell-related assays. According to the Collins Lab, the project has discovered or designed nine antibiotic candidates. The report notes that candidates must still undergo further development work, including toxicity studies, pharmacokinetic evaluation, animal studies and human clinical trials; it does not state that all nine candidates have completed or remain at the same stage of validation.
The project shows AI being used at scale to identify and design potential antibiotics, while the report emphasizes that laboratory activity does not by itself establish a usable medicine. Candidate synthesis, validation and clinical development still require substantial time and resources.
面对越来越难以治疗的耐药菌感染,科学家正在尝试一种新的抗生素研发方式:让 AI 从数百亿个分子中寻找潜在药物,甚至直接设计新的化学结构。
近日,麻省理工学院(MIT)宣布,该校教授 Regina Barzilay 和 James J. Collins 获得 2026 年普利兹克 AI 科学研究卓越奖。该奖项由芝加哥大学于 2024 年设立,旨在表彰利用人工智能推动自然科学和工程研究的杰出学者,此次也是首次授予合作研究团队。
Barzilay 是 MIT 工程学院 AI 与健康方向的杰出教授,长期从事机器学习及其医学应用研究;Collins 则是 MIT 生物工程系及医学工程与科学研究所教授,也是合成生物学领域的先驱之一。多年来,两人一直合作探索如何利用 AI 加速抗生素研发。