斯坦福大学研究人员利用基因组语言模型Evo 2，从零开始设计了16种功能性噬菌体，这标志着人工智能首次生成了完整、可行的病毒基因组[1]。这项研究成果于2024年8月6日发表在《科学》杂志上，基于超过200万个噬菌体基因组的数据集进行训练[1]。
设计的病毒成功感染了已对天然噬菌体产生耐药性的大肠杆菌，展示了AI病毒设计在噬菌体疗法中的潜在应用价值[1]。从302个候选基因组中，研究团队合成了285个，最终仅有16个转化为功能性噬菌体[1]。这些AI设计的病毒与原型ΦX174的序列同源性平均达到97%，但在蛋白质三维结构、生长动力学和感染动力学方面存在差异[1]。
然而，这一突破也引起了广泛的生物安全担忧。斯坦福计算生物学家Brian Hie表示，制造危险人类病原体"仍需大量工作，肯定需要一个非常有才华的跨学科团队"[1]。约翰·霍普金斯健康安全中心的生物安全专家指出，"问题不再是生成式病毒基因组设计是否会存在，而是社会能否建立监管以允许其益处展开同时防止其造成严重伤害"[1]。据估计，DNA合成成本在10至20万美元之间[1]。
Stanford University researchers have used a genomic language model called Evo 2 to design 16 functional bacteriophages from scratch, marking the first time artificial intelligence has generated complete, viable viral genomes.[1] The AI-designed viruses were capable of infecting strains of E. coli that had developed resistance to naturally occurring bacteriophages, demonstrating the potential of AI-driven viral design for phage therapy applications.[1] The findings were published in Science on August 6, 2024.[1]
The research involved training the AI model on a dataset of more than 2 million bacteriophage genomes.[1] Of 302 candidate genomes generated by the model, 285 could be synthesized, yet only 16 ultimately proved functional as viable viruses.[1] The designed viruses shared an average sequence similarity of 97 percent with the prototype ΦX174, despite exhibiting differences in protein three-dimensional structure, growth kinetics, and infection dynamics.[1]
The achievement has raised biosafety concerns even as it opens new therapeutic possibilities. Brian Hie, a computational biologist at Stanford, noted that creating dangerous human pathogens "would still require a lot of work, and definitely a very talented interdisciplinary team."[1] The estimated cost for DNA synthesis stands between $100,000 and $200,000.[1] According to biosafety experts at the Johns Hopkins Center for Health Security, "the question is no longer whether generative viral genome design will exist, but whether society can establish governance to allow its benefits to unfold while preventing serious harm."[1]