控制生物学：Fabian Theis 谈虚拟细胞与多尺度生物数据
原标题：Controlling Biology - Fabian Theis, Helmholtz Munich
AI 摘要
Latent Space 播客采访了 Helmholtz Munich 的 Fabian Theis，讨论虚拟细胞（Virtual Cell）研究现状。文章指出，2025 年 Arc Virtual Cell Challenge 显示，在扰动预测任务中，简单基线常优于复杂模型，部分原因是评估指标校准不佳；Theis 认为现有数据集上的扰动预测问题不够难，应转向利用 Biohub 等机构投入 18 亿美元收集的大规模、高精度、多尺度生物数据，预测分子到细胞、组织、器官的演化。Helmholtz 的计算健康中心与计算生物学研究所正用 AI 结合分子、影像、基因组和临床数据开发精准医学模型。
正文节选
Fabian Theis has been one of the key players in the various iterations of the Virtual Cell concept and initiatives. It is not hyperbole at all to say that his lab has been pushing the frontier of algorithms to understand multi-scale, multi-modal (”multiomics”), time-varying biological data. Indeed, he has shaped the large scale vision and data collection through his leadership at Helmholtz, various startups and probably at Biohub itself. Much of the recent attention in AI x Bio has been on prote