Controlling Biology - Fabian Theis, Helmholtz Munich

Virtual cell models are transitioning from static mapping tools to predictive world models capable of steering biological experimentation. By leveraging large-scale perturbation data, these models simulate cellular responses to environmental changes or genetic interventions, effectively acting as an input-output system for complex biological states. While early efforts relied on linear models, modern approaches increasingly utilize non-linear, data-driven architectures that scale with massive datasets, such as those found in single-cell atlases. Fabian Theis, a leader in computational health at Helmholtz Munich, emphasizes that the next frontier involves "lab-in-the-loop" systems where inverse generative models, like LabCompass, iteratively guide experimental design. Success in this field requires deep integration between computational methods and wet-lab workflows, supported by interdisciplinary training models that bridge the gap between data science and biological domain expertise to solve complex, real-world health challenges.