Despite advances in computational and experimental biology, integrating these approaches to predict and validate molecular regulators of cell differentiation remains challenging. I propose combining multiplexed in vivo perturbation, untargeted molecular profiling, and probabilistic modeling to iteratively refine molecular networks governing cell state. Focusing on lung and liver macrophages—where dozens of genes can be perturbed in hundreds of thousands of cells in vivo—this project will develop probabilistic models that generate actionable hypotheses on regulators of cell state, gene expression, and environmental sensing. Preliminary data show the approach is sensitive, scalable, and hypothesis-generating. Recursive experiments will validate predictions and refine models. This "experiment-in-the-loop" strategy aims to reveal upstream regulators of differentiation and serve as a blueprint for decoding diverse cell populations in health and disease.