Abstract
Engineering predictable and robust gene-regulatory networks requires a quantitative understanding of how their dynamics depend on genetic design and environmental conditions. To investigate these relationships, we introduce BARRACUDA, a high-throughput platform that combines combinatorial assembly and programmable perturbations with lineage-resolved time-lapse imaging and optical pooled screening. Experiments using this platform follow approximately one million cell lineages, collectively sampling tens of thousands of circuit designs. We used BARRACUDA to investigate how the robustness of circuit behavior depends on molecular components and their interactions, combining quantitative dynamic measurements with precise genetic and environmental control. Antithetic integral feedback provided a model system linking a defined molecular architecture to theoretical predictions of robust adaptation. Through a hierarchy of increasingly complex libraries, we characterized individual components and used those measurements to model and interpret the dynamics of complete feedback circuits. Circuits built from wild-type inhibitory proteins rejected disturbances but also exhibited sustained overadaptation and fast disturbance-induced oscillations unexplained by a reduced control model. These discrepancies motivated a screen of de novo binders that retained inhibitory function while reducing host-stress response. Selected de novo binders supported strong feedback regulation and robust performance in the rebuilt circuits. Across more than 20,000 recovered feedback-circuit variants, binder identity and controller expression organized regimes of strong regulation and sustained oscillation. Counterintuitively, disturbances sometimes improved compensation rather than compromising it, consistent with nonlinear changes in feedback engagement. Together, these results show how structured libraries and massively parallel measurements of single-cell dynamics can test mechanistic theories across scales of circuit complexity, reveal missing biological interactions, and map the design space of gene-regulatory networks.