BME PhD Dissertation Defense: James Robson

Title: "Decoding sequence-structure-function relationships of programmable RNA sensors"

Advisory Committee: Alexander Green, PhD – BU BME (Research Advisor) Catherine Klapperich, PhD – BU BME (Chair) Mary Dunlop, PhD – BU BME Mo Khalil, PhD – BU BME Trevor Siggers, PhD – BU Biology

Abstract: RNA can simultaneously encode genetic information and fold into precise structures that perform regulatory, catalytic, and sensing functions. Synthetic biologists have harnessed this principle to engineer programmable RNA sensors that switch gene expression on or off upon recognizing specific RNA targets. However, designing high-performance RNA sensors remains time-intensive and poorly predictable, as the sequence-structure-function relationships governing sensor behavior are incompletely understood. Here we show that these barriers can be overcome by combining robotic laboratory automation, target-aware machine learning, and community-driven structural exploration. An automated liquid-handling pipeline enables high-throughput plasmid assembly with performance equivalent to manual methods and substantially reduced hands-on time. VISTA, a machine learning framework that jointly models sensor and target RNA biophysical properties, substantially outperforms existing design tools and successfully guides RNA sensor design against SARS-CoV-2 RNA. A community-driven riboregulator study reveals that distributed designers access structural solutions beyond existing paradigms, and that co-transcriptional folding kinetics are significant determinants of sensor performance. Ultimately, these principles are combined to develop novel sequence-independent transcriptional regulatory architectures, Sequence-Independent Enhanced Synthetic Transcriptional Attenuators (SIESTAs), and Fully Independent Engineered Synthetic Transcriptional Activators (FIESTAs). SIESTA-FIESTA represent two novel transcriptional regulators that overcome the sequence constraints of prior transcriptional riboregulators by physically decoupling target recognition from the intrinsic terminator hairpin, enabling programmable attenuation and activation in response to any arbitrary RNA sequence with up to 400-fold dynamic range. Together, these advances establish a comprehensive framework for decoding sequence-structure-function relationships in programmable RNA sensors, with direct implications for pathogen diagnostics, genetic circuit construction, and cellular engineering. Integrating automation, machine learning, and community science can substantially accelerate the design-build-test-learn cycle in RNA synthetic biology.