For realizing large-scale synthetic genetic systems, modules need to be designed to specs
One of the major indicators we look at is robustness to disturbances
Based on the quantitative models, we design and test multiple versions of the same module
For every design, we assess their performance considering the design specifications
We rank different designs and assess their behavior with respect to model predictions
Considering the data and the predictions, we revise and refine our quantitative models
Cellular processes share building blocks and machinery that have limited availability
Competition for shared resources introduces coupling among "independent" processes
By accounting for the limited availability of resources, we can develop predictive models
Based on models explicitly accounting for coupling, we can predict emergent behaviors
Our approach is validated experimentally in the lab using our customized Chi.Bio platform
Leveraging the predictive models guides the redesign of components for increased modularity
Biological systems are complex nonlinear systems, thus we rely on data about them
Leveraging data about their behavior, we use machine learning to obtain a model of cells
Model predictive control enables us to steer cellular behavior towards our goal
Relying optimization methods, we can prioritize some objectives over others
Our approach is validated experimentally in the lab using our customized Chi.Bio platform
All the data we collect forms the foundation of learning the biophsycal system dynamics