Chemical Engineering Seminar
DNA is Nature's programming language and its sequences control how organisms sense their environment, perform decision-making, and catalytically produce valuable chemical products. The Salis lab uses statistical thermodynamics and kinetics to develop predictive biophysical models of gene expression that are systematically validated across thousands of experiments. These sequence-to-function models are then combined with computational optimization to automatically design non-natural genetic systems with targeted functions inside cells, reprogramming an organism to sense new chemicals, process signals, and maximally synthesize targeted chemical products. Collectively, our web-accessible design methods have been used by 6000 world-wide researchers to design over 50,000 synthetic genetic parts and systems, highlighting the growing expansion of automated expert systems that have greatly accelerated the design of non-natural DNA sequences within large genetic systems.
In this talk, we present the Riboswitch Calculator, a physics-based sequence-structure-function model that automatically designs non-natural riboswitch sensors (ligand-binding, shape-changing mRNAs) from any ligand-binding RNA aptamer. Using results from 52 automatically designed synthetic riboswitches, we show how changing the aptamer's switching free energy and ligand-binding affinity, its co-transcriptional ligand binding, the ligand and mRNA concentrations, and the surrounding mRNA sequences controlled riboswitch activation ratios, up to 383-fold. Second, we show how to rationally design analog genetic circuits to process chemical signals, using a new approach to experimentally determine the binding free energies of six orthogonal transcription factors. Using the model, we present a new dimensionless unit, the Ptashne number, that encapsulates several co-dependent variables into a single prediction. Finally, we demonstrate a new approach to metabolic pathway engineering that efficiently maps the sequence-expression-activity space for a multi-enzyme metabolic pathway and correctly predicts optimally balanced enzyme expression levels. Overall, our work illustrates how the same chemical engineering principles can be applied to both understand and design a wide range of biological parts with desired functions.
One Model to Rule them All, One Model to Design Them,
And in the Genome where They Belong, One Model to Predict What Binds Them