What is the problem we are trying to solve?
The fundamental thought process behind this project is exploration beyond conventionally-used methods and how these can be made accessible to the everyday researcher. E. coli emerged as a feasible chassis organism in the early days of synthetic biology, but this led to a sunken cost feedback loop in which research investigating E. coli was favored disproportionately over looking for new and better chassis. Over time, simply because the body of information on E. coli aggregated to be too large for any investigative effort outside it to be practically unfavorable, synthetic biology fell into a rut of having a singular chassis for the overwhelming majority of its workflow.
How are we solving it?
As students, driven by an interest in synthetic biology, we have relatively more freedom to question the fundamentals of the nature of our work, explore novel chassis and develop better genetic engineering techniques. Our laboratory efforts have been driven towards characterizing Lactococcus lactis better, which requires significantly less effort with regards to downstream processing as there is no additional purification required to remove an endotoxin layer. The most significant hurdles encountered when working with a new organism is an absence of genetic and regulatory information, making circuit design and fine-tuning gene expression challenging and extremely labor intensive. Our wet lab team constructed a library of randomized RBS sequences characterized by their expression levels, providing a solid base of initial information for anyone using this chassis.
Our RBS Predictor-Optimizer: Synthopedia!
Our predictor-optimizer addresses the issue of fine tuning protein expression on a broader level, and contributes to the groundwork to an in silico solution to better control over expression dynamics. Harnessing machine learning principles over a wider set of parameters, including both thermodynamic and biophysical data points, our team has designed a machine learning model for protein expression prediction and optimization, leveraging which, we are able to predict expression levels for RBS sequences as well as optimize them for a particular expression level. The results of this model have been cross-tested against state of the art models.