Accelerating organism engineering through the application of AI

The engineering of biological systems requires the integration of experimental inputs and output
data across different levels of biodesign and implementation. These can readily be classed into
either intrinsic factors, which relate to the biology, or external factors that relate to physical
conditions such as media and temperature. When considering the design and implementation of a
microbial biosynthetic pathway, they do not follow a set of prior-known rules, but we can break
them down into 3 classes: firstly, the genetic design of the pathway itself can be modulated by
enzyme identity, and genetic elements used for its instantiation, including promoter, 5'UTR, RBS,
gene order, copy number. Secondly, the host metabolic network, which usually provides the input
precursor and energy for the biosynthetic process and the metabolic load that is exerted on the
system by the new pathway. Finally, the external factors such as cell strain, media composition, time
and temperature.

This project will focus on extending the Robot Synthetic Biologist to more ambitious biosynthetic
pathways for the production of high value chemicals in E. coli. The first approach will be to develop
CRISPR based tools that can modulate the host metabolic network, thus providing the system with
interactive nodes that can thus learn how to optimise the host for the specific bioproduction task.
Secondly, it will integrate Artificial Metabolic Networks (AMN), which are a white box neural net
implementation of the E. coli Flux Balance Analysis (FBA) model of metabolic reactions. Thirdly,
these will be implemented in biosynthetic pathways: initially for Violacein, a model 5-gene
biosynthetic pathway with a pigment output and subsequently for butyl acetate, a volatile flavour
compound that naturally accumulates external to the cell and can easily be captured using a solvent
overlay.