Project Reference: ICS27-BIO-SB1

Project title: Using AI-based methods for rational design of mRNA coding sequences

Project outline

Our ability to express and purify proteins in other species is of critical importance for both research and biomedical applications such as the production of human insulin in E. coli. However, whether a human protein is expressed well in bacteria depends on a range of factors, including its tRNA demand as determined by its coding sequence. A commonly used strategy to increase expression is to replace rare codons by more common synonymous codons but this approach is unreliable and sometimes has detrimental consequences. For instance, it may alter co-translational protein folding, leading to protein aggregation and difficulties purifying the target protein.

In this project, you will address this challenge by developing AI-based approaches to guide sequence design. Our aim is to reliably enhance heterologous protein expression in E. coli, while also gaining new insights into molecular mechanisms that currently limit protein production.

To achieve this, you will:

  • Benchmark existing foundation models for heterologous protein expression prediction in E. coli.
  • Build a framework that optimises coding sequences for expression in E. coli.
  • Validate your model(s) by performing wet-lab experiments to measure protein output.

The project combines expertise in the Bornelöv lab and at Qkine in predicting and measuring protein output. You will have access to state-of-the-art foundation models, challenging expression targets, and wet-lab validation assays. The project will also use in-house data previously collected at Qkine, combined with a publicly available dataset comprising 244,000 partial coding sequences expressed in E. coli. If needed, you will also be able to generate your own high-throughput data to use for model training. To complement data collected in E. coli, transfer learning will enable you to use insights obtained in other species. For this, you will use both in-house and publicly available foundation models such as Evo2, ESM2 and Encodon.

Principal Supervisor

Dr Susanne Bornelöv

Email address: smb208@cam.ac.uk

Department / Partner Institute: Biochemistry

Website: https://www.sblab.uk

Industrial Partner

Qkine Ltd

BBSRC strategic theme

Transformative technologies