Data-driven design of complex functional proteins

Second Supervisor: Chris Wells Wood, School of Biological Sciences, [Email Address Removed]

Proteins are the building blocks of life as they are responsible for all the functions in a cell. Rapid advances in computational protein design now means that, given enough compute, it is relatively routine to design amino-acid sequences that fold into a desired 3D structure. As a result, it has been said that protein design has “come of age” [1]. However, in order unlock the full potential of protein design, structure alone is not enough. We must develop tools to guide the design process in order to create complex functional proteins with properties tuned to their intended application.

Here we propose to improve the reliability of the design process and target specific functions using a machine learning approach, specifically deep-generative models, such as Bayesian deep learning and variational autoencoders. We aim to create an algorithm that can generate amino-acid sequences with targeted functional properties. The design strategy will be tested at scale using state-of-the-art experimental automation, and applied to tackle real-world challenges in synthetic biology.

The ideal candidate either has a background in mathematics, computer science, bioinformatics, statistics, physics or related fields; or biochemistry/molecular biology but can demonstrate a strong mathematical background. He/she is strongly motivated to develop a competitive profile in protein engineering and machine learning and likes to work in a fast pace environment. We put a strong emphasis on reproducible research; the candidate will receive training in advanced research software engineering and in workflows for data analyses. The ideal candidate is expected to have good knowledge of either Python or C.

www.stracquadaniolab.org and www.wellswoodresearchgroup.com