Dr Harvinder Virk

CASE: Exploring the utility of AI-designed binders for the secretome and discovery proteomics

University of Registration: University of Leicester

Non-academic partner: YCharOS Inc.

Secondary Supervisor(s): Dr Carl Laflamme (YCharOS Inc.), Professor Martin Dyer, Professor Jeffrey E. Lee (YCharOS Inc.), Dr Michael Biddle

Project Outline

Recent work has shown that AI can design small proteins, called mini-binders, that stick tightly to a chosen target. There is real excitement that these could transform research, diagnostics and therapeutics. But as with antibodies, which are already established research and clinical tools, binding tightly is only half the problem. A reagent is only useful if it binds the right target and not the thousands of others in a real biological sample. Nobody has yet shown that AI-designed binders can be made selective at that level, at scale.

This PhD will find out, focusing on the proteins that cells secrete. The secretome is one of the least well served parts of the human proteome - many secreted proteins have no reliable reagent at all - and it is where new tools would open the most doors, from discovery science through to future diagnostic tests.

You will use existing proteomic datasets to choose which secreted proteins to target, design binders against them with our partners in Canada, produce them, and then test how selectively they behave. Selectivity is measured against knockout cell lines, to the standards set by YCharOS, the international consortium Leicester belongs to, which has already tested thousands of antibodies with industry. Where good antibodies already exist you will benchmark the new binders against them; where none exist, you will be building the first trustworthy tool for that protein.

You will be based in the UK's only YCharOS laboratory, and will spend at least three months with YCharOS Inc. in Canada - training in AI binder design with Professor Jeff Lee in Toronto, and in selectivity testing with Dr Carl Laflamme in Ottawa.

You will finish with an unusual combination of skills - computational design, protein production, and rigorous experimental validation - and an openly shared toolbox that laboratories anywhere can pick up and use.