James McCabe

Hi, I'm James, a second-year PhD student working where evolutionary biology, structural biology and artificial intelligence meet.

Cytokines are the proteins cells use to talk to one another, and that conversation sits at the centre of many diseases: viruses and other pathogens manipulate cytokines to hijack immune signalling, inherited faults in the same proteins cause immunodeficiency and chronic inflammation, and tumours subvert them to escape immune control.

My research asks how evolutionary and structural context, combined with state of the art AI, can elevate drug design for immune signalling. I characterise the evolutionary landscape of immunity across 500 million years of vertebrate phylogeny, then measure protein language models against it. These models learn what a protein can tolerate, but from databases that cover some branches of life far better than others. Decomposing which signals they absorb during pretraining shows where they are blind, whether that blindness is predictable, and how to improve them for therapeutic design. The aim is a map of immune signalling: the sites pathogens exploit, the ones that drive disease when they go wrong, and the ones worth engineering into better therapeutics.

Alongside the PhD I enjoy creating open-source software, for example SCEPTR, a tool for characterising transcriptomes which I built during my time as a computational biologist at the Institute of Microbiology of the Czech Academy of Sciences. It lives on my GitHub.