A Self-Supervised AI Approach to Biological Discovery in Lung Cancer
Supervisors:
Prof John Le Quesne, School of Cancer Sciences
Dr Ke Yuan, School of Cancer Sciences
Summary:
Histology images of lung cancer contain a universe of biological information that dictates patient survival, but much of it remains invisible to the human eye. This project will use cutting-edge self-supervised AI to decode this world at single cell resolution. Building on our recently published work in Nature Communications, you will adapt our powerful, annotation-free deep learning framework to analyse individual cells across unique multimodal datasets, including H&E, high-plex immunofluorescence, and spatial transcriptomics. Your aims will be to discover novel trajectories of tumour cell plasticity and create a detailed spatial map of the tumour microenvironment, identifying the functional states of immune cells and fibroblasts. The ultimate goal is to train a transformative AI model capable of predicting these complex cellular states directly from a standard H&E slide alone, with huge potential for future diagnostic tools.
Hosted between the labs of a clinical pathologist (Prof. Le Quesne) and a machine learning scientist (Dr. Yuan), this project offers an unparalleled interdisciplinary training environment at the University of Glasgow and CRUK Scotland Institute. We are seeking a highly motivated candidate to pioneer the next generation of computational pathology.