Integrative pathology modelling of epithelial-to-mesenchymal transformation in cancer

The epithelial to mesenchymal transition (EMT) is a key process in cancer that leads to tumours spreading from their original site and resisting various treatments. Tumours can be classified into three states within this process based on their genetic data.

Identifying these states in tumour tissue that is routinely collected during surgery would enable us to predict whether a cancer patient is at high risk for developing metastases and responses to therapies.

While artificial intelligence approaches have revolutionised our ability to detect patterns of interest in cancer tissue images in recent years, it has not yet been shown whether EMT states can be identified in tumour images.

Furthermore, how cells in different EMT states organise spatially and interact with other non-cancer cells in their environment is not known. We will predict EMT states from pathology images by developing deep learning based methods. We will also analyse the spatial arrangement of the tissue in these different states from the tumour images.

This will be carried out using graph-based methods and will enable us to further characterise EMT.

Overall, this project will provide a greater understanding of how cancer cells transform and progress towards more advanced stages by interacting with their environment.