Our mechanistic studies are designed to understand the biological process, the pathophysiology of the disease or the mechanism behind an effect.
With these projects, our aim is to gain a new and better understanding of the biological mechanisms behind the difference between indolent and lethal tumours. We will focus on tumour heterogeneity and gene expression, but also investigate in-depth why our biomarkers work as well as they do.
Learning from Deep Learning
We are thrilled to launch our project “Learning from Deep Learning” in 2021, with the goal to open the black box of deep learning and identify the most important parameters that the network uses so well to prognosticate cancer. This project will assess several different biomarkers on a cell-by-cell basis using several new technologies recently developed at ICGI.
Characterising image regions
Through this, we will characterise the image regions defined by the network as holding prognostic information and relate the network predictions to biomedical features. We have shown how tumour heterogeneity impacts biomarkers in prostate cancer (BJC 2018) and will perform similar studies to describe heterogeneity in the endometrial, lung- and colorectal cancers. We have ongoing studies on mRNA expression of a large number of candidate genes in prostate and colorectal cancer, and through AI-guided immunohistochemistry, we are further studying the protein expression of the same candidate genes.