< Back to previous page

Project

Improving immuno-oncology diagnostics by high-dimensional correlation analysis of histopathology and proteomics data in breast cancer samples

Breast cancer remains the most frequent malignancy in women and is characterized by complex underlying molecular mechanisms. While major steps have been taken to manage the disease, patients still die, explaining why more research is needed to better characterize, diagnose and treat patients with tailored-made approaches. Defining the right pathological diagnosis remains a key feature in this process as it largely defines the right prognosis and therapeutic strategy. In the current clinical routine, pathological diagnostics are still largely based on morphological features, for which hematoxylin and eosin (H&E)-stained cancer tissue slides remain key. In contrast to most molecular assays, tissue sections maintain the original tissue architecture and as such contain information regarding cell structure/shape and their position within their original spatial environment – features that are becoming increasingly relevant . Next to H&E analysis, a handful of molecular markers (e.g. PgR, ER, HER2, and Kl-67) are evaluated using immunohistochemistry (IHC) using slow, one-marker-at-the-time techniques. We recently established a revolutionary multiplex IHC analysis method (MILAN) allowing the evaluation of 70+ markers in a single tissue section thereby enabling us to map and dissect the tissue architecture at unseen resolution. However, in the current MILAN assessment, the morphological features as defined by H&E are not integrated yet, even though this could add significant information to evaluate the tissue. In this project, I will therefore develop computational tools to integrate insights obtained by both technologies and correlate the observations to clinical features (e.g. response to therapy). As test set, I will use data collected through H&E and MILAN from samples collected in the context of breast specific biomarker studies such as the window of opportunity trial BIOKEY in which breast cancer patients with diverse surrogate molecular subtypes have been treated with immune-check point inhibitory immunotherapy in combination or not with standard chemotherapy. To achieve this goal, I will make use of state-of-the-art deep learning and artificial intelligence methodologies through which I will develop novel bioinformatics pipelines that can evaluate how features from H&E correlate to those observed with MILAN and vice versa, and determine how the combination of both methods could improve diagnostics. This project will therefore generate one of the most detailed pathological frameworks in breast cancer, which, by linking my observations to responsiveness to therapy, could become highly instrumental in therapy selection as well.

Date:4 Sep 2020 →  22 Feb 2022
Keywords:Breast cancer, immuno-oncology diagnostics
Disciplines:Single-cell data analysis, Structural bioinformatics and computational proteomics, Data visualisation and high-throughput image analysis, Computational biomodelling and machine learning, Bioinformatics of disease
Project type:PhD project