Querying and exploiting the spatial colorectal tumor ecosystem for novel biological, diagnostic and therapeutic insights

Inter and intra-tumor heterogeneity complicates cancer diagnostics and therapeutics. The use of machine learning on simple tissue stains could already reveal and quantify such heterogeneity, and such algorithms are deployed in diagnostics. We have identified single cell determinants of heterogeneity in colorectal cancer and have performed spatial analyses to address these in situ. To exploit this knowledge diagnostically and therapeutically, we aim to integrate the data from different modalities to reduce the complex data in simple tissue measurements feasible in diagnostics, discriminate macrostructures and identify cell communities and their central players by graph analyses. We will use these parameters in predictive models for stratification and prognosis. The project results will advance KULeuven’s portfolio in three ways: 1) predictive models and macrostructure features can be further tested and validated for predictions of outcome or treatment effect, 2) cell cell interactions and their heterogeneity will help prioritize therapeutic regimens, and 3) consolidation of this pipeline will be applicable to exploit deep spatial datasets in which KULeuven researchers have largely invested.

Keywords:Colorectal cancer, Pathology, Pattern recognition, Diagnostics, Cancer biology

Disciplines:Cancer diagnosis, Gastro-enterology, Computational biomodelling and machine learning